diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml new file mode 100644 index 000000000..c3ad89568 --- /dev/null +++ b/.github/workflows/build.yml @@ -0,0 +1,93 @@ +name: Build Probackup + +on: + push: + branches: + - "**" + # Runs triggered by pull requests are disabled to prevent executing potentially unsafe code from public pull requests + # pull_request: + # branches: + # - main + + # Allows you to run this workflow manually from the Actions tab + workflow_dispatch: + +jobs: + + build-win2019: + + runs-on: + - windows-2019 + + env: + zlib_dir: C:\dep\zlib + + steps: + + - uses: actions/checkout@v2 + + - name: Install pacman packages + run: | + $env:PATH += ";C:\msys64\usr\bin" + pacman -S --noconfirm --needed bison flex + + - name: Make zlib + run: | + git clone -b v1.2.11 --depth 1 https://github.com/madler/zlib.git + cd zlib + cmake -DCMAKE_INSTALL_PREFIX:PATH=C:\dep\zlib -G "Visual Studio 16 2019" . + cmake --build . --config Release --target ALL_BUILD + cmake --build . --config Release --target INSTALL + copy C:\dep\zlib\lib\zlibstatic.lib C:\dep\zlib\lib\zdll.lib + copy C:\dep\zlib\bin\zlib.dll C:\dep\zlib\lib + + - name: Get Postgres sources + run: git clone -b REL_14_STABLE https://github.com/postgres/postgres.git + + # Copy ptrack to contrib to build the ptrack extension + # Convert line breaks in the patch file to LF otherwise the patch doesn't apply + - name: Get Ptrack sources + run: | + git clone -b master --depth 1 https://github.com/postgrespro/ptrack.git + Copy-Item -Path ptrack -Destination postgres\contrib -Recurse + (Get-Content ptrack\patches\REL_14_STABLE-ptrack-core.diff -Raw).Replace("`r`n","`n") | Set-Content ptrack\patches\REL_14_STABLE-ptrack-core.diff -Force -NoNewline + cd postgres + git apply -3 ../ptrack/patches/REL_14_STABLE-ptrack-core.diff + + - name: Build Postgres + run: | + $env:PATH += ";C:\msys64\usr\bin" + cd postgres\src\tools\msvc + (Get-Content config_default.pl) -Replace "zlib *=>(.*?)(?=,? *#)", "zlib => '${{ env.zlib_dir }}'" | Set-Content config.pl + cmd.exe /s /c "`"C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvarsall.bat`" amd64 && .\build.bat" + + - name: Build Probackup + run: cmd.exe /s /c "`"C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\VC\Auxiliary\Build\vcvarsall.bat`" amd64 && perl .\gen_probackup_project.pl `"${{ github.workspace }}`"\postgres" + + - name: Install Postgres + run: | + cd postgres + src\tools\msvc\install.bat postgres_install + + - name: Install Testgres + run: | + git clone -b no-port-for --single-branch --depth 1 https://github.com/postgrespro/testgres.git + pip3 install psycopg2 ./testgres + + # Grant the Github runner user full control of the workspace for initdb to successfully process the data folder + - name: Test Probackup + run: | + icacls.exe "${{ github.workspace }}" /grant "${env:USERNAME}:(OI)(CI)F" + $env:PATH += ";${{ github.workspace }}\postgres\postgres_install\lib;${{ env.zlib_dir }}\lib" + $Env:LC_MESSAGES = "English" + $Env:PG_CONFIG = "${{ github.workspace }}\postgres\postgres_install\bin\pg_config.exe" + $Env:PGPROBACKUPBIN = "${{ github.workspace }}\postgres\Release\pg_probackup\pg_probackup.exe" + $Env:PG_PROBACKUP_PTRACK = "ON" + If (!$Env:MODE -Or $Env:MODE -Eq "basic") { + $Env:PG_PROBACKUP_TEST_BASIC = "ON" + python -m unittest -v tests + python -m unittest -v tests.init_test + } else { + python -m unittest -v tests.$Env:MODE + } + diff --git a/.gitignore b/.gitignore index 4fc21d916..97d323ceb 100644 --- a/.gitignore +++ b/.gitignore @@ -21,6 +21,9 @@ # Binaries /pg_probackup +# Generated translated file +/po/ru.mo + # Generated by test suite /regression.diffs /regression.out @@ -34,9 +37,6 @@ # Extra files /src/pg_crc.c -/src/datapagemap.c -/src/datapagemap.h -/src/logging.h /src/receivelog.c /src/receivelog.h /src/streamutil.c @@ -44,3 +44,24 @@ /src/xlogreader.c /src/walmethods.c /src/walmethods.h +/src/instr_time.h + +# Doc files +/doc/*html + +# Docker files +/docker-compose.yml +/Dockerfile +/Dockerfile.in +/make_dockerfile.sh +/backup_restore.sh + +# Packaging +/build +/packaging/pkg/tarballs/pgpro.tar.bz2 +/packaging/repo/pg_probackup +/packaging/repo/pg_probackup-forks + +# Misc +.python-version +.vscode diff --git a/.travis.yml b/.travis.yml index 35b49ec5b..074ae3d02 100644 --- a/.travis.yml +++ b/.travis.yml @@ -1,7 +1,102 @@ -sudo: required +os: linux -services: -- docker +dist: jammy + +language: c + +cache: ccache + +addons: + apt: + packages: + - sudo + - libc-dev + - bison + - flex + - libreadline-dev + - zlib1g-dev + - libzstd-dev + - libssl-dev + - perl + - libperl-dev + - libdbi-perl + - cpanminus + - locales + - python3 + - python3-dev + - python3-pip + - libicu-dev + - libgss-dev + - libkrb5-dev + - libxml2-dev + - libxslt1-dev + - libldap2-dev + - tcl-dev + - diffutils + - gdb + - gettext + - lcov + - openssh-client + - openssh-server + - libipc-run-perl + - libtime-hires-perl + - libtimedate-perl + - libdbd-pg-perl + +before_install: + - sudo travis/before-install.sh + +install: + - travis/install.sh + +before_script: + - sudo travis/before-script.sh + - travis/before-script-user.sh script: -- docker run -v $(pwd):/tests --rm centos:7 /tests/travis/backup_restore.sh + - travis/script.sh + +notifications: + email: + on_success: change + on_failure: always + +# Default MODE is basic, i.e. all tests with PG_PROBACKUP_TEST_BASIC=ON +env: + - PG_VERSION=16 PG_BRANCH=master PTRACK_PATCH_PG_BRANCH=master + - PG_VERSION=15 PG_BRANCH=REL_15_STABLE PTRACK_PATCH_PG_BRANCH=REL_15_STABLE + - PG_VERSION=14 PG_BRANCH=REL_14_STABLE PTRACK_PATCH_PG_BRANCH=REL_14_STABLE + - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=REL_13_STABLE + - PG_VERSION=12 PG_BRANCH=REL_12_STABLE PTRACK_PATCH_PG_BRANCH=REL_12_STABLE + - PG_VERSION=11 PG_BRANCH=REL_11_STABLE PTRACK_PATCH_PG_BRANCH=REL_11_STABLE + - PG_VERSION=10 PG_BRANCH=REL_10_STABLE + - PG_VERSION=9.6 PG_BRANCH=REL9_6_STABLE + - PG_VERSION=9.5 PG_BRANCH=REL9_5_STABLE + - PG_VERSION=15 PG_BRANCH=REL_15_STABLE PTRACK_PATCH_PG_BRANCH=OFF MODE=backup_test.BackupTest.test_full_backup + - PG_VERSION=15 PG_BRANCH=REL_15_STABLE PTRACK_PATCH_PG_BRANCH=OFF MODE=backup_test.BackupTest.test_full_backup_stream +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=REL_13_STABLE MODE=backup +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=REL_13_STABLE MODE=catchup +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=REL_13_STABLE MODE=checkdb +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=OFF MODE=compression +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=OFF MODE=delta +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=OFF MODE=locking +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=REL_13_STABLE MODE=merge +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=OFF MODE=option +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=OFF MODE=page +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=REL_13_STABLE MODE=ptrack +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=REL_13_STABLE MODE=replica +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=OFF MODE=retention +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=REL_13_STABLE MODE=restore +# - PG_VERSION=13 PG_BRANCH=REL_13_STABLE PTRACK_PATCH_PG_BRANCH=REL_13_STABLE MODE=time_consuming + +jobs: + allow_failures: + - if: env(PG_BRANCH) = master + - if: env(PG_BRANCH) = REL9_5_STABLE +# - if: env(MODE) IN (archive, backup, delta, locking, merge, replica, retention, restore) + +# Only run CI for master branch commits to limit our travis usage +#branches: +# only: +# - master + diff --git a/Documentation.md b/Documentation.md deleted file mode 100644 index 544e936a7..000000000 --- a/Documentation.md +++ /dev/null @@ -1,1331 +0,0 @@ -# pg_probackup - -pg_probackup is a utility to manage backup and recovery of PostgreSQL database clusters. It is designed to perform periodic backups of the PostgreSQL instance that enable you to restore the server in case of a failure. pg_probackup supports PostgreSQL 9.5 or higher. - -Current version - 2.1.4 - -1. [Synopsis](#synopsis) -2. [Versioning](#versioning) -3. [Overview](#overview) - * [Limitations](#limitations) - -4. [Installation and Setup](#installation-and-setup) - * [Initializing the Backup Catalog](#initializing-the-backup-catalog) - * [Adding a New Backup Instance](#adding-a-new-backup-instance) - * [Configuring the Database Cluster](#configuring-the-database-cluster) - * [Setting up STREAM Backups](#setting-up-stream-backups) - * [Setting up Continuous WAL Archiving](#setting-up-continuous-wal-archiving) - * [Setting up Backup from Standby](#backup-from-standby) - * [Setting up Cluster Verification](#setting-up-cluster-verification) - * [Setting up PTRACK Backups](#setting-up-ptrack-backups) - -5. [Command-Line Reference](#command-line-reference) - * [Commands](#commands) - * [version](#version) - * [help](#help) - * [init](#init) - * [add-instance](#add-instance) - * [del-instance](#del-instance) - * [set-config](#set-config) - * [show-config](#show-config) - * [show](#show) - * [backup](#backup) - * [restore](#restore) - * [checkdb](#checkdb) - * [validate](#validate) - * [merge](#merge) - * [delete](#delete) - * [archive-push](#archive-push) - * [archive-get](#archive-get) - - * [Options](#options) - * [Common Options](#common-options) - * [Backup Options](#backup-options) - * [Restore Options](#restore-options) - * [Checkdb Options](#checkdb-options) - * [Recovery Target Options](#recovery-target-options) - * [Retention Options](#retention-options) - * [Logging Options](#logging-options) - * [Connection Options](#connection-options) - * [Compression Options](#compression-options) - * [Archiving Options](#archiving-options) - * [Remote Mode Options](#remote-mode-options) - * [Replica Options](#replica-options) - -6. [Usage](#usage) - * [Creating a Backup](#creating-a-backup) - * [Verifying a Cluster](#verifying-a-cluster) - * [Validating a Backup](#validating-a-backup) - * [Restoring a Cluster](#restoring-a-cluster) - * [Performing Point-in-Time (PITR) Recovery](#performing-point-in-time-pitr-recovery) - * [Using pg_probackup in the Remote Mode](#using-pg_probackup-in-the-remote-mode) - * [Running pg_probackup on Parallel Threads](#running-pg_probackup-on-parallel-threads) - * [Configuring pg_probackup](#configuring-pg_probackup) - * [Managing the Backup Catalog](#managing-the-backup-Catalog) - * [Configuring Backup Retention Policy](#configuring-backup-retention-policy) - * [Merging Backups](#merging-backups) - * [Deleting Backups](#deleting-backups) - -7. [Authors](#authors) -8. [Credits](#credits) - - -## Synopsis - -`pg_probackup version` - -`pg_probackup help [command]` - -`pg_probackup init -B backup_dir` - -`pg_probackup add-instance -B backup_dir -D data_dir --instance instance_name` - -`pg_probackup del-instance -B backup_dir --instance instance_name` - -`pg_probackup set-config -B backup_dir --instance instance_name [option...]` - -`pg_probackup show-config -B backup_dir --instance instance_name [--format=format]` - -`pg_probackup show -B backup_dir [option...]` - -`pg_probackup backup -B backup_dir --instance instance_name -b backup_mode [option...]` - -`pg_probackup restore -B backup_dir --instance instance_name [option...]` - -`pg_probackup checkdb -B backup_dir --instance instance_name [-D data_dir] [option...]` - -`pg_probackup validate -B backup_dir [option...]` - -`pg_probackup merge -B backup_dir --instance instance_name -i backup_id [option...]` - -`pg_probackup delete -B backup_dir --instance instance_name { -i backup_id | --delete-wal | --delete-expired | --merge-expired }` - -`pg_probackup archive-push -B backup_dir --instance instance_name --wal-file-path %p --wal-file-name %f [option...]` - -`pg_probackup archive-get -B backup_dir --instance instance_name --wal-file-path %p --wal-file-name %f` - - -## Versioning - -pg_probackup is following [semantic](https://semver.org/) versioning. - -## Overview - -As compared to other backup solutions, pg_probackup offers the following benefits that can help you implement different backup strategies and deal with large amounts of data: - -- Incremental backup: page-level incremental backup allows you to save disk space, speed up backup and restore. With three different incremental modes you can plan the backup strategy in accordance with your data flow -- Validation: Automatic data consistency checks and on-demand backup validation without actual data recovery -- Verification: On-demand verification of PostgreSQL instance via dedicated command `checkdb` -- Retention: Managing backups in accordance with retention policies - Time and/or Redundancy based, with two retention methods: `delete expired` and `merge expired` -- Parallelization: Running backup, restore, merge, delete, verificaton and validation processes on multiple parallel threads -- Compression: Storing backup data in a compressed state to save disk space -- Deduplication: Saving disk space by not copying the not changed non-data files ('_vm', '_fsm', etc) -- Remote operations: Backup PostgreSQL instance located on remote machine or restore backup on it -- Backup from replica: Avoid extra load on the master server by taking backups from a standby -- External directories: Add to backup content of directories located outside of the PostgreSQL data directory (PGDATA), such as scripts, configs, logs and pg_dump files -- Backup Catalog: get list of backups and corresponding meta information in `plain` or `json` formats - -To manage backup data, pg_probackup creates a `backup catalog`. This is a directory that stores all backup files with additional meta information, as well as WAL archives required for point-in-time recovery. You can store backups for different instances in separate subdirectories of a single backup catalog. - -Using pg_probackup, you can take full or incremental [backups](#creating-a-backup): - -- FULL backups contain all the data files required to restore the database cluster. -- Incremental backups only store the data that has changed since the previous backup. It allows to decrease the backup size and speed up backup and restore operations. pg_probackup supports the following modes of incremental backups: - - DELTA backup. In this mode, pg_probackup reads all data files in the data directory and copies only those pages that has changed since the previous backup. Note that this mode can impose read-only I/O pressure equal to a full backup. - - PAGE backup. In this mode, pg_probackup scans all WAL files in the archive from the moment the previous full or incremental backup was taken. Newly created backups contain only the pages that were mentioned in WAL records. This requires all the WAL files since the previous backup to be present in the WAL archive. If the size of these files is comparable to the total size of the database cluster files, speedup is smaller, but the backup still takes less space. You have to configure WAL archiving as explained in the section [Setting up continuous WAL archiving](#setting-up-continuous-wal-archiving) to make PAGE backups. - - PTRACK backup. In this mode, PostgreSQL tracks page changes on the fly. Continuous archiving is not necessary for it to operate. Each time a relation page is updated, this page is marked in a special PTRACK bitmap for this relation. As one page requires just one bit in the PTRACK fork, such bitmaps are quite small. Tracking implies some minor overhead on the database server operation, but speeds up incremental backups significantly. - -pg_probackup can take only physical online backups, and online backups require WAL for consistent recovery. So regardless of the chosen backup mode (FULL, PAGE or DELTA), any backup taken with pg_probackup must use one of the following `WAL delivery methods`: - -- [ARCHIVE](#archive-mode). Such backups rely on [continuous archiving](#setting-up-continuous-wal-archiving) to ensure consistent recovery. This is the default WAL delivery method. -- [STREAM](#stream-mode). Such backups include all the files required to restore the cluster to a consistent state at the time the backup was taken. Regardless of [continuous archiving](#setting-up-continuous-wal-archiving) been set up or not, the WAL segments required for consistent recovery are streamed (hence STREAM) via replication protocol during backup and included into the backup files. - -### Limitations - -pg_probackup currently has the following limitations: - -- Creating backups from a remote server is currently not supported on Windows systems. -- The PostgreSQL server from which the backup was taken and the restored server must be compatible by the [block_size](https://www.postgresql.org/docs/current/runtime-config-preset.html#GUC-BLOCK-SIZE) and [wal_block_size](https://www.postgresql.org/docs/current/runtime-config-preset.html#GUC-WAL-BLOCK-SIZE) parameters and have the same major release number. - -## Installation and Setup - -Once you have pg_probackup installed, complete the following setup: - -- Initialize the backup catalog. -- Add a new backup instance to the backup catalog. -- Configure the database cluster to enable pg_probackup backups. - -### Initializing the Backup Catalog -pg_probackup stores all WAL and backup files in the corresponding subdirectories of the backup catalog. - -To initialize the backup catalog, run the following command: - - pg_probackup init -B backup_dir - -Where *backup_dir* is the path to backup catalog. If the *backup_dir* already exists, it must be empty. Otherwise, pg_probackup returns an error. - -pg_probackup creates the backup_dir backup catalog, with the following subdirectories: - -- wal/ — directory for WAL files. -- backups/ — directory for backup files. - -Once the backup catalog is initialized, you can add a new backup instance. - -### Adding a New Backup Instance -pg_probackup can store backups for multiple database clusters in a single backup catalog. To set up the required subdirectories, you must add a backup instance to the backup catalog for each database cluster you are going to back up. - -To add a new backup instance, run the following command: - - pg_probackup add-instance -B backup_dir -D data_dir --instance instance_name [remote_options] - -Where: - -- *data_dir* is the data directory of the cluster you are going to back up. To set up and use pg_probackup, write access to this directory is required. -- *instance_name* is the name of the subdirectories that will store WAL and backup files for this cluster. -- The optional parameters [remote_options](#remote-mode-options) should be used if *data_dir* is located on remote machine. - -pg_probackup creates the *instance_name* subdirectories under the 'backups/' and 'wal/' directories of the backup catalog. The 'backups/*instance_name*' directory contains the 'pg_probackup.conf' configuration file that controls pg_probackup settings for this backup instance. If you run this command with the [remote_options](#remote-mode-options), used parameters will be added to 'pg_probackup.conf'. - -For details on how to fine-tune pg_probackup configuration, see the section [Configuring pg_probackup](#configuring-pg_probackup). - -The user launching pg_probackup must have full access to *backup_dir* directory and at least read-only access to *data_dir* directory. If you specify the path to the backup catalog in the `BACKUP_PATH` environment variable, you can omit the corresponding option when running pg_probackup commands. - ->NOTE: For PostgreSQL >= 11 it is recommended to use [allow-group-access](https://www.postgresql.org/docs/11/app-initdb.html#APP-INITDB-ALLOW-GROUP-ACCESS) feature, so backup can be done by OS user with read-only permissions. - -### Configuring the Database Cluster - -Although pg_probackup can be used by a superuser, it is recommended to create a separate role with the minimum permissions required for the chosen backup strategy. In these configuration instructions, the *backup* role is used as an example. - -To perform [backup](#backup), the following permissions are required: - -For PostgreSQL 9.5: -``` -BEGIN; -CREATE ROLE backup WITH LOGIN; -GRANT USAGE ON SCHEMA pg_catalog TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; -COMMIT; -``` - -For PostgreSQL 9.6: -``` -BEGIN; -CREATE ROLE backup WITH LOGIN; -GRANT USAGE ON SCHEMA pg_catalog TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_xlog_replay_location() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; -COMMIT; -``` - -For PostgreSQL >= 10: -``` -BEGIN; -CREATE ROLE backup WITH LOGIN; -GRANT USAGE ON SCHEMA pg_catalog TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; -GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; -COMMIT; -``` - ->NOTE: In PostgreSQL 9.5 functions `pg_create_restore_point(text)` and `pg_switch_xlog()` can be executed only by superuser role. So during backup of PostgreSQL 9.5 pg_probackup will use them only if backup role is superuser, although it is NOT recommended to run backup under superuser. - -Since pg_probackup needs to read cluster files directly, pg_probackup must be started on behalf of an OS user that has read access to all files and directories inside the data directory (PGDATA) you are going to back up. - -Depending on whether you are plan to take STREAM and/or ARCHIVE backups, PostgreSQL cluster configuration will differ, as specified in the sections below. To back up the database cluster from a standby server or create PTRACK backups, additional setup is required. - -For details, see the sections [Setting up STREAM Backups](#setting-up-stream-backups), [Setting up continuous WAL archiving](#setting-up-continuous-wal-archiving), [Setting up Backup from Standby](#backup-from-standby) and [Setting up PTRACK Backups](#setting-up-ptrack-backups). - -### Setting up STREAM Backups - -To set up the cluster for [STREAM](#stream-mode) backups, complete the following steps: - -- Grant the REPLICATION privilege to the backup role: - - ALTER ROLE backup WITH REPLICATION; - -- In the [pg_hba.conf](https://www.postgresql.org/docs/current/auth-pg-hba-conf.html) file, allow replication on behalf of the *backup* role. -- Make sure the parameter [max_wal_senders](https://www.postgresql.org/docs/current/runtime-config-replication.html#GUC-MAX-WAL-SENDERS) is set high enough to leave at least one session available for the backup process. -- Set the parameter [wal_level](https://www.postgresql.org/docs/current/runtime-config-wal.html#GUC-WAL-LEVEL) to be higher than `minimal`. - -If you are planning to take PAGE backups in STREAM mode, you still have to configure WAL archiving as explained in the section [Setting up continuous WAL archiving](#setting-up-continuous-wal-archiving). - -Once these steps are complete, you can start taking FULL, PAGE, DELTA and PTRACK backups with [STREAM](#stream-mode) WAL mode. - -### Setting up continuous WAL archiving - -Making backups in PAGE backup mode, performing [PITR](#performing-point-in-time-pitr-recovery) and making backups with [ARCHIVE](#archive-mode) WAL delivery mode require [continious WAL archiving](https://www.postgresql.org/docs/current/continuous-archiving.html) to be enabled. To set up continious archiving in the cluster, complete the following steps: - -- Make sure the [wal_level](https://www.postgresql.org/docs/current/runtime-config-wal.html#GUC-WAL-LEVEL) parameter is higher than `minimal`. -- If you are configuring archiving on master, [archive_mode](https://www.postgresql.org/docs/current/runtime-config-wal.html#GUC-ARCHIVE-MODE) must be set to `on`. To perform archiving on standby, set this parameter to `always`. -- Set the [archive_command](https://www.postgresql.org/docs/current/runtime-config-wal.html#GUC-ARCHIVE-COMMAND) parameter, as follows: - - archive_command = 'pg_probackup archive-push -B backup_dir --instance instance_name --wal-file-path %p --wal-file-name %f [remote_options]' - -Where *backup_dir* and *instance_name* refer to the already initialized backup catalog instance for this database cluster and optional parameters [remote_options](#remote-mode-options) should be used to archive WAL to the remote host. For details about all possible `archive-push` parameters, see the section [archive-push](#archive-push). - -Once these steps are complete, you can start making backups with ARCHIVE WAL mode, backups in PAGE backup mode and perform [PITR](#performing-point-in-time-pitr-recovery). - -If you are planning to make PAGE backups and/or backups with [ARCHIVE](#archive-mode) WAL mode from a standby of a server, that generates small amount of WAL traffic, without long waiting for WAL segment to fill up, consider setting [archive_timeout](https://www.postgresql.org/docs/current/runtime-config-wal.html#GUC-ARCHIVE-TIMEOUT) PostgreSQL parameter **on master**. It is advisable to synchronize the value of this setting with pg_probackup option `--archive-timeout`. - ->NOTE: using pg_probackup command [archive-push](#archive-push) for continious archiving is optional. You can use any other tool you like as long as it delivers WAL segments into '*backup_dir*/wal/*instance_name*' directory. If compression is used, it should be `gzip`, and '.gz' suffix in filename is mandatory. - ->NOTE: Instead of `archive_mode`+`archive_command` method you may opt to use the utility [pg_receivewal](https://www.postgresql.org/docs/current/app-pgreceivewal.html). In this case pg_receivewal `-D directory` option should point to '*backup_dir*/wal/*instance_name*' directory. WAL compression that could be done by pg_receivewal is supported by pg_probackup. `Zero Data Loss` archive strategy can be achieved only by using pg_receivewal. - -### Backup from Standby - -For PostgreSQL 9.6 or higher, pg_probackup can take backups from a standby server. This requires the following additional setup: - -- On the standby server, set the parameter [hot_standby](https://www.postgresql.org/docs/current/runtime-config-replication.html#GUC-HOT-STANDBY) to `on`. -- On the master server, set the parameter [full_page_writes](https://www.postgresql.org/docs/current/runtime-config-wal.html#GUC-FULL-PAGE-WRITES) to `on`. -- To perform STREAM backup on standby, complete all steps in section [Setting up STREAM Backups](#setting-up-stream-backups) -- To perform ARCHIVE backup on standby, complete all steps in section [Setting up continuous WAL archiving](#setting-up-continuous-wal-archiving) - -Once these steps are complete, you can start taking FULL, PAGE, DELTA or PTRACK backups with appropriate WAL delivery mode: ARCHIVE or STREAM, from the standby server. - -Backup from the standby server has the following limitations: - -- If the standby is promoted to the master during backup, the backup fails. -- All WAL records required for the backup must contain sufficient full-page writes. This requires you to enable `full_page_writes` on the master, and not to use a tools like pg_compresslog as [archive_command](https://www.postgresql.org/docs/current/runtime-config-wal.html#GUC-ARCHIVE-COMMAND) to remove full-page writes from WAL files. - -### Setting up Cluster Verification - -Logical verification of database cluster requires the following additional setup. Role *backup* is used as an example: - -- Install extension `amcheck` or `amcheck_next` in every database of the cluster: - - CREATE EXTENSION amcheck; - -- To perform logical verification the following permissions are requiared: - -``` -GRANT SELECT ON TABLE pg_catalog.pg_am TO backup; -GRANT SELECT ON TABLE pg_catalog.pg_class TO backup; -GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; -GRANT SELECT ON TABLE pg_catalog.pg_namespace TO backup; -GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; -GRANT EXECUTE ON FUNCTION bt_index_check(oid) TO backup; -GRANT EXECUTE ON FUNCTION bt_index_check(oid, bool) TO backup; -``` - -### Setting up PTRACK Backups - -If you are going to use PTRACK backups, complete the following additional steps: - -- Set the parameter `ptrack_enable` to `on`. -- Grant the rights to execute `ptrack` functions to the *backup* role: - - GRANT EXECUTE ON FUNCTION pg_catalog.pg_ptrack_clear() TO backup; - GRANT EXECUTE ON FUNCTION pg_catalog.pg_ptrack_get_and_clear(oid, oid) TO backup; -- The *backup* role must have access to all the databases of the cluster. - -## Command-Line Reference -### Commands - -This section describes pg_probackup commands. Some commands require mandatory parameters and can take additional options. Optional parameters encased in square brackets. For detailed descriptions of options, see the section [Options](#options). - -#### version - - pg_probackup version - -Prints pg_probackup version. - -#### help - - pg_probackup help [command] - -Displays the synopsis of pg_probackup commands. If one of the pg_probackup commands is specified, shows detailed information about the options that can be used with this command. - -#### init - - pg_probackup init -B backup_dir [--help] - -Initializes the backup catalog in *backup_dir* that will store backup copies, WAL archive and meta information for the backed up database clusters. If the specified *backup_dir* already exists, it must be empty. Otherwise, pg_probackup displays a corresponding error message. - -For details, see the secion [Initializing the Backup Catalog](#initializing-the-backup-catalog). - -#### add-instance - - pg_probackup add-instance -B backup_dir -D data_dir --instance instance_name - [--help] [--external-dirs=external_directory_path] - -Initializes a new backup instance inside the backup catalog *backup_dir* and generates the pg_probackup.conf configuration file that controls pg_probackup settings for the cluster with the specified *data_dir* data directory. - -For details, see the section [Adding a New Backup Instance](#adding-a-new-backup-instance). - -#### del-instance - - pg_probackup del-instance -B backup_dir --instance instance_name - [--help] - -Deletes all backups and WAL files associated with the specified instance. - -#### set-config - - pg_probackup set-config -B backup_dir --instance instance_name - [--help] [--pgdata=pgdata-path] - [--retention-redundancy=redundancy][--retention-window=window] - [--compress-algorithm=compression_algorithm] [--compress-level=compression_level] - [-d dbname] [-h host] [-p port] [-U username] - [--archive-timeout=timeout] [--external-dirs=external_directory_path] - [remote_options] [logging_options] - -Adds the specified connection, compression, retention, logging and external directory settings into the pg_probackup.conf configuration file, or modifies the previously defined values. - -It is **not recommended** to edit pg_probackup.conf manually. - -#### show-config - - pg_probackup show-config -B backup_dir --instance instance_name [--format=plain|json] - -Displays the contents of the pg_probackup.conf configuration file located in the '*backup_dir*/backups/*instance_name*' directory. You can specify the `--format=json` option to return the result in the JSON format. By default, configuration settings are shown as plain text. - -To edit pg_probackup.conf, use the [set-config](#set-config) command. - -#### show - - pg_probackup show -B backup_dir - [--help] [--instance instance_name [-i backup_id]] [--format=plain|json] - -Shows the contents of the backup catalog. If *instance_name* and *backup_id* are specified, shows detailed information about this backup. You can specify the `--format=json` option to return the result in the JSON format. - -By default, the contents of the backup catalog is shown as plain text. - -#### backup - - pg_probackup backup -B backup_dir -b backup_mode --instance instance_name - [--help] [-j num_threads] [--progress] - [-C] [--stream [-S slot_name] [--temp-slot]] [--backup-pg-log] - [--no-validate] [--skip-block-validation] - [-w --no-password] [-W --password] - [--archive-timeout=timeout] [--external-dirs=external_directory_path] - [connection_options] [compression_options] [remote_options] - [retention_options] [logging_options] - -Creates a backup copy of the PostgreSQL instance. The *backup_mode* option specifies the backup mode to use. - -For details, see the sections [Backup Options](#backup-options) and [Creating a Backup](#creating-a-backup). - -#### restore - - pg_probackup restore -B backup_dir --instance instance_name - [--help] [-D data_dir] [-i backup_id] - [-j num_threads] [--progress] - [-T OLDDIR=NEWDIR] [--external-mapping=OLDDIR=NEWDIR] [--skip-external-dirs] - [-R | --restore-as-replica] [--no-validate] [--skip-block-validation] - [recovery_options] [logging_options] [remote_options] - -Restores the PostgreSQL instance from a backup copy located in the *backup_dir* backup catalog. If you specify a recovery target option, pg_probackup will find the closest backup and restores it to the specified recovery target. Otherwise, the most recent backup is used. - -For details, see the sections [Restore Options](#restore-options), [Recovery Target Options](#recovery-target-options) and [Restoring a Cluster](#restoring-a-cluster). - -#### checkdb - - pg_probackup checkdb - [-B backup_dir] [--instance instance_name] [-D data_dir] - [--help] [-j num_threads] [--progress] - [--skip-block-validation] [--amcheck] [--heapallindexed] - [connection_options] [logging_options] - -Verifyes the PostgreSQL database cluster correctness by detecting physical and logical corruption. - -For details, see the sections [Checkdb Options](#checkdb-options) and [Verifying a Cluster](#verifying-a-cluster). - -#### validate - - pg_probackup validate -B backup_dir - [--help] [--instance instance_name] [-i backup_id] - [-j num_threads] [--progress] - [--skip-block-validation] - [recovery_options] [logging_options] - -Verifies that all the files required to restore the cluster are present and not corrupted. If *instance_name* is not specified, pg_probackup validates all backups available in the backup catalog. If you specify the *instance_name* without any additional options, pg_probackup validates all the backups available for this backup instance. If you specify the *instance_name* with a [recovery target option](#recovery-target-options) and/or a *backup_id*, pg_probackup checks whether it is possible to restore the cluster using these options. - -For details, see the section [Validating a Backup](#validating-a-backup). - -#### merge - - pg_probackup merge -B backup_dir --instance instance_name -i backup_id - [--help] [-j num_threads][--progress] - [logging_options] - -Merges the specified incremental backup to its parent full backup, together with all incremental backups between them, if any. As a result, the full backup takes in all the merged data, and the incremental backups are removed as redundant. - -For details, see the section [Merging Backups](#merging-backups). - -#### delete - - pg_probackup delete -B backup_dir --instance instance_name - [--help] [-j num_threads] [--progress] - [--delete-wal] {-i backup_id | --delete-expired [--merge-expired] | --merge-expired} - [--dry-run] - [logging_options] - -Deletes backup with specified *backip_id* or launches the retention purge of backups and archived WAL that do not satisfy the current retention policies. - -For details, see the sections [Deleting Backups](#deleting-backups), [Retention Options](#retention-otions) and [Configuring Backup Retention Policy](#configuring-backup-retention-policy). - -#### archive-push - - pg_probackup archive-push -B backup_dir --instance instance_name - --wal-file-path %p --wal-file-name %f - [--help] [--compress] [--compress-algorithm=compression_algorithm] - [--compress-level=compression_level] [--overwrite] - [remote_options] [logging_options] - -Copies WAL files into the corresponding subdirectory of the backup catalog and validates the backup instance by *instance_name* and *system-identifier*. If parameters of the backup instance and the cluster do not match, this command fails with the following error message: “Refuse to push WAL segment segment_name into archive. Instance parameters mismatch.” For each WAL file moved to the backup catalog, you will see the following message in PostgreSQL logfile: “pg_probackup archive-push completed successfully”. - -If the files to be copied already exist in the backup catalog, pg_probackup computes and compares their checksums. If the checksums match, archive-push skips the corresponding file and returns successful execution code. Otherwise, archive-push fails with an error. If you would like to replace WAL files in the case of checksum mismatch, run the archive-push command with the `--overwrite` option. - -Copying is done to temporary file with `.partial` suffix or, if [compression](#compression-options) is used, with `.gz.partial` suffix. After copy is done, atomic rename is performed. This algorihtm ensures that failed archive-push will not stall continuous archiving and that concurrent archiving from multiple sources into single WAL archive has no risk of archive corruption. -Copied to archive WAL segments are synced to disk. - -You can use `archive-push` in [archive_command](https://www.postgresql.org/docs/current/runtime-config-wal.html#GUC-ARCHIVE-COMMAND) PostgreSQL parameter to set up [continous WAl archiving](#setting-up-continuous-wal-archiving). - -For details, see sections [Archiving Options](#archiving-options) and [Compression Options](#compression-options). - -#### archive-get - - pg_probackup archive-get -B backup_dir --instance instance_name --wal-file-path %p --wal-file-name %f - [--help] [remote_options] [logging_options] - -Copies WAL files from the corresponding subdirectory of the backup catalog to the cluster's write-ahead log location. This command is automatically set by pg_probackup as part of the `restore_command` in 'recovery.conf' when restoring backups using a WAL archive. You do not need to set it manually. - -### Options - -This section describes all command-line options for pg_probackup commands. If the option value can be derived from an environment variable, this variable is specified below the command-line option, in the uppercase. Some values can be taken from the pg_probackup.conf configuration file located in the backup catalog. - -For details, see the section [Configuring pg_probackup](#configuring-pg_probackup). - -If an option is specified using more than one method, command-line input has the highest priority, while the pg_probackup.conf settings have the lowest priority. - -#### Common Options -The list of general options. - - -B directory - --backup-path=directory - BACKUP_PATH -Specifies the absolute path to the backup catalog. Backup catalog is a directory where all backup files and meta information are stored. Since this option is required for most of the pg_probackup commands, you are recommended to specify it once in the BACKUP_PATH environment variable. In this case, you do not need to use this option each time on the command line. - - -D directory - --pgdata=directory - PGDATA -Specifies the absolute path to the data directory of the database cluster. This option is mandatory only for the [add-instance](#add-instance) command. Other commands can take its value from the PGDATA environment variable, or from the pg_probackup.conf configuration file. - - -i backup_id - -backup-id=backup_id -Specifies the unique identifier of the backup. - - -j num_threads - --threads=num_threads -Sets the number of parallel threads for backup, restore, merge, validation and verification processes. - - --progress -Shows the progress of operations. - - --help -Shows detailed information about the options that can be used with this command. - -#### Backup Options - -The following options can be used together with the [backup](#backup) command. - -Additionally [Connection Options](#connection-options), [Retention Options](#retention-options), [Remote Mode Options](#remote-mode-options), [Compression Options](#compression-options), [Logging Options](#logging-options) and [Common Options](#common-options) can be used. - - -b mode - --backup-mode=mode - -Specifies the backup mode to use. Possible values are: - -- FULL — creates a full backup that contains all the data files of the cluster to be restored. -- DELTA — reads all data files in the data directory and creates an incremental backup for pages that have changed since the previous backup. -- PAGE — creates an incremental PAGE backup based on the WAL files that have changed since the previous full or incremental backup was taken. -- PTRACK — creates an incremental PTRACK backup tracking page changes on the fly. - -For details, see the section [Creating a Backup](#creating-a-backup). - - -C - --smooth-checkpoint - SMOOTH_CHECKPOINT -Spreads out the checkpoint over a period of time. By default, pg_probackup tries to complete the checkpoint as soon as possible. - - --stream -Makes an STREAM backup that includes all the necessary WAL files by streaming them from the database server via replication protocol. - - -S slot_name - --slot=slot_name -Specifies the replication slot for WAL streaming. This option can only be used together with the `--stream` option. - - --temp-slot -Creates a temporary physical replication slot for streaming WAL from the backed up PostgreSQL instance. It ensures that all the required WAL segments remain available if WAL is rotated while the backup is in progress. This option can only be used together with the `--stream` option. Default slot name is `pg_probackup_slot`, which can be changed via option `--slot/-S`. - - --backup-pg-log -Includes the log directory into the backup. This directory usually contains log messages. By default, log directory is excluded. - - -E external_directory_path - --external-dirs=external_directory_path -Includes the specified directory into the backup. This option is useful to back up scripts, sql dumps and configuration files located outside of the data directory. If you would like to back up several external directories, separate their paths by a colon on Unix and a semicolon on Windows. - - --archive-timeout=wait_time -Sets in seconds the timeout for WAL segment archiving and streaming. By default pg_probackup waits 300 seconds. - - --skip-block-validation -Disables block-level checksum verification to speed up backup. - - --no-validate -Skips automatic validation after successfull backup. You can use this option if you validate backups regularly and would like to save time when running backup operations. - -#### Restore Options - -The following options can be used together with the [restore](#restore) command. - -Additionally [Recovery Target Options](#recovery-target-options), [Remote Mode Options](#remote-mode-options), [Logging Options](#logging-options) and [Common Options](#common-options) can be used. - - -R | --restore-as-replica -Writes a minimal recovery.conf in the output directory to facilitate setting up a standby server. The password is not included. If the replication connection requires a password, you must specify the password manually. - - -T OLDDIR=NEWDIR - --tablespace-mapping=OLDDIR=NEWDIR - -Relocates the tablespace from the OLDDIR to the NEWDIR directory at the time of recovery. Both OLDDIR and NEWDIR must be absolute paths. If the path contains the equals sign (=), escape it with a backslash. This option can be specified multiple times for multiple tablespaces. - - --external-mapping=OLDDIR=NEWDIR -Relocates an external directory included into the backup from the OLDDIR to the NEWDIR directory at the time of recovery. Both OLDDIR and NEWDIR must be absolute paths. If the path contains the equals sign (=), escape it with a backslash. This option can be specified multiple times for multiple directories. - - --skip-external-dirs -Skip external directories included into the backup with the `--external-dirs` option. The contents of these directories will not be restored. - - --skip-block-validation -Disables block-level checksum verification to speed up validation. During automatic validation before restore only file-level checksums will be verified. - - --no-validate -Skips backup validation. You can use this option if you validate backups regularly and would like to save time when running restore operations. - -#### Checkdb Options - -The following options can be used together with the [checkdb](#checkdb) command. - -For details on verifying PostgreSQL database cluster, see section [Verifying a Cluster](#verifying-a-cluster). - - --amcheck -Performs logical verification of indexes for the specified PostgreSQL instance if no corruption was found while checking data files. You must have the `amcheck` extention or the `amcheck_next` extension installed in the database to check its indexes. For databases without amcheck, index verification will be skipped. - - --skip-block-validation -Skip validation of data files. Can be used only with `--amcheck` option, so only logical verification of indexes is performed. - - --heapallindexed -Checks that all heap tuples that should be indexed are actually indexed. You can use this option only together with the `--amcheck` option. Can be used only with `amcheck` extension of version 2.0 and `amcheck_next` extension of any version. - -#### Recovery Target Options - -If [continuous WAL archiving](#setting-up-continuous-wal-archiving) is configured, you can use one of these options together with [restore](#restore) or [validate](#validate) commands to specify the moment up to which the database cluster must be restored or validated. - - --recovery-target=immediate|latest -Defines when to stop the recovery: - -- `immediate` value stops the recovery after reaching the consistent state of the specified backup, or the latest available backup if the `-i/--backup_id` option is omitted. -- `latest` value continues the recovery until all WAL segments available in the archive are applied. - -Default value of `--recovery-target` depends on WAL delivery method of restored backup, `immediate` for STREAM backup and `latest` for ARCHIVE. - - --recovery-target-timeline=timeline -Specifies a particular timeline to which recovery will proceed. By default, the timeline of the specified backup is used. - - --recovery-target-lsn=lsn -Specifies the LSN of the write-ahead log location up to which recovery will proceed. Can be used only when restoring database cluster of major version 10 or higher. - - --recovery-target-name=recovery_target_name -Specifies a named savepoint up to which to restore the cluster data. - - --recovery-target-time=time -Specifies the timestamp up to which recovery will proceed. - - --recovery-target-xid=xid -Specifies the transaction ID up to which recovery will proceed. - - --recovery-target-inclusive=boolean -Specifies whether to stop just after the specified recovery target (true), or just before the recovery target (false). This option can only be used together with `--recovery-target-name`, `--recovery-target-time`, `--recovery-target-lsn` or `--recovery-target-xid` options. The default depends on [recovery_target_inclusive](https://www.postgresql.org/docs/current/recovery-target-settings.html#RECOVERY-TARGET-INCLUSIVE) parameter. - - --recovery-target-action=pause|promote|shutdown - Default: pause -Specifies [the action](https://www.postgresql.org/docs/current/recovery-target-settings.html#RECOVERY-TARGET-ACTION) the server should take when the recovery target is reached. - -#### Retention Options -You can use these options together with [backup](#backup) and [delete](#delete) commands. - -For details on configuring retention policy, see the section [Configuring Backup Retention Policy](#configuring-backup-retention-policy). - - --retention-redundancy=redundancy - Default: 0 -Specifies the number of full backup copies to keep in the data directory. Must be a positive integer. The zero value disables this setting. - - --retention-window=window - Default: 0 -Number of days of recoverability. Must be a positive integer. The zero value disables this setting. - - --delete-wal -Deletes WAL files that are no longer required to restore the cluster from any of the existing backups. - - --delete-expired -Deletes backups that do not conform to the retention policy defined in the pg_probackup.conf configuration file. - - --merge-expired -Merges the oldest incremental backup that satisfies the requirements of retention policy with its parent backups that have already expired. - - --dry-run -Displays the current status of all the available backups, without deleting or merging expired backups, if any. - -#### Logging Options -You can use these options with any command. - - --log-level-console=log_level - Default: info -Controls which message levels are sent to the console log. Valid values are `verbose`, `log`, `info`, `warning`, `error` and `off`. Each level includes all the levels that follow it. The later the level, the fewer messages are sent. The `off` level disables console logging. - ->NOTE: all console log messages are going to stderr, so output from [show](#show) and [show-config](#show-config) commands do not mingle with log messages. - - --log-level-file=log_level - Default: off -Controls which message levels are sent to a log file. Valid values are `verbose`, `log`, `info`, `warning`, `error` and `off`. Each level includes all the levels that follow it. The later the level, the fewer messages are sent. The `off` level disables file logging. - - --log-filename=log_filename - Default: pg_probackup.log -Defines the filenames of the created log files. The filenames are treated as a strftime pattern, so you can use %-escapes to specify time-varying filenames, as explained in *log_filename*. - -For example, if you specify the 'pg_probackup-%u.log' pattern, pg_probackup generates a separate log file for each day of the week, with %u replaced by the corresponding decimal number: pg_probackup-1.log for Monday, pg_probackup-2.log for Tuesday, and so on. - -This option takes effect if file logging is enabled by the `log-level-file` option. - - --error-log-filename=error_log_filename - Default: none -Defines the filenames of log files for error messages only. The filenames are treated as a strftime pattern, so you can use %-escapes to specify time-varying filenames, as explained in *error_log_filename*. - -For example, if you specify the 'error-pg_probackup-%u.log' pattern, pg_probackup generates a separate log file for each day of the week, with %u replaced by the corresponding decimal number: error-pg_probackup-1.log for Monday, error-pg_probackup-2.log for Tuesday, and so on. - -This option is useful for troubleshooting and monitoring. - - --log-directory=log_directory - Default: $BACKUP_PATH/log/ -Defines the directory in which log files will be created. You must specify the absolute path. This directory is created lazily, when the first log message is written. - - --log-rotation-size=log_rotation_size - Default: 0 -Maximum size of an individual log file. If this value is reached, the log file is rotated once a pg_probackup command is launched, except help and version commands. The zero value disables size-based rotation. Supported units: kB, MB, GB, TB (kB by default). - - --log-rotation-age=log_rotation_age - Default: 0 -Maximum lifetime of an individual log file. If this value is reached, the log file is rotated once a pg_probackup command is launched, except help and version commands. The time of the last log file creation is stored in $BACKUP_PATH/log/log_rotation. The zero value disables time-based rotation. Supported units: ms, s, min, h, d (min by default). - -#### Connection Options -You can use these options together with [backup](#backup) and [checkdb](#checkdb) commands. - -All [libpq environment variables](https://www.postgresql.org/docs/current/libpq-envars.html) are supported. - - -d dbname - --dbname=dbname - PGDATABASE -Specifies the name of the database to connect to. The connection is used only for managing backup process, so you can connect to any existing database. If this option is not provided on the command line, PGDATABASE environment variable, or the pg_probackup.conf configuration file, pg_probackup tries to take this value from the PGUSER environment variable, or from the current user name if PGUSER variable is not set. - - -h host - --host=host - PGHOST - Default: local socket -Specifies the host name of the system on which the server is running. If the value begins with a slash, it is used as a directory for the Unix domain socket. - - -p port - --port=port - PGPORT - Default: 5432 -Specifies the TCP port or the local Unix domain socket file extension on which the server is listening for connections. - - -U username - --username=username - PGUSER -User name to connect as. - - -w - --no-password - Disables a password prompt. If the server requires password authentication and a password is not available by other means such as a [.pgpass](https://www.postgresql.org/docs/current/libpq-pgpass.html) file or PGPASSWORD environment variable, the connection attempt will fail. This option can be useful in batch jobs and scripts where no user is present to enter a password. - - -W - --password -Forces a password prompt. - -#### Compression Options - -You can use these options together with [backup](#backup) and [archive-push](#archive-push) commands. - - --compress-algorithm=compression_algorithm - Default: none -Defines the algorithm to use for compressing data files. Possible values are `zlib`, `pglz`, and `none`. If set to zlib or pglz, this option enables compression. By default, compression is disabled. -For the [archive-push](#archive-push) command, the pglz compression algorithm is not supported. - - --compress-level=compression_level - Default: 1 -Defines compression level (0 through 9, 0 being no compression and 9 being best compression). This option can be used together with `--compress-algorithm` option. - - --compress -Alias for `--compress-algorithm=zlib` and `--compress-level=1`. - -#### Archiving Options - -These options can be used with [archive-push](#archive-push) command in [archive_command](https://www.postgresql.org/docs/current/runtime-config-wal.html#GUC-ARCHIVE-COMMAND) setting and [archive-get](#archive-get) command in [restore_command](https://www.postgresql.org/docs/current/archive-recovery-settings.html#RESTORE-COMMAND) setting. - -Additionally [Remote Mode Options](#remote-mode-options) and [Logging Options](#logging-options) can be used. - - --wal-file-path=wal_file_path %p -Provides the path to the WAL file in `archive_command` and `restore_command`. The %p variable is required for correct processing. - - --wal-file-name=wal_file_name %f -Provides the name of the WAL file in `archive_command` and `restore_command`. The %f variable is required for correct processing. - - --overwrite -Overwrites archived WAL file. Use this option together with the archive-push command if the specified subdirectory of the backup catalog already contains this WAL file and it needs to be replaced with its newer copy. Otherwise, archive-push reports that a WAL segment already exists, and aborts the operation. If the file to replace has not changed, archive-push skips this file regardless of the `--overwrite` option. - -#### Remote Mode Options - -This section describes the options related to running pg_probackup operations remotely via SSH. These options can be used with [add-instance](#add-instance), [set-config](#set-config), [backup](#backup), [restore](#restore), [archive-push](#archive-push) and [archive-get](#archive-get) commands. - -For details on configuring remote operation mode, see the section [Using pg_probackup in the Remote Mode](#using-pg_probackup-in-the-remote-mode). - - --remote-proto -Specifies the protocol to use for remote operations. Currently only the SSH protocol is supported. Possible values are: - -- `ssh` enables the remote backup mode via SSH. This is the Default value. -- `none` explicitly disables the remote mode. - -You can omit this option if the `--remote-host` option is specified. - - --remote-host -Specifies the remote host IP address or hostname to connect to. - - --remote-port - Default: 22 -Specifies the remote host port to connect to. - - --remote-user - Default: current user -Specifies remote host user for SSH connection. If you omit this option, the current user initiating the SSH connection is used. - - --remote-path -Specifies pg_probackup installation directory on the remote system. - - --ssh-options -Specifies a string of SSH command-line options. For example, the following options can used to set keep-alive for ssh connections opened by pg_probackup: `--ssh-options='-o ServerAliveCountMax=5 -o ServerAliveInterval=60'`. Full list of possible options can be found here: (https://linux.die.net/man/5/ssh_config)[https://linux.die.net/man/5/ssh_config] - -#### Replica Options - -This section describes the options related to taking a backup from standby. - ->NOTE: Starting from pg_probackup 2.0.24, backups can be taken from standby without connecting to the master server, so these options are no longer required. In lower versions, pg_probackup had to connect to the master to determine recovery time — the earliest moment for which you can restore a consistent state of the database cluster. - - --master-db=dbname - Default: postgres, the default PostgreSQL database. -Deprecated. Specifies the name of the database on the master server to connect to. The connection is used only for managing the backup process, so you can connect to any existing database. Can be set in the pg_probackup.conf using the [set-config](#set-config) command. - - --master-host=host -Deprecated. Specifies the host name of the system on which the master server is running. - - --master-port=port - Default: 5432, the PostgreSQL default port. -Deprecated. Specifies the TCP port or the local Unix domain socket file extension on which the master server is listening for connections. - - --master-user=username - Default: postgres, the PostgreSQL default user name. -Deprecated. User name to connect as. - - --replica-timeout=timeout - Default: 300 sec -Deprecated. Wait time for WAL segment streaming via replication, in seconds. By default, pg_probackup waits 300 seconds. You can also define this parameter in the pg_probackup.conf configuration file using the [set-config](#set-config) command. - -## Usage - -- [Creating a Backup](#creating-a-backup) -- [Verifying a Cluster](#verifying-a-cluster) -- [Validating a Backup](#vaklidating-a-backup) -- [Restoring a Cluster](#restoring-a-cluster) -- [Performing Point-in-Time (PITR) Recovery](#performing-point-in-time-pitr-recovery) -- [Using pg_probackup in the Remote Mode](#using-pg_probackup-in-the-remote-mode) -- [Running pg_probackup on Parallel Threads](#running-pg_probackup-on-parallel-threads) -- [Configuring pg_probackup](#configuring-pg_probackup) -- [Managing the Backup Catalog](#managing-the-backup-catalog) -- [Configuring Backup Retention Policy](#configuring-backup-retention-policy) -- [Merging Backups](#merging-backups) -- [Deleting Backups](#deleting-backups) - -### Creating a Backup - -To create a backup, run the following command: - - pg_probackup backup -B backup_dir --instance instance_name -b backup_mode - -Where *backup_mode* can take one of the following values: - -- FULL — creates a full backup that contains all the data files of the cluster to be restored. -- DELTA — reads all data files in the data directory and creates an incremental backup for pages that have changed since the previous backup. -- PAGE — creates an incremental PAGE backup based on the WAL files that have changed since the previous full or incremental backup was taken. -- PTRACK — creates an incremental PTRACK backup tracking page changes on the fly. - -When restoring a cluster from an incremental backup, pg_probackup relies on the parent full backup and all the incremental backups between them, which is called `backup chain`. You must create at least one full backup before taking incremental ones. - -#### ARCHIVE mode - -ARCHIVE is the default WAL delivery mode. - -For example, to make a FULL backup in ARCHIVE mode, run: - - pg_probackup backup -B backup_dir --instance instance_name -b FULL - -Unlike backup in STREAM mode, ARCHIVE backup rely on [continuous archiving](#setting-up-continuous-wal-archiving) to provide WAL segments required to restore the cluster to a consistent state at the time the backup was taken. - -During [backup](#backup) pg_probackup ensures that WAL files containing WAL records between START LSN and STOP LSN are actually exists in '*backup_dir*/wal/*instance_name*' directory. Also pg_probackup ensures that WAL records between START LSN and STOP LSN can be parsed. This precations eliminates the risk of silent WAL corruption. - -#### STREAM mode - -STREAM is the optional WAL delivery mode. - -For example, to make a FULL backup in STREAM mode, add the `--stream` option to the command from the previous example: - - pg_probackup backup -B backup_dir --instance instance_name -b FULL --stream --temp-slot - -The optional `--temp-slot` flag ensures that the required segments remain available if the WAL is rotated before the backup is complete. - -Unlike backup in ARCHIVE mode, STREAM backup include all the WAL segments required to restore the cluster to a consistent state at the time the backup was taken. - -During [backup](#backup) pg_probackup streams WAL files containing WAL records between START LSN and STOP LSN in '*backup_dir*/backups/*instance_name*/*BACKUP ID*/database/pg_wal' directory. Also pg_probackup ensures that WAL records between START LSN and STOP LSN can be parsed. This precations eliminates the risk of silent WAL corruption. - -Even if you are using [continuous archiving](#setting-up-continuous-wal-archiving), STREAM backups can still be useful in the following cases: - -- STREAM backups can be restored on the server that has no file access to WAL archive. -- STREAM backups enable you to restore the cluster state at the point in time for which WAL files are no longer available. -- Backup in STREAM mode can be taken from standby of a server, that generates small amount of WAL traffic, without long waiting for WAL segment to fill up. - -#### Page validation - -If [data checksums](https://www.postgresql.org/docs/current/runtime-config-preset.html#GUC-DATA-CHECKSUMS) are enabled in the database cluster, pg_probackup uses this information to check correctness of data files during backup. While reading each page, pg_probackup checks whether the calculated checksum coincides with the checksum stored in the page header. This guarantees that the PostgreSQL instance and backup itself are free of corrupted pages. -Note that pg_probackup reads database files directly from filesystem, so under heavy write load during backup it can show false positive checksum failures because of partial writes. In case of page checksumm mismatch, page is readed again and checksumm comparison repeated. - -Page is considered corrupted if checksumm comparison failed more than 100 times, in this case backup is aborted. - -Redardless of data checksums been enabled or not, pg_probackup always check page header "sanity". - -#### External directories - -To back up a directory located outside of the data directory, use the optional `--external-dirs` parameter that specifies the path to this directory. If you would like to add more than one external directory, provide several paths separated by colons. - -For example, to include '/etc/dir1/' and '/etc/dir2/' directories into the full backup of your *instance_name* instance that will be stored under the *backup_dir* directory, run: - - pg_probackup backup -B backup_dir --instance instance_name -b FULL --external-dirs=/etc/dir1:/etc/dir2 - -pg_probackup creates a separate subdirectory in the backup directory for each external directory. Since external directories included into different backups do not have to be the same, when you are restoring the cluster from an incremental backup, only those directories that belong to this particular backup will be restored. Any external directories stored in the previous backups will be ignored. - -To include the same directories into each backup of your instance, you can specify them in the pg_probackup.conf configuration file using the [set-config](#set-config) command with the `--external-dirs` option. - -### Verifying a Cluster - -To verify that PostgreSQL database cluster is free of corruption, run the following command: - - pg_probackup checkdb [-B backup_dir [--instance instance_name]] [-D data_dir] - -This physical verification works similar to [page validation](#page-validation) that is done during backup with several differences: - -- `checkdb` is read-only -- if corrupted page is detected, `checkdb` is not aborted, but carry on, until all pages in the cluster are validated -- `checkdb` do not strictly require the backup catalog, so it can be used to verify database clusters that are **not** [added to the backup catalog](#adding-a-new-backup-instance). - -If *backup_dir* and *instance_name* are omitted, then [connection options](#connection-options) and *data_dir* must be provided via environment variables or command-line options. - -Physical verification cannot detect logical inconsistencies, missing and nullified blocks or entire files, repercussions from PostgreSQL bugs and other wicked anomalies. -Extensions [amcheck](https://www.postgresql.org/docs/current/amcheck.html) and [amcheck_next](https://github.com/petergeoghegan/amcheck) provide a partial solution to these problems. - -If you would like, in addition to physical verification, to verify all indexes in all databases using these extensions, you can specify `--amcheck` option when running [checkdb](#checkdb) command: - - pg_probackup checkdb -D data_dir --amcheck - -Physical verification can be skipped if `--skip-block-validation` option is used. For logical only verification *backup_dir* and *data_dir* are optional, only [connection options](#connection-options) are mandatory: - - pg_probackup checkdb --amcheck --skip-block-validation [connection options] - -Logical verification can be done more thoroughly with option `--heapallindexed` by checking that all heap tuples that should be indexed are actually indexed, but at the higher cost of CPU, memory and I/O comsumption. - -### Validating Backups - -pg_probackup calculates checksums for each file in a backup during backup process. The process of checking checksumms of backup data files is called `backup validation`. By default validation is run immediately after backup is taken and right before restore, to detect possible backup corruptions. - -If you would like to skip backup validation, you can specify the `--no-validate` option when running [backup](#backup) and [restore](#restore) commands. - -To ensure that all the required backup files are present and can be used to restore the database cluster, you can run the [validate](#validate) command with the exact recovery target options you are going to use for recovery. - -For example, to check that you can restore the database cluster from a backup copy up to the specified xid transaction ID, run this command: - - pg_probackup validate -B backup_dir --instance instance_name --recovery-target-xid=xid - -If validation completes successfully, pg_probackup displays the corresponding message. If validation fails, you will receive an error message with the exact time, transaction ID and LSN up to which the recovery is possible. - -If you specify *backup_id* via `-i/--backup-id` option, then only backup copy with specified backup ID will be validated. If *backup_id* belong to incremental backup, then all its parents starting from FULL backup will be validated. - -If you omit all the parameters, all backups are validated. - -### Restoring a Cluster - -To restore the database cluster from a backup, run the restore command with at least the following options: - - pg_probackup restore -B backup_dir --instance instance_name -i backup_id - -Where: - -- *backup_dir* is the backup catalog that stores all backup files and meta information. -- *instance_name* is the backup instance for the cluster to be restored. -- *backup_id* specifies the backup to restore the cluster from. If you omit this option, pg_probackup uses the latest valid backup available for the specified instance. If you specify an incremental backup to restore, pg_probackup automatically restores the underlying full backup and then sequentially applies all the necessary increments. - -If the cluster to restore contains tablespaces, pg_probackup restores them to their original location by default. To restore tablespaces to a different location, use the `--tablespace-mapping/-T` option. Otherwise, restoring the cluster on the same host will fail if tablespaces are in use, because the backup would have to be written to the same directories. - -When using the `--tablespace-mapping/-T` option, you must provide absolute paths to the old and new tablespace directories. If a path happens to contain an equals sign (=), escape it with a backslash. This option can be specified multiple times for multiple tablespaces. For example: - - pg_probackup restore -B backup_dir --instance instance_name -D data_dir -j 4 -i backup_id -T tablespace1_dir=tablespace1_newdir -T tablespace2_dir=tablespace2_newdir - -Once the restore command is complete, start the database service. - -If you are restoring an STREAM backup, the restore is complete at once, with the cluster returned to a self-consistent state at the point when the backup was taken. For ARCHIVE backups, PostgreSQL replays all available archived WAL segments, so the cluster is restored to the latest state possible. You can change this behavior by using the [recovery target options](#recovery-target-options) with the `restore` command. Note that using the [recovery target options](#recovery-target-options) when restoring STREAM backup is possible if the WAL archive is available at least starting from the time the STREAM backup was taken. - ->NOTE: By default, the [restore](#restore) command validates the specified backup before restoring the cluster. If you run regular backup validations and would like to save time when restoring the cluster, you can specify the `--no-validate` option to skip validation and speed up the recovery. - -### Performing Point-in-Time (PITR) Recovery - -If you have enabled [continuous WAL archiving](#setting-up-continuous-wal-archiving) before taking backups, you can restore the cluster to its state at an arbitrary point in time (recovery target) using [recovery target options](#recovery-target-options) with the [restore](#restore) and [validate](#validate) commands. - -If `-i/--backup-id` option is omitted, pg_probackup automatically chooses the backup that is the closest to the specified recovery target and starts the restore process, otherwise pg_probackup will try to restore *backup_id* to the specified recovery target. - -- To restore the cluster state at the exact time, specify the recovery-target-time option, in the timestamp format. For example: - - pg_probackup restore -B backup_dir --instance instance_name --recovery-target-time='2017-05-18 14:18:11+03' - -- To restore the cluster state up to a specific transaction ID, use the recovery-target-xid option: - - pg_probackup restore -B backup_dir --instance instance_name --recovery-target-xid=687 - -- To restore the cluster state up to a specific LSN, use recovery-target-lsn: - - pg_probackup restore -B backup_dir --instance instance_name --recovery-target-lsn=16/B374D848 - -- To restore the cluster state up to a specific named restore point, use recovery-target-name: - - pg_probackup restore -B backup_dir --instance instance_name --recovery-target-name='before_app_upgrade' - -- To restore the cluster to the latest state available in archive, use recovery-target option: - - pg_probackup restore -B backup_dir --instance instance_name --recovery-target='latest' - -### Using pg_probackup in the Remote Mode - -pg_probackup supports the remote mode that allows to perform `backup` and `restore` operations remotely via SSH. In this mode, the backup catalog is stored on a local system, while PostgreSQL instance to be backed up is located on a remote system. You must have pg_probackup installed on both systems. - -Do note that pg_probackup rely on passwordless SSH connection for communication between the hosts. - -The typical workflow is as follows: - - - On your backup host, configure pg_probackup as explained in the section [Installation and Setup](#installation-and-setup). For the [add-instance](#add-instance) and [set-config](#set-config) commands, make sure to specify [remote options](#remote-mode-options) that point to the database host with the PostgreSQL instance. - -- If you would like to take remote backup in [PAGE](#creating-a-backup) mode, or rely on [ARCHIVE](#archive-mode) WAL delivery method, or use [PITR](#performing-point-in-time-pitr-recovery), then configure continuous WAL archiving from database host to the backup host as explained in the section [Setting up continuous WAL archiving](#setting-up-continuous-wal-archiving). For the [archive-push](#archive-push) and [archive-get](#archive-get) commands, you must specify the [remote options](#remote-mode-options) that point to backup host with backup catalog. - -- Run [backup](#backup) or [restore](#restore) commands with [remote options](#remote-mode-options) **on backup host**. pg_probackup connects to the remote system via SSH and creates a backup locally or restores the previously taken backup on the remote system, respectively. - ->NOTE: The remote backup mode is currently unavailable for Windows systems. - -### Running pg_probackup on Parallel Threads - -[Backup](#backup), [restore](#restore), [merge](#merge), [delete](#delete), [checkdb](#checkdb) and [validate](#validate) processes can be executed on several parallel threads. This can significantly speed up pg_probackup operation given enough resources (CPU cores, disk, and network throughput). - -Parallel execution is controlled by the `-j/--threads` command line option. For example, to create a backup using four parallel threads, run: - - pg_probackup backup -B backup_dir --instance instance_name -b FULL -j 4 - ->NOTE: Parallel restore applies only to copying data from the backup catalog to the data directory of the cluster. When PostgreSQL server is started, WAL records need to be replayed, and this cannot be done in parallel. - -### Configuring pg_probackup - -Once the backup catalog is initialized and a new backup instance is added, you can use the pg_probackup.conf configuration file located in the '*backup_dir*/backups/*instance_name*' directory to fine-tune pg_probackup configuration. - -For example, [backup](#backup) and [checkdb](#checkdb) commands uses a regular PostgreSQL connection. To avoid specifying these options each time on the command line, you can set them in the pg_probackup.conf configuration file using the [set-config](#set-config) command. - ->NOTE: It is **not recommended** to edit pg_probackup.conf manually. - -Initially, pg_probackup.conf contains the following settings: - -- PGDATA — the path to the data directory of the cluster to back up. -- system-identifier — the unique identifier of the PostgreSQL instance. - -Additionally, you can define [remote](#remote-mode-options), [retention](#retention-options), [logging](#logging-options) and [compression](#compression-options) settings using the `set-config` command: - - pg_probackup set-config -B backup_dir --instance instance_name - [--external-dirs=external_directory_path] [remote_options] [connection_options] [retention_options] [logging_options] - -To view the current settings, run the following command: - - pg_probackup show-config -B backup_dir --instance instance_name - -You can override the settings defined in pg_probackup.conf when running pg_probackups [commands](#commands) via corresponding environment variables and/or command line options. - -### Specifying Connection Settings - -If you define connection settings in the 'pg_probackup.conf' configuration file, you can omit connection options in all the subsequent pg_probackup commands. However, if the corresponding environment variables are set, they get higher priority. The options provided on the command line overwrite both environment variables and configuration file settings. - -If nothing is given, the default values are taken. By default pg_probackup tries to use local connection via Unix domain socket (localhost on Windows) and tries to get the database name and the user name from the PGUSER environment variable or the current OS user name. - -### Managing the Backup Catalog - -With pg_probackup, you can manage backups from the command line: - -- View available backups -- Validate backups -- Merge backups -- Delete backups -- Viewing Backup Information - -To view the list of existing backups for every instance, run the command: - - pg_probackup show -B backup_dir - -pg_probackup displays the list of all the available backups. For example: - -``` -BACKUP INSTANCE 'node' -============================================================================================================================================ - Instance Version ID Recovery time Mode WAL Current/Parent TLI Time Data Start LSN Stop LSN Status -============================================================================================================================================ - node 10 P7XDQV 2018-04-29 05:32:59+03 DELTA STREAM 1 / 1 11s 19MB 0/15000060 0/15000198 OK - node 10 P7XDJA 2018-04-29 05:28:36+03 PTRACK STREAM 1 / 1 21s 32MB 0/13000028 0/13000198 OK - node 10 P7XDHU 2018-04-29 05:27:59+03 PAGE STREAM 1 / 1 31s 33MB 0/11000028 0/110001D0 OK - node 10 P7XDHB 2018-04-29 05:27:15+03 FULL STREAM 1 / 0 11s 39MB 0/F000028 0/F000198 OK -``` - -For each backup, the following information is provided: - -- Instance — the instance name. -- Version — PostgreSQL major version. -- ID — the backup identifier. -- Recovery time — the earliest moment for which you can restore the state of the database cluster. -- Mode — the method used to take this backup. Possible values: FULL, PAGE, DELTA, PTRACK. -- WAL — the WAL delivery method. Possible values: STREAM and ARCHIVE. -- Current/Parent TLI — timeline identifiers of current backup and its parent. -- Time — the time it took to perform the backup. -- Data — the size of the data files in this backup. This value does not include the size of WAL files. -- Start LSN — WAL log sequence number corresponding to the start of the backup process. -- Stop LSN — WAL log sequence number corresponding to the end of the backup process. -- Status — backup status. Possible values: - - - OK — the backup is complete and valid. - - DONE — the backup is complete, but was not validated. - - RUNNING — the backup is in progress. - - MERGING — the backup is being merged. - - DELETING — the backup files are being deleted. - - CORRUPT — some of the backup files are corrupted. - - ERROR — the backup was aborted because of an unexpected error. - - ORPHAN — the backup is invalid because one of its parent backups is corrupt or missing. - -You can restore the cluster from the backup only if the backup status is OK or DONE. - -To get more detailed information about the backup, run the show with the backup ID: - - pg_probackup show -B backup_dir --instance instance_name -i backup_id - -The sample output is as follows: - -``` -#Configuration -backup-mode = FULL -stream = false -compress-alg = zlib -compress-level = 1 -from-replica = false - -#Compatibility -block-size = 8192 -wal-block-size = 8192 -checksum-version = 1 -program-version = 2.1.3 -server-version = 10 - -#Result backup info -timelineid = 1 -start-lsn = 0/04000028 -stop-lsn = 0/040000f8 -start-time = '2017-05-16 12:57:29' -end-time = '2017-05-16 12:57:31' -recovery-xid = 597 -recovery-time = '2017-05-16 12:57:31' -data-bytes = 22288792 -wal-bytes = 16777216 -status = OK -parent-backup-id = 'PT8XFX' -primary_conninfo = 'user=backup passfile=/var/lib/pgsql/.pgpass port=5432 sslmode=disable sslcompression=1 target_session_attrs=any' -``` - -To get more detailed information about the backup in json format, run the show with the backup ID: - - pg_probackup show -B backup_dir --instance instance_name --format=json -i backup_id - -The sample output is as follows: - -``` -[ - { - "instance": "node", - "backups": [ - { - "id": "PT91HZ", - "parent-backup-id": "PT8XFX", - "backup-mode": "DELTA", - "wal": "ARCHIVE", - "compress-alg": "zlib", - "compress-level": 1, - "from-replica": false, - "block-size": 8192, - "xlog-block-size": 8192, - "checksum-version": 1, - "program-version": "2.1.3", - "server-version": "10", - "current-tli": 16, - "parent-tli": 2, - "start-lsn": "0/8000028", - "stop-lsn": "0/8000160", - "start-time": "2019-06-17 18:25:11+03", - "end-time": "2019-06-17 18:25:16+03", - "recovery-xid": 0, - "recovery-time": "2019-06-17 18:25:15+03", - "data-bytes": 106733, - "wal-bytes": 16777216, - "primary_conninfo": "user=backup passfile=/var/lib/pgsql/.pgpass port=5432 sslmode=disable sslcompression=1 target_session_attrs=any", - "status": "OK" - } - ] - } -] -``` - -### Configuring Backup Retention Policy - -By default, all backup copies created with pg_probackup are stored in the specified backup catalog. To save disk space, you can configure retention policy and periodically clean up redundant backup copies accordingly. - -To configure retention policy, set one or more of the following variables in the pg_probackup.conf file via [set-config](#set-config): - -- retention-redundancy — specifies the number of full backup copies to keep in the backup catalog. -- retention-window — defines the earliest point in time for which pg_probackup can complete the recovery. This option is set in the number of days from the current moment. For example, if `retention-window=7`, pg_probackup must keep at least one full backup copy that is older than seven days, with all the corresponding WAL files. - -If both `retention-redundancy` and `retention-window` options are set, pg_probackup keeps backup copies that satisfy at least one condition. For example, if you set `retention-redundancy=2` and `retention-window=7`, pg_probackup cleans up the backup directory to keep only two full backup copies and all backups that are newer than seven days. - -To clean up the backup catalog in accordance with retention policy, run: - - pg_probackup delete -B backup_dir --instance instance_name --delete-expired - -pg_probackup deletes all backup copies that do not conform to the defined retention policy. - -If you would like to also remove the WAL files that are no longer required for any of the backups, add the `--delete-wal` option: - - pg_probackup delete -B backup_dir --instance instance_name --delete-expired --delete-wal - ->NOTE: Alternatively, you can use the `--delete-expired`, `--merge-expired`, `--delete-wal` flags and the `--retention-window` and `--retention-redundancy` options together with the [backup](#backup) command to remove and merge the outdated backup copies once the new backup is created. - -Since incremental backups require that their parent full backup and all the preceding incremental backups are available, if any of such backups expire, they still cannot be removed while at least one incremental backup in this chain satisfies the retention policy. To avoid keeping expired backups that are still required to restore an active incremental one, you can merge them with this backup using the `--merge-expired` option when running [backup](#backup) or [delete](#delete) commands. - -Suppose you have backed up the *node* instance in the *backup_dir* directory, with the `retention-window` option is set to 7, and you have the following backups available on April 10, 2019: - -``` -BACKUP INSTANCE 'node' -=========================================================================================================================================== - Instance Version ID Recovery time Mode WAL Current/Parent TLI Time Data Start LSN Stop LSN Status -=========================================================================================================================================== - node 10 P7XDHR 2019-04-10 05:27:15+03 FULL STREAM 1 / 0 11s 200MB 0/18000059 0/18000197 OK - node 10 P7XDQV 2019-04-08 05:32:59+03 PAGE STREAM 1 / 0 11s 19MB 0/15000060 0/15000198 OK - node 10 P7XDJA 2019-04-03 05:28:36+03 DELTA STREAM 1 / 0 21s 32MB 0/13000028 0/13000198 OK - ---------------------------------------retention window------------------------------------------------------------- - node 10 P7XDHU 2019-04-02 05:27:59+03 PAGE STREAM 1 / 0 31s 33MB 0/11000028 0/110001D0 OK - node 10 P7XDHB 2019-04-01 05:27:15+03 FULL STREAM 1 / 0 11s 200MB 0/F000028 0/F000198 OK - node 10 P7XDFT 2019-03-29 05:26:25+03 FULL STREAM 1 / 0 11s 200MB 0/D000028 0/D000198 OK -``` - -Even though P7XDHB and P7XDHU backups are outside the retention window, they cannot be removed as it invalidates the succeeding incremental backups P7XDJA and P7XDQV that are still required, so, if you run the [delete](#delete) command with the `--delete-expired` option, only the P7XDFT full backup will be removed. - -With the `--merge-expired` option, the P7XDJA backup is merged with the underlying P7XDHU and P7XDHB backups and becomes a full one, so there is no need to keep these expired backups anymore: - - pg_probackup delete -B backup_dir --instance node --delete-expired --merge-expired - pg_probackup show -B backup_dir - -``` -BACKUP INSTANCE 'node' -============================================================================================================================================ - Instance Version ID Recovery time Mode WAL Current/Parent TLI Time Data Start LSN Stop LSN Status -============================================================================================================================================ - node 10 P7XDHR 2019-04-10 05:27:15+03 FULL STREAM 1 / 0 11s 200MB 0/18000059 0/18000197 OK - node 10 P7XDQV 2019-04-08 05:32:59+03 DELTA STREAM 1 / 0 11s 19MB 0/15000060 0/15000198 OK - node 10 P7XDJA 2019-04-04 05:28:36+03 FULL STREAM 1 / 0 5s 200MB 0/13000028 0/13000198 OK -``` - ->NOTE: The Time field for the merged backup displays the time required for the merge. - -### Merging Backups - -As you take more and more incremental backups, the total size of the backup catalog can substantially grow. To save disk space, you can merge incremental backups to their parent full backup by running the merge command, specifying the backup ID of the most recent incremental backup you would like to merge: - - pg_probackup merge -B backup_dir --instance instance_name -i backup_id - -This command merges the specified incremental backup to its parent full backup, together with all incremental backups between them. Once the merge is complete, the incremental backups are removed as redundant. Thus, the merge operation is virtually equivalent to retaking a full backup and removing all the outdated backups, but it allows to save much time, especially for large data volumes, I/O and network traffic in case of [remote](#using-pg_probackup-in-the-remote-mode) backup. - -Before the merge, pg_probackup validates all the affected backups to ensure that they are valid. You can check the current backup status by running the [show](#show) command with the backup ID: - - pg_probackup show -B backup_dir --instance instance_name -i backup_id - -If the merge is still in progress, the backup status is displayed as MERGING. The merge is idempotent, so you can restart the merge if it was interrupted. - -### Deleting Backups - -To delete a backup that is no longer required, run the following command: - - pg_probackup delete -B backup_dir --instance instance_name -i backup_id - -This command will delete the backup with the specified *backup_id*, together with all the incremental backups that descend from *backup_id* if any. This way you can delete some recent incremental backups, retaining the underlying full backup and some of the incremental backups that follow it. - -To delete obsolete WAL files that are not necessary to restore any of the remaining backups, use the `--delete-wal` option: - - pg_probackup delete -B backup_dir --instance instance_name --delete-wal - -To delete backups that are expired according to the current retention policy, use the `--delete-expired` option: - - pg_probackup delete -B backup_dir --instance instance_name --delete-expired - -Note that expired backups cannot be removed while at least one incremental backup that satisfies the retention policy is based on them. If you would like to minimize the number of backups still required to keep incremental backups valid, specify the `--merge-expired` option when running this command: - - pg_probackup delete -B backup_dir --instance instance_name --delete-expired --merge-expired - -In this case, pg_probackup searches for the oldest incremental backup that satisfies the retention policy and merges this backup with the underlying full and incremental backups that have already expired, thus making it a full backup. Once the merge is complete, the remaining expired backups are deleted. - -Before merging or deleting backups, you can run the delete command with the `--dry-run` option, which displays the status of all the available backups according to the current retention policy, without performing any irreversible actions. - -## Authors -Postgres Professional, Moscow, Russia. - -## Credits -pg_probackup utility is based on pg_arman, that was originally written by NTT and then developed and maintained by Michael Paquier. diff --git a/LICENSE b/LICENSE index dc4e8b8d5..66476e8a9 100644 --- a/LICENSE +++ b/LICENSE @@ -1,4 +1,4 @@ -Copyright (c) 2015-2019, Postgres Professional +Copyright (c) 2015-2023, Postgres Professional Portions Copyright (c) 2009-2013, NIPPON TELEGRAPH AND TELEPHONE CORPORATION Portions Copyright (c) 1996-2016, PostgreSQL Global Development Group diff --git a/Makefile b/Makefile index 499255f8b..f93cc37a4 100644 --- a/Makefile +++ b/Makefile @@ -1,4 +1,7 @@ PROGRAM = pg_probackup +WORKDIR ?= $(CURDIR) +BUILDDIR = $(WORKDIR)/build/ +PBK_GIT_REPO = https://github.com/postgrespro/pg_probackup # utils OBJS = src/utils/configuration.o src/utils/json.o src/utils/logger.o \ @@ -6,90 +9,81 @@ OBJS = src/utils/configuration.o src/utils/json.o src/utils/logger.o \ OBJS += src/archive.o src/backup.o src/catalog.o src/checkdb.o src/configure.o src/data.o \ src/delete.o src/dir.o src/fetch.o src/help.o src/init.o src/merge.o \ - src/parsexlog.o src/pg_probackup.o src/restore.o src/show.o src/util.o \ - src/validate.o + src/parsexlog.o src/ptrack.o src/pg_probackup.o src/restore.o src/show.o src/stream.o \ + src/util.o src/validate.o src/datapagemap.o src/catchup.o # borrowed files -OBJS += src/pg_crc.o src/datapagemap.o src/receivelog.o src/streamutil.o \ +OBJS += src/pg_crc.o src/receivelog.o src/streamutil.o \ src/xlogreader.o -EXTRA_CLEAN = src/pg_crc.c src/datapagemap.c src/datapagemap.h src/logging.h \ +EXTRA_CLEAN = src/pg_crc.c \ src/receivelog.c src/receivelog.h src/streamutil.c src/streamutil.h \ - src/xlogreader.c + src/xlogreader.c src/instr_time.h -INCLUDES = src/datapagemap.h src/logging.h src/streamutil.h src/receivelog.h +ifdef top_srcdir +srchome := $(abspath $(top_srcdir)) +else +top_srcdir=../.. +ifneq (,$(wildcard ../../../contrib/pg_probackup)) +# separate build directory support +srchome := $(abspath $(top_srcdir)/..) +else +srchome := $(abspath $(top_srcdir)) +endif +endif + +# OBJS variable must be finally defined before invoking the include directive +ifneq (,$(wildcard $(srchome)/src/bin/pg_basebackup/walmethods.c)) +OBJS += src/walmethods.o +EXTRA_CLEAN += src/walmethods.c src/walmethods.h +endif ifdef USE_PGXS PG_CONFIG = pg_config PGXS := $(shell $(PG_CONFIG) --pgxs) include $(PGXS) -# !USE_PGXS else subdir=contrib/pg_probackup top_builddir=../.. include $(top_builddir)/src/Makefile.global include $(top_srcdir)/contrib/contrib-global.mk -endif # USE_PGXS - -ifeq ($(top_srcdir),../..) - ifeq ($(LN_S),ln -s) - srchome=$(top_srcdir)/.. - endif -else -srchome=$(top_srcdir) endif -ifeq (,$(filter 9.5 9.6,$(MAJORVERSION))) -OBJS += src/walmethods.o -EXTRA_CLEAN += src/walmethods.c src/walmethods.h -INCLUDES += src/walmethods.h -endif - -PG_CPPFLAGS = -I$(libpq_srcdir) ${PTHREAD_CFLAGS} -Isrc -I$(top_srcdir)/$(subdir)/src +PG_CPPFLAGS = -I$(libpq_srcdir) ${PTHREAD_CFLAGS} -Isrc -I$(srchome)/$(subdir)/src override CPPFLAGS := -DFRONTEND $(CPPFLAGS) $(PG_CPPFLAGS) PG_LIBS_INTERNAL = $(libpq_pgport) ${PTHREAD_CFLAGS} -all: checksrcdir $(INCLUDES); - -$(PROGRAM): $(OBJS) +src/utils/configuration.o: src/datapagemap.h +src/archive.o: src/instr_time.h +src/backup.o: src/receivelog.h src/streamutil.h -src/datapagemap.c: $(top_srcdir)/src/bin/pg_rewind/datapagemap.c - rm -f $@ && $(LN_S) $(srchome)/src/bin/pg_rewind/datapagemap.c $@ -src/datapagemap.h: $(top_srcdir)/src/bin/pg_rewind/datapagemap.h - rm -f $@ && $(LN_S) $(srchome)/src/bin/pg_rewind/datapagemap.h $@ -src/logging.h: $(top_srcdir)/src/bin/pg_rewind/logging.h - rm -f $@ && $(LN_S) $(srchome)/src/bin/pg_rewind/logging.h $@ -src/pg_crc.c: $(top_srcdir)/src/backend/utils/hash/pg_crc.c +src/instr_time.h: $(srchome)/src/include/portability/instr_time.h + rm -f $@ && $(LN_S) $(srchome)/src/include/portability/instr_time.h $@ +src/pg_crc.c: $(srchome)/src/backend/utils/hash/pg_crc.c rm -f $@ && $(LN_S) $(srchome)/src/backend/utils/hash/pg_crc.c $@ -src/receivelog.c: $(top_srcdir)/src/bin/pg_basebackup/receivelog.c +src/receivelog.c: $(srchome)/src/bin/pg_basebackup/receivelog.c rm -f $@ && $(LN_S) $(srchome)/src/bin/pg_basebackup/receivelog.c $@ -src/receivelog.h: $(top_srcdir)/src/bin/pg_basebackup/receivelog.h +ifneq (,$(wildcard $(srchome)/src/bin/pg_basebackup/walmethods.c)) +src/receivelog.h: src/walmethods.h $(srchome)/src/bin/pg_basebackup/receivelog.h +else +src/receivelog.h: $(srchome)/src/bin/pg_basebackup/receivelog.h +endif rm -f $@ && $(LN_S) $(srchome)/src/bin/pg_basebackup/receivelog.h $@ -src/streamutil.c: $(top_srcdir)/src/bin/pg_basebackup/streamutil.c +src/streamutil.c: $(srchome)/src/bin/pg_basebackup/streamutil.c rm -f $@ && $(LN_S) $(srchome)/src/bin/pg_basebackup/streamutil.c $@ -src/streamutil.h: $(top_srcdir)/src/bin/pg_basebackup/streamutil.h +src/streamutil.h: $(srchome)/src/bin/pg_basebackup/streamutil.h rm -f $@ && $(LN_S) $(srchome)/src/bin/pg_basebackup/streamutil.h $@ -src/xlogreader.c: $(top_srcdir)/src/backend/access/transam/xlogreader.c +src/xlogreader.c: $(srchome)/src/backend/access/transam/xlogreader.c rm -f $@ && $(LN_S) $(srchome)/src/backend/access/transam/xlogreader.c $@ - -ifeq (,$(filter 9.5 9.6,$(MAJORVERSION))) -src/walmethods.c: $(top_srcdir)/src/bin/pg_basebackup/walmethods.c +src/walmethods.c: $(srchome)/src/bin/pg_basebackup/walmethods.c rm -f $@ && $(LN_S) $(srchome)/src/bin/pg_basebackup/walmethods.c $@ -src/walmethods.h: $(top_srcdir)/src/bin/pg_basebackup/walmethods.h +src/walmethods.h: $(srchome)/src/bin/pg_basebackup/walmethods.h rm -f $@ && $(LN_S) $(srchome)/src/bin/pg_basebackup/walmethods.h $@ -endif ifeq ($(PORTNAME), aix) CC=xlc_r endif -# This rule's only purpose is to give the user instructions on how to pass -# the path to PostgreSQL source tree to the makefile. -.PHONY: checksrcdir -checksrcdir: -ifndef top_srcdir - @echo "You must have PostgreSQL source tree available to compile." - @echo "Pass the path to the PostgreSQL source tree to make, in the top_srcdir" - @echo "variable: \"make top_srcdir=\"" - @exit 1 -endif +include packaging/Makefile.pkg +include packaging/Makefile.repo +include packaging/Makefile.test diff --git a/README.md b/README.md index fae9ea171..2279b97a4 100644 --- a/README.md +++ b/README.md @@ -1,24 +1,29 @@ +[![GitHub release](https://img.shields.io/github/v/release/postgrespro/pg_probackup?include_prereleases)](https://github.com/postgrespro/pg_probackup/releases/latest) +[![Build Status](https://travis-ci.com/postgrespro/pg_probackup.svg?branch=master)](https://travis-ci.com/postgrespro/pg_probackup) + # pg_probackup `pg_probackup` is a utility to manage backup and recovery of PostgreSQL database clusters. It is designed to perform periodic backups of the PostgreSQL instance that enable you to restore the server in case of a failure. The utility is compatible with: -* PostgreSQL 9.5, 9.6, 10, 11; +* PostgreSQL 11, 12, 13, 14, 15, 16 As compared to other backup solutions, `pg_probackup` offers the following benefits that can help you implement different backup strategies and deal with large amounts of data: -* Choosing between full and page-level incremental backups to speed up backup and recovery -* Implementing a single backup strategy for multi-server PostgreSQL clusters -* Automatic data consistency checks and on-demand backup validation without actual data recovery -* Managing backups in accordance with retention policy -* Merging incremental into full backups without actual data recovery -* Running backup, restore, merge and validation processes on multiple parallel threads -* Storing backup data in a compressed state to save disk space -* Taking backups from a standby server to avoid extra load on the master server -* Extended logging settings -* Custom commands to simplify WAL log archiving -* External to PGDATA directories, such as directories with config files and scripts, can be included in backup -* Remote backup, restore, add-instance, archive-push and archive-get operations via ssh (beta) -* Checking running PostgreSQL instance for the sights of corruption in read-only mode via `checkdb` command. +* Incremental backup: page-level incremental backup allows you to save disk space, speed up backup and restore. With three different incremental modes, you can plan the backup strategy in accordance with your data flow. +* Incremental restore: page-level incremental restore allows you dramatically speed up restore by reusing valid unchanged pages in destination directory. +* Merge: using this feature allows you to implement "incrementally updated backups" strategy, eliminating the need to do periodical full backups. +* Validation: automatic data consistency checks and on-demand backup validation without actual data recovery +* Verification: on-demand verification of PostgreSQL instance with the `checkdb` command. +* Retention: managing WAL archive and backups in accordance with retention policy. You can configure retention policy based on recovery time or the number of backups to keep, as well as specify `time to live` (TTL) for a particular backup. Expired backups can be merged or deleted. +* Parallelization: running backup, restore, merge, delete, verificaton and validation processes on multiple parallel threads +* Compression: storing backup data in a compressed state to save disk space +* Deduplication: saving disk space by not copying unchanged non-data files, such as `_vm` or `_fsm` +* Remote operations: backing up PostgreSQL instance located on a remote system or restoring a backup remotely +* Backup from standby: avoid extra load on master by taking backups from a standby server +* External directories: backing up files and directories located outside of the PostgreSQL `data directory` (PGDATA), such as scripts, configuration files, logs, or SQL dump files. +* Backup Catalog: get list of backups and corresponding meta information in plain text or JSON formats +* Archive catalog: getting the list of all WAL timelines and the corresponding meta information in plain text or JSON formats +* Partial Restore: restore only the specified databases or exclude the specified databases from restore. To manage backup data, `pg_probackup` creates a backup catalog. This directory stores all backup files with additional meta information, as well as WAL archives required for [point-in-time recovery](https://postgrespro.com/docs/postgresql/current/continuous-archiving.html). You can store backups for different instances in separate subdirectories of a single backup catalog. @@ -33,72 +38,63 @@ Regardless of the chosen backup type, all backups taken with `pg_probackup` supp * `Autonomous backups` streams via replication protocol all the WAL files required to restore the cluster to a consistent state at the time the backup was taken. Even if continuous archiving is not set up, the required WAL segments are included into the backup. * `Archive backups` rely on continuous archiving. +## ptrack support + `PTRACK` backup support provided via following options: -* vanilla PostgreSQL compiled with ptrack patch. Currently there are patches for [PostgreSQL 9.6](https://gist.githubusercontent.com/gsmol/5b615c971dfd461c76ef41a118ff4d97/raw/e471251983f14e980041f43bea7709b8246f4178/ptrack_9.6.6_v1.5.patch) and [PostgreSQL 10](https://gist.githubusercontent.com/gsmol/be8ee2a132b88463821021fd910d960e/raw/de24f9499f4f314a4a3e5fae5ed4edb945964df8/ptrack_10.1_v1.5.patch) -* Postgres Pro Standard 9.5, 9.6, 10, 11 -* Postgres Pro Enterprise 9.5, 9.6, 10 +* vanilla PostgreSQL 11, 12, 13, 14, 15, 16 with [ptrack extension](https://github.com/postgrespro/ptrack) +* Postgres Pro Standard 11, 12, 13, 14, 15, 16 +* Postgres Pro Enterprise 11, 12, 13, 14, 15, 16 ## Limitations `pg_probackup` currently has the following limitations: * The server from which the backup was taken and the restored server must be compatible by the [block_size](https://postgrespro.com/docs/postgresql/current/runtime-config-preset#GUC-BLOCK-SIZE) and [wal_block_size](https://postgrespro.com/docs/postgresql/current/runtime-config-preset#GUC-WAL-BLOCK-SIZE) parameters and have the same major release number. -* Remote mode is in beta stage. -* Incremental chain can span only within one timeline. So if you have backup incremental chain taken from replica and it gets promoted, you would be forced to take another FULL backup. +* Remote backup via ssh on Windows currently is not supported. +* When running remote operations via ssh, remote and local pg_probackup versions must be the same. + +## Documentation + +Documentation can be found at [github](https://postgrespro.github.io/pg_probackup) and [Postgres Professional documentation](https://postgrespro.com/docs/postgrespro/current/app-pgprobackup) + +## Development -## Current release +* Stable version state can be found under the respective [release tag](https://github.com/postgrespro/pg_probackup/releases). +* `master` branch contains minor fixes that are planned to the nearest minor release. +* Upcoming major release is developed in a release branch i.e. `release_2_6`. -[2.1.4](https://github.com/postgrespro/pg_probackup/releases/tag/2.1.4) +For detailed release plans check [Milestones](https://github.com/postgrespro/pg_probackup/milestones) ## Installation and Setup ### Windows Installation -[Installers download link](https://oc.postgrespro.ru/index.php/s/CGsjXlc5NmhRI0L) +Installers are available in release **assets**. [Latests](https://github.com/postgrespro/pg_probackup/releases/latest). ### Linux Installation -```shell -#DEB Ubuntu|Debian Packages -echo "deb [arch=amd64] http://repo.postgrespro.ru/pg_probackup/deb/ $(lsb_release -cs) main-$(lsb_release -cs)" > /etc/apt/sources.list.d/pg_probackup.list -wget -O - http://repo.postgrespro.ru/pg_probackup/keys/GPG-KEY-PG_PROBACKUP | apt-key add - && apt-get update -apt-get install pg-probackup-{11,10,9.6,9.5} -apt-get install pg-probackup-{11,10,9.6,9.5}-dbg - -#DEB-SRC Packages -echo "deb-src [arch=amd64] http://repo.postgrespro.ru/pg_probackup/deb/ $(lsb_release -cs) main-$(lsb_release -cs)" >>\ - /etc/apt/sources.list.d/pg_probackup.list -apt-get source pg-probackup-{11,10,9.6,9.5} - -#RPM Centos Packages -rpm -ivh http://repo.postgrespro.ru/pg_probackup/keys/pg_probackup-repo-centos.noarch.rpm -yum install pg_probackup-{11,10,9.6,9.5} -yum install pg_probackup-{11,10,9.6,9.5}-debuginfo - -#RPM RHEL Packages -rpm -ivh http://repo.postgrespro.ru/pg_probackup/keys/pg_probackup-repo-rhel.noarch.rpm -yum install pg_probackup-{11,10,9.6,9.5} -yum install pg_probackup-{11,10,9.6,9.5}-debuginfo - -#RPM Oracle Linux Packages -rpm -ivh http://repo.postgrespro.ru/pg_probackup/keys/pg_probackup-repo-oraclelinux.noarch.rpm -yum install pg_probackup-{11,10,9.6,9.5} -yum install pg_probackup-{11,10,9.6,9.5}-debuginfo - -#SRPM Packages -yumdownloader --source pg_probackup-{11,10,9.6,9.5} -``` -Once you have `pg_probackup` installed, complete [the setup](https://github.com/postgrespro/pg_probackup/blob/master/Documentation.md#installation-and-setup). +See the [Installation](https://postgrespro.github.io/pg_probackup/#pbk-install) section in the documentation. + +Once you have `pg_probackup` installed, complete [the setup](https://postgrespro.github.io/pg_probackup/#pbk-setup). + +For users of Postgres Pro products, commercial editions of pg_probackup are available for installation from the corresponding Postgres Pro product repository. ## Building from source ### Linux -To compile `pg_probackup`, you must have a PostgreSQL installation and raw source tree. To install `pg_probackup`, execute this in the module's directory: +To compile `pg_probackup`, you must have a PostgreSQL installation and raw source tree. Execute this in the module's directory: ```shell make USE_PGXS=1 PG_CONFIG= top_srcdir= ``` + +The alternative way, without using the PGXS infrastructure, is to place `pg_probackup` source directory into `contrib` directory and build it there. Example: + +```shell +cd && git clone https://github.com/postgrespro/pg_probackup contrib/pg_probackup && cd contrib/pg_probackup && make +``` + ### Windows Currently pg_probackup can be build using only MSVC 2013. -Build PostgreSQL using [pgwininstall](https://github.com/postgrespro/pgwininstall) or [PostgreSQL instruction](https://www.postgresql.org/docs/10/install-windows-full.html) with MSVC 2013. +Build PostgreSQL using [pgwininstall](https://github.com/postgrespro/pgwininstall) or [PostgreSQL instruction](https://www.postgresql.org/docs/current/install-windows-full.html) with MSVC 2013. If zlib support is needed, src/tools/msvc/config.pl must contain path to directory with compiled zlib. [Example](https://gist.githubusercontent.com/gsmol/80989f976ce9584824ae3b1bfb00bd87/raw/240032950d4ac4801a79625dd00c8f5d4ed1180c/gistfile1.txt) ```shell @@ -108,10 +104,6 @@ SET PATH=%PATH%;C:\msys64\usr\bin gen_probackup_project.pl C:\path_to_postgresql_source_tree ``` -## Documentation - -Currently the latest documentation can be found at [github](https://postgrespro.github.io/pg_probackup) and [Postgres Pro Enterprise documentation](https://postgrespro.com/docs/postgrespro/current/app-pgprobackup). - ## License This module available under the [license](LICENSE) similar to [PostgreSQL](https://www.postgresql.org/about/license/). @@ -127,3 +119,17 @@ Postgres Professional, Moscow, Russia. ## Credits `pg_probackup` utility is based on `pg_arman`, that was originally written by NTT and then developed and maintained by Michael Paquier. + + +### Localization files (*.po) + +Description of how to add new translation languages. +1. Add a flag --enable-nls in configure. +2. Build postgres. +3. Adding to nls.mk in folder pg_probackup required files in GETTEXT_FILES. +4. In folder pg_probackup do 'make update-po'. +5. As a result, the progname.pot file will be created. Copy the content and add it to the file with the desired language. +6. Adding to nls.mk in folder pg_probackup required language in AVAIL_LANGUAGES. + +For more information, follow the link below: +https://postgrespro.ru/docs/postgresql/12/nls-translator diff --git a/doc/Readme.md b/doc/Readme.md new file mode 100644 index 000000000..0e1d64590 --- /dev/null +++ b/doc/Readme.md @@ -0,0 +1,8 @@ +# Generating documentation +``` +xmllint --noout --valid probackup.xml +xsltproc stylesheet.xsl probackup.xml >pg-probackup.html +``` +> [!NOTE] +>Install ```docbook-xsl``` if you got +>``` "xsl:import : unable to load http://docbook.sourceforge.net/release/xsl/current/xhtml/docbook.xsl"``` \ No newline at end of file diff --git a/doc/pgprobackup.xml b/doc/pgprobackup.xml new file mode 100644 index 000000000..10e766239 --- /dev/null +++ b/doc/pgprobackup.xml @@ -0,0 +1,6327 @@ + + + + pg_probackup + + + + pg_probackup + 1 + Application + + + + pg_probackup + manage backup and recovery of PostgreSQL database clusters + + + + + + pg_probackup + + + + pg_probackup + + command + + + pg_probackup + + backup_dir + + + pg_probackup + + backup_dir + data_dir + instance_name + + + pg_probackup + + backup_dir + instance_name + + + pg_probackup + + backup_dir + instance_name + option + + + pg_probackup + + backup_dir + instance_name + backup_id + option + + + pg_probackup + + backup_dir + instance_name + + + + pg_probackup + + backup_dir + option + + + pg_probackup + + backup_dir + instance_name + backup_mode + option + + + pg_probackup + + backup_dir + instance_name + option + + + pg_probackup + + backup_dir + instance_name + data_dir + + + + pg_probackup + + backup_dir + option + + + pg_probackup + + backup_dir + instance_name + backup_id + option + + + pg_probackup + + backup_dir + instance_name + + backup_id + + + + + option + + + pg_probackup + + backup_dir + instance_name + wal_file_path + wal_file_name + option + + + pg_probackup + + backup_dir + instance_name + wal_file_path + wal_file_name + option + + + pg_probackup + + catchup_mode + =path_to_pgdata_on_remote_server + =path_to_local_dir + option + + + + + + + Description + + + pg_probackup is a utility to manage backup and + recovery of PostgreSQL database clusters. + It is designed to perform periodic backups of the PostgreSQL + instance that enable you to restore the server in case of a failure. + pg_probackup supports PostgreSQL 11 or higher. + + + + + Overview + + + Quick Start + + + Installation + + + Setup + + + Command-Line Reference + + + Usage + + + + + + + + Overview + + + + As compared to other backup solutions, pg_probackup offers the + following benefits that can help you implement different backup + strategies and deal with large amounts of data: + + + + + Incremental backup: with three different incremental modes, + you can plan the backup strategy in accordance with your data flow. + Incremental backups allow you to save disk space + and speed up backup as compared to taking full backups. + It is also faster to restore the cluster by applying incremental + backups than by replaying WAL files. + + + + + Incremental restore: speed up restore from backup by reusing + valid unchanged pages available in PGDATA. + + + + + Validation: automatic data consistency checks and on-demand + backup validation without actual data recovery. + + + + + Verification: on-demand verification of PostgreSQL instance + with the checkdb command. + + + + + Retention: managing WAL archive and backups in accordance with + retention policy. You can configure retention policy based on recovery time + or the number of backups to keep, as well as specify time to live (TTL) + for a particular backup. Expired backups can be merged or deleted. + + + + + Parallelization: running backup, + restore, merge, + delete, validate, + and checkdb processes on multiple parallel threads. + + + + + Compression: storing backup data in a compressed state to save + disk space. + + + + + Deduplication: saving disk space by excluding non-data + files (such as _vm or _fsm) + from incremental backups if these files have not changed since + they were copied into one of the previous backups in this incremental chain. + + + + + Remote operations: backing up PostgreSQL + instance located on a remote system or restoring a backup remotely. + + + + + Backup from standby: avoiding extra load on master by + taking backups from a standby server. + + + + + External directories: backing up files and directories + located outside of the PostgreSQL data + directory (PGDATA), such as scripts, configuration + files, logs, or SQL dump files. + + + + + Backup catalog: getting the list of backups and the corresponding meta + information in plain text or + JSON formats. + + + + + Archive catalog: getting the list of all WAL timelines and + the corresponding meta information in plain text or + JSON formats. + + + + + Partial restore: restoring only the specified databases. + + + + + Catchup: cloning a PostgreSQL instance for a fallen-behind standby server to catch up with master. + + + + + To manage backup data, pg_probackup creates a + backup catalog. This is a directory that stores + all backup files with additional meta information, as well as WAL + archives required for point-in-time recovery. You can store + backups for different instances in separate subdirectories of a + single backup catalog. + + + Using pg_probackup, you can take full or incremental + backups: + + + + + FULL backups contain all the data files required to restore + the database cluster. + + + + + Incremental backups operate at the page level, only storing the data that has changed since + the previous backup. It allows you to save disk space + and speed up the backup process as compared to taking full backups. + It is also faster to restore the cluster by applying incremental + backups than by replaying WAL files. pg_probackup supports + the following modes of incremental backups: + + + + + DELTA backup. In this mode, pg_probackup reads all data + files in the data directory and copies only those pages + that have changed since the previous backup. This + mode can impose read-only I/O pressure equal to a full + backup. + + + + + PAGE backup. In this mode, pg_probackup scans all WAL + files in the archive from the moment the previous full or + incremental backup was taken. Newly created backups + contain only the pages that were mentioned in WAL records. + This requires all the WAL files since the previous backup + to be present in the WAL archive. If the size of these + files is comparable to the total size of the database + cluster files, speedup is smaller, but the backup still + takes less space. You have to configure WAL archiving as + explained in Setting + up continuous WAL archiving to make PAGE backups. + + + + + PTRACK backup. In this mode, PostgreSQL tracks page + changes on the fly. Continuous archiving is not necessary + for it to operate. Each time a relation page is updated, + this page is marked in a special PTRACK bitmap. Tracking implies some + minor overhead on the database server operation, but + speeds up incremental backups significantly. + + + + + + + pg_probackup can take only physical online backups, and online + backups require WAL for consistent recovery. So regardless of the + chosen backup mode (FULL, PAGE or DELTA), any backup taken with + pg_probackup must use one of the following + WAL delivery modes: + + + + + ARCHIVE. Such backups rely + on + continuous + archiving to ensure consistent recovery. This is the + default WAL delivery mode. + + + + + STREAM. Such backups + include all the files required to restore the cluster to a + consistent state at the time the backup was taken. Regardless + of + continuous + archiving having been set up or not, the WAL segments required + for consistent recovery are streamed via + replication protocol during backup and included into the + backup files. That's why such backups are called + autonomous, or standalone. + + + + + Limitations + + pg_probackup currently has the following limitations: + + + + pg_probackup only supports PostgreSQL 9.5 and higher. + + + + + The remote mode is not supported on Windows systems. + + + + + On Unix systems, for PostgreSQL 11, + a backup can be made only by the same OS user that has started the PostgreSQL + server. For example, if PostgreSQL server is started by + user postgres, the backup command must also be run + by user postgres. To satisfy this requirement when taking backups in the + remote mode using SSH, you must set + option to postgres. + + + + + For PostgreSQL 9.5, functions + pg_create_restore_point(text) and + pg_switch_xlog() can be executed only if + the backup role is a superuser, so backup of a + cluster with low amount of WAL traffic by a non-superuser + role can take longer than the backup of the same cluster by + a superuser role. + + + + + The PostgreSQL server from which the backup was taken and + the restored server must be compatible by the + block_size + and + wal_block_size + parameters and have the same major release number. + Depending on cluster configuration, PostgreSQL itself may + apply additional restrictions, such as CPU architecture + or libc/icu versions. + + + + + + + + Quick Start + + To quickly get started with pg_probackup, complete the steps below. This will set up FULL and DELTA backups in the remote mode and demonstrate some + basic pg_probackup operations. In the following, these terms are used: + + + + + backup — PostgreSQL + role used to connect to the PostgreSQL + cluster. + + + + + backupdb — database used to connect to the + PostgreSQL cluster. + + + + + backup_host — host with the backup catalog. + + + + + backup_user — user on + backup_host running all pg_probackup + operations. + + + + + /mnt/backups — directory on + backup_host where the backup catalog is stored. + + + + + postgres_host — host with the + PostgreSQL cluster. + + + + + postgres — user on + postgres_host under which + PostgreSQL cluster processes are running. + + + + + /var/lib/postgresql/16/main — + PostgreSQL data directory on + postgres_host. + + + + + Steps to perform: + + + Install pg_probackup on both backup_host and postgres_host. + + + Set up an SSH connection from backup_host to postgres_host. + + + Configure your database cluster for STREAM backups. + + + Initialize the backup catalog: + +backup_user@backup_host:~$ pg_probackup-16 init -B /mnt/backups +INFO: Backup catalog '/mnt/backups' successfully initialized + + + + Add a backup instance called mydb to the backup catalog: + +backup_user@backup_host:~$ pg_probackup-16 add-instance \ + -B /mnt/backups \ + -D /var/lib/pgpro/std-16/data \ + --instance=node \ + --remote-host=postgres_host \ + --remote-user=postgres +INFO: Instance 'node' successfully initialized + + + + Make a FULL backup: + +backup_user@backup_host:~$ pg_probackup-16 backup \ + -B /mnt/backups \ + -b FULL \ + --instance=node \ + --stream \ + --compress-algorithm=zlib \ + --remote-host=postgres_host \ + --remote-user=postgres \ + -U backup \ + -d backupdb +INFO: Backup start, pg_probackup version: 2.5.15, instance: node, backup ID: SCUN1Q, backup mode: FULL, wal mode: STREAM, remote: true, compress-algorithm: zlib, compress-level: 1 +INFO: This PostgreSQL instance was initialized with data block checksums. Data block corruption will be detected +INFO: Database backup start +INFO: wait for pg_backup_start() +INFO: Wait for WAL segment /mnt/backups/backups/node/SCUN1Q/database/pg_wal/000000010000000000000008 to be streamed +INFO: PGDATA size: 96MB +INFO: Current Start LSN: 0/8000028, TLI: 1 +INFO: Start transferring data files +INFO: Data files are transferred, time elapsed: 1s +INFO: wait for pg_stop_backup() +INFO: pg_stop backup() successfully executed +INFO: stop_lsn: 0/800BBD0 +INFO: Getting the Recovery Time from WAL +INFO: Syncing backup files to disk +INFO: Backup files are synced, time elapsed: 1s +INFO: Validating backup SCUN1Q +INFO: Backup SCUN1Q data files are valid +INFO: Backup SCUN1Q resident size: 56MB +INFO: Backup SCUN1Q completed + + + + List the backups of the instance: + +backup_user@backup_host:~$ pg_probackup-16 show \ + -B /mnt/backups \ + --instance=node +================================================================================================================================ + Instance Version ID Recovery Time Mode WAL Mode TLI Time Data WAL Zratio Start LSN Stop LSN Status +================================================================================================================================ + node 16 SCUN1Q 2024-05-02 11:17:53+03 FULL STREAM 1/0 12s 40MB 16MB 2.42 0/8000028 0/800BBD0 OK + + + + Make an incremental backup in the DELTA mode: + +backup_user@backup_host:~$ pg_probackup-16 backup \ + -B /mnt/backups \ + -b DELTA \ + --instance=node \ + --stream \ + --compress-algorithm=zlib \ + --remote-host=postgres_host \ + --remote-user=postgres \ + -U backup \ + -d backupdb +INFO: Backup start, pg_probackup version: 2.5.15, instance: node, backup ID: SCUN22, backup mode: DELTA, wal mode: STREAM, remote: true, compress-algorithm: zlib, compress-level: 1 +INFO: This PostgreSQL instance was initialized with data block checksums. Data block corruption will be detected +INFO: Database backup start +INFO: wait for pg_backup_start() +INFO: Parent backup: SCUN1Q +INFO: Wait for WAL segment /mnt/backups/backups/node/SCUN22/database/pg_wal/000000010000000000000009 to be streamed +INFO: PGDATA size: 96MB +INFO: Current Start LSN: 0/9000028, TLI: 1 +INFO: Parent Start LSN: 0/8000028, TLI: 1 +INFO: Start transferring data files +INFO: Data files are transferred, time elapsed: 1s +INFO: wait for pg_stop_backup() +INFO: pg_stop backup() successfully executed +INFO: stop_lsn: 0/9000168 +INFO: Getting the Recovery Time from WAL +INFO: Syncing backup files to disk +INFO: Backup files are synced, time elapsed: 1s +INFO: Validating backup SCUN22 +INFO: Backup SCUN22 data files are valid +INFO: Backup SCUN22 resident size: 34MB +INFO: Backup SCUN22 completed + + + + Add or modify some parameters in the pg_probackup + configuration file, so that you do not have to specify them each time on the command line: + +backup_user@backup_host:~$ pg_probackup-16 set-config \ + -B /mnt/backups \ + --instance=node \ + --remote-host=postgres_host \ + --remote-user=postgres \ + -U backup \ + -d backupdb + + + + Check the configuration of the instance: + +backup_user@backup_host:~$ pg_probackup-16 show-config \ + -B /mnt/backups \ + --instance=node +# Backup instance information +pgdata = /var/lib/pgpro/std-16/data +system-identifier = 7364313570668255886 +xlog-seg-size = 16777216 +# Connection parameters +pgdatabase = backupdb +pghost = postgres_host +pguser = backup +# Replica parameters +replica-timeout = 5min +# Archive parameters +archive-timeout = 5min +# Logging parameters +log-level-console = INFO +log-level-file = OFF +log-format-console = PLAIN +log-format-file = PLAIN +log-filename = pg_probackup.log +log-rotation-size = 0TB +log-rotation-age = 0d +# Retention parameters +retention-redundancy = 0 +retention-window = 0 +wal-depth = 0 +# Compression parameters +compress-algorithm = none +compress-level = 1 +# Remote access parameters +remote-proto = ssh +remote-host = postgres_host +remote-user = postgres + + + Note that the parameters not modified via set-config retain their default values. + + + + Make another incremental backup in the DELTA mode, omitting + the parameters stored in the configuration file earlier: + +backup_user@backup_host:~$ pg_probackup-16 backup \ + -B /mnt/backups \ + -b DELTA \ + --instance=node \ + --stream \ + --compress-algorithm=zlib +INFO: Backup start, pg_probackup version: 2.5.15, instance: node, backup ID: SCUN2C, backup mode: DELTA, wal mode: STREAM, remote: true, compress-algorithm: zlib, compress-level: 1 +INFO: This PostgreSQL instance was initialized with data block checksums. Data block corruption will be detected +INFO: Database backup start +INFO: wait for pg_backup_start() +INFO: Parent backup: SCUN22 +INFO: Wait for WAL segment /mnt/backups/backups/node/SCUN2C/database/pg_wal/00000001000000000000000B to be streamed +INFO: PGDATA size: 96MB +INFO: Current Start LSN: 0/B000028, TLI: 1 +INFO: Parent Start LSN: 0/9000028, TLI: 1 +INFO: Start transferring data files +INFO: Data files are transferred, time elapsed: 0 +INFO: wait for pg_stop_backup() +INFO: pg_stop backup() successfully executed +INFO: stop_lsn: 0/B000168 +INFO: Getting the Recovery Time from WAL +INFO: Syncing backup files to disk +INFO: Backup files are synced, time elapsed: 0 +INFO: Validating backup SCUN2C +INFO: Backup SCUN2C data files are valid +INFO: Backup SCUN2C resident size: 17MB +INFO: Backup SCUN2C completed + + + + List the backups of the instance again: + +backup_user@backup_host:~$ pg_probackup-16 show \ + -B /mnt/backups \ + --instance=node +=================================================================================================================================== + Instance Version ID Recovery Time Mode WAL Mode TLI Time Data WAL Zratio Start LSN Stop LSN Status +=================================================================================================================================== + node 16 SCUN2C 2024-05-02 11:18:13+03 DELTA STREAM 1/1 10s 1139kB 16MB 1.00 0/B000028 0/B000168 OK + node 16 SCUN22 2024-05-02 11:18:04+03 DELTA STREAM 1/1 10s 2357kB 32MB 1.02 0/9000028 0/9000168 OK + node 16 SCUN1Q 2024-05-02 11:17:53+03 FULL STREAM 1/0 12s 40MB 16MB 2.42 0/8000028 0/800BBD0 OK + + + + Restore the data from the latest available backup to an arbitrary location: + +backup_user@backup_host:~$ pg_probackup-16 restore \ + -B /mnt/backups \ + -D /var/lib/pgpro/std-16/staging-data \ + --instance=node +INFO: Validating parents for backup SCUN2C +INFO: Validating backup SCUN1Q +INFO: Backup SCUN1Q data files are valid +INFO: Validating backup SCUN22 +INFO: Backup SCUN22 data files are valid +INFO: Validating backup SCUN2C +INFO: Backup SCUN2C data files are valid +INFO: Backup SCUN2C WAL segments are valid +INFO: Backup SCUN2C is valid. +INFO: Restoring the database from backup SCUN2C on localhost +INFO: Start restoring backup files. PGDATA size: 112MB +INFO: Backup files are restored. Transfered bytes: 112MB, time elapsed: 0 +INFO: Restore incremental ratio (less is better): 100% (112MB/112MB) +INFO: Syncing restored files to disk +INFO: Restored backup files are synced, time elapsed: 2s +INFO: Restore of backup SCUN2C completed. + + + + + + + Installation + + Installation on Debian family systems (Debian, Ubuntu etc.) + + You may need to use apt-get instead of apt on older systems in the commands below. + + + + + Add the pg_probackup repository GPG key + + +sudo apt install gpg wget +wget -qO - https://repo.postgrespro.ru/pg_probackup/keys/GPG-KEY-PG-PROBACKUP | \ +sudo tee /etc/apt/trusted.gpg.d/pg_probackup.asc + + + + + Setup the binary package repository + + +. /etc/os-release +echo "deb [arch=amd64] https://repo.postgrespro.ru/pg_probackup/deb $VERSION_CODENAME main-$VERSION_CODENAME" | \ +sudo tee /etc/apt/sources.list.d/pg_probackup.list + + + + + Optionally setup the source package repository for rebuilding the binaries + + +echo "deb-src [arch=amd64] https://repo.postgrespro.ru/pg_probackup/deb $VERSION_CODENAME main-$VERSION_CODENAME" | \ +sudo tee -a /etc/apt/sources.list.d/pg_probackup.list + + + + + List the available pg_probackup packages + + + + + Using apt: + + +sudo apt update +apt search pg_probackup + + + + + Using apt-get: + + +sudo apt-get update +apt-cache search pg_probackup + + + + + + + Install or upgrade a pg_probackup version of your choice + + +sudo apt install pg-probackup-16 + + + + + Optionally install the debug package + + +sudo apt install pg-probackup-16-dbg + + + + + Optionally install the source package (provided you have set up the source package repository as described above) + + +sudo apt install dpkg-dev +sudo apt source pg-probackup-16 + + + + + + Installation on Red Hat family systems (CentOS, Oracle Linux etc.) + + You may need to use yum instead of dnf on older systems in the commands below. + + + + + Install the pg_probackup repository + + +dnf install https://repo.postgrespro.ru/pg_probackup/keys/pg_probackup-repo-centos.noarch.rpm + + + + + List the available pg_probackup packages + + +dnf search pg_probackup + + + + + Install or upgrade a pg_probackup version of your choice + + +dnf install pg_probackup-16 + + + + + Optionally install the debug package + + +dnf install pg_probackup-16-debuginfo + + + + + Optionally install the source package for rebuilding the binaries + + + + + Using dnf: + + +dnf install 'dnf-command(download)' +dnf download --source pg_probackup-16 + + + + + Using yum: + + +yumdownloader --source pg_probackup-16 + + + + + + + + Installation on ALT Linux + + + + Setup the repository + + + + + On ALT Linux 10: + + +. /etc/os-release +echo "rpm http://repo.postgrespro.ru/pg_probackup/rpm/latest/altlinux-p$VERSION_ID x86_64 vanilla" | \ +sudo tee /etc/apt/sources.list.d/pg_probackup.list + + + + + On ALT Linux 8 and 9: + + +. /etc/os-release +echo "rpm http://repo.postgrespro.ru/pg_probackup/rpm/latest/altlinux-$VERSION_ID x86_64 vanilla" | \ +sudo tee /etc/apt/sources.list.d/pg_probackup.list + + + + + + + List the available pg_probackup packages + + +sudo apt-get update +apt-cache search pg_probackup + + + + + Install or upgrade a pg_probackup version of your choice + + +sudo apt-get install pg_probackup-16 + + + + + Optionally install the debug package + + +sudo apt-get install pg_probackup-16-debuginfo + + + + + + Installation on SUSE Linux + + + + Add the pg_probackup repository GPG key + + +zypper in -y gpg wget +wget -O GPG-KEY-PG_PROBACKUP https://repo.postgrespro.ru/pg_probackup/keys/GPG-KEY-PG_PROBACKUP +rpm --import GPG-KEY-PG_PROBACKUP + + + + + Setup the repository + + +zypper in https://repo.postgrespro.ru/pg_probackup/keys/pg_probackup-repo-suse.noarch.rpm + + + + + List the available pg_probackup packages + + +zypper se pg_probackup + + + + + Install or upgrade a pg_probackup version of your choice + + +zypper in pg_probackup-16 + + + + + Optionally install the source package for rebuilding the binaries + + +zypper si pg_probackup-16 + + + + + + + Setup + + Once you have pg_probackup installed, complete the following + setup: + + + + + Initialize the backup catalog. + + + + + Add a new backup instance to the backup catalog. + + + + + Configure the database cluster to enable pg_probackup backups. + + + + + Optionally, configure SSH for running pg_probackup operations + in the remote mode. + + + + + Initializing the Backup Catalog + + pg_probackup stores all WAL and backup files in the + corresponding subdirectories of the backup catalog. + + + To initialize the backup catalog, run the following command: + + +pg_probackup init -B backup_dir + + + where backup_dir is the path to the backup + catalog. If the backup_dir already exists, + it must be empty. Otherwise, pg_probackup returns an error. + + + The user launching pg_probackup must have full access to + the backup_dir directory. + + + pg_probackup creates the backup_dir backup + catalog, with the following subdirectories: + + + + + wal/ — directory for WAL files. + + + + + backups/ — directory for backup files. + + + + + Once the backup catalog is initialized, you can add a new backup + instance. + + + + Adding a New Backup Instance + + pg_probackup can store backups for multiple database clusters in + a single backup catalog. To set up the required subdirectories, + you must add a backup instance to the backup catalog for each + database cluster you are going to back up. + + + To add a new backup instance, run the following command: + + +pg_probackup add-instance -B backup_dir -D data_dir --instance=instance_name [remote_options] + + + Where: + + + + + data_dir is the data directory of the + cluster you are going to back up. To set up and use + pg_probackup, write access to this directory is required. + + + + + instance_name is the name of the + subdirectories that will store WAL and backup files for this + cluster. + + + + + remote_options + are optional parameters that need to be specified only if + data_dir is located + on a remote system. + + + + + pg_probackup creates the instance_name + subdirectories under the backups/ and wal/ directories of + the backup catalog. The + backups/instance_name directory contains + the pg_probackup.conf configuration file that controls + pg_probackup settings for this backup instance. To add + remote_options to the configuration file, use the + command. + + + For details on how to fine-tune pg_probackup configuration, see + . + + + The user launching pg_probackup must have full access to + backup_dir directory and at least read-only + access to data_dir directory. If you + specify the path to the backup catalog in the + BACKUP_PATH environment variable, you can + omit the corresponding option when running pg_probackup + commands. + + + + For PostgreSQL 11 or higher, it is recommended to use the + allow-group-access + feature, so that backup can be done by any OS user in the same + group as the cluster owner. In this case, the user should have + read permissions for the cluster directory. + + + + + Configuring the Database Cluster + + Although pg_probackup can be used by a superuser, it is + recommended to create a separate role with the minimum + permissions required for the chosen backup strategy. In these + configuration instructions, the backup role + is used as an example. + + + For security reasons, it is recommended to run the configuration SQL queries below + in a separate database. + + +postgres=# CREATE DATABASE backupdb; +postgres=# \c backupdb + + + To perform a , the following + permissions for role backup are required + only in the database used for + connection to the PostgreSQL server. + + + For PostgreSQL versions 11 — 14: + + +BEGIN; +CREATE ROLE backup WITH LOGIN; +GRANT USAGE ON SCHEMA pg_catalog TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_checkpoint() TO backup; +COMMIT; + + + For PostgreSQL 15 or higher: + + +BEGIN; +CREATE ROLE backup WITH LOGIN; +GRANT USAGE ON SCHEMA pg_catalog TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_start(text, boolean) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_stop(boolean) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; +GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_checkpoint() TO backup; +COMMIT; + + + In the + pg_hba.conf + file, allow connection to the database cluster on behalf of the + backup role. + + + Since pg_probackup needs to read cluster files directly, + pg_probackup must be started by (or connected to, + if used in the remote mode) the OS user that has read access to all files and + directories inside the data directory (PGDATA) you are going to + back up. + + + Depending on whether you plan to take + standalone or + archive backups, PostgreSQL + cluster configuration will differ, as specified in the sections + below. To back up the database cluster from a standby server, + run pg_probackup in the remote mode, or create PTRACK backups, + additional setup is required. + + + For details, see the sections + Setting up STREAM + Backups, + Setting up + continuous WAL archiving, + Setting up Backup + from Standby, + Configuring the + Remote Mode, + Setting up Partial + Restore, and + Setting up PTRACK + Backups. + + + + Setting up STREAM Backups + + To set up the cluster for + STREAM backups, complete the + following steps: + + + + + If the backup role does not exist, create it with + the REPLICATION privilege when + Configuring the + Database Cluster: + + +CREATE ROLE backup WITH LOGIN REPLICATION; + + + + + If the backup role already exists, grant it with the REPLICATION privilege: + + +ALTER ROLE backup WITH REPLICATION; + + + + + In the + pg_hba.conf + file, allow replication on behalf of the + backup role. + + + + + Make sure the parameter + max_wal_senders + is set high enough to leave at least one session available + for the backup process. + + + + + Set the parameter + wal_level + to be higher than minimal. + + + + + If you are planning to take PAGE backups in the STREAM mode or + perform PITR with STREAM backups, you still have to configure + WAL archiving, as explained in the section + Setting up + continuous WAL archiving. + + + Once these steps are complete, you can start taking FULL, PAGE, + DELTA, and PTRACK backups in the + STREAM WAL mode. + + + + If you are planning to rely on + .pgpass + for authentication when running backup in STREAM mode, + then .pgpass must contain credentials for replication database, + used to establish connection via replication protocol. Example: + pghost:5432:replication:backup_user:my_strong_password + + + + + Setting up Continuous WAL Archiving + + Making backups in PAGE backup mode, performing + PITR and + making backups with + ARCHIVE WAL delivery mode + require + continuous + WAL archiving to be enabled. To set up continuous + archiving in the cluster, complete the following steps: + + + + + Make sure the + wal_level + parameter is higher than minimal. + + + + + If you are configuring archiving on master, + archive_mode + must be set to on or + always. To perform archiving on standby, + set this parameter to always. + + + + + Set the + archive_command + parameter, as follows: + + +archive_command = '"install_dir/pg_probackup" archive-push -B "backup_dir" --instance=instance_name --wal-file-name=%f [remote_options]' + + + + + where install_dir is the + installation directory of the pg_probackup + version you are going to use, backup_dir and + instance_name refer to the already + initialized backup catalog instance for this database cluster, + and remote_options + only need to be specified to archive WAL on a remote host. For details about all + possible archive-push parameters, see the + section . + + + Once these steps are complete, you can start making backups in the + ARCHIVE WAL mode, backups in + the PAGE backup mode, as well as perform + PITR. + + + You can view the current state of the WAL archive using the + command. For details, see + . + + + If you are planning to make PAGE backups and/or backups with + ARCHIVE WAL mode from a + standby server that generates a small amount of WAL traffic, + without long waiting for WAL segment to fill up, consider + setting the + archive_timeout + PostgreSQL parameter on + master. The value of this parameter should be slightly + lower than the setting (5 minutes by default), + so that there is enough time for the rotated + segment to be streamed to standby and sent to WAL archive before the + backup is aborted because of . + + + + Instead of using the + command provided by pg_probackup, you can use + any other tool to set up continuous archiving as long as it delivers WAL segments into + backup_dir/wal/instance_name + directory. If compression is used, it should be + gzip, and .gz suffix in filename is + mandatory. + + + + + Instead of configuring continuous archiving by setting the + archive_mode and archive_command + parameters, you can opt for using the + pg_receivewal + utility. In this case, pg_receivewal -D directory + option should point to + backup_dir/wal/instance_name + directory. pg_probackup supports WAL compression + that can be done by pg_receivewal. + Zero Data Loss archive strategy can be + achieved only by using pg_receivewal. + + + + + Setting up Backup from Standby + + pg_probackup can take backups from + a standby server. This requires the following additional setup: + + + + + On the standby server, set the + hot_standby + parameter to on. + + + + + On the master server, set the + full_page_writes + parameter to on. + + + + + To perform standalone backups on standby, complete all steps + in section Setting + up STREAM Backups. + + + + + To perform archive backups on standby, complete all steps in + section + Setting + up continuous WAL archiving. + + + + + Once these steps are complete, you can start taking FULL, PAGE, + DELTA, or PTRACK backups with appropriate WAL delivery mode: + ARCHIVE or STREAM, from the standby server. + + + Backup from the standby server has the following limitations: + + + + + If the standby is promoted to the master during backup, the + backup fails. + + + + + All WAL records required for the backup must contain + sufficient full-page writes. This requires you to enable + full_page_writes on the master, and not + to use tools like pg_compresslog as + archive_command + to remove full-page writes from WAL files. + + + + + + Setting up Cluster Verification + + Logical verification of a database cluster requires the following + additional setup. Role backup is used as an + example: + + + + + Install the + amcheck + or + amcheck_next extension + in every database of the + cluster: + + +CREATE EXTENSION amcheck; + + + + + Grant the following permissions to the backup + role in every database of the cluster: + + + + +GRANT SELECT ON TABLE pg_catalog.pg_am TO backup; +GRANT SELECT ON TABLE pg_catalog.pg_class TO backup; +GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; +GRANT SELECT ON TABLE pg_catalog.pg_namespace TO backup; +GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; +GRANT EXECUTE ON FUNCTION bt_index_check(regclass) TO backup; +GRANT EXECUTE ON FUNCTION bt_index_check(regclass, bool) TO backup; +GRANT EXECUTE ON FUNCTION bt_index_check(regclass, bool, bool) TO backup; + + + + Setting up Partial Restore + + If you are planning to use partial restore, complete the + following additional step: + + + + + Grant the read-only access to pg_catalog.pg_database to the + backup role only in the database + used for connection to + PostgreSQL server: + + +GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; + + + + + + Configuring the Remote Mode + + pg_probackup supports the remote mode that + allows you to perform backup, restore and WAL archiving operations remotely. + In this mode, the backup catalog is stored on a local system, while + PostgreSQL instance to backup and/or to restore + is located on a remote system. Currently the only supported remote + protocol is SSH. + + + Set up SSH + + If you are going to use pg_probackup in remote mode via SSH, + complete the following steps: + + + + + Install pg_probackup on both systems: + backup_host and + postgres_host. + + + + + For communication between the hosts set up a passwordless + SSH connection between the backup_user user on + backup_host and the + postgres user on + postgres_host: + + +[backup_user@backup_host] ssh-copy-id postgres@postgres_host + + + Where: + + + + + backup_host is the system with + backup catalog. + + + + + postgres_host is the system with the PostgreSQL + cluster. + + + + + backup_user is the OS user on + backup_host used to run pg_probackup. + + + + + postgres is the user on + postgres_host under which + PostgreSQL cluster processes are running. + For PostgreSQL 11 or higher a + more secure approach can be used thanks to + allow-group-access feature. + + + + + + + If you are going to rely on + continuous + WAL archiving, set up a passwordless SSH + connection between the postgres user on + postgres_host and the backup + user on backup_host: + + +[postgres@postgres_host] ssh-copy-id backup_user@backup_host + + + + + Make sure pg_probackup on postgres_host + can be located when a connection via SSH is made. For example, for Bash, you can + modify PATH in ~/.bashrc of the postgres user + (above the line in bashrc that exits the script for non-interactive shells). + Alternatively, for pg_probackup commands, specify the path to the directory + containing the pg_probackup binary on postgres_host via + the --remote-path option. + + + + + pg_probackup in the remote mode via SSH works + as follows: + + + + + Only the following commands can be launched in the remote + mode: , + , + , + , + , and + . + + + + + Operating in remote mode requires pg_probackup + binary to be installed on both local and remote systems. + The versions of local and remote binary must be the same. + + + + + When started in the remote mode, the main pg_probackup process + on the local system connects to the remote system via SSH and + launches one or more agent processes on the remote system, which are called + remote agents. The number of remote agents + is equal to the / setting. + + + + + The main pg_probackup process uses remote agents to access + remote files and transfer data between local and remote + systems. + + + + + Remote agents try to minimize the network traffic and the number of + round-trips between hosts. + + + + + The main process is usually started on + backup_host and connects to + postgres_host, but in case of + archive-push and + archive-get commands the main process + is started on postgres_host and connects to + backup_host. + + + + + Once data transfer is complete, remote agents are + terminated and SSH connections are closed. + + + + + If an error condition is encountered by a remote agent, + then all agents are terminated and error details are + reported by the main pg_probackup process, which exits + with an error. + + + + + Compression is always done on + postgres_host, while decompression is always done on + backup_host. + + + + + + You can impose + additional + restrictions on SSH settings to protect the system + in the event of account compromise. + + + + + + Setting up PTRACK Backups + + The PTRACK backup mode can be used only for Postgres Pro Standard and + Postgres Pro Enterprise installations, or patched vanilla + PostgreSQL. Links to PTRACK patches can be found + here. + + + + PTRACK versions lower than 2.0 are deprecated and not supported. Postgres Pro Standard and Postgres Pro Enterprise + versions starting with 11.9.1 contain PTRACK 2.0. Upgrade your server to avoid issues in backups + that you will take in future and be sure to take fresh backups of your clusters with the upgraded + PTRACK since the backups taken with PTRACK 1.x might be corrupt. + + + + If you are going to use PTRACK backups, complete the following + additional steps. The role that will perform PTRACK backups + (the backup role in the examples below) must have + access to all the databases of the cluster. + + + For PostgreSQL 11 or higher: + + + + + Create PTRACK extension: + +CREATE EXTENSION ptrack; + + + + + + To enable tracking page updates, set ptrack.map_size + parameter to a positive integer and restart the server. + + + For optimal performance, it is recommended to set + ptrack.map_size to + N / 1024, where + N is the size of the + PostgreSQL cluster, in MB. If you set this + parameter to a lower value, PTRACK is more likely to map several blocks + together, which leads to false-positive results when tracking changed + blocks and increases the incremental backup size as unchanged blocks + can also be copied into the incremental backup. + Setting ptrack.map_size to a higher value does not + affect PTRACK operation, but it is not recommended to set this parameter + to a value higher than 1024. + + + + + + + If you change the ptrack.map_size parameter value, + the previously created PTRACK map file is cleared, + and tracking newly changed blocks starts from scratch. Thus, you have + to retake a full backup before taking incremental PTRACK backups after + changing ptrack.map_size. + + + + + + + + Usage + + Creating a Backup + + To create a backup, run the following command: + + +pg_probackup backup -B backup_dir --instance=instance_name -b backup_mode + + + Where backup_mode can take one of the + following values: + FULL, + DELTA, + PAGE, and + PTRACK. + + + When restoring a cluster from an incremental backup, + pg_probackup relies on the parent full backup and all the + incremental backups between them, which is called + the backup chain. You must create at least + one full backup before taking incremental ones. + + + ARCHIVE Mode + + ARCHIVE is the default WAL delivery mode. + + + For example, to make a FULL backup in ARCHIVE mode, run: + + +pg_probackup backup -B backup_dir --instance=instance_name -b FULL + + + ARCHIVE backups rely on + continuous + archiving to get WAL segments required to restore + the cluster to a consistent state at the time the backup was + taken. + + + When a backup is taken, pg_probackup + ensures that WAL files containing WAL records between Start + LSN and Stop LSN actually exist in + backup_dir/wal/instance_name + directory. pg_probackup also ensures that WAL records between + Start LSN and Stop LSN can be parsed. This precaution + eliminates the risk of silent WAL corruption. + + + + STREAM Mode + + STREAM is the optional WAL delivery mode. + + + For example, to make a FULL backup in the STREAM mode, add the + flag to the command from the + previous example: + + +pg_probackup backup -B backup_dir --instance=instance_name -b FULL --stream --temp-slot + + + The optional flag ensures that + the required segments remain available if the WAL is rotated + before the backup is complete. + + + Unlike backups in ARCHIVE mode, STREAM backups include all the + WAL segments required to restore the cluster to a consistent + state at the time the backup was taken. + + + During pg_probackup + streams WAL files containing WAL records between Start LSN and + Stop LSN to + backup_dir/backups/instance_name/backup_id/database/pg_wal directory. To eliminate the risk + of silent WAL corruption, pg_probackup also + checks that WAL records between Start LSN and + Stop LSN can be parsed. + + + Even if you are using + continuous + archiving, STREAM backups can still be useful in the + following cases: + + + + + STREAM backups can be restored on the server that has no + file access to WAL archive. + + + + + STREAM backups enable you to restore the cluster state at + the point in time for which WAL files in archive are no + longer available. + + + + + Backup in STREAM mode can be taken from a standby of a + server that generates small amount of WAL traffic, + without long waiting for WAL segment to fill up. + + + + + + Page Validation + + If + data + checksums are enabled in the database cluster, + pg_probackup uses this information to check correctness of + data files during backup. While reading each page, + pg_probackup checks whether the calculated checksum coincides + with the checksum stored in the page header. This guarantees + that the PostgreSQL instance and the backup itself have no + corrupt pages. Note that pg_probackup reads database files + directly from the filesystem, so under heavy write load during + backup it can show false-positive checksum mismatches because of + partial writes. If a page checksum mismatch occurs, the page is + re-read and checksum comparison is repeated. + + + A page is considered corrupt if checksum comparison has failed + more than 100 times. In this case, the backup is aborted. + + + Even if data checksums are not enabled, pg_probackup + always performs sanity checks for page headers. + + + + External Directories + + To back up a directory located outside of the data directory, + use the optional parameter + that specifies the path to this directory. If you would like + to add more than one external directory, you can provide several paths + separated by colons on Linux systems or semicolons on Windows systems. + + + For example, to include /etc/dir1 and + /etc/dir2 directories into the full + backup of your instance_name instance + that will be stored under the backup_dir + directory on Linux, run: + + +pg_probackup backup -B backup_dir --instance=instance_name -b FULL --external-dirs=/etc/dir1:/etc/dir2 + + + Similarly, to include C:\dir1 and + C:\dir2 directories into the full backup + on Windows, run: + + +pg_probackup backup -B backup_dir --instance=instance_name -b FULL --external-dirs=C:\dir1;C:\dir2 + + + pg_probackup recursively copies the contents + of each external directory into a separate subdirectory in the backup + catalog. Since external directories included into different backups + do not have to be the same, when you are restoring the cluster from an + incremental backup, only those directories that belong to this + particular backup will be restored. Any external directories + stored in the previous backups will be ignored. + + + To include the same directories into each backup of your + instance, you can specify them in the pg_probackup.conf + configuration file using the + command with the + option. + + + + + + Performing Cluster Verification + + To verify that PostgreSQL database cluster is + not corrupt, run the following command: + + +pg_probackup checkdb [-B backup_dir [--instance=instance_name]] [-D data_dir] [connection_options] + + + + This command performs physical verification of all data files + located in the specified data directory by running page header + sanity checks, as well as block-level checksum verification if checksums are enabled. + If a corrupt page is detected, checkdb + continues cluster verification until all pages in the cluster + are validated. + + + + By default, similar page validation + is performed automatically while a backup is taken by + pg_probackup. The checkdb + command enables you to perform such page validation + on demand, without taking any backup copies, even if the cluster + is not backed up using pg_probackup at all. + + + + To perform cluster verification, pg_probackup + needs to connect to the cluster to be verified. In general, it is + enough to specify the backup instance of this cluster for + pg_probackup to determine the required + connection options. However, if -B and + --instance options are omitted, you have to provide + connection options and + data_dir via environment + variables or command-line options. + + + + Physical verification cannot detect logical inconsistencies, + missing or nullified blocks and entire files, or similar anomalies. Extensions + amcheck + and + amcheck_next + provide a partial solution to these problems. + + + If you would like, in addition to physical verification, to + verify all indexes in all databases using these extensions, you + can specify the flag when running + the command: + + +pg_probackup checkdb -D data_dir --amcheck [connection_options] + + + You can skip physical verification by specifying the + flag. In this case, + you can omit backup_dir and + data_dir options, only + connection options are + mandatory: + + +pg_probackup checkdb --amcheck --skip-block-validation [connection_options] + + + Logical verification can be done more thoroughly with the + flag by checking that all heap + tuples that should be indexed are actually indexed, but at the + higher cost of CPU, memory, and I/O consumption. + + + + + Validating a Backup + + pg_probackup calculates checksums for each file in a backup + during the backup process. The process of checking checksums of + backup data files is called + the backup validation. By default, validation + is run immediately after the backup is taken and right before the + restore, to detect possible backup corruption. + + + If you would like to skip backup validation, you can specify the + flag when running + and + commands. + + + To ensure that all the required backup files are present and can + be used to restore the database cluster, you can run the + command with the exact + recovery target + options you are going to use for recovery. + + + For example, to check that you can restore the database cluster + from a backup copy up to transaction ID 4242, run + this command: + + +pg_probackup validate -B backup_dir --instance=instance_name --recovery-target-xid=4242 + + + If validation completes successfully, pg_probackup displays the + corresponding message. If validation fails, you will receive an + error message with the exact time, transaction ID, and LSN up to + which the recovery is possible. + + + If you specify backup_id via + -i/--backup-id option, then only the backup copy + with specified backup ID will be validated. If + backup_id is specified with + recovery target + options, the validate command will check whether it is possible + to restore the specified backup to the specified + recovery target. + + + For example, to check that you can restore the database cluster + from a backup copy with the SCUN2C backup ID up to the + specified timestamp, run this command: + + +pg_probackup validate -B backup_dir --instance=instance_name -i SCUN2C --recovery-target-time="2024-05-03 11:18:13+03" + + + If you specify the backup_id of an incremental backup, + all its parents starting from FULL backup will be + validated. + + + If you omit all the parameters, all backups are validated. + + + + Restoring a Cluster + + To restore the database cluster from a backup, run the + command with at least the following options: + + +pg_probackup restore -B backup_dir --instance=instance_name -i backup_id + + + Where: + + + + + backup_dir is the backup catalog that + stores all backup files and meta information. + + + + + instance_name is the backup instance + for the cluster to be restored. + + + + + backup_id specifies the backup to + restore the cluster from. If you omit this option, + pg_probackup uses the latest valid backup available for the + specified instance. If you specify an incremental backup to + restore, pg_probackup automatically restores the underlying + full backup and then sequentially applies all the necessary + increments. + + + + + + Once the restore command is complete, start + the database service. + + + + If you restore ARCHIVE backups, + perform PITR, + or specify the --restore-as-replica flag with the + restore command to set up a standby server, + pg_probackup creates a recovery configuration + file once all data files are copied into the target directory. This file + includes the minimal settings required for recovery, except for the password in the + primary_conninfo + parameter; you have to add the password manually or use + the --primary-conninfo option, if required. + For PostgreSQL 11, + recovery settings are written into the recovery.conf + file. Starting from PostgreSQL 12, + pg_probackup writes these settings into + the probackup_recovery.conf file and then includes + it into postgresql.auto.conf. + + + + If you are restoring a STREAM backup, the restore is complete + at once, with the cluster returned to a self-consistent state at + the point when the backup was taken. For ARCHIVE backups, + PostgreSQL replays all available archived WAL + segments, so the cluster is restored to the latest state possible + within the current timeline. You can change this behavior by using the + recovery target + options with the restore command, + as explained in . + + + + If the cluster to restore contains tablespaces, pg_probackup + restores them to their original location by default. To restore + tablespaces to a different location, use the + / option. Otherwise, + restoring the cluster on the same host will fail if tablespaces + are in use, because the backup would have to be written to the + same directories. + + + When using the / + option, you must provide absolute paths to the old and new + tablespace directories. If a path happens to contain an equals + sign (=), escape it with a backslash. This option can be + specified multiple times for multiple tablespaces. For example: + + +pg_probackup restore -B backup_dir --instance=instance_name -D data_dir -j 4 -i backup_id -T tablespace1_dir=tablespace1_newdir -T tablespace2_dir=tablespace2_newdir + + + + To restore the cluster on a remote host, follow the instructions in + . + + + + By default, the + command validates the specified backup before restoring the + cluster. If you run regular backup validations and would like + to save time when restoring the cluster, you can specify the + flag to skip validation and + speed up the recovery. + + + + Incremental Restore + + The speed of restore from backup can be significantly improved + by replacing only invalid and changed pages in already + existing PostgreSQL data directory using + incremental + restore options with the + command. + + + To restore the database cluster from a backup in incremental mode, + run the command with the following options: + + +pg_probackup restore -B backup_dir --instance=instance_name -D data_dir -I incremental_mode + + + Where incremental_mode can take one of the + following values: + + + + + CHECKSUM — read all data files in the data directory, validate + header and checksum in every page and replace only invalid + pages and those with checksum and LSN not matching with + corresponding page in backup. This is the simplest, + the most fool-proof incremental mode. Recommended to use by default. + + + + + LSN — read the pg_control in the + data directory to obtain redo LSN and redo TLI, which allows you + to determine a point in history(shiftpoint), where data directory + state shifted from target backup chain history. If shiftpoint is not within + reach of backup chain history, then restore is aborted. + If shiftpoint is within reach of backup chain history, then read + all data files in the data directory, validate header and checksum in + every page and replace only invalid pages and those with LSN greater + than shiftpoint. + This mode offers a greater speed up compared to CHECKSUM, but rely + on two conditions to be met. First, + + data checksums parameter must be enabled in data directory (to avoid corruption + due to hint bits). This condition will be checked at the start of + incremental restore and the operation will be aborted if checksums are disabled. + Second, the pg_control file must be + synched with state of data directory. This condition cannot checked + at the start of restore, so it is a user responsibility to ensure + that pg_control contain valid information. + Therefore it is not recommended to use LSN mode in any situation, + where pg_control cannot be trusted or has been tampered with: + after pg_resetxlog execution, + after restore from backup without recovery been run, etc. + + + + + NONE — regular restore without any incremental optimizations. + + + + + + Regardless of chosen incremental mode, pg_probackup will check, that postmaster + in given destination directory is not running and system-identifier is + the same as in the backup. + + + + Suppose you want to return an old master as replica after switchover + using incremental restore in LSN mode: + + +====================================================================================================================================== + Instance Version ID Recovery Time Mode WAL Mode TLI Time Data WAL Zratio Start LSN Stop LSN Status +====================================================================================================================================== + node 16 SCUN3Y 2024-05-02 11:19:16+03 DELTA STREAM 16/15 7s 92MB 208MB 2.27 0/3C0043A8 0/46159C70 OK + node 16 SCUN3M 2024-05-02 11:19:01+03 PTRACK STREAM 15/15 10s 30MB 16MB 2.23 0/32000028 0/32005ED0 OK + node 16 SCUN39 2024-05-02 11:18:50+03 PAGE STREAM 15/15 12s 46MB 32MB 1.44 0/2A000028 0/2B0000B8 OK + node 16 SCUN2V 2024-05-02 11:18:38+03 FULL STREAM 15/0 11s 154MB 16MB 2.32 0/23000028 0/23000168 OK + +backup_user@backup_host:~$ pg_probackup-16 restore -B /mnt/backups --instance=node -R -I lsn +INFO: Destination directory and tablespace directories are empty, disable incremental restore +INFO: Validating parents for backup SCUN3Y +INFO: Validating backup SCUN2V +INFO: Backup SCUN2V data files are valid +INFO: Validating backup SCUN39 +INFO: Backup SCUN39 data files are valid +INFO: Validating backup SCUN3M +INFO: Backup SCUN3M data files are valid +INFO: Validating backup SCUN3Y +INFO: Backup SCUN3Y data files are valid +INFO: Backup SCUN3Y WAL segments are valid +INFO: Backup SCUN3Y is valid. +INFO: Restoring the database from backup SCUN3Y +INFO: Start restoring backup files. PGDATA size: 759MB +INFO: Backup files are restored. Transfered bytes: 759MB, time elapsed: 3s +INFO: Restore incremental ratio (less is better): 100% (759MB/759MB) +INFO: Syncing restored files to disk +INFO: Restored backup files are synced, time elapsed: 1s +INFO: Restore of backup SCUN3Y completed. + + + + Incremental restore is possible only for backups with + program_version equal or greater than 2.4.0. + + + + + Partial Restore + + If you have enabled + partial + restore before taking backups, you can restore + only some of the databases using + partial restore + options with the + commands. + + + To restore the specified databases only, run the command + with the following options: + + +pg_probackup restore -B backup_dir --instance=instance_name --db-include=database_name + + + The option can be specified + multiple times. For example, to restore only databases + db1 and db2, run the + following command: + + +pg_probackup restore -B backup_dir --instance=instance_name --db-include=db1 --db-include=db2 + + + To exclude one or more databases from restore, use + the option: + + +pg_probackup restore -B backup_dir --instance=instance_name --db-exclude=database_name + + + The option can be specified + multiple times. For example, to exclude the databases + db1 and db2 from + restore, run the following command: + + +pg_probackup restore -B backup_dir --instance=instance_name --db-exclude=db1 --db-exclude=db2 + + + Partial restore relies on lax behavior of PostgreSQL recovery + process toward truncated files. For recovery to work properly, files of excluded databases + are restored as files of zero size. After the PostgreSQL cluster is successfully started, + you must drop the excluded databases using + DROP DATABASE command. + + + To decouple a single cluster containing multiple databases into separate clusters with minimal downtime, + you can do partial restore of the cluster as a standby using the option + for specific databases. + + + + The template0 and + template1 databases are always restored. + + + + + Due to recovery specifics of PostgreSQL versions earlier than 12, + it is advisable that you set the + hot_standby + parameter to off when running partial + restore of a PostgreSQL cluster of version earlier than 12. + Otherwise the recovery may fail. + + + + + + Performing Point-in-Time (PITR) Recovery + + If you have enabled + continuous + WAL archiving before taking backups, you can restore the + cluster to its state at an arbitrary point in time (recovery + target) using recovery + target options with the + command. + + + + You can use both STREAM and ARCHIVE backups for point in time + recovery as long as the WAL archive is available at least starting + from the time the backup was taken. + If / option is omitted, + pg_probackup automatically chooses the backup that is the + closest to the specified recovery target and starts the restore + process, otherwise pg_probackup will try to restore + the specified backup to the specified recovery target. + + + + + To restore the cluster state at the exact time, specify the + option, in the + timestamp format. For example: + + +pg_probackup restore -B backup_dir --instance=instance_name --recovery-target-time="2024-05-03 11:18:13+03" + + + + + To restore the cluster state up to a specific transaction + ID, use the option: + + +pg_probackup restore -B backup_dir --instance=instance_name --recovery-target-xid=687 + + + + + To restore the cluster state up to the specific LSN, use + option: + + +pg_probackup restore -B backup_dir --instance=instance_name --recovery-target-lsn=16/B374D848 + + + + + To restore the cluster state up to the specific named restore + point, use option: + + +pg_probackup restore -B backup_dir --instance=instance_name --recovery-target-name="before_app_upgrade" + + + + + To restore the backup to the latest state available in + the WAL archive, use option + with latest value: + + +pg_probackup restore -B backup_dir --instance=instance_name --recovery-target="latest" + + + + + To restore the cluster to the earliest point of consistency, + use option with the + immediate value: + + +pg_probackup restore -B backup_dir --instance=instance_name --recovery-target='immediate' + + + + + + Using <application>pg_probackup</application> in the Remote Mode + + pg_probackup supports the remote mode that allows you to perform + backup and restore + operations remotely via SSH. In this mode, the backup catalog is + stored on a local system, while PostgreSQL instance to be backed + up is located on a remote system. You must have pg_probackup + installed on both systems. + + + + pg_probackup relies on passwordless SSH connection + for communication between the hosts. + + + + + In addition to SSH connection, pg_probackup uses + a regular connection to the database to manage the remote operation. + See the section Configuring + the Database Cluster for details of how to set up + a database connection. + + + + The typical workflow is as follows: + + + + + On your backup host, configure pg_probackup as explained in + the section + Setup. For the + and + commands, make + sure to specify remote + options that point to the database host with the + PostgreSQL instance. + + + + + If you would like to take remote backups in + PAGE mode, or rely + on ARCHIVE WAL delivery + mode, or use + PITR, + configure continuous WAL archiving from the database host + to the backup host as explained in the section + Setting + up continuous WAL archiving. For the + and + commands, you + must specify the remote + options that point to the backup host with the backup + catalog. + + + + + Run or + commands with + remote options + on the backup host. + pg_probackup connects to the remote system via SSH and + creates a backup locally or restores the previously taken + backup on the remote system, respectively. + + + + + For example, to create an archive full backup of a + PostgreSQL cluster located on + a remote system with host address 192.168.0.2 + on behalf of the postgres user via SSH connection + through port 2302, run: + + +pg_probackup backup -B backup_dir --instance=instance_name -b FULL --remote-user=postgres --remote-host=192.168.0.2 --remote-port=2302 + + + To restore the latest available backup on a remote system with host address + 192.168.0.2 on behalf of the postgres + user via SSH connection through port 2302, run: + + +pg_probackup restore -B backup_dir --instance=instance_name --remote-user=postgres --remote-host=192.168.0.2 --remote-port=2302 + + + Restoring an ARCHIVE backup or performing PITR in the remote mode + require additional information: destination address, port and + username for establishing an SSH connection + from the host with database + to the host with the backup + catalog. This information will be used by the + restore_command to copy WAL segments + from the archive to the PostgreSQL pg_wal directory. + + + To solve this problem, you can use + Remote WAL Archive + Options. + + + For example, to restore latest backup on remote system using + remote mode through SSH connection to user + postgres on host with address + 192.168.0.2 via port 2302 + and user backup on backup catalog host with + address 192.168.0.3 via port + 2303, run: + + +pg_probackup restore -B backup_dir --instance=instance_name --remote-user=postgres --remote-host=192.168.0.2 --remote-port=2302 --archive-host=192.168.0.3 --archive-port=2303 --archive-user=backup + + + Provided arguments will be used to construct the restore_command: + + +restore_command = '"install_dir/pg_probackup" archive-get -B "backup_dir" --instance=instance_name --wal-file-path=%p --wal-file-name=%f --remote-host=192.168.0.3 --remote-port=2303 --remote-user=backup' + + + Alternatively, you can use the + option to provide the entire restore_command: + + +pg_probackup restore -B backup_dir --instance=instance_name --remote-user=postgres --remote-host=192.168.0.2 --remote-port=2302 --restore-command='"install_dir/pg_probackup" archive-get -B "backup_dir" --instance=instance_name --wal-file-path=%p --wal-file-name=%f --remote-host=192.168.0.3 --remote-port=2303 --remote-user=backup' + + + + The remote mode is currently unavailable for + Windows systems. + + + + + Running <application>pg_probackup</application> on Parallel Threads + + , + , + , + , + , + , and + processes can be + executed on several parallel threads. This can significantly + speed up pg_probackup operation given enough resources (CPU + cores, disk, and network bandwidth). + + + Parallel execution is controlled by the + -j/--threads command-line option. For + example, to create a backup using four parallel threads, run: + + +pg_probackup backup -B backup_dir --instance=instance_name -b FULL -j 4 + + + + Parallel restore applies only to copying data from the + backup catalog to the data directory of the cluster. When + PostgreSQL server is started, WAL records need to be replayed, + and this cannot be done in parallel. + + + + + Configuring <application>pg_probackup</application> + + Once the backup catalog is initialized and a new backup instance + is added, you can use the pg_probackup.conf configuration file + located in the + backup_dir/backups/instance_name + directory to fine-tune pg_probackup configuration. + + + For example, and + commands use a regular + PostgreSQL connection. To avoid specifying + connection options + each time on the command line, you can set them in the + pg_probackup.conf configuration file using the + command. + + + + It is not recommended + to edit pg_probackup.conf manually. + + + + Initially, pg_probackup.conf contains the following settings: + + + + + PGDATA — the path to the data directory of the cluster to + back up. + + + + + system-identifier — the unique identifier of the PostgreSQL + instance. + + + + + Additionally, you can define + remote, + retention, + logging, and + compression settings + using the set-config command: + + +pg_probackup set-config -B backup_dir --instance=instance_name +[--external-dirs=external_directory_path] [remote_options] [connection_options] [retention_options] [logging_options] + + + To view the current settings, run the following command: + + +pg_probackup show-config -B backup_dir --instance=instance_name + + + You can override the settings defined in pg_probackup.conf when + running pg_probackup commands + via the corresponding environment variables and/or command line + options. + + + + Specifying Connection Settings + + If you define connection settings in the pg_probackup.conf + configuration file, you can omit connection options in all the + subsequent pg_probackup commands. However, if the corresponding + environment variables are set, they get higher priority. The + options provided on the command line overwrite both environment + variables and configuration file settings. + + + If nothing is given, the default values are taken. By default + pg_probackup tries to use local connection via Unix domain + socket (localhost on Windows) and tries to get the database name + and the user name from the PGUSER environment variable or the + current OS user name. + + + + Managing the Backup Catalog + + With pg_probackup, you can manage backups from the command line: + + + + + View backup + information + + + + + View WAL + Archive Information + + + + + Validate backups + + + + + Merge backups + + + + + Delete backups + + + + + Viewing Backup Information + + To view the list of existing backups for every instance, run + the command: + + +pg_probackup show -B backup_dir + + + pg_probackup displays the list of all the available backups. + For example: + + +BACKUP INSTANCE 'node' +====================================================================================================================================== + Instance Version ID Recovery Time Mode WAL Mode TLI Time Data WAL Zratio Start LSN Stop LSN Status +====================================================================================================================================== + node 16 SCUN4E 2024-05-02 11:19:37+03 FULL ARCHIVE 1/0 13s 239MB 16MB 2.31 0/4C000028 0/4D0000B8 OK + node 16 SCUN3Y 2024-05-02 11:19:16+03 DELTA STREAM 1/1 7s 92MB 208MB 2.27 0/3C0043A8 0/46159C70 OK + node 16 SCUN3M 2024-05-02 11:19:01+03 PTRACK STREAM 1/1 10s 30MB 16MB 2.23 0/32000028 0/32005ED0 OK + node 16 SCUN39 2024-05-02 11:18:50+03 PAGE STREAM 1/1 12s 46MB 32MB 1.44 0/2A000028 0/2B0000B8 OK + node 16 SCUN2V 2024-05-02 11:18:38+03 FULL STREAM 1/0 11s 154MB 16MB 2.32 0/23000028 0/23000168 OK + + + For each backup, the following information is provided: + + + + + Instance — the instance name. + + + + + Version — PostgreSQL major version. + + + + + ID — the backup identifier. + + + + + Recovery time — the earliest moment for which you can + restore the state of the database cluster. + + + + + Mode — the method used to take this backup. Possible + values: FULL, PAGE, DELTA, PTRACK. + + + + + WAL Mode — WAL delivery mode. Possible values: STREAM + and ARCHIVE. + + + + + TLI — timeline identifiers of the current backup and its + parent. + + + + + Time — the time it took to perform the backup. + + + + + Data — the size of the data files in this backup. This + value does not include the size of WAL files. For + STREAM backups, the total size of the backup can be calculated + as Data + WAL. + + + + + WAL — the uncompressed size of WAL files + that need to be applied during recovery for the backup to reach a consistent state. + + + + + Zratio — compression ratio calculated as + uncompressed-bytes / data-bytes. + + + + + Start LSN — WAL log sequence number corresponding to the + start of the backup process. REDO point for PostgreSQL + recovery process to start from. + + + + + Stop LSN — WAL log sequence number corresponding to the + end of the backup process. Consistency point for + PostgreSQL recovery process. + + + + + Status — backup status. Possible values: + + + + + OK — the backup is complete and valid. + + + + + DONE — the backup is complete, but was not validated. + + + + + RUNNING — the backup is in progress. + + + + + MERGING — the backup is being merged. + + + + + MERGED — the backup data files were + successfully merged, but its metadata is in the process + of being updated. Only full backups can have this status. + + + + + DELETING — the backup files are being deleted. + + + + + CORRUPT — some of the backup files are corrupt. + + + + + ERROR — the backup was aborted because of an + unexpected error. + + + + + ORPHAN — the backup is invalid because one of its + parent backups is corrupt or missing. + + + + + + + You can restore the cluster from the backup only if the backup + status is OK or DONE. + + + To get more detailed information about the backup, run the + show command with the backup ID: + + +pg_probackup show -B backup_dir --instance=instance_name -i backup_id + + + The sample output is as follows: + + +#Configuration +backup-mode = FULL +stream = false +compress-alg = zlib +compress-level = 1 +from-replica = false + +#Compatibility +block-size = 8192 +xlog-block-size = 8192 +checksum-version = 1 +program-version = 2.5.15 +server-version = 16 + +#Result backup info +timelineid = 1 +start-lsn = 0/4C000028 +stop-lsn = 0/4D0000B8 +start-time = '2024-05-02 11:19:26+03' +end-time = '2024-05-02 11:19:39+03' +recovery-xid = 743 +recovery-time = '2024-05-02 11:19:37+03' +data-bytes = 250827955 +wal-bytes = 16777216 +uncompressed-bytes = 578216425 +pgdata-bytes = 578216107 +status = OK +primary_conninfo = 'user=backup channel_binding=prefer host=localhost port=5432 sslmode=prefer sslcompression=0 sslcertmode=allow sslsni=1 ssl_min_protocol_version=TLSv1.2 gssencmode=prefer krbsrvname=postgres gssdelegation=0 target_session_attrs=any load_balance_hosts=disable' +content-crc = 802820606 + + + Detailed output has additional attributes: + + + + compress-alg — compression algorithm used during backup. Possible values: + zlib, pglz, none. + + + + + compress-level — compression level used during backup. + + + + + from-replica — was this backup taken on standby? Possible values: + 1, 0. + + + + + block-size — the block_size + setting of PostgreSQL cluster at the backup start. + + + + + checksum-version — are + data + block checksums enabled in the backed up PostgreSQL cluster? Possible values: 1, 0. + + + + + program-version — full version of pg_probackup binary used to create the backup. + + + + + start-time — the backup start time. + + + + + end-time — the backup end time. + + + + + expire-time — the point in time + when a pinned backup can be removed in accordance with retention + policy. This attribute is only available for pinned backups. + + + + + uncompressed-bytes — the size of data files before adding page headers and applying + compression. You can evaluate the effectiveness of compression + by comparing uncompressed-bytes to data-bytes if + compression if used. + + + + + pgdata-bytes — the size of PostgreSQL + cluster data files at the time of backup. You can evaluate the + effectiveness of an incremental backup by comparing + pgdata-bytes to uncompressed-bytes. + + + + + recovery-xid — transaction ID at the backup end time. + + + + + parent-backup-id — ID of the parent backup. Available only + for incremental backups. + + + + + primary_conninfo — libpq connection parameters + used to connect to the PostgreSQL cluster to take this backup. The + password is not included. + + + + + note — text note attached to backup. + + + + + content-crc — CRC32 checksum of backup_content.control file. + It is used to detect corruption of backup metainformation. + + + + + + You can also get the detailed information about the backup + in the JSON format: + + +pg_probackup show -B backup_dir --instance=instance_name --format=json -i backup_id + + + The sample output is as follows: + + +[ + { + "instance": "node", + "backups": [ + { + "id": "SCUN4E", + "backup-mode": "FULL", + "wal": "ARCHIVE", + "compress-alg": "zlib", + "compress-level": 1, + "from-replica": "false", + "block-size": 8192, + "xlog-block-size": 8192, + "checksum-version": 1, + "program-version": "2.5.15", + "server-version": "16", + "current-tli": 16, + "parent-tli": 2, + "start-lsn": "0/4C000028", + "stop-lsn": "0/4D0000B8", + "start-time": "2024-05-02 11:19:26+03", + "end-time": "2024-05-02 11:19:39+03", + "recovery-xid": 743, + "recovery-time": "2024-05-02 11:19:37+03", + "data-bytes": 250827955, + "wal-bytes": 16777216, + "uncompressed-bytes": 578216425, + "pgdata-bytes": 578216107, + "primary_conninfo": "user=backup channel_binding=prefer host=localhost port=5432 sslmode=prefer sslcompression=0 sslcertmode=allow sslsni=1 ssl_min_protocol_version=TLSv1.2 gssencmode=prefer krbsrvname=postgres gssdelegation=0 target_session_attrs=any load_balance_hosts=disable", + "status": "OK", + "content-crc": 802820606 + } + ] + } +] + + + + Viewing WAL Archive Information + + To view the information about WAL archive for every instance, + run the command: + + +pg_probackup show -B backup_dir [--instance=instance_name] --archive + + + pg_probackup displays the list of all the available WAL files + grouped by timelines. For example: + + + +ARCHIVE INSTANCE 'node' +================================================================================================================================ + TLI Parent TLI Switchpoint Min Segno Max Segno N segments Size Zratio N backups Status +================================================================================================================================ + 1 0 0/0 000000010000000000000019 00000001000000000000004D 53 848MB 1.00 5 OK + + + For each timeline, the following information is provided: + + + + + TLI — timeline identifier. + + + + + Parent TLI — identifier of the timeline from which this timeline branched off. + + + + + Switchpoint — LSN of the moment when the timeline branched + off from its parent timeline. + + + + + Min Segno — the first WAL segment + belonging to the timeline. + + + + + Max Segno — the last WAL segment + belonging to the timeline. + + + + + N segments — number of WAL segments belonging to the + timeline. + + + + + Size — the size that files take on disk. + + + + + Zratio — compression ratio calculated as N segments * + wal_segment_size * wal_block_size / Size. + + + + + N backups — number of backups belonging to the timeline. + To get the details about backups, use the JSON format. + + + + + Status — status of the WAL archive for this timeline. Possible + values: + + + + + OK — all WAL segments between Min Segno + and Max Segno are present. + + + + + DEGRADED — some WAL segments between Min Segno + and Max Segno are missing. To find out which files are lost, + view this report in the JSON format. + + + + + + + To get more detailed information about the WAL archive in the JSON + format, run the command: + + +pg_probackup show -B backup_dir [--instance=instance_name] --archive --format=json + + + The sample output is as follows: + + +[ + { + "instance": "node", + "timelines": [ + { + "tli": 1, + "parent-tli": 0, + "switchpoint": "0/0", + "min-segno": "000000010000000000000019", + "max-segno": "00000001000000000000004D", + "n-segments": 53, + "size": 889192448, + "zratio": 1.00, + "closest-backup-id": "", + "status": "OK", + "lost-segments": [], + "backups": [ + { + "id": "SCUN4E", + "backup-mode": "FULL", + "wal": "ARCHIVE", + "compress-alg": "zlib", + "compress-level": 1, + "from-replica": "false", + "block-size": 8192, + "xlog-block-size": 8192, + "checksum-version": 1, + "program-version": "2.5.15", + "server-version": "16", + "current-tli": 1, + "parent-tli": 0, + "start-lsn": "0/4C000028", + "stop-lsn": "0/4D0000B8", + "start-time": "2024-05-02 11:19:26+03", + "end-time": "2024-05-02 11:19:39+03", + "recovery-xid": 743, + "recovery-time": "2024-05-02 11:19:37+03", + "data-bytes": 250827955, + "wal-bytes": 16777216, + "uncompressed-bytes": 578216425, + "pgdata-bytes": 578216107, + "primary_conninfo": "user=backup channel_binding=prefer host=localhost port=5432 sslmode=prefer sslcompression=0 sslcertmode=allow sslsni=1 ssl_min_protocol_version=TLSv1.2 gssencmode=prefer krbsrvname=postgres gssdelegation=0 target_session_attrs=any load_balance_hosts=disable", + "status": "OK", + "content-crc": 802820606 + }, + { + "id": "SCUN3Y", + "parent-backup-id": "SCUN3M", + "backup-mode": "DELTA", + "wal": "STREAM", + "compress-alg": "zlib", + "compress-level": 1, + "from-replica": "false", + "block-size": 8192, + "xlog-block-size": 8192, + "checksum-version": 1, + "program-version": "2.5.15", + "server-version": "16", + "current-tli": 1, + "parent-tli": 1, + "start-lsn": "0/3C0043A8", + "stop-lsn": "0/46159C70", + "start-time": "2024-05-02 11:19:10+03", + "end-time": "2024-05-02 11:19:17+03", + "recovery-xid": 743, + "recovery-time": "2024-05-02 11:19:16+03", + "data-bytes": 96029293, + "wal-bytes": 218103808, + "uncompressed-bytes": 217639806, + "pgdata-bytes": 578216107, + "primary_conninfo": "user=backup channel_binding=prefer host=localhost port=5432 sslmode=prefer sslcompression=0 sslcertmode=allow sslsni=1 ssl_min_protocol_version=TLSv1.2 gssencmode=prefer krbsrvname=postgres gssdelegation=0 target_session_attrs=any load_balance_hosts=disable", + "status": "OK", + "content-crc": 3074300814 + }, + { + "id": "SCUN3M", + "parent-backup-id": "SCUN39", + "backup-mode": "PTRACK", + "wal": "STREAM", + "compress-alg": "zlib", + "compress-level": 1, + "from-replica": "false", + "block-size": 8192, + "xlog-block-size": 8192, + "checksum-version": 1, + "program-version": "2.5.15", + "server-version": "16", + "current-tli": 1, + "parent-tli": 1, + "start-lsn": "0/32000028", + "stop-lsn": "0/32005ED0", + "start-time": "2024-05-02 11:18:58+03", + "end-time": "2024-05-02 11:19:08+03", + "recovery-xid": 742, + "recovery-time": "2024-05-02 11:19:01+03", + "data-bytes": 31205704, + "wal-bytes": 16777216, + "uncompressed-bytes": 69585790, + "pgdata-bytes": 509927595, + "primary_conninfo": "user=backup channel_binding=prefer host=localhost port=5432 sslmode=prefer sslcompression=0 sslcertmode=allow sslsni=1 ssl_min_protocol_version=TLSv1.2 gssencmode=prefer krbsrvname=postgres gssdelegation=0 target_session_attrs=any load_balance_hosts=disable", + "status": "OK", + "content-crc": 3446949708 + }, + { + "id": "SCUN39", + "parent-backup-id": "SCUN2V", + "backup-mode": "PAGE", + "wal": "STREAM", + "compress-alg": "pglz", + "compress-level": 1, + "from-replica": "false", + "block-size": 8192, + "xlog-block-size": 8192, + "checksum-version": 1, + "program-version": "2.5.15", + "server-version": "16", + "current-tli": 1, + "parent-tli": 1, + "start-lsn": "0/2A000028", + "stop-lsn": "0/2B0000B8", + "start-time": "2024-05-02 11:18:45+03", + "end-time": "2024-05-02 11:18:57+03", + "recovery-xid": 741, + "recovery-time": "2024-05-02 11:18:50+03", + "data-bytes": 48381612, + "wal-bytes": 33554432, + "uncompressed-bytes": 69569406, + "pgdata-bytes": 441639083, + "primary_conninfo": "user=backup channel_binding=prefer host=localhost port=5432 sslmode=prefer sslcompression=0 sslcertmode=allow sslsni=1 ssl_min_protocol_version=TLSv1.2 gssencmode=prefer krbsrvname=postgres gssdelegation=0 target_session_attrs=any load_balance_hosts=disable", + "status": "OK", + "content-crc": 3492989773 + }, + { + "id": "SCUN2V", + "backup-mode": "FULL", + "wal": "STREAM", + "compress-alg": "zlib", + "compress-level": 1, + "from-replica": "false", + "block-size": 8192, + "xlog-block-size": 8192, + "checksum-version": 1, + "program-version": "2.5.15", + "server-version": "16", + "current-tli": 1, + "parent-tli": 0, + "start-lsn": "0/23000028", + "stop-lsn": "0/23000168", + "start-time": "2024-05-02 11:18:31+03", + "end-time": "2024-05-02 11:18:42+03", + "recovery-xid": 740, + "recovery-time": "2024-05-02 11:18:38+03", + "data-bytes": 161084290, + "wal-bytes": 16777216, + "uncompressed-bytes": 373359081, + "pgdata-bytes": 373358763, + "primary_conninfo": "user=backup channel_binding=prefer host=localhost port=5432 sslmode=prefer sslcompression=0 sslcertmode=allow sslsni=1 ssl_min_protocol_version=TLSv1.2 gssencmode=prefer krbsrvname=postgres gssdelegation=0 target_session_attrs=any load_balance_hosts=disable", + "status": "OK", + "content-crc": 1621343133 + } + ] + } + ] + } +] + + + Most fields are consistent with the plain format, with some + exceptions: + + + + + The size is in bytes. + + + + + The closest-backup-id attribute + contains the ID of the most recent valid backup that belongs to + one of the previous timelines. You can use this backup to perform + point-in-time recovery to this timeline. If + such a backup does not exist, this string is empty. + + + + + The lost-segments array provides with + information about intervals of missing segments in DEGRADED timelines. In OK + timelines, the lost-segments array is empty. + + + + + The backups array lists all backups + belonging to the timeline. If the timeline has no backups, this array is empty. + + + + + + + Configuring Retention Policy + + With pg_probackup, you can configure + retention policy to remove redundant backups, clean up unneeded + WAL files, as well as pin specific backups to ensure they are + kept for the specified time, as explained in the sections below. + All these actions can be combined together in any way. + + + + Removing Redundant Backups + + By default, all backup copies created with pg_probackup are + stored in the specified backup catalog. To save disk space, + you can configure retention policy to remove redundant backup copies. + + + To configure retention policy, set one or more of the + following variables in the pg_probackup.conf file via + : + + +--retention-redundancy=redundancy + + + Specifies the number of full backup + copies to keep in the backup catalog. + + +--retention-window=window + + + Defines the earliest point in time for which pg_probackup can + complete the recovery. This option is set in + the number of days from the + current moment. For example, if + retention-window=7, pg_probackup must + keep at least one backup copy that is older than seven days, with + all the corresponding WAL files, and all the backups that follow. + + + If both and + options are set, both these + conditions have to be taken into account when purging the backup + catalog. For example, if you set --retention-redundancy=2 + and --retention-window=7, + pg_probackup has to keep two full backup + copies, as well as all the backups required to ensure recoverability + for the last seven days: + + +pg_probackup set-config -B backup_dir --instance=instance_name --retention-redundancy=2 --retention-window=7 + + + + To clean up the backup catalog in accordance with retention policy, + you have to run the command with + retention flags, as shown + below, or use the command with + these flags to process the outdated backup copies right when the new + backup is created. + + + + For example, to remove all backup copies that no longer satisfy the + defined retention policy, run the following command with the + --delete-expired flag: + + +pg_probackup delete -B backup_dir --instance=instance_name --delete-expired + + + If you would like to also remove the WAL files that are no + longer required for any of the backups, you should also specify the + flag: + + +pg_probackup delete -B backup_dir --instance=instance_name --delete-expired --delete-wal + + + + You can also set or override the current retention policy by + specifying and + options directly when + running delete or backup + commands: + + +pg_probackup delete -B backup_dir --instance=instance_name --delete-expired --retention-window=7 --retention-redundancy=2 + + + Since incremental backups require that their parent full + backup and all the preceding incremental backups are + available, if any of such backups expire, they still cannot be + removed while at least one incremental backup in this chain + satisfies the retention policy. To avoid keeping expired + backups that are still required to restore an active + incremental one, you can merge them with this backup using the + flag when running + or + commands. + + + + Suppose you have backed up the node + instance in the backup_dir directory, + with the option set + to 7, and you have the following backups + available on May 02, 2024: + + +BACKUP INSTANCE 'node' +===================================================================================================================================== + Instance Version ID Recovery Time Mode WAL Mode TLI Time Data WAL Zratio Start LSN Stop LSN Status +===================================================================================================================================== + node 16 SCUN6L 2024-05-02 11:20:48+03 FULL ARCHIVE 1/0 5s 296MB 16MB 2.30 0/46000028 0/470000B8 OK + node 16 SCQXUI 2024-04-30 11:20:45+03 PAGE ARCHIVE 1/1 5s 6280kB 16MB 1.00 0/44000028 0/450000F0 OK + node 16 SCFTUG 2024-04-24 11:20:43+03 DELTA ARCHIVE 1/1 5s 6280kB 16MB 1.00 0/42000028 0/430000B8 OK +----------------------------------------------------------retention window----------------------------------------------------------- + node 16 SCDZ6D 2024-04-23 11:20:40+03 PAGE ARCHIVE 1/1 5s 6280kB 16MB 1.00 0/40000028 0/410000B8 OK + node 16 SCC4HX 2024-04-22 11:20:24+03 FULL ARCHIVE 1/0 5s 296MB 16MB 2.30 0/3E000028 0/3F0000F0 OK + node 16 SC8F5G 2024-04-20 11:20:07+03 FULL ARCHIVE 1/0 5s 296MB 16MB 2.30 0/3C0000D8 0/3D00BB58 OK + + + Even though SCC4HX and SCDZ6D backups are outside the + retention window, they cannot be removed as it invalidates the + succeeding incremental backups SCFTUG and SCQXUI that are + still required, so, if you run the + command with the + flag, only the SC8F5G full + backup will be removed. + + + With the option, the SCFTUG + backup is merged with the underlying SCDZ6D and SCC4HX backups + and becomes a full one, so there is no need to keep these + expired backups anymore: + + +pg_probackup delete -B backup_dir --instance=node --delete-expired --merge-expired +pg_probackup show -B backup_dir + + +BACKUP INSTANCE 'node' +===================================================================================================================================== + Instance Version ID Recovery Time Mode WAL Mode TLI Time Data WAL Zratio Start LSN Stop LSN Status +===================================================================================================================================== + node 16 SCUN6L 2024-05-02 11:20:48+03 FULL ARCHIVE 1/0 5s 296MB 16MB 2.30 0/46000028 0/470000B8 OK + node 16 SCQXUI 2024-04-30 11:20:45+03 PAGE ARCHIVE 1/1 5s 6280kB 16MB 1.00 0/44000028 0/450000F0 OK + node 16 SCFTUG 2024-04-24 11:20:43+03 FULL ARCHIVE 1/1 5s 296MB 16MB 1.00 0/42000028 0/430000B8 OK + + + The Time field for the merged backup displays the time + required for the merge. + + + + + Pinning Backups + + If you need to keep certain backups longer than the + established retention policy allows, you can pin them + for arbitrary time. For example: + + +pg_probackup set-backup -B backup_dir --instance=instance_name -i backup_id --ttl=30d + + + This command sets the expiration time of the + specified backup to 30 days starting from the time + indicated in its recovery-time attribute. + + + You can also explicitly set the expiration time for a backup + using the option. For example: + + +pg_probackup set-backup -B backup_dir --instance=instance_name -i backup_id --expire-time="2027-05-02 11:21:00+00" + + + Alternatively, you can use the and + options with the + command to pin the newly + created backup: + + +pg_probackup backup -B backup_dir --instance=instance_name -b FULL --ttl=30d +pg_probackup backup -B backup_dir --instance=instance_name -b FULL --expire-time="2027-05-02 11:21:00+00" + + + To check if the backup is pinned, + run the command: + +pg_probackup show -B backup_dir --instance=instance_name -i backup_id + + + + If the backup is pinned, it has the expire-time + attribute that displays its expiration time: + +... +recovery-time = '2024-05-02 11:21:00+00' +expire-time = '2027-05-02 11:21:00+00' +data-bytes = 22288792 +... + + + + You can unpin the backup by setting the option to zero: + + +pg_probackup set-backup -B backup_dir --instance=instance_name -i backup_id --ttl=0 + + + + + A pinned incremental backup implicitly pins all + its parent backups. If you unpin such a backup later, + its implicitly pinned parents will also be automatically unpinned. + + + + + + Configuring WAL Archive Retention Policy + + When continuous + WAL archiving is enabled, archived WAL segments can take a lot + of disk space. Even if you delete old backup copies from time to time, + the --delete-wal flag can + purge only those WAL segments that do not apply to any of the + remaining backups in the backup catalog. However, if point-in-time + recovery is critical only for the most recent backups, you can + configure WAL archive retention policy to keep WAL archive of + limited depth and win back some more disk space. + + + + To configure WAL archive retention policy, you have to run the + command with the + --wal-depth option that specifies the number + of backups that can be used for PITR. + This setting applies to all the timelines, so you should be able to perform + PITR for the same number of backups on each timeline, if available. + Pinned backups are + not included into this count: if one of the latest backups + is pinned, pg_probackup ensures that + PITR is possible for one extra backup. + + + + To remove WAL segments that do not satisfy the defined WAL archive + retention policy, you simply have to run the + or command with the --delete-wal + flag. For archive backups, WAL segments between Start LSN + and Stop LSN are always kept intact, so such backups + remain valid regardless of the --wal-depth setting + and can still be restored, if required. + + + + You can also use the option + with the and + commands to override the previously defined WAL archive retention + policy and purge old WAL segments on the fly. + + + + Suppose you have backed up the node + instance in the backup_dir directory and + configured + continuous WAL + archiving: + + +pg_probackup show -B backup_dir --instance=node + + +====================================================================================================================================== + Instance Version ID Recovery Time Mode WAL Mode TLI Time Data WAL Zratio Start LSN Stop LSN Status +====================================================================================================================================== + node 16 SCUN92 2024-05-02 11:22:16+03 DELTA STREAM 1/1 9s 1162kB 32MB 1.08 0/7C000028 0/7C000168 OK + node 16 SCUN8N 2024-05-02 11:22:09+03 FULL STREAM 1/0 12s 296MB 16MB 2.30 0/7A000028 0/7A009A08 OK + node 16 SCUN8I 2024-05-02 11:21:55+03 DELTA STREAM 1/1 5s 1148kB 32MB 1.01 0/78000028 0/78000168 OK + node 16 SCUN86 2024-05-02 11:21:47+03 DELTA STREAM 1/1 11s 120MB 16MB 2.27 0/76000028 0/760001A0 OK + node 16 SCUN7I 2024-05-02 11:21:29+03 FULL STREAM 1/0 22s 296MB 288MB 2.30 0/63012FE8 0/74E7ADA0 OK + node 16 SCUN71 2024-05-02 11:21:12+03 FULL STREAM 1/0 13s 296MB 272MB 2.30 0/49000028 0/573683B8 OK + + + You can check the state of the WAL archive by running the + command with the + flag: + + +pg_probackup show -B backup_dir --instance=node --archive + + + +ARCHIVE INSTANCE 'node' +================================================================================================================================ + TLI Parent TLI Switchpoint Min Segno Max Segno N segments Size Zratio N backups Status +================================================================================================================================ + 1 0 0/0 000000010000000000000048 00000001000000000000007C 53 848MB 1.00 6 OK + + + WAL purge without cannot + achieve much, only one segment is removed: + + +pg_probackup delete -B backup_dir --instance=node --delete-wal + + + +ARCHIVE INSTANCE 'node' +================================================================================================================================ + TLI Parent TLI Switchpoint Min Segno Max Segno N segments Size Zratio N backups Status +================================================================================================================================ + 1 0 0/0 000000010000000000000049 00000001000000000000007C 52 832MB 1.00 6 OK + + + If you would like, for example, to keep only those WAL + segments that can be applied to the latest valid backup, set the + option to 1: + + +pg_probackup delete -B backup_dir --instance=node --delete-wal --wal-depth=1 + + + +ARCHIVE INSTANCE 'node' +=============================================================================================================================== + TLI Parent TLI Switchpoint Min Segno Max Segno N segments Size Zratio N backups Status +=============================================================================================================================== + 1 0 0/0 00000001000000000000007C 00000001000000000000007C 1 16MB 1.00 6 OK + + + Alternatively, you can use the + option with the command: + + +pg_probackup backup -B backup_dir --instance=node -b DELTA --wal-depth=1 --delete-wal + + + +ARCHIVE INSTANCE 'node' +=============================================================================================================================== + TLI Parent TLI Switchpoint Min Segno Max Segno N segments Size Zratio N backups Status +=============================================================================================================================== + 1 0 0/0 00000001000000000000007E 00000001000000000000007E 1 16MB 1.00 7 OK + + + + + Merging Backups + + As you take more and more incremental backups, the total size of + the backup catalog can substantially grow. To save disk space, + you can merge incremental backups to their parent full backup by + running the merge command, specifying the backup ID of the most + recent incremental backup you would like to merge: + + +pg_probackup merge -B backup_dir --instance=instance_name -i backup_id + + + This command merges backups that belong to a common incremental backup + chain. If you specify a full backup, it will be merged with its first + incremental backup. If you specify an incremental backup, it will be + merged to its parent full backup, together with all incremental backups + between them. Once the merge is complete, the full backup takes in all + the merged data, and the incremental backups are removed as redundant. + Thus, the merge operation is virtually equivalent to retaking a full + backup and removing all the outdated backups, but it allows you to save much + time, especially for large data volumes, as well as I/O and network + traffic if you are using pg_probackup in the + remote mode. + + + Before the merge, pg_probackup validates all the affected + backups to ensure that they are valid. You can check the current + backup status by running the + command with the backup ID: + + +pg_probackup show -B backup_dir --instance=instance_name -i backup_id + + + If the merge is still in progress, the backup status is + displayed as MERGING. For full backups, + it can also be shown as MERGED while the + metadata is being updated at the final stage of the merge. + The merge is idempotent, so you can + restart the merge if it was interrupted. + + + + Deleting Backups + + To delete a backup that is no longer required, run the following + command: + + +pg_probackup delete -B backup_dir --instance=instance_name -i backup_id + + + This command will delete the backup with the specified + backup_id, together with all the + incremental backups that descend from + backup_id, if any. This way you can delete + some recent incremental backups, retaining the underlying full + backup and some of the incremental backups that follow it. + + + To delete obsolete WAL files that are not necessary to restore + any of the remaining backups, use the + flag: + + +pg_probackup delete -B backup_dir --instance=instance_name --delete-wal + + + To delete backups that are expired according to the current + retention policy, use the + flag: + + +pg_probackup delete -B backup_dir --instance=instance_name --delete-expired + + + Expired backups cannot be removed while at least one + incremental backup that satisfies the retention policy is based + on them. If you would like to minimize the number of backups + still required to keep incremental backups valid, specify the + flag when running this + command: + + +pg_probackup delete -B backup_dir --instance=instance_name --delete-expired --merge-expired + + + In this case, pg_probackup searches for the oldest incremental + backup that satisfies the retention policy and merges this + backup with the underlying full and incremental backups that + have already expired, thus making it a full backup. Once the + merge is complete, the remaining expired backups are deleted. + + + Before merging or deleting backups, you can run the + delete command with the + flag, which displays the status of + all the available backups according to the current retention + policy, without performing any irreversible actions. + + + To delete all backups with specific status, use the : + + +pg_probackup delete -B backup_dir --instance=instance_name --status=ERROR + + + + Deleting backups by status ignores established retention policies. + + + + + + Cloning and Synchronizing <productname>PostgreSQL</productname> Instance + + pg_probackup can create a copy of a PostgreSQL + instance directly, without using the backup catalog. To do this, you can run the command. + It can be useful in the following cases: + + + To add a new standby server. + Usually, pg_basebackup + is used to create a copy of a PostgreSQL instance. If the data directory of the destination instance + is empty, the catchup command works similarly, but it can be faster if run in parallel mode. + + + To have a fallen-behind standby server catch up with master. + Under write-intensive load, replicas may fail to replay WAL fast enough to keep up with master and hence may lag behind. + A usual solution to create a new replica and switch to it requires a lot of extra space and data transfer. The catchup + command allows you to update an existing replica much faster by bringing differences from master. + + + + + + catchup is different from other pg_probackup + operations: + + + + The backup catalog is not required. + + + + + STREAM WAL delivery mode is only supported. + + + + + Copying external directories + is not supported. + + + + + DDL commands + CREATE TABLESPACE/DROP TABLESPACE + cannot be run simultaneously with catchup. + + + + + catchup takes configuration files, such as + postgresql.conf, postgresql.auto.conf, + or pg_hba.conf, from the source server and overwrites them + on the target server. The option allows you to keep + the configuration files intact. + + + + + + + To prepare for cloning/synchronizing a PostgreSQL instance, + set up the source server as follows: + + + + Configure + the database cluster for the instance to copy. + + + + + To copy from a remote server, configure the remote mode. + + + + + To use the PTRACK catchup mode, set up PTRACK backups. + + + + + + + Before cloning/synchronizing a PostgreSQL instance, ensure that the source + server is running and accepting connections. To clone/sync a PostgreSQL instance, + on the server with the destination instance, you can run + the command as follows: + + +pg_probackup catchup -b catchup_mode --source-pgdata=path_to_pgdata_on_remote_server --destination-pgdata=path_to_local_dir --stream [connection_options] [remote_options] + + + Where catchup_mode can take one of the + following values: + + + + + FULL — creates a full copy of the PostgreSQL instance. + The data directory of the destination instance must be empty for this mode. + + + + + DELTA — reads all data files in the data directory and + creates an incremental copy for pages that have changed + since the destination instance was shut down. + + + + + PTRACK — tracking page changes on the fly, + only reads and copies pages that have changed since the point of divergence + of the source and destination instances. + + + PTRACK catchup mode requires PTRACK + not earlier than 2.0 and hence, PostgreSQL not earlier than 11. + + + + + + + By specifying the option, you can set + STREAM WAL delivery mode + of copying, which will include all the necessary WAL files by streaming them from + the server via replication protocol. + + + You can use connection_options to specify + the connection to the source database cluster. If it is located on a different server, + also specify remote_options. + + + If the source database cluster contains tablespaces that must be located in + a different directory, additionally specify the + option: + +pg_probackup catchup -b catchup_mode --source-pgdata=path_to_pgdata_on_remote_server --destination-pgdata=path_to_local_dir --stream --tablespace-mapping=OLDDIR=NEWDIR + + To run the catchup command on parallel threads, specify the number + of threads with the option: + +pg_probackup catchup -b catchup_mode --source-pgdata=path_to_pgdata_on_remote_server --destination-pgdata=path_to_local_dir --stream --threads=num_threads + + + + Before cloning/synchronising a PostgreSQL instance, you can run the + catchup command with the flag + to estimate the size of data files to be transferred, but make no changes on disk: + +pg_probackup catchup -b catchup_mode --source-pgdata=path_to_pgdata_on_remote_server --destination-pgdata=path_to_local_dir --stream --dry-run + + + + For example, assume that a remote standby server with the PostgreSQL instance having /replica-pgdata data directory has fallen behind. To sync this instance with the one in /master-pgdata data directory, you can run + the catchup command in the PTRACK mode on four parallel threads as follows: + +pg_probackup catchup --source-pgdata=/master-pgdata --destination-pgdata=/replica-pgdata -p 5432 -d postgres -U remote-postgres-user --stream --backup-mode=PTRACK --remote-host=remote-hostname --remote-user=remote-unix-username -j 4 --exclude-path=postgresql.conf --exclude-path=postgresql.auto.conf --exclude-path=pg_hba.conf --exclude-path=pg_ident.conf + + Note that in this example, the configuration files will not be overwritten during synchronization. + + + Another example shows how you can add a new remote standby server with the PostgreSQL data directory /replica-pgdata by running the catchup command in the FULL mode + on four parallel threads: + +pg_probackup catchup --source-pgdata=/master-pgdata --destination-pgdata=/replica-pgdata -p 5432 -d postgres -U remote-postgres-user --stream --backup-mode=FULL --remote-host=remote-hostname --remote-user=remote-unix-username -j 4 + + + + + + +Command-Line Reference + + Commands + + This section describes pg_probackup commands. + Optional parameters are enclosed in square brackets. For detailed + parameter descriptions, see the section Options. + + + version + +pg_probackup version + + + Prints pg_probackup version. + + + + help + +pg_probackup help [command] + + + Displays the synopsis of pg_probackup commands. If one of the + pg_probackup commands is specified, shows detailed information + about the options that can be used with this command. + + + + init + +pg_probackup init -B backup_dir [--help] + + + Initializes the backup catalog in + backup_dir that will store backup copies, + WAL archive, and meta information for the backed up database + clusters. If the specified backup_dir + already exists, it must be empty. Otherwise, pg_probackup + displays a corresponding error message. + + + For details, see the section + Initializing + the Backup Catalog. + + + + add-instance + +pg_probackup add-instance -B backup_dir -D data_dir --instance=instance_name [--help] + + + Initializes a new backup instance inside the backup catalog + backup_dir and generates the + pg_probackup.conf configuration file that controls + pg_probackup settings for the cluster with the specified + data_dir data directory. + + + For details, see the section + Adding a New + Backup Instance. + + + + del-instance + +pg_probackup del-instance -B backup_dir --instance=instance_name [--help] + + + Deletes all backups and WAL files associated with the + specified instance. + + + + set-config + +pg_probackup set-config -B backup_dir --instance=instance_name +[--help] [--pgdata=pgdata-path] +[--retention-redundancy=redundancy][--retention-window=window][--wal-depth=wal_depth] +[--compress-algorithm=compression_algorithm] [--compress-level=compression_level] +[-d dbname] [-h host] [-p port] [-U username] +[--archive-timeout=timeout] [--external-dirs=external_directory_path] +[--restore-command=cmdline] +[remote_options] [remote_wal_archive_options] [logging_options] + + + Adds the specified connection, compression, retention, logging, + and external directory settings into the pg_probackup.conf + configuration file, or modifies the previously defined values. + + + For all available settings, see the + Options section. + + + It is not recommended to + edit pg_probackup.conf manually. + + + + set-backup + +pg_probackup set-backup -B backup_dir --instance=instance_name -i backup_id +{--ttl=ttl | --expire-time=time} +[--note=backup_note] [--help] + + + Sets the provided backup-specific settings into the + backup.control configuration file, or modifies the previously + defined values. + + + + + + + Sets the text note for backup copy. + If backup_note contain newline characters, + then only substring before first newline character will be saved. + Max size of text note is 1 KB. + The 'none' value removes current note. + + + + + + For all available pinning settings, see the section + Pinning Options. + + + + show-config + +pg_probackup show-config -B backup_dir --instance=instance_name [--format=plain|json] +[--no-scale-units] [logging_options] + + + Displays the contents of the pg_probackup.conf configuration + file located in the + backup_dir/backups/instance_name + directory. You can specify the + --format=json option to get the result + in the JSON format. By default, configuration settings are + shown as plain text. + + + You can also specify the + option to display time and memory configuration settings in their base (unscaled) units. + Otherwise, the values are scaled to larger units for optimal display. + For example, if archive-timeout is 300, then + 5min is displayed, but if archive-timeout + is 301, then 301s is displayed. + Also, if the option is specified, configuration + settings are displayed without units and for the JSON format, + numeric and boolean values are not enclosed in quotes. This facilitates parsing + the output. + + + To edit pg_probackup.conf, use the + command. + + + + show + +pg_probackup show -B backup_dir +[--help] [--instance=instance_name [-i backup_id | --archive]] [--format=plain|json] [--no-color] + + + Shows the contents of the backup catalog. If + instance_name and + backup_id are specified, shows detailed + information about this backup. If the option is + specified, shows the contents of WAL archive of the backup + catalog. + + + By default, the contents of the backup catalog is shown as + plain text. You can specify the + --format=json option to get the result + in the JSON format. + If --no-color flag is used, + then the output is not colored. + + + For details on usage, see the sections + Managing the + Backup Catalog and + Viewing WAL + Archive Information. + + + + backup + +pg_probackup backup -B backup_dir -b backup_mode --instance=instance_name +[--help] [-j num_threads] [--progress] +[-C] [--stream [-S slot_name] [--temp-slot]] [--backup-pg-log] +[--no-validate] [--skip-block-validation] +[-w --no-password] [-W --password] +[--archive-timeout=timeout] [--external-dirs=external_directory_path] +[--no-sync] [--note=backup_note] +[connection_options] [compression_options] [remote_options] +[retention_options] [pinning_options] [logging_options] + + + Creates a backup copy of the PostgreSQL instance. + + + + + + + + Specifies the backup mode to use. Possible values are: + FULL, + DELTA, + PAGE, and + PTRACK. + + + + + + + + + + Spreads out the checkpoint over a period of time. By default, + pg_probackup tries to complete the checkpoint as soon as + possible. + + + + + + + + + Makes a STREAM backup, which + includes all the necessary WAL files by streaming them from + the database server via replication protocol. + + + + + + + + + Creates a temporary physical replication slot for streaming + WAL from the backed up PostgreSQL instance. It ensures that + all the required WAL segments remain available if WAL is + rotated while the backup is in progress. This flag can only be + used together with the flag. + The default slot name is pg_probackup_slot, + which can be changed using the / option. + + + + + + + + + + Specifies the replication slot for WAL streaming. This option + can only be used together with the + flag. + + + + + + + + + Includes the log directory into the backup. This directory + usually contains log messages. By default, log directory is + excluded. + + + + + + + + + + Includes the specified directory into the backup by recursively + copying its contents into a separate subdirectory in the backup catalog. This option + is useful to back up scripts, SQL dump files, and configuration + files located outside of the data directory. If you would like + to back up several external directories, separate their paths + by a colon on Unix and a semicolon on Windows. + + + + + + + + + Sets the timeout for WAL segment archiving and + streaming, in seconds. By default, pg_probackup waits 300 seconds. + + + + + + + + + Disables block-level checksum verification to speed up + the backup process. + + + + + + + + + Skips automatic validation after the backup is taken. You can + use this flag if you validate backups regularly and would like + to save time when running backup operations. + + + + + + + + + Do not sync backed up files to disk. You can use this flag to speed + up the backup process. Using this flag can result in data + corruption in case of operating system or hardware crash. + If you use this option, it is recommended to run the + command once the backup is complete + to detect possible issues. + + + + + + + + Sets the text note for backup copy. + If backup_note contain newline characters, + then only substring before first newline character will be saved. + Max size of text note is 1 KB. + The 'none' value removes current note. + + + + + + + + + Additionally, connection + options, retention + options, pinning + options, remote + mode options, + compression + options, logging + options, and common + options can be used. + + + For details on usage, see the section + Creating a Backup. + + + + restore + +pg_probackup restore -B backup_dir --instance=instance_name +[--help] [-D data_dir] [-i backup_id] +[-j num_threads] [--progress] +[-T OLDDIR=NEWDIR] [--external-mapping=OLDDIR=NEWDIR] [--skip-external-dirs] +[-R | --restore-as-replica] [--no-validate] [--skip-block-validation] +[--force] [--no-sync] +[--restore-command=cmdline] +[--primary-conninfo=primary_conninfo] +[-S | --primary-slot-name=slot_name] +[-X wal_dir | --waldir=wal_dir] +[recovery_target_options] [logging_options] [remote_options] +[partial_restore_options] [remote_wal_archive_options] + + + Restores the PostgreSQL instance from a backup copy located in + the backup_dir backup catalog. If you + specify a recovery + target option, pg_probackup finds the closest + backup and restores it to the specified recovery target. + If neither the backup ID nor recovery target options are provided, + pg_probackup uses the most recent backup + to perform the recovery. + + + + + + + + + Creates a minimal recovery configuration file to facilitate setting up a + standby server. If the replication connection requires a password, + you must specify the password manually in the + primary_conninfo + parameter as it is not included. + For PostgreSQL 11 or lower, + recovery settings are written into the recovery.conf + file. Starting from PostgreSQL 12, + pg_probackup writes these settings into + the probackup_recovery.conf file in the data + directory, and then includes them into the + postgresql.auto.conf when the cluster is + is started. + + + + + + + + + Sets the + primary_conninfo + parameter to the specified value. + This option will be ignored unless the flag is specified. + + + Example: --primary-conninfo="host=192.168.1.50 port=5432 user=foo password=foopass" + + + + + + + + + + Sets the + primary_slot_name + parameter to the specified value. + This option will be ignored unless the flag is specified. + + + + + + + + + + Relocates the tablespace from the OLDDIR to the NEWDIR + directory at the time of recovery. Both OLDDIR and NEWDIR must + be absolute paths. If the path contains the equals sign (=), + escape it with a backslash. This option can be specified + multiple times for multiple tablespaces. + + + + + + + + + Relocates an external directory included into the backup from + the OLDDIR to the NEWDIR directory at the time of recovery. + Both OLDDIR and NEWDIR must be absolute paths. If the path + contains the equals sign (=), escape it with a backslash. This + option can be specified multiple times for multiple + directories. + + + + + + + + + Skip external directories included into the backup with the + option. The contents of + these directories will not be restored. + + + + + + + + + Disables block-level checksum verification to speed up + validation. During automatic validation before the restore only + file-level checksums will be verified. + + + + + + + + + Skips backup validation. You can use this flag if you validate + backups regularly and would like to save time when running + restore operations. + + + + + + + + + Sets the + restore_command + parameter to the specified command. For example: + --restore-command='cp /mnt/server/archivedir/%f "%p"' + + + + + + + + + Allows to ignore an invalid status of the backup. You can use + this flag if you need to restore the + PostgreSQL cluster from a corrupt or an invalid backup. + Use with caution. + If PGDATA contains a non-empty directory with system ID different from that + of the backup being restored, incremental restore + with this flag overwrites the directory contents (while an error occurs without the flag). If tablespaces + are remapped through the --tablespace-mapping option into non-empty directories, + the contents of such directories will be deleted. + + + + + + + + + Do not sync restored files to disk. You can use this flag to speed + up restore process. Using this flag can result in data + corruption in case of operating system or hardware crash. + If it happens, you have to run the + command again. + + + + + + + + + + Specifies the directory where WAL should be stored. + + + + + + + + Additionally, recovery + target options, + remote mode + options, + remote WAL archive + options, logging + options, partial + restore options, and common + options can be used. + + + For details on usage, see the section + Restoring a + Cluster. + + + + checkdb + +pg_probackup checkdb +[-B backup_dir] [--instance=instance_name] [-D data_dir] +[--help] [-j num_threads] [--progress] +[--amcheck [--skip-block-validation] [--checkunique] [--heapallindexed]] +[connection_options] [logging_options] + + + Verifies the PostgreSQL database cluster correctness by + detecting physical and logical corruption. + + + + + + + + Performs logical verification of indexes for the specified + PostgreSQL instance if no corruption was found while checking + data files. You must have the amcheck + extension or the amcheck_next extension + installed in the database to check its indexes. For databases + without amcheck, index verification will be skipped. + Additional options and + are effective depending on the version of amcheck installed. + + + + + + + + + Verifies unique constraints during logical verification of indexes. + You can use this flag only together with the flag when + the amcheck extension is + installed in the database. + + + The verification of unique constraints is only possible if in the version of the + amcheck extension you are using, the + bt_index_check function takes the + checkunique parameter. + + + + + + + + + Checks that all heap tuples that should be indexed are + actually indexed. You can use this flag only together with the + flag. + + + This check is only possible if in the version of the + amcheck/amcheck_next extension + you are using, the bt_index_check + function takes the heapallindexed parameter. + + + + + + + + + Skip validation of data files. You can use this flag only + together with the flag, so that only logical + verification of indexes is performed. + + + + + + + + Additionally, connection + options and logging + options can be used. + + + For details on usage, see the section + Verifying a + Cluster. + + + + validate + +pg_probackup validate -B backup_dir +[--help] [--instance=instance_name] [-i backup_id] +[-j num_threads] [--progress] +[--skip-block-validation] +[recovery_target_options] [logging_options] + + + Verifies that all the files required to restore the cluster + are present and are not corrupt. If + instance_name is not specified, + pg_probackup validates all backups available in the backup + catalog. If you specify the instance_name + without any additional options, pg_probackup validates all the + backups available for this backup instance. If you specify the + instance_name with a + recovery target + option and/or a backup_id, + pg_probackup checks whether it is possible to restore the + cluster using these options. + + + For details, see the section + Validating a + Backup. + + + + merge + +pg_probackup merge -B backup_dir --instance=instance_name -i backup_id +[--help] [-j num_threads] [--progress] [--no-validate] [--no-sync] +[logging_options] + + + Merges backups that belong to a common incremental backup + chain. If you specify a full backup, it will be merged with its first + incremental backup. If you specify an incremental backup, it will be + merged to its parent full backup, together with all incremental backups + between them. Once the merge is complete, the full backup takes in all + the merged data, and the incremental backups are removed as redundant. + + + + + + + + + Skips automatic validation before and after merge. + + + + + + + + Do not sync merged files to disk. You can use this flag to speed + up the merge process. Using this flag can result in data + corruption in case of operating system or hardware crash. + + + + + + + + For details, see the section + Merging Backups. + + + + delete + +pg_probackup delete -B backup_dir --instance=instance_name +[--help] [-j num_threads] [--progress] +[--retention-redundancy=redundancy][--retention-window=window][--wal-depth=wal_depth] [--delete-wal] +{-i backup_id | --delete-expired [--merge-expired] | --merge-expired | --status=backup_status} +[--dry-run] [--no-validate] [--no-sync] [logging_options] + + + Deletes backup with specified backup_id + or launches the retention purge of backups and archived WAL + that do not satisfy the current retention policies. + + + + + + + + + Skips automatic validation before and after retention merge. + + + + + + + + + Do not sync merged files to disk. You can use this flag to speed + up the retention merge process. Using this flag can result in data + corruption in case of operating system or hardware crash. + + + + + + + + For details, see the sections + Deleting Backups, + Retention Options, and + Configuring + Retention Policy. + + + + archive-push + +pg_probackup archive-push -B backup_dir --instance=instance_name +--wal-file-name=wal_file_name [--wal-file-path=wal_file_path] +[--help] [--no-sync] [--compress] [--no-ready-rename] [--overwrite] +[-j num_threads] [--batch-size=batch_size] +[--archive-timeout=timeout] +[--compress-algorithm=compression_algorithm] +[--compress-level=compression_level] +[remote_options] [logging_options] + + + Copies WAL files into the corresponding subdirectory of the + backup catalog and validates the backup instance by + instance_name and + system-identifier. If parameters of the + backup instance and the cluster do not match, this command + fails with the following error message: Refuse to push WAL + segment segment_name into archive. Instance parameters + mismatch. + + + If the files to be copied already exists in the backup catalog, + pg_probackup computes and compares their checksums. If the + checksums match, archive-push skips the corresponding file and + returns a successful execution code. Otherwise, archive-push + fails with an error. If you would like to replace WAL files in + the case of checksum mismatch, run the archive-push command + with the flag. + + + Each file is copied to a temporary file with the + .part suffix. If the temporary file already + exists, pg_probackup will wait + seconds before discarding it. + After the copy is done, atomic rename is performed. + This algorithm ensures that a failed archive-push + will not stall continuous archiving and that concurrent archiving from + multiple sources into a single WAL archive has no risk of archive + corruption. + + + To speed up archiving, you can specify the option + to copy WAL segments in batches of the specified size. + If option is used, then you can also specify + the option to copy the batch of WAL segments on multiple threads. + + + WAL segments copied to the archive are synced to disk unless + the flag is used. + + + You can use archive-push in the + archive_command + PostgreSQL parameter to set up + continuous + WAL archiving. + + + For details, see sections + Archiving Options and + Compression + Options. + + + + archive-get + +pg_probackup archive-get -B backup_dir --instance=instance_name --wal-file-path=wal_file_path --wal-file-name=wal_file_name +[-j num_threads] [--batch-size=batch_size] +[--prefetch-dir=prefetch_dir_path] [--no-validate-wal] +[--help] [remote_options] [logging_options] + + + Copies WAL files from the corresponding subdirectory of the + backup catalog to the cluster's write-ahead log location. This + command is automatically set by pg_probackup as part of the + restore_command when + restoring backups using a WAL archive. You do not need to set + it manually. + + + + To speed up recovery, you can specify the option + to copy WAL segments in batches of the specified size. + If option is used, then you can also specify + the option to copy the batch of WAL segments on multiple threads. + + + + For details, see section Archiving Options. + + + + + catchup + +pg_probackup catchup -b catchup_mode +--source-pgdata=path_to_pgdata_on_remote_server +--destination-pgdata=path_to_local_dir +[--help] [-j | --threads=num_threads] [--stream] [--dry-run] +[--temp-slot] [-P | --perm-slot] [-S | --slot=slot_name] +[--exclude-path=PATHNAME] +[-T OLDDIR=NEWDIR] +[connection_options] [remote_options] + + + Creates a copy of a PostgreSQL + instance without using the backup catalog. + + + + + + + + Specifies the catchup mode to use. Possible values are: + FULL, + DELTA, and + PTRACK. + + + + + + + + + Specifies the path to the data directory of the instance to be copied. The path can be local or remote. + + + + + + + + + Specifies the path to the local data directory to copy to. + + + + + + + + + + Sets the number of parallel threads for + catchup process. + + + + + + + + + Copies the instance in STREAM WAL delivery mode, + including all the necessary WAL files by streaming them from + the server via replication protocol. + + + + + + + + + Displays the total size of the files to be transferred by catchup. + This flag initiates a trial run of catchup, which does + not actually create, delete or move files on disk. WAL streaming is skipped with . + This flag also allows you to check that + all the options are correct and cloning/synchronising is ready to run. + + + + + +=path_prefix +=path_prefix + + + Specifies a prefix for files to exclude from the synchronization of PostgreSQL + instances during copying. The prefix must contain a path relative to the data directory of an instance. + If the prefix specifies a directory, + all files in this directory will not be synchronized. + + + This option is dangerous since excluding files from synchronization can result in + incomplete synchronization; use with care. + + + + + + + + + + + Creates a temporary physical replication slot for streaming + WAL from the PostgreSQL instance being copied. It ensures that + all the required WAL segments remain available if WAL is + rotated while the backup is in progress. This flag can only be + used together with the flag and + cannot be used together with the flag. + The default slot name is pg_probackup_slot, + which can be changed using the / option. + + + + + + + + + + Creates a permanent physical replication slot for streaming + WAL from the PostgreSQL instance being copied. This flag can only be + used together with the flag and + cannot be used together with the flag. + The default slot name is pg_probackup_perm_slot, + which can be changed using the / option. + + + + + + + + + + Specifies the replication slot for WAL streaming. This option + can only be used together with the + flag. + + + + + + + + + + Relocates the tablespace from the OLDDIR to the NEWDIR + directory at the time of recovery. Both OLDDIR and NEWDIR must + be absolute paths. If the path contains the equals sign (=), + escape it with a backslash. This option can be specified + multiple times for multiple tablespaces. + + + + + + + + + Additionally, connection + options, remote + mode options can be used. + + + For details on usage, see the section + Cloning and Synchronizing PostgreSQL Instance. + + + + + Options + + This section describes command-line options for pg_probackup + commands. If the option value can be derived from an environment + variable, this variable is specified below the command-line + option, in the uppercase. Some values can be taken from the + pg_probackup.conf configuration file located in the backup + catalog. + + + For details, see . + + + If an option is specified using more than one method, + command-line input has the highest priority, while the + pg_probackup.conf settings have the lowest priority. + + + Common Options + + The list of general options. + + + + + + +BACKUP_PATH + + + Specifies the absolute path to the backup catalog. Backup + catalog is a directory where all backup files and meta + information are stored. Since this option is required for most + of the pg_probackup commands, you are recommended to specify + it once in the BACKUP_PATH environment variable. In this case, + you do not need to use this option each time on the command + line. + + + + + + + +PGDATA + + + Specifies the absolute path to the data directory of the + database cluster. This option is mandatory only for the + command. + Other commands can take its value from the PGDATA environment + variable, or from the pg_probackup.conf configuration file. + + + + + + + + + + Specifies the unique identifier of the backup. + + + + + + + + + + Sets the number of parallel threads for backup, + restore, merge, + validate, checkdb, and + archive-push processes. + + + + + + + + + Shows the progress of operations. + + + + + + + + + Shows detailed information about the options that can be used + with this command. + + + + + + + + Recovery Target Options + + If + continuous + WAL archiving is configured, you can use one of these + options together with + or commands to + specify the moment up to which the database cluster must be + restored or validated. + + + + + + + + Defines when to stop the recovery: + + + + The immediate value stops the recovery + after reaching the consistent state of the specified + backup, or the latest available backup if the + / option is omitted. + This is the default behavior for STREAM backups. + + + + + The latest value continues the recovery + until all WAL segments available in the archive are + applied. This is the default behavior for ARCHIVE backups. + + + + + + + + + + + + Specifies a particular timeline to be used for recovery. + By default, the timeline of the specified backup is used. + + + + + + + + + Specifies the LSN of the write-ahead log location up to which + recovery will proceed. + + + + + + + + + Specifies a named savepoint up to which to restore the cluster. + + + + + + + + + Specifies the timestamp up to which recovery will proceed. + If the time zone offset is not specified, the local time zone is used. + + + Example: --recovery-target-time="2027-05-02 11:21:00+00" + + + + + + + + + Specifies the transaction ID up to which recovery will + proceed. + + + + + + + + + + Specifies whether to stop just after the specified recovery + target (true), or just before the recovery target (false). + This option can only be used together with + , + , + or + options. The default + depends on the + recovery_target_inclusive + parameter. + + + + + + + + + Specifies + the + action the server should take when the recovery target + is reached. + + + Default: pause + + + + + + + + Retention Options + + You can use these options together with + and + commands. + + + For details on configuring retention policy, see the section + Configuring + Retention Policy. + + + + + + + + + + Specifies the number of full backup copies to keep in the data + directory. Must be a non-negative integer. The zero value disables + this setting. + + + Default: 0 + + + + + + + + + + Number of days of recoverability. Must be a non-negative integer. + The zero value disables this setting. + + + Default: 0 + + + + + + + + + Number of latest valid backups on every timeline that must + retain the ability to perform PITR. Must be a non-negative + integer. The zero value disables this setting. + + + Default: 0 + + + + + + + + + Deletes WAL files that are no longer required to restore the + cluster from any of the existing backups. + + + + + + + + + Deletes backups that do not conform to the retention policy + defined in the pg_probackup.conf configuration file. + + + + + + + + + + Merges the oldest incremental backup that satisfies the + requirements of retention policy with its parent backups that + have already expired. + + + + + + + + + Displays the current status of all the available backups, + without deleting or merging expired backups, if any. + + + + + + + + Pinning Options + + You can use these options together with + and + commands. + + + For details on backup pinning, see the section + Backup Pinning. + + + + + + + + Specifies the amount of time the backup should be pinned. + Must be a non-negative integer. The zero value unpins the already + pinned backup. Supported units: ms, s, min, h, d (s by + default). + + + Example: --ttl=30d + + + + + + + + + Specifies the timestamp up to which the backup will stay + pinned. Must be an ISO-8601 complaint timestamp. + If the time zone offset is not specified, the local time zone is used. + + + Example: --expire-time="2027-05-02 11:21:00+00" + + + + + + + + Logging Options + + You can use these options with any command. + + + + + + + + + Disable coloring for console log messages of warning and error levels. + + + + + + + + + Controls which message levels are sent to the console log. + Valid values are verbose, + log, info, + warning, error and + off. Each level includes all the levels + that follow it. The later the level, the fewer messages are + sent. The off level disables console + logging. + + + Default: info + + + + All console log messages are going to stderr, so + the output of and + commands does + not mingle with log messages. + + + + + + + + + + Controls which message levels are sent to a log file. Valid + values are verbose, log, + info, warning, + error, and off. Each + level includes all the levels that follow it. The later the + level, the fewer messages are sent. The off + level disables file logging. + + + Default: off + + + + + + + + + Defines the filenames of the created log files. The filenames + are treated as a strftime pattern, so you can use %-escapes to + specify time-varying filenames. + + + Default: pg_probackup.log + + + For example, if you specify the pg_probackup-%u.log pattern, + pg_probackup generates a separate log file for each day of the + week, with %u replaced by the corresponding decimal number: + pg_probackup-1.log for Monday, pg_probackup-2.log for Tuesday, + and so on. + + + This option takes effect if file logging is enabled by the + option. + + + + + + + + + Defines the filenames of log files for error messages only. + The filenames are treated as a strftime pattern, so you can + use %-escapes to specify time-varying filenames. + + + Default: none + + + For example, if you specify the error-pg_probackup-%u.log + pattern, pg_probackup generates a separate log file for each + day of the week, with %u replaced by the corresponding decimal + number: error-pg_probackup-1.log for Monday, + error-pg_probackup-2.log for Tuesday, and so on. + + + This option is useful for troubleshooting and monitoring. + + + + + + + + + Defines the directory in which log files will be created. You + must specify the absolute path. This directory is created + lazily, when the first log message is written. + + + Default: $BACKUP_PATH/log/ + + + + + + + + + Defines the format of the console log. Only set from the command line. Note that you cannot + specify this option in the pg_probackup.conf configuration file through + the command and that the + command also treats this option specified in the configuration file as an error. + Possible values are: + + + + + plain — sets the plain-text format of the console log. + + + + + json — sets the JSON format of the console log. + + + + + + Default: plain + + + + + + + + + Defines the format of log files used. Possible values are: + + + + + plain — sets the plain-text format of log files. + + + + + json — sets the JSON format of log files. + + + + + + Default: plain + + + + + + + + + Maximum size of an individual log file. If this value is + reached, the log file is rotated once a pg_probackup command + is launched, except help and version commands. The zero value + disables size-based rotation. Supported units: kB, MB, GB, TB + (kB by default). + + + Default: 0 + + + + + + + + + Maximum lifetime of an individual log file. If this value is + reached, the log file is rotated once a pg_probackup command + is launched, except help and version commands. The time of the + last log file creation is stored in + $BACKUP_PATH/log/log_rotation. The zero value disables + time-based rotation. Supported units: ms, s, min, h, d (min by + default). + + + Default: 0 + + + + + + + + Connection Options + + You can use these options together with + , + , and + commands. + + + All + libpq + environment variables are supported. + + + + + + +PGDATABASE + + + Specifies the name of the database to connect to. The + connection is used only for managing backup process, so you + can connect to any existing database. If this option is not + provided on the command line, PGDATABASE environment variable, + or the pg_probackup.conf configuration file, pg_probackup + tries to take this value from the PGUSER environment variable, + or from the current user name if PGUSER variable is not set. + + + + + + + +PGHOST + + + Specifies the host name of the system on which the server is + running. If the value begins with a slash, it is used as a + directory for the Unix domain socket. + + + Default: localhost + + + + + + + +PGPORT + + + Specifies the TCP port or the local Unix domain socket file + extension on which the server is listening for connections. + + + Default: 5432 + + + + + + + +PGUSER + + + User name to connect as. + + + + + + + + + + Disables a password prompt. If the server requires password + authentication and a password is not available by other means + such as a + .pgpass + file or PGPASSWORD environment variable, the connection + attempt will fail. This flag can be useful in batch jobs and + scripts where no user is present to enter a password. + + + + + + + + + + + Forces a password prompt. (Deprecated) + + + + + + + + Compression Options + + You can use these options together with + and + commands. + + + + + + + + Defines the algorithm to use for compressing data files. + Possible values are zlib, + pglz, and none. If set + to zlib or pglz, this option enables compression. By default, + compression is disabled. For the + command, the + pglz compression algorithm is not supported. + + + Default: none + + + + + + + + + Defines compression level (0 through 9, 0 being no compression + and 9 being best compression). This option can be used + together with the option. + + + Default: 1 + + + + + + + + + Alias for --compress-algorithm=zlib and + --compress-level=1. + + + + + + + + Archiving Options + + These options can be used with the + command in the + archive_command + setting and the + command in the + restore_command + setting. + + + Additionally, remote mode + options and logging + options can be used. + + + + + + + + Provides the path to the WAL file in + archive_command and + restore_command. Use the %p + variable as the value for this option or explicitly specify the path to a file + outside of the data directory. If you skip this option, the path + specified in pg_probackup.conf will be used. + + + + + + + + + Provides the name of the WAL file in + archive_command and + restore_command. Use the %f + variable as the value for this option for correct processing. + If the value of is a path + outside of the data directory, explicitly specify the filename. + + + + + + + + + Overwrites archived WAL file. Use this flag together with the + command if + the specified subdirectory of the backup catalog already + contains this WAL file and it needs to be replaced with its + newer copy. Otherwise, archive-push reports that a WAL segment + already exists, and aborts the operation. If the file to + replace has not changed, archive-push skips this file + regardless of the flag. + + + + + + + + + Sets the maximum number of files that can be copied into the archive + by a single archive-push process, or from + the archive by a single archive-get process. + + + + + + + + + Sets the timeout for considering existing .part + files to be stale. By default, pg_probackup + waits 300 seconds. + This option can be used only with command. + + + + + + + + + Do not rename status files in the archive_status directory. + This option should be used only if archive_command + contains multiple commands. + This option can be used only with command. + + + + + + + + + Do not sync copied WAL files to disk. You can use this flag to speed + up archiving process. Using this flag can result in WAL archive + corruption in case of operating system or hardware crash. + This option can be used only with command. + + + + + + + + + Directory used to store prefetched WAL segments if option is used. + Directory must be located on the same filesystem and on the same mountpoint the + PGDATA/pg_wal is located. + By default files are stored in PGDATA/pg_wal/pbk_prefetch directory. + This option can be used only with command. + + + + + + + + + Do not validate prefetched WAL file before using it. + Use this option if you want to increase the speed of recovery. + This option can be used only with command. + + + + + + + + + Remote Mode Options + + This section describes the options related to running + pg_probackup operations remotely via SSH. These options can be + used with , + , + , + , + , + , and + commands. + + + For details on configuring and using the remote mode, + see and + . + + + + + + + + Specifies the protocol to use for remote operations. Currently + only the SSH protocol is supported. Possible values are: + + + + + ssh enables the remote mode via + SSH. This is the default value. + + + + + none explicitly disables the remote + mode. + + + + + You can omit this option if the + option is specified. + + + + + + + + + + Specifies the remote host IP address or hostname to connect + to. + + + + + + + + + Specifies the remote host port to connect to. + + + Default: 22 + + + + + + + + + Specifies remote host user for SSH connection. If you omit + this option, the current user initiating the SSH connection is + used. + + + + + + + + + Specifies pg_probackup installation directory on the remote + system. + + + + + + + + + Provides a string of SSH command-line options. For example, + the following options can be used to set keep-alive for SSH + connections opened by pg_probackup: + --ssh-options="-o ServerAliveCountMax=5 -o ServerAliveInterval=60". + For the full list of possible options, see + ssh_config + manual page. + + + + + + + + Remote WAL Archive Options + + This section describes the options used to provide the + arguments for remote mode + options in + used in the + restore_command + command when restoring ARCHIVE backups or performing PITR. + + + + + + + + Provides the argument for the + option in the archive-get command. + + + + + + + + + Provides the argument for the + option in the archive-get command. + + + Default: 22 + + + + + + + + + Provides the argument for the + option in the archive-get command. If you omit + this option, the user that has started the PostgreSQL cluster is used. + + + Default: PostgreSQL user + + + + + + + + Incremental Restore Options + + This section describes the options for incremental cluster restore. + These options can be used with the + command. + + + + + + + + + Specifies the incremental mode to be used. Possible values are: + + + + + CHECKSUM — replace only pages with mismatched checksum and LSN. + + + + + LSN — replace only pages with LSN greater than point of divergence. + + + + + NONE — regular restore. + + + + + + + + + + + Partial Restore Options + + This section describes the options for partial cluster restore. + These options can be used with the + command. + + + + + + + + Specifies the name of the database to exclude from restore. All other + databases in the cluster will be restored as usual, including + template0 and template1. + This option can be specified multiple times for multiple + databases. + + + + + + + + + Specifies the name of the database to restore from a backup. All other + databases in the cluster will not be restored, with the exception + of template0 and + template1. This option can be specified + multiple times for multiple databases. + + + + + + + + + + + Versioning + + pg_probackup follows + semantic versioning. + + + + + Authors + + + Postgres Professional, Moscow, Russia. + + + + Credits + + pg_probackup utility is based on pg_arman, + which was originally written by NTT and then developed and maintained by Michael Paquier. + + + + + + + diff --git a/doc/probackup.xml b/doc/probackup.xml new file mode 100644 index 000000000..8ea3cdc46 --- /dev/null +++ b/doc/probackup.xml @@ -0,0 +1,12 @@ + +]> + + + + pg_probackup Documentation +&pgprobackup; + + \ No newline at end of file diff --git a/doc/stylesheet.css b/doc/stylesheet.css new file mode 100644 index 000000000..31464154b --- /dev/null +++ b/doc/stylesheet.css @@ -0,0 +1,421 @@ +@import url('https://fonts.googleapis.com/css?family=Roboto:300,400,500,700&subset=cyrillic'); + +body { + font-family: 'Roboto',Arial,sans-serif; +} + +body { + font-size: 18px; + font-weight: 300; +} + +/* ../media/css/docs.css */ +.navheader th { text-align: center; } /* anti-bootstrap */ + +.navheader tbody tr:nth-child(1) th { /* временно убрать ненужную строчку */ + display: none; +} + +/* PostgreSQL.org Documentation Style */ + +.book div.NAVHEADER table { + margin-left: 0; +} + +.book div.NAVHEADER th { + text-align: center; +} + +.book { + font-size: 15px; + line-height: 1.6; +} + +/* Heading Definitions */ + +.book h1, +.book h2, +.book h3 { + font-weight: bold; + margin-top: 2ex; +} + +.book h1 a, +.book h2 a, +.book h3 a, +.book h4 a + { + color: #EC5800; +} + +/* EKa --> */ +.book h1 { + font-size: 1.4em; +} + +.book h2 { + font-size: 1.25em; +} + +.book h3 { + font-size: 1.2em; +} + +.book h4 { + font-size: 1.15em; +} + +.book h5 { + font-size: 1.1em; +} + +.book h6 { + font-size: 1.0em; +} +/* <-- EKa */ + +.book h1 a:hover { + color: #EC5800; + text-decoration: none; +} + +.book h2 a:hover, +.book h3 a:hover, +.book h4 a:hover { + color: #666666; + text-decoration: none; +} + + + +/* Text Styles */ + +.book div.SECT2 { + margin-top: 4ex; +} + +.book div.SECT3 { + margin-top: 3ex; + margin-left: 3ex; +} + +.book .txtCurrentLocation { + font-weight: bold; +} + +.book p, +.book ol, +.book ul, +.book li { + line-height: 1.5em; +} + +.book code { + font-size: 1em; + padding: 0px; + color: #525f6c; + background-color: #FFF; + border-radius: 0px; +} + +.book code, kbd, pre, samp { + font-family: monospace,monospace; + font-size: 90%; +} + +.book .txtCommentsWrap { + border: 2px solid #F5F5F5; + width: 100%; +} + +.book .txtCommentsContent { + background: #F5F5F5; + padding: 3px; +} + +.book .txtCommentsPoster { + float: left; +} + +.book .txtCommentsDate { + float: right; +} + +.book .txtCommentsComment { + padding: 3px; +} + +.book #docContainer pre code, +.book #docContainer pre tt, +.book #docContainer pre pre, +.book #docContainer tt tt, +.book #docContainer tt code, +.book #docContainer tt pre { + font-size: 1em; +} + +.book pre.LITERALLAYOUT, +.book .SCREEN, +.book .SYNOPSIS, +.book .PROGRAMLISTING, +.book .REFSYNOPSISDIV p, +.book table.CAUTION, +.book table.WARNING, +.book blockquote.NOTE, +.book blockquote.TIP, +.book div.note, +.book div.tip, +.book table.CALSTABLE { + -moz-box-shadow: 3px 3px 5px #DFDFDF; + -webkit-box-shadow: 3px 3px 5px #DFDFDF; + -khtml-box-shadow: 3px 3px 5px #DFDFDF; + -o-box-shadow: 3px 3px 5px #DFDFDF; + box-shadow: 3px 3px 5px #DFDFDF; +} + +.book pre.LITERALLAYOUT, +.book .SCREEN, +.book .SYNOPSIS, +.book .PROGRAMLISTING, +.book .REFSYNOPSISDIV p, +.book table.CAUTION, +.book table.WARNING, +.book blockquote.NOTE, +.book blockquote.TIP +.book div.note, +.book div.tip { + color: black; + border-width: 1px; + border-style: solid; + padding: 2ex; + margin: 2ex 0 2ex 2ex; + overflow: auto; + -moz-border-radius: 8px; + -webkit-border-radius: 8px; + -khtml-border-radius: 8px; + border-radius: 8px; +} + +.book div.note, +.book div.tip { + -moz-border-radius: 8px !important; + -webkit-border-radius: 8px !important; + -khtml-border-radius: 8px !important; + border-radius: 8px !important; +} + + +.book pre.LITERALLAYOUT, +.book pre.SYNOPSIS, +.book pre.PROGRAMLISTING, +.book .REFSYNOPSISDIV p, +.book .SCREEN { + border-color: #CFCFCF; + background-color: #F7F7F7; +} + +.book blockquote.NOTE, +.book blockquote.TIP, +.book div.note, +.book div.tip { + border-color: #DBDBCC; + background-color: #EEEEDD; + padding: 14px; + width: 572px; +/* font-size: 12px; */ +} + +.book blockquote.NOTE, +.book blockquote.TIP, +.book table.CAUTION, +.book table.WARNING { + margin: 4ex auto; +} + +.book div.note, +.book div.tip { + margin: 4ex auto !important; +} + + +.book blockquote.NOTE p, +.book blockquote.TIP p, +.book div.note p, +.book div.tip p { + margin: 0; +} + +.book blockquote.NOTE pre, +.book blockquote.NOTE code, +.book div.note pre, +.book div.note code, +.book blockquote.TIP pre, +.book blockquote.TIP code, +.book div.tip pre, +.book div.tio code { + margin-left: 0; + margin-right: 0; + -moz-box-shadow: none; + -webkit-box-shadow: none; + -khtml-box-shadow: none; + -o-box-shadow: none; + box-shadow: none; +} + +.book .emphasis, +.book .c2 { + font-weight: bold; +} + +.book .REPLACEABLE { + font-style: italic; +} + +/* Table Styles */ + +.book table { + margin-left: 2ex; +} + +.book table.CALSTABLE td, +.book table.CALSTABLE th, +.book table.CAUTION td, +.book table.CAUTION th, +.book table.WARNING td, +.book table.WARNING th { + border-style: solid; +} + +.book table.CALSTABLE, +.book table.CAUTION, +.book table.WARNING { + border-spacing: 0; + border-collapse: collapse; +} + +.book table.CALSTABLE +{ + margin: 2ex 0 2ex 2ex; + background-color: #E0ECEF; + border: 2px solid #A7C6DF; +} + +.book table.CALSTABLE tr:hover td +{ + background-color: #EFEFEF; +} + +.book table.CALSTABLE td { + background-color: #FFF; +} + +.book table.CALSTABLE td, +.book table.CALSTABLE th { + border: 1px solid #A7C6DF; + padding: 0.5ex 0.5ex; +} + +table.CAUTION, +.book table.WARNING { + border-collapse: separate; + display: block; + padding: 0; + max-width: 600px; +} + +.book table.CAUTION { + background-color: #F5F5DC; + border-color: #DEDFA7; +} + +.book table.WARNING { + background-color: #FFD7D7; + border-color: #DF421E; +} + +.book table.CAUTION td, +.book table.CAUTION th, +.book table.WARNING td, +.book table.WARNING th { + border-width: 0; + padding-left: 2ex; + padding-right: 2ex; +} + +.book table.CAUTION td, +.book table.CAUTION th { + border-color: #F3E4D5 +} + +.book table.WARNING td, +.book table.WARNING th { + border-color: #FFD7D7; +} + +.book td.c1, +.book td.c2, +.book td.c3, +.book td.c4, +.book td.c5, +.book td.c6 { + font-size: 1.1em; + font-weight: bold; + border-bottom: 0px solid #FFEFEF; + padding: 1ex 2ex 0; +} + +.book .table thead { + background: #E0ECEF; + border-bottom: 1px solid #000; +} +.book .table > thead > tr > th { + border-bottom: 1px solid #000; +} + +.book td, th { + padding: 0.1ex 0.5ex; +} + +.book .book table tr:hover td { + background-color: #EFEFEF; +} + +/* Link Styles */ + +.book #docNav a { + font-weight: bold; +} + +.book code.FUNCTION tt { + font-size: 1em; +} + +.book table.docs-compare { + align: center; + width: 90%; + border: 2px solid #999; + border-collapse: collapse; +} + +.book table.docs-compare td { + padding: 12px; + border: 1px solid #DDD; +} + +.book dd { + margin-left: 40px; +} + + +.book .sidebar { + padding: 8px; + background: #FFF; + width: auto; +} + +.book pre { + background: #f5f5f5; + padding: 10px; + border: 1px solid #ccc; + border-radius: 4px; +} diff --git a/doc/stylesheet.xsl b/doc/stylesheet.xsl new file mode 100644 index 000000000..466127e9c --- /dev/null +++ b/doc/stylesheet.xsl @@ -0,0 +1,50 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +book/reference toc, title + + + diff --git a/doit.cmd b/doit.cmd deleted file mode 100644 index 7830a7ead..000000000 --- a/doit.cmd +++ /dev/null @@ -1,3 +0,0 @@ -CALL "C:\Program Files (x86)\Microsoft Visual Studio 12.0\VC\vcvarsall" amd64 -SET PERL5LIB=. -perl gen_probackup_project.pl C:\Shared\Postgresql\myPostgres\11\postgrespro \ No newline at end of file diff --git a/doit96.cmd b/doit96.cmd deleted file mode 100644 index 94d242c99..000000000 --- a/doit96.cmd +++ /dev/null @@ -1 +0,0 @@ -perl win32build96.pl "C:\PgPro96" "C:\PgProject\pg96ee\postgrespro\src" \ No newline at end of file diff --git a/gen_probackup_project.pl b/gen_probackup_project.pl index 511e7e07c..8143b7d0d 100644 --- a/gen_probackup_project.pl +++ b/gen_probackup_project.pl @@ -13,11 +13,11 @@ BEGIN { $pgsrc = shift @ARGV; if($pgsrc eq "--help"){ - print STDERR "Usage $0 pg-source-dir \n"; - print STDERR "Like this: \n"; - print STDERR "$0 C:/PgProject/postgresql.10dev/postgrespro \n"; - print STDERR "May be need input this before: \n"; - print STDERR "CALL \"C:\\Program Files (x86)\\Microsoft Visual Studio 12.0\\VC\\vcvarsall\" amd64\n"; + print STDERR "Usage $0 pg-source-dir\n"; + print STDERR "Like this:\n"; + print STDERR "$0 C:/PgProject/postgresql.10dev/postgrespro\n"; + print STDERR "May need to run this first:\n"; + print STDERR "CALL \"C:\\Program Files (x86)\\Microsoft Visual Studio\\2019\\Community\\VC\\Auxiliary\\Build\\vcvarsall.bat\" amd64\n"; exit 1; } } @@ -61,6 +61,16 @@ BEGIN my $libpq; my @unlink_on_exit; +if (-d "src/fe_utils") +{ + $libpgfeutils = 1; +} +else +{ + $libpgfeutils = 0; +} + + use lib "src/tools/msvc"; @@ -123,13 +133,16 @@ sub build_pgprobackup unless (-d 'src/tools/msvc' && -d 'src'); # my $vsVersion = DetermineVisualStudioVersion(); - my $vsVersion = '12.00'; + my $vsVersion = '16.00'; $solution = CreateSolution($vsVersion, $config); $libpq = $solution->AddProject('libpq', 'dll', 'interfaces', 'src/interfaces/libpq'); - $libpgfeutils = $solution->AddProject('libpgfeutils', 'lib', 'misc'); + if ($libpgfeutils) + { + $libpgfeutils = $solution->AddProject('libpgfeutils', 'lib', 'misc'); + } $libpgcommon = $solution->AddProject('libpgcommon', 'lib', 'misc'); $libpgport = $solution->AddProject('libpgport', 'lib', 'misc'); @@ -142,6 +155,7 @@ sub build_pgprobackup 'archive.c', 'backup.c', 'catalog.c', + 'catchup.c', 'configure.c', 'data.c', 'delete.c', @@ -154,9 +168,11 @@ sub build_pgprobackup 'pg_probackup.c', 'restore.c', 'show.c', + 'stream.c', 'util.c', 'validate.c', - 'checkdb.c' + 'checkdb.c', + 'ptrack.c' ); $probackup->AddFiles( "$currpath/src/utils", @@ -193,15 +209,21 @@ sub build_pgprobackup $probackup->AddIncludeDir("$pgsrc/src/interfaces/libpq"); $probackup->AddIncludeDir("$pgsrc/src"); $probackup->AddIncludeDir("$pgsrc/src/port"); + $probackup->AddIncludeDir("$pgsrc/src/include/portability"); $probackup->AddIncludeDir("$currpath"); $probackup->AddIncludeDir("$currpath/src"); $probackup->AddIncludeDir("$currpath/src/utils"); - # $probackup->AddReference($libpq, $libpgfeutils, $libpgcommon, $libpgport); - $probackup->AddReference($libpq, $libpgcommon, $libpgport); + if ($libpgfeutils) + { + $probackup->AddReference($libpq, $libpgfeutils, $libpgcommon, $libpgport); + } + else + { + $probackup->AddReference($libpq, $libpgcommon, $libpgport); + } $probackup->AddLibrary('ws2_32.lib'); - $probackup->AddLibrary('advapi32.lib'); $probackup->Save(); return $solution->{vcver}; diff --git a/nls.mk b/nls.mk new file mode 100644 index 000000000..981c1c4fe --- /dev/null +++ b/nls.mk @@ -0,0 +1,6 @@ +# contrib/pg_probackup/nls.mk +CATALOG_NAME = pg_probackup +AVAIL_LANGUAGES = ru +GETTEXT_FILES = src/help.c +GETTEXT_TRIGGERS = $(FRONTEND_COMMON_GETTEXT_TRIGGERS) +GETTEXT_FLAGS = $(FRONTEND_COMMON_GETTEXT_FLAGS) diff --git a/packaging/Dockerfiles/Dockerfile-altlinux_8 b/packaging/Dockerfiles/Dockerfile-altlinux_8 new file mode 100644 index 000000000..961aa43dd --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-altlinux_8 @@ -0,0 +1,5 @@ +FROM alt:p8 +RUN ulimit -n 1024 && apt-get update -y && apt-get install -y tar wget rpm-build +RUN ulimit -n 1024 && apt-get install -y make perl libicu-devel glibc-devel bison flex +RUN ulimit -n 1024 && apt-get install -y git perl-devel readline-devel libxml2-devel libxslt-devel python-devel zlib-devel openssl-devel libkrb5 libkrb5-devel +RUN ulimit -n 1024 && apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-altlinux_9 b/packaging/Dockerfiles/Dockerfile-altlinux_9 new file mode 100644 index 000000000..a75728720 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-altlinux_9 @@ -0,0 +1,5 @@ +FROM alt:p9 +RUN ulimit -n 1024 && apt-get update -y && apt-get install -y tar wget rpm-build +RUN ulimit -n 1024 && apt-get install -y make perl libicu-devel glibc-devel bison flex +RUN ulimit -n 1024 && apt-get install -y git perl-devel readline-devel libxml2-devel libxslt-devel python-devel zlib-devel openssl-devel libkrb5 libkrb5-devel +RUN ulimit -n 1024 && apt-get dist-upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-astra_1.11 b/packaging/Dockerfiles/Dockerfile-astra_1.11 new file mode 100644 index 000000000..7db4999cd --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-astra_1.11 @@ -0,0 +1,7 @@ +FROM pgpro/astra:1.11 +RUN apt-get update -y +RUN apt-get install -y devscripts +RUN apt-get install -y dpkg-dev lsb-release git equivs wget vim +RUN apt-get install -y cmake bison flex libboost-all-dev +RUN apt-get install reprepro -y +RUN apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-centos_7 b/packaging/Dockerfiles/Dockerfile-centos_7 new file mode 100644 index 000000000..363440e85 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-centos_7 @@ -0,0 +1,5 @@ +FROM centos:7 +RUN yum install -y tar wget rpm-build yum-utils +RUN yum install -y gcc make perl libicu-devel glibc-devel bison flex +RUN yum install -y git +RUN yum upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-centos_8 b/packaging/Dockerfiles/Dockerfile-centos_8 new file mode 100644 index 000000000..9de1d31b1 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-centos_8 @@ -0,0 +1,5 @@ +FROM centos:8 +RUN yum install -y tar wget rpm-build yum-utils +RUN yum install -y gcc make perl libicu-devel glibc-devel bison flex +RUN yum install -y git +RUN yum upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-createrepo1C b/packaging/Dockerfiles/Dockerfile-createrepo1C new file mode 100644 index 000000000..d987c4f5f --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-createrepo1C @@ -0,0 +1,4 @@ +FROM ubuntu:17.10 +RUN apt-get -qq update -y +RUN apt-get -qq install -y reprepro rpm createrepo gnupg rsync perl less wget expect rsync dpkg-dev +RUN apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-debian_10 b/packaging/Dockerfiles/Dockerfile-debian_10 new file mode 100644 index 000000000..f25ceeac5 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-debian_10 @@ -0,0 +1,7 @@ +FROM debian:10 +RUN apt-get update -y +RUN apt-get install -y devscripts +RUN apt-get install -y dpkg-dev lsb-release git equivs wget vim +RUN apt-get install -y cmake bison flex libboost-all-dev +RUN apt-get install -y reprepro +RUN apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-debian_11 b/packaging/Dockerfiles/Dockerfile-debian_11 new file mode 100644 index 000000000..db736c193 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-debian_11 @@ -0,0 +1,7 @@ +FROM debian:11 +RUN apt-get update -y +RUN apt-get install -y devscripts +RUN apt-get install -y dpkg-dev lsb-release git equivs wget vim +RUN apt-get install -y cmake bison flex libboost-all-dev +RUN apt-get install -y reprepro +RUN apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-debian_8 b/packaging/Dockerfiles/Dockerfile-debian_8 new file mode 100644 index 000000000..0be9528bb --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-debian_8 @@ -0,0 +1,7 @@ +FROM debian:8 +RUN apt-get update -y +RUN apt-get install -y devscripts +RUN apt-get install -y dpkg-dev lsb-release git equivs wget vim +RUN apt-get install -y cmake bison flex libboost-all-dev +RUN apt-get install -y reprepro +RUN apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-debian_9 b/packaging/Dockerfiles/Dockerfile-debian_9 new file mode 100644 index 000000000..6ca10faa8 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-debian_9 @@ -0,0 +1,7 @@ +FROM debian:9 +RUN apt-get update -y +RUN apt-get install -y devscripts +RUN apt-get install -y dpkg-dev lsb-release git equivs wget vim +RUN apt-get install -y cmake bison flex libboost-all-dev +RUN apt-get install -y reprepro +RUN apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-oraclelinux_6 b/packaging/Dockerfiles/Dockerfile-oraclelinux_6 new file mode 100644 index 000000000..04325e869 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-oraclelinux_6 @@ -0,0 +1,5 @@ +FROM oraclelinux:6 +RUN yum install -y tar wget rpm-build yum-utils +RUN yum install -y gcc make perl libicu-devel glibc-devel bison flex +RUN yum install -y git openssl +RUN yum upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-oraclelinux_7 b/packaging/Dockerfiles/Dockerfile-oraclelinux_7 new file mode 100644 index 000000000..871d920eb --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-oraclelinux_7 @@ -0,0 +1,5 @@ +FROM oraclelinux:7 +RUN yum install -y tar wget rpm-build yum-utils +RUN yum install -y gcc make perl libicu-devel glibc-devel bison flex +RUN yum install -y git openssl +RUN yum upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-oraclelinux_8 b/packaging/Dockerfiles/Dockerfile-oraclelinux_8 new file mode 100644 index 000000000..32e7cb03f --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-oraclelinux_8 @@ -0,0 +1,5 @@ +FROM oraclelinux:8 +RUN yum install -y tar wget rpm-build yum-utils +RUN yum install -y gcc make perl libicu-devel glibc-devel bison flex +RUN yum install -y git openssl +RUN yum upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-rhel_7 b/packaging/Dockerfiles/Dockerfile-rhel_7 new file mode 100644 index 000000000..f64819e13 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-rhel_7 @@ -0,0 +1,9 @@ +FROM registry.access.redhat.com/ubi7 +RUN yum install -y http://mirror.centos.org/centos/7/os/x86_64/Packages/elfutils-0.176-5.el7.x86_64.rpm +RUN yum install -y http://mirror.centos.org/centos/7/os/x86_64/Packages/rpm-build-4.11.3-45.el7.x86_64.rpm +RUN yum install -y tar wget yum-utils +RUN yum install -y gcc make perl libicu-devel glibc-devel +RUN yum install -y git +RUN yum upgrade -y +RUN yum install -y http://mirror.centos.org/centos/7/os/x86_64/Packages/bison-3.0.4-2.el7.x86_64.rpm +RUN yum install -y http://mirror.centos.org/centos/7/os/x86_64/Packages/flex-2.5.37-6.el7.x86_64.rpm diff --git a/packaging/Dockerfiles/Dockerfile-rhel_8 b/packaging/Dockerfiles/Dockerfile-rhel_8 new file mode 100644 index 000000000..82385785b --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-rhel_8 @@ -0,0 +1,7 @@ +FROM registry.access.redhat.com/ubi8 +RUN yum install -y tar wget rpm-build yum-utils +RUN yum install -y gcc make perl libicu-devel glibc-devel +RUN yum install -y git +RUN yum upgrade -y +RUN yum install -y http://mirror.centos.org/centos/8/AppStream/x86_64/os/Packages/bison-3.0.4-10.el8.x86_64.rpm +RUN yum install -y http://mirror.centos.org/centos/8/AppStream/x86_64/os/Packages/flex-2.6.1-9.el8.x86_64.rpm diff --git a/packaging/Dockerfiles/Dockerfile-rosa_6 b/packaging/Dockerfiles/Dockerfile-rosa_6 new file mode 100644 index 000000000..42fa913e1 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-rosa_6 @@ -0,0 +1,5 @@ +FROM pgpro/rosa-6 +RUN yum install -y tar wget rpm-build yum-utils +RUN yum install -y gcc make perl libicu-devel glibc-devel bison flex +RUN yum install -y git +RUN yum upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-suse_15.1 b/packaging/Dockerfiles/Dockerfile-suse_15.1 new file mode 100644 index 000000000..afc9434a2 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-suse_15.1 @@ -0,0 +1,4 @@ +FROM opensuse/leap:15.1 +RUN ulimit -n 1024 && zypper install -y tar wget rpm-build +RUN ulimit -n 1024 && zypper install -y gcc make perl libicu-devel glibc-devel bison flex +RUN ulimit -n 1024 && zypper install -y git rsync diff --git a/packaging/Dockerfiles/Dockerfile-suse_15.2 b/packaging/Dockerfiles/Dockerfile-suse_15.2 new file mode 100644 index 000000000..7e56e299a --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-suse_15.2 @@ -0,0 +1,4 @@ +FROM opensuse/leap:15.2 +RUN ulimit -n 1024 && zypper install -y tar wget rpm-build +RUN ulimit -n 1024 && zypper install -y gcc make perl libicu-devel glibc-devel bison flex +RUN ulimit -n 1024 && zypper install -y git rsync diff --git a/packaging/Dockerfiles/Dockerfile-ubuntu_14.04 b/packaging/Dockerfiles/Dockerfile-ubuntu_14.04 new file mode 100644 index 000000000..10132f826 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-ubuntu_14.04 @@ -0,0 +1,7 @@ +FROM ubuntu:14.04 +RUN apt-get update -y +RUN apt-get install -y devscripts +RUN apt-get install -y dpkg-dev lsb-release git equivs wget vim +RUN apt-get install -y cmake bison flex libboost-all-dev +RUN apt-get install -y reprepro +RUN apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-ubuntu_16.04 b/packaging/Dockerfiles/Dockerfile-ubuntu_16.04 new file mode 100644 index 000000000..d511829c0 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-ubuntu_16.04 @@ -0,0 +1,7 @@ +FROM ubuntu:16.04 +RUN apt-get update -y +RUN apt-get install -y devscripts +RUN apt-get install -y dpkg-dev lsb-release git equivs wget vim +RUN apt-get install -y cmake bison flex libboost-all-dev +RUN apt-get install -y reprepro +RUN apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-ubuntu_18.04 b/packaging/Dockerfiles/Dockerfile-ubuntu_18.04 new file mode 100644 index 000000000..20a8567e0 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-ubuntu_18.04 @@ -0,0 +1,7 @@ +FROM ubuntu:18.04 +RUN apt-get update -y +RUN apt-get install -y devscripts +RUN apt-get install -y dpkg-dev lsb-release git equivs wget vim +RUN apt-get install -y cmake bison flex libboost-all-dev +RUN apt-get install -y reprepro +RUN apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-ubuntu_18.10 b/packaging/Dockerfiles/Dockerfile-ubuntu_18.10 new file mode 100644 index 000000000..66cefff16 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-ubuntu_18.10 @@ -0,0 +1,7 @@ +FROM ubuntu:18.10 +RUN apt-get update -y +RUN apt-get install -y devscripts +RUN apt-get install -y dpkg-dev lsb-release git equivs wget vim +RUN apt-get install -y cmake bison flex libboost-all-dev +RUN apt-get install -y reprepro +RUN apt-get upgrade -y diff --git a/packaging/Dockerfiles/Dockerfile-ubuntu_20.04 b/packaging/Dockerfiles/Dockerfile-ubuntu_20.04 new file mode 100644 index 000000000..79288d308 --- /dev/null +++ b/packaging/Dockerfiles/Dockerfile-ubuntu_20.04 @@ -0,0 +1,8 @@ +FROM ubuntu:20.04 +ENV DEBIAN_FRONTEND noninteractive +RUN ulimit -n 1024 && apt-get update -y +RUN ulimit -n 1024 && apt-get install -y devscripts +RUN ulimit -n 1024 && apt-get install -y dpkg-dev lsb-release git equivs wget vim +RUN ulimit -n 1024 && apt-get install -y cmake bison flex libboost-all-dev +RUN ulimit -n 1024 && apt-get install -y reprepro +RUN ulimit -n 1024 && apt-get upgrade -y diff --git a/packaging/Makefile.pkg b/packaging/Makefile.pkg new file mode 100644 index 000000000..e17243614 --- /dev/null +++ b/packaging/Makefile.pkg @@ -0,0 +1,185 @@ +ifeq ($(PBK_EDITION),std) + PBK_PKG_REPO = pg_probackup-forks + PBK_EDITION_FULL = Standard + PKG_NAME_SUFFIX = std- +else ifeq ($(PBK_EDITION),ent) + PBK_PKG_REPO = pg_probackup-forks + PBK_EDITION_FULL = Enterprise + PKG_NAME_SUFFIX = ent- +else + PBK_PKG_REPO = pg_probackup + PBK_EDITION_FULL = + PBK_EDITION = + PKG_NAME_SUFFIX = +endif + +check_env: + @if [ -z ${PBK_VERSION} ] ; then \ + echo "Env variable PBK_VERSION is not set" ; \ + false ; \ + fi + + @if [ -z ${PBK_RELEASE} ] ; then \ + echo "Env variable PBK_RELEASE is not set" ; \ + false ; \ + fi + + @if [ -z ${PBK_HASH} ] ; then \ + echo "Env variable PBK_HASH is not set" ; \ + false ; \ + fi + +pkg: check_env build/prepare build/all + @echo Build for all platform: done + +build/prepare: + mkdir -p build + +build/clean: build/prepare + find $(BUILDDIR) -maxdepth 1 -type f -exec rm -f {} \; + +build/all: build/debian build/ubuntu build/centos build/oraclelinux build/alt build/suse build/rhel + @echo Packaging is done + +### DEBIAN +build/debian: build/debian_9 build/debian_10 build/debian_11 + @echo Debian: done + +build/debian_9: build/debian_9_9.6 build/debian_9_10 build/debian_9_11 build/debian_9_12 build/debian_9_13 build/debian_9_14 + @echo Debian 9: done + +build/debian_10: build/debian_10_9.6 build/debian_10_10 build/debian_10_11 build/debian_10_12 build/debian_10_13 build/debian_10_14 + @echo Debian 10: done + +build/debian_11: build/debian_11_9.6 build/debian_11_10 build/debian_11_11 build/debian_11_12 build/debian_11_13 build/debian_11_14 + @echo Debian 11: done + +### UBUNTU +build/ubuntu: build/ubuntu_18.04 build/ubuntu_20.04 + @echo Ubuntu: done + +build/ubuntu_18.04: build/ubuntu_18.04_9.6 build/ubuntu_18.04_10 build/ubuntu_18.04_11 build/ubuntu_18.04_12 build/ubuntu_18.04_13 build/ubuntu_18.04_14 + @echo Ubuntu 18.04: done + +build/ubuntu_20.04: build/ubuntu_20.04_9.6 build/ubuntu_20.04_10 build/ubuntu_20.04_11 build/ubuntu_20.04_12 build/ubuntu_20.04_13 build/ubuntu_20.04_14 + @echo Ubuntu 20.04: done + +define build_deb + docker rm -f $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION) >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/pkg:/app/in \ + -v $(WORKDIR)/build/data/$(PBK_PKG_REPO)/$1/$2/pg-probackup-$(PKG_NAME_SUFFIX)$4/$(PBK_VERSION):/app/out \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" -e "PG_VERSION=$4" -e "PG_FULL_VERSION=$5" \ + -e "PKG_HASH=$(PBK_HASH)" -e "PKG_URL=$(PBK_GIT_REPO)" -e "PKG_RELEASE=$(PBK_RELEASE)" -e "PKG_NAME=pg-probackup-$(PKG_NAME_SUFFIX)$4" \ + -e "PKG_VERSION=$(PBK_VERSION)" -e "PBK_EDITION=$(PBK_EDITION)" -e "PBK_EDITION_FULL=$(PBK_EDITION_FULL)" \ + --name $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION)_pg_$5 \ + --rm pgpro/$1:$2 /app/in/scripts/deb.sh +endef + +include packaging/pkg/Makefile.debian +include packaging/pkg/Makefile.ubuntu + +# CENTOS +build/centos: build/centos_7 build/centos_8 #build/rpm_repo_package_centos + @echo Centos: done + +build/centos_7: build/centos_7_9.6 build/centos_7_10 build/centos_7_11 build/centos_7_12 build/centos_7_13 build/centos_7_14 + @echo Centos 7: done + +# pgpro-9.6@centos-8 doesn't exist +build/centos_8: build/centos_8_10 build/centos_8_11 build/centos_8_12 build/centos_8_13 build/centos_8_14 #build/centos_8_9.6 + @echo Centos 8: done + +# Oracle Linux +build/oraclelinux: build/oraclelinux_6 build/oraclelinux_7 build/oraclelinux_8 #build/rpm_repo_package_oraclelinux + @echo Oraclelinux: done + +build/oraclelinux_6: build/oraclelinux_6_9.6 build/oraclelinux_6_10 build/oraclelinux_6_11 build/oraclelinux_6_12 build/oraclelinux_6_13 build/oraclelinux_6_14 + @echo Oraclelinux 6: done + +build/oraclelinux_7: build/oraclelinux_7_9.6 build/oraclelinux_7_10 build/oraclelinux_7_11 build/oraclelinux_7_12 build/oraclelinux_7_13 build/oraclelinux_7_14 + @echo Oraclelinux 7: done + +# pgpro-9.6@oraclelinux-8 doesn't exist +build/oraclelinux_8: build/oraclelinux_8_10 build/oraclelinux_8_11 build/oraclelinux_8_12 build/oraclelinux_8_13 build/oraclelinux_8_14 #build/oraclelinux_8_9.6 + @echo Oraclelinux 8: done + +# RHEL +build/rhel: build/rhel_7 build/rhel_8 #build/rpm_repo_package_rhel + @echo Rhel: done + +build/rhel_7: build/rhel_7_9.6 build/rhel_7_10 build/rhel_7_11 build/rhel_7_12 build/rhel_7_13 build/rhel_7_14 + @echo Rhel 7: done + +build/rhel_8: build/rhel_8_9.6 build/rhel_8_10 build/rhel_8_11 build/rhel_8_12 build/rhel_8_13 build/rhel_8_14 + @echo Rhel 8: done + + +define build_rpm + docker rm -f $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION) >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/pkg:/app/in \ + -v $(WORKDIR)/build/data/$(PBK_PKG_REPO)/$1/$2/pg_probackup-$(PKG_NAME_SUFFIX)$4/$(PBK_VERSION):/app/out \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" -e "PG_VERSION=$4" -e "PG_FULL_VERSION=$5" \ + -e "PKG_HASH=$(PBK_HASH)" -e "PKG_URL=$(PBK_GIT_REPO)" -e "PKG_RELEASE=$(PBK_RELEASE)" -e "PKG_NAME=pg_probackup-$(PKG_NAME_SUFFIX)$4" \ + -e "PKG_VERSION=$(PBK_VERSION)" -e "PBK_EDITION=$(PBK_EDITION)" -e "PBK_EDITION_FULL=$(PBK_EDITION_FULL)" \ + --name $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION)_pg_$5 \ + --rm pgpro/$1:$2 /app/in/scripts/rpm.sh +endef + +include packaging/pkg/Makefile.centos +include packaging/pkg/Makefile.rhel +include packaging/pkg/Makefile.oraclelinux + + +# Alt Linux +build/alt: build/alt_7 build/alt_8 build/alt_9 + @echo Alt Linux: done + +build/alt_7: build/alt_7_9.6 build/alt_7_10 build/alt_7_11 build/alt_7_12 build/alt_7_13 build/alt_7_14 + @echo Alt Linux 7: done + +build/alt_8: build/alt_8_9.6 build/alt_8_10 build/alt_8_11 build/alt_8_12 build/alt_8_13 build/alt_8_14 + @echo Alt Linux 8: done + +build/alt_9: build/alt_9_9.6 build/alt_9_10 build/alt_9_11 build/alt_9_12 build/alt_9_13 build/alt_9_14 + @echo Alt Linux 9: done + +define build_alt + docker rm -f $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION) >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/pkg:/app/in \ + -v $(WORKDIR)/build/data/$(PBK_PKG_REPO)/$1/$2/pg_probackup-$(PKG_NAME_SUFFIX)$4/$(PBK_VERSION):/app/out \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" -e "PG_VERSION=$4" -e "PG_FULL_VERSION=$5" \ + -e "PKG_HASH=$(PBK_HASH)" -e "PKG_URL=$(PBK_GIT_REPO)" -e "PKG_RELEASE=$(PBK_RELEASE)" -e "PKG_NAME=pg_probackup-$(PKG_NAME_SUFFIX)$4" \ + -e "PKG_VERSION=$(PBK_VERSION)" -e "PBK_EDITION=$(PBK_EDITION)" -e "PBK_EDITION_FULL=$(PBK_EDITION_FULL)" \ + --name $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION)_pg_$5 \ + --rm pgpro/$1:$2 /app/in/scripts/alt.sh +endef + +include packaging/pkg/Makefile.alt + +# SUSE Linux +build/suse: build/suse_15.1 build/suse_15.2 + @echo Suse: done + +# there is no PG-14 in suse-15.1 repositories (test fails) +build/suse_15.1: build/suse_15.1_9.6 build/suse_15.1_10 build/suse_15.1_11 build/suse_15.1_12 build/suse_15.1_13 + @echo Rhel 15.1: done + +build/suse_15.2: build/suse_15.2_9.6 build/suse_15.2_10 build/suse_15.2_11 build/suse_15.2_12 build/suse_15.2_13 build/suse_15.2_14 + @echo Rhel 15.1: done + +define build_suse + docker rm -f $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION) >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/pkg:/app/in \ + -v $(WORKDIR)/build/data/$(PBK_PKG_REPO)/$1/$2/pg_probackup-$(PKG_NAME_SUFFIX)$4/$(PBK_VERSION):/app/out \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" -e "PG_VERSION=$4" -e "PG_FULL_VERSION=$5" \ + -e "PKG_HASH=$(PBK_HASH)" -e "PKG_URL=$(PBK_GIT_REPO)" -e "PKG_RELEASE=$(PBK_RELEASE)" -e "PKG_NAME=pg_probackup-$(PKG_NAME_SUFFIX)$4" \ + -e "PKG_VERSION=$(PBK_VERSION)" -e "PBK_EDITION=$(PBK_EDITION)" -e "PBK_EDITION_FULL=$(PBK_EDITION_FULL)" \ + --name $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION)_pg_$5 \ + --rm pgpro/$1:$2 /app/in/scripts/suse.sh +endef + +include packaging/pkg/Makefile.suse diff --git a/packaging/Makefile.repo b/packaging/Makefile.repo new file mode 100644 index 000000000..10fb27137 --- /dev/null +++ b/packaging/Makefile.repo @@ -0,0 +1,156 @@ +#### REPO BUILD #### +repo: check_env repo/debian repo/ubuntu repo/centos repo/oraclelinux repo/rhel repo/alt repo/suse repo_finish + @echo Build repo for all platform: done + +# Debian +repo/debian: build/repo_debian_9 build/repo_debian_10 build/repo_debian_11 + @echo Build repo for debian platforms: done + +build/repo_debian_9: + $(call build_repo_deb,debian,9,stretch) + touch build/repo_debian_9 + +build/repo_debian_10: + $(call build_repo_deb,debian,10,buster) + touch build/repo_debian_10 + +build/repo_debian_11: + $(call build_repo_deb,debian,11,bullseye) + touch build/repo_debian_11 + +# Ubuntu +repo/ubuntu: build/repo_ubuntu_18.04 build/repo_ubuntu_20.04 + @echo Build repo for ubuntu platforms: done + +build/repo_ubuntu_18.04: + $(call build_repo_deb,ubuntu,18.04,bionic) + touch build/repo_ubuntu_18.04 + +build/repo_ubuntu_20.04: + $(call build_repo_deb,ubuntu,20.04,focal) + touch build/repo_ubuntu_20.04 + +# Centos +repo/centos: build/repo_centos_7 build/repo_centos_8 + @echo Build repo for centos platforms: done + +build/repo_centos_7: + $(call build_repo_rpm,centos,7,,) + touch build/repo_centos_7 + +build/repo_centos_8: + $(call build_repo_rpm,centos,8,,) + touch build/repo_centos_8 + +# Oraclelinux +repo/oraclelinux: build/repo_oraclelinux_6 build/repo_oraclelinux_7 build/repo_oraclelinux_8 + @echo Build repo for oraclelinux platforms: done + +build/repo_oraclelinux_6: + $(call build_repo_rpm,oraclelinux,6,6Server) + touch build/repo_oraclelinux_6 + +build/repo_oraclelinux_7: + $(call build_repo_rpm,oraclelinux,7,7Server) + touch build/repo_oraclelinux_7 + +build/repo_oraclelinux_8: + $(call build_repo_rpm,oraclelinux,8,,) + touch build/repo_oraclelinux_8 + +# RHEL +repo/rhel: build/repo_rhel_7 build/repo_rhel_8 + @echo Build repo for rhel platforms: done + +build/repo_rhel_7: + $(call build_repo_rpm,rhel,7,7Server) + touch build/repo_rhel_7 + +build/repo_rhel_8: + $(call build_repo_rpm,rhel,8,,) + touch build/repo_rhel_8 + +# ALT +repo/alt: build/repo_alt_7 build/repo_alt_8 build/repo_alt_9 + @echo Build repo for alt platforms: done + +build/repo_alt_7: + $(call build_repo_alt,alt,7,,) + touch build/repo_alt_7 + +build/repo_alt_8: + $(call build_repo_alt,alt,8,,) + touch build/repo_alt_8 + +build/repo_alt_9: + $(call build_repo_alt,alt,9,,) + touch build/repo_alt_9 + +# SUSE +repo/suse: build/repo_suse_15.1 build/repo_suse_15.2 + @echo Build repo for suse platforms: done + +build/repo_suse_15.1: + $(call build_repo_suse,suse,15.1,,) + touch build/repo_suse_15.1 + +build/repo_suse_15.2: + $(call build_repo_suse,suse,15.2,,) + touch build/repo_suse_15.2 + +repo_finish: +# cd build/data/www/$(PBK_PKG_REPO)/ + cd $(BUILDDIR)/data/www/$(PBK_PKG_REPO)/rpm && sudo ln -nsf $(PBK_VERSION) latest + # following line only for vanilla + cd $(BUILDDIR)/data/www/$(PBK_PKG_REPO)/srpm && sudo ln -nsf $(PBK_VERSION) latest + +# sudo ln -rfs build/data/www/$(PBK_PKG_REPO)/rpm/${PBK_VERSION} build/data/www/$(PBK_PKG_REPO)/rpm/latest +# sudo ln -rfs build/data/www/$(PBK_PKG_REPO)/srpm/${PBK_VERSION} build/data/www/$(PBK_PKG_REPO)/srpm/latest + +define build_repo_deb + docker rm -f $1_$2_pbk_repo >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/repo:/app/repo \ + -v $(WORKDIR)/build/data/www:/app/www \ + -v $(WORKDIR)/build/data/$(PBK_PKG_REPO)/$1/$2:/app/in \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" \ + -e "PBK_PKG_REPO=$(PBK_PKG_REPO)" -e "PBK_EDITION=$(PBK_EDITION)" \ + --name $1_$2_pbk_repo \ + --rm pgpro/repo /app/repo/scripts/deb.sh +endef + +define build_repo_rpm + docker rm -f $1_$2_pbk_repo >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/repo:/app/repo \ + -v $(WORKDIR)/build/data/www:/app/www \ + -v $(WORKDIR)/build/data/$(PBK_PKG_REPO)/$1/$2:/app/in \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" \ + -e "PBK_PKG_REPO=$(PBK_PKG_REPO)" -e "PBK_EDITION=$(PBK_EDITION)" \ + --name $1_$2_pbk_repo \ + --rm pgpro/repo /app/repo/scripts/rpm.sh +endef + +define build_repo_alt + docker rm -f $1_$2_pbk_repo >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/repo:/app/repo \ + -v $(WORKDIR)/build/data/www:/app/www \ + -v $(WORKDIR)/build/data/$(PBK_PKG_REPO)/$1/$2:/app/in \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" \ + -e "PBK_PKG_REPO=$(PBK_PKG_REPO)" -e "PBK_EDITION=$(PBK_EDITION)" \ + --name $1_$2_pbk_repo \ + --rm pgpro/$1:$2 /app/repo/scripts/alt.sh +endef + +define build_repo_suse + docker rm -f $1_$2_pbk_repo >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/repo:/app/repo \ + -v $(WORKDIR)/build/data/www:/app/www \ + -v $(WORKDIR)/build/data/$(PBK_PKG_REPO)/$1/$2:/app/in \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" \ + -e "PBK_PKG_REPO=$(PBK_PKG_REPO)" -e "PBK_EDITION=$(PBK_EDITION)" \ + --name $1_$2_pbk_repo \ + --rm pgpro/$1:$2 /app/repo/scripts/suse.sh +endef diff --git a/packaging/Makefile.test b/packaging/Makefile.test new file mode 100644 index 000000000..11c63619a --- /dev/null +++ b/packaging/Makefile.test @@ -0,0 +1,150 @@ +ifeq ($(PBK_EDITION),std) + SCRIPT_SUFFIX = _forks +else ifeq ($(PBK_EDITION),ent) + SCRIPT_SUFFIX = _forks +else + SCRIPT_SUFFIX = +endif + +test: build/test_all + @echo Test for all platform: done + +build/test_all: build/test_debian build/test_ubuntu build/test_centos build/test_oraclelinux build/test_alt build/test_suse #build/test_rhel + @echo Package testing is done + +### DEBIAN +build/test_debian: build/test_debian_9 build/test_debian_10 build/test_debian_11 + @echo Debian: done + +build/test_debian_9: build/test_debian_9_9.6 build/test_debian_9_10 build/test_debian_9_11 build/test_debian_9_12 build/test_debian_9_13 build/test_debian_9_14 + @echo Debian 9: done + +build/test_debian_10: build/test_debian_10_9.6 build/test_debian_10_10 build/test_debian_10_11 build/test_debian_10_12 build/test_debian_10_13 build/test_debian_10_14 + @echo Debian 10: done + +build/test_debian_11: build/test_debian_11_9.6 build/test_debian_11_10 build/test_debian_11_11 build/test_debian_11_12 build/test_debian_11_13 build/test_debian_11_14 + @echo Debian 11: done + +### UBUNTU +build/test_ubuntu: build/test_ubuntu_18.04 build/test_ubuntu_20.04 + @echo Ubuntu: done + +build/test_ubuntu_18.04: build/test_ubuntu_18.04_9.6 build/test_ubuntu_18.04_10 build/test_ubuntu_18.04_11 build/test_ubuntu_18.04_12 build/test_ubuntu_18.04_13 build/test_ubuntu_18.04_14 + @echo Ubuntu 18.04: done + +build/test_ubuntu_20.04: build/test_ubuntu_20.04_9.6 build/test_ubuntu_20.04_10 build/test_ubuntu_20.04_11 build/test_ubuntu_20.04_12 build/test_ubuntu_20.04_13 build/test_ubuntu_20.04_14 + @echo Ubuntu 20.04: done + +define test_deb + docker rm -f $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION) >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/test:/app/in \ + -v $(BUILDDIR)/data/www:/app/www \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" -e "PG_VERSION=$4" -e "PG_FULL_VERSION=$5" \ + -e "PKG_HASH=$(PBK_HASH)" -e "PKG_URL=$(PBK_GIT_REPO)" -e "PKG_RELEASE=$(PBK_RELEASE)" -e "PKG_NAME=pg-probackup-$(PKG_NAME_SUFFIX)$4" \ + -e "PKG_VERSION=$(PBK_VERSION)" -e "PBK_EDITION=$(PBK_EDITION)" -e "PBK_EDITION_FULL=$(PBK_EDITION_FULL)" \ + --name $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION)_pg_$5 \ + --rm pgpro/$1:$2 /app/in/scripts/deb$(SCRIPT_SUFFIX).sh +endef + +include packaging/test/Makefile.debian +include packaging/test/Makefile.ubuntu + +# CENTOS +build/test_centos: build/test_centos_7 build/test_centos_8 + @echo Centos: done + +build/test_centos_7: build/test_centos_7_9.6 build/test_centos_7_10 build/test_centos_7_11 build/test_centos_7_12 build/test_centos_7_13 #build/test_centos_7_14 + @echo Centos 7: done + +# pgpro-9.6@centos-8 doesn't exist +build/test_centos_8: build/test_centos_8_10 build/test_centos_8_11 build/test_centos_8_12 build/test_centos_8_13 #build/test_centos_8_14 build/test_centos_8_9.6 + @echo Centos 8: done + +# Oracle Linux +build/test_oraclelinux: build/test_oraclelinux_7 build/test_oraclelinux_8 + @echo Oraclelinux: done + +build/test_oraclelinux_7: build/test_oraclelinux_7_9.6 build/test_oraclelinux_7_10 build/test_oraclelinux_7_11 build/test_oraclelinux_7_12 build/test_oraclelinux_7_13 #build/test_oraclelinux_7_14 + @echo Oraclelinux 7: done + +# pgpro-9.6@oraclelinux-8 doesn't exist +build/test_oraclelinux_8: build/test_oraclelinux_8_10 build/test_oraclelinux_8_11 build/test_oraclelinux_8_12 build/test_oraclelinux_8_13 #build/test_oraclelinux_8_14 build/test_oraclelinux_8_9.6 + @echo Oraclelinux 8: done + +# RHEL +build/test_rhel: build/test_rhel_7 #build/test_rhel_8 + @echo Rhel: done + +build/test_rhel_7: build/test_rhel_7_9.6 build/test_rhel_7_10 build/test_rhel_7_11 build/test_rhel_7_12 build/test_rhel_7_13 #build/test_rhel_7_14 + @echo Rhel 7: done + +build/test_rhel_8: build/test_rhel_8_9.6 build/test_rhel_8_10 build/test_rhel_8_11 build/test_rhel_8_12 build/test_rhel_8_13 build/test_rhel_8_14 + @echo Rhel 8: done + +define test_rpm + docker rm -f $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION) >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/test:/app/in \ + -v $(BUILDDIR)/data/www:/app/www \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" -e "PG_VERSION=$4" -e "PG_FULL_VERSION=$5" \ + -e "PKG_HASH=$(PBK_HASH)" -e "PKG_URL=$(PBK_GIT_REPO)" -e "PKG_RELEASE=$(PBK_RELEASE)" -e "PKG_NAME=pg_probackup-$(PKG_NAME_SUFFIX)$4" \ + -e "PKG_VERSION=$(PBK_VERSION)" -e "PBK_EDITION=$(PBK_EDITION)" -e "PBK_EDITION_FULL=$(PBK_EDITION_FULL)" \ + --name $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION)_pg_$5 \ + --rm pgpro/$1:$2 /app/in/scripts/rpm$(SCRIPT_SUFFIX).sh +endef + +include packaging/test/Makefile.centos +include packaging/test/Makefile.rhel +include packaging/test/Makefile.oraclelinux + +# Alt Linux +build/test_alt: build/test_alt_8 build/test_alt_9 + @echo Alt Linux: done + +# nginx@alt7 fall with 'nginx: [alert] sysctl(KERN_RTSIGMAX) failed (1: Operation not permitted)' +# within docker on modern host linux kernels (this nginx build require Linux between 2.2.19 and 2.6.17) + +build/test_alt_8: build/test_alt_8_9.6 build/test_alt_8_10 build/test_alt_8_11 build/test_alt_8_12 build/test_alt_8_13 build/test_alt_8_14 + @echo Alt Linux 8: done + +build/test_alt_9: build/test_alt_9_9.6 build/test_alt_9_10 build/test_alt_9_11 build/test_alt_9_12 build/test_alt_9_13 build/test_alt_9_14 + @echo Alt Linux 9: done + +define test_alt + docker rm -f $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION) >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/test:/app/in \ + -v $(BUILDDIR)/data/www:/app/www \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" -e "PG_VERSION=$4" -e "PG_FULL_VERSION=$5" \ + -e "PKG_HASH=$(PBK_HASH)" -e "PKG_URL=$(PBK_GIT_REPO)" -e "PKG_RELEASE=$(PBK_RELEASE)" -e "PKG_NAME=pg_probackup-$(PKG_NAME_SUFFIX)$4" \ + -e "PKG_VERSION=$(PBK_VERSION)" -e "PBK_EDITION=$(PBK_EDITION)" -e "PBK_EDITION_FULL=$(PBK_EDITION_FULL)" \ + --name $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION)_pg_$5 \ + --rm pgpro/$1:$2 /app/in/scripts/alt$(SCRIPT_SUFFIX).sh +endef + +include packaging/test/Makefile.alt + +# SUSE Linux +build/test_suse: build/test_suse_15.1 build/test_suse_15.2 + @echo Suse: done + +build/test_suse_15.1: build/test_suse_15.1_9.6 build/test_suse_15.1_10 build/test_suse_15.1_11 build/test_suse_15.1_12 build/test_suse_15.1_13 + @echo Suse 15.1: done + +build/test_suse_15.2: build/test_suse_15.2_9.6 build/test_suse_15.2_10 build/test_suse_15.2_11 build/test_suse_15.2_12 build/test_suse_15.2_13 build/test_suse_15.2_14 + @echo Suse 15.2: done + +define test_suse + docker rm -f $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION) >> /dev/null 2>&1 ; \ + docker run \ + -v $(WORKDIR)/packaging/test:/app/in \ + -v $(BUILDDIR)/data/www:/app/www \ + -e "DISTRIB=$1" -e "DISTRIB_VERSION=$2" -e "CODENAME=$3" -e "PG_VERSION=$4" -e "PG_FULL_VERSION=$5" \ + -e "PKG_HASH=$(PBK_HASH)" -e "PKG_URL=$(PBK_GIT_REPO)" -e "PKG_RELEASE=$(PBK_RELEASE)" -e "PKG_NAME=pg_probackup-$(PKG_NAME_SUFFIX)$4" \ + -e "PKG_VERSION=$(PBK_VERSION)" -e "PBK_EDITION=$(PBK_EDITION)" -e "PBK_EDITION_FULL=$(PBK_EDITION_FULL)" \ + --name $1_$2_probackup_$(PKG_NAME_SUFFIX)$(PBK_VERSION)_pg_$5 \ + --rm pgpro/$1:$2 /app/in/scripts/suse$(SCRIPT_SUFFIX).sh +endef + +include packaging/test/Makefile.suse diff --git a/packaging/Readme.md b/packaging/Readme.md new file mode 100644 index 000000000..f4437d838 --- /dev/null +++ b/packaging/Readme.md @@ -0,0 +1,23 @@ +Example: +``` +export PBK_VERSION=2.4.17 +export PBK_HASH=57f871accce2604 +export PBK_RELEASE=1 +export PBK_EDITION=std|ent +make --keep-going pkg +``` + +To build binaries for PostgresPro Standard or Enterprise, a pgpro.tar.bz2 with latest git tree must be preset in `packaging/pkg/tarballs` directory: +``` +cd packaging/pkg/tarballs +git clone pgpro_repo pgpro +tar -cjSf pgpro.tar.bz2 pgpro +``` + +To build repo the gpg keys for package signing must be present ... +Repo must be build using 1 thread (due to debian bullshit): +``` +make repo -j1 +``` + + diff --git a/packaging/pkg/Makefile.alt b/packaging/pkg/Makefile.alt new file mode 100644 index 000000000..28eabf53f --- /dev/null +++ b/packaging/pkg/Makefile.alt @@ -0,0 +1,87 @@ +# ALT 7 +build/alt_7_9.5: + $(call build_alt,alt,7,,9.5,9.5.25) + touch build/alt_7_9.5 + +build/alt_7_9.6: + $(call build_alt,alt,7,,9.6,9.6.24) + touch build/alt_7_9.6 + +build/alt_7_10: + $(call build_alt,alt,7,,10,10.19) + touch build/alt_7_10 + +build/alt_7_11: + $(call build_alt,alt,7,,11,11.14) + touch build/alt_7_11 + +build/alt_7_12: + $(call build_alt,alt,7,,12,12.9) + touch build/alt_7_12 + +build/alt_7_13: + $(call build_alt,alt,7,,13,13.5) + touch build/alt_7_13 + +build/alt_7_14: + $(call build_alt,alt,7,,14,14.1) + touch build/alt_7_14 + +# ALT 8 +build/alt_8_9.5: + $(call build_alt,alt,8,,9.5,9.5.25) + touch build/alt_8_9.5 + +build/alt_8_9.6: + $(call build_alt,alt,8,,9.6,9.6.24) + touch build/alt_8_9.6 + +build/alt_8_10: + $(call build_alt,alt,8,,10,10.19) + touch build/alt_8_10 + +build/alt_8_11: + $(call build_alt,alt,8,,11,11.14) + touch build/alt_8_11 + +build/alt_8_12: + $(call build_alt,alt,8,,12,12.9) + touch build/alt_8_12 + +build/alt_8_13: + $(call build_alt,alt,8,,13,13.5) + touch build/alt_8_13 + +build/alt_8_14: + $(call build_alt,alt,8,,14,14.1) + touch build/alt_8_14 + +# ALT 9 +build/alt_9_9.5: + $(call build_alt,alt,9,,9.5,9.5.25) + touch build/alt_9_9.5 + +build/alt_9_9.6: + $(call build_alt,alt,9,,9.6,9.6.24) + touch build/alt_9_9.6 + +build/alt_9_10: + $(call build_alt,alt,9,,10,10.19) + touch build/alt_9_10 + +build/alt_9_11: + $(call build_alt,alt,9,,11,11.14) + touch build/alt_9_11 + +build/alt_9_12: + $(call build_alt,alt,9,,12,12.9) + touch build/alt_9_12 + +build/alt_9_13: + $(call build_alt,alt,9,,13,13.5) + touch build/alt_9_13 + +build/alt_9_14: + $(call build_alt,alt,9,,14,14.1) + touch build/alt_9_14 + diff --git a/packaging/pkg/Makefile.centos b/packaging/pkg/Makefile.centos new file mode 100644 index 000000000..fb537d0a6 --- /dev/null +++ b/packaging/pkg/Makefile.centos @@ -0,0 +1,57 @@ +# CENTOS 7 +build/centos_7_9.5: + $(call build_rpm,centos,7,,9.5,9.5.25) + touch build/centos_7_9.5 + +build/centos_7_9.6: + $(call build_rpm,centos,7,,9.6,9.6.24) + touch build/centos_7_9.6 + +build/centos_7_10: + $(call build_rpm,centos,7,,10,10.19) + touch build/centos_7_10 + +build/centos_7_11: + $(call build_rpm,centos,7,,11,11.14) + touch build/centos_7_11 + +build/centos_7_12: + $(call build_rpm,centos,7,,12,12.9) + touch build/centos_7_12 + +build/centos_7_13: + $(call build_rpm,centos,7,,13,13.5) + touch build/centos_7_13 + +build/centos_7_14: + $(call build_rpm,centos,7,,14,14.1) + touch build/centos_7_14 + +# CENTOS 8 +build/centos_8_9.5: + $(call build_rpm,centos,8,,9.5,9.5.25) + touch build/centos_8_9.5 + +build/centos_8_9.6: + $(call build_rpm,centos,8,,9.6,9.6.24) + touch build/centos_8_9.6 + +build/centos_8_10: + $(call build_rpm,centos,8,,10,10.19) + touch build/centos_8_10 + +build/centos_8_11: + $(call build_rpm,centos,8,,11,11.14) + touch build/centos_8_11 + +build/centos_8_12: + $(call build_rpm,centos,8,,12,12.9) + touch build/centos_8_12 + +build/centos_8_13: + $(call build_rpm,centos,8,,13,13.5) + touch build/centos_8_13 + +build/centos_8_14: + $(call build_rpm,centos,8,,14,14.1) + touch build/centos_8_14 diff --git a/packaging/pkg/Makefile.debian b/packaging/pkg/Makefile.debian new file mode 100644 index 000000000..d9c885d3a --- /dev/null +++ b/packaging/pkg/Makefile.debian @@ -0,0 +1,86 @@ +# DEBIAN 9 +build/debian_9_9.5: + $(call build_deb,debian,9,stretch,9.5,9.5.25) + touch build/debian_9_9.5 + +build/debian_9_9.6: + $(call build_deb,debian,9,stretch,9.6,9.6.24) + touch build/debian_9_9.6 + +build/debian_9_10: + $(call build_deb,debian,9,stretch,10,10.19) + touch build/debian_9_10 + +build/debian_9_11: + $(call build_deb,debian,9,stretch,11,11.14) + touch build/debian_9_11 + +build/debian_9_12: + $(call build_deb,debian,9,stretch,12,12.9) + touch build/debian_9_12 + +build/debian_9_13: + $(call build_deb,debian,9,stretch,13,13.5) + touch build/debian_9_13 + +build/debian_9_14: + $(call build_deb,debian,9,stretch,14,14.1) + touch build/debian_9_14 + +# DEBIAN 10 +build/debian_10_9.5: + $(call build_deb,debian,10,buster,9.5,9.5.25) + touch build/debian_10_9.5 + +build/debian_10_9.6: + $(call build_deb,debian,10,buster,9.6,9.6.24) + touch build/debian_10_9.6 + +build/debian_10_10: + $(call build_deb,debian,10,buster,10,10.19) + touch build/debian_10_10 + +build/debian_10_11: + $(call build_deb,debian,10,buster,11,11.14) + touch build/debian_10_11 + +build/debian_10_12: + $(call build_deb,debian,10,buster,12,12.9) + touch build/debian_10_12 + +build/debian_10_13: + $(call build_deb,debian,10,buster,13,13.5) + touch build/debian_10_13 + +build/debian_10_14: + $(call build_deb,debian,10,buster,14,14.1) + touch build/debian_10_14 + +# DEBIAN 11 +build/debian_11_9.5: + $(call build_deb,debian,11,bullseye,9.5,9.5.25) + touch build/debian_11_9.5 + +build/debian_11_9.6: + $(call build_deb,debian,11,bullseye,9.6,9.6.24) + touch build/debian_11_9.6 + +build/debian_11_10: + $(call build_deb,debian,11,bullseye,10,10.19) + touch build/debian_11_10 + +build/debian_11_11: + $(call build_deb,debian,11,bullseye,11,11.14) + touch build/debian_11_11 + +build/debian_11_12: + $(call build_deb,debian,11,bullseye,12,12.9) + touch build/debian_11_12 + +build/debian_11_13: + $(call build_deb,debian,11,bullseye,13,13.5) + touch build/debian_11_13 + +build/debian_11_14: + $(call build_deb,debian,11,bullseye,14,14.1) + touch build/debian_11_14 diff --git a/packaging/pkg/Makefile.oraclelinux b/packaging/pkg/Makefile.oraclelinux new file mode 100644 index 000000000..127a578f1 --- /dev/null +++ b/packaging/pkg/Makefile.oraclelinux @@ -0,0 +1,87 @@ +# ORACLE LINUX 6 +build/oraclelinux_6_9.5: + $(call build_rpm,oraclelinux,6,,9.5,9.5.25) + touch build/oraclelinux_6_9.5 + +build/oraclelinux_6_9.6: + $(call build_rpm,oraclelinux,6,,9.6,9.6.24) + touch build/oraclelinux_6_9.6 + +build/oraclelinux_6_10: + $(call build_rpm,oraclelinux,6,,10,10.19) + touch build/oraclelinux_6_10 + +build/oraclelinux_6_11: + $(call build_rpm,oraclelinux,6,,11,11.14) + touch build/oraclelinux_6_11 + +build/oraclelinux_6_12: + $(call build_rpm,oraclelinux,6,,12,12.9) + touch build/oraclelinux_6_12 + +build/oraclelinux_6_13: + $(call build_rpm,oraclelinux,6,,13,13.5) + touch build/oraclelinux_6_13 + +build/oraclelinux_6_14: + $(call build_rpm,oraclelinux,6,,14,14.1) + touch build/oraclelinux_6_14 + +# ORACLE LINUX 7 +build/oraclelinux_7_9.5: + $(call build_rpm,oraclelinux,7,,9.5,9.5.25) + touch build/oraclelinux_7_9.5 + +build/oraclelinux_7_9.6: + $(call build_rpm,oraclelinux,7,,9.6,9.6.24) + touch build/oraclelinux_7_9.6 + +build/oraclelinux_7_10: + $(call build_rpm,oraclelinux,7,,10,10.19) + touch build/oraclelinux_7_10 + +build/oraclelinux_7_11: + $(call build_rpm,oraclelinux,7,,11,11.14) + touch build/oraclelinux_7_11 + +build/oraclelinux_7_12: + $(call build_rpm,oraclelinux,7,,12,12.9) + touch build/oraclelinux_7_12 + +build/oraclelinux_7_13: + $(call build_rpm,oraclelinux,7,,13,13.5) + touch build/oraclelinux_7_13 + +build/oraclelinux_7_14: + $(call build_rpm,oraclelinux,7,,14,14.1) + touch build/oraclelinux_7_14 + +# ORACLE LINUX 8 +build/oraclelinux_8_9.5: + $(call build_rpm,oraclelinux,8,,9.5,9.5.25) + touch build/oraclelinux_8_9.5 + +build/oraclelinux_8_9.6: + $(call build_rpm,oraclelinux,8,,9.6,9.6.24) + touch build/oraclelinux_8_9.6 + +build/oraclelinux_8_10: + $(call build_rpm,oraclelinux,8,,10,10.19) + touch build/oraclelinux_8_10 + +build/oraclelinux_8_11: + $(call build_rpm,oraclelinux,8,,11,11.14) + touch build/oraclelinux_8_11 + +build/oraclelinux_8_12: + $(call build_rpm,oraclelinux,8,,12,12.9) + touch build/oraclelinux_8_12 + +build/oraclelinux_8_13: + $(call build_rpm,oraclelinux,8,,13,13.5) + touch build/oraclelinux_8_13 + +build/oraclelinux_8_14: + $(call build_rpm,oraclelinux,8,,14,14.1) + touch build/oraclelinux_8_14 + diff --git a/packaging/pkg/Makefile.rhel b/packaging/pkg/Makefile.rhel new file mode 100644 index 000000000..8c1b0687b --- /dev/null +++ b/packaging/pkg/Makefile.rhel @@ -0,0 +1,57 @@ +# RHEL 7 +build/rhel_7_9.5: + $(call build_rpm,rhel,7,7Server,9.5,9.5.25) + touch build/rhel_7_9.5 + +build/rhel_7_9.6: + $(call build_rpm,rhel,7,7Server,9.6,9.6.24) + touch build/rhel_7_9.6 + +build/rhel_7_10: + $(call build_rpm,rhel,7,7Server,10,10.19) + touch build/rhel_7_10 + +build/rhel_7_11: + $(call build_rpm,rhel,7,7Server,11,11.14) + touch build/rhel_7_11 + +build/rhel_7_12: + $(call build_rpm,rhel,7,7Server,12,12.9) + touch build/rhel_7_12 + +build/rhel_7_13: + $(call build_rpm,rhel,7,7Server,13,13.5) + touch build/rhel_7_13 + +build/rhel_7_14: + $(call build_rpm,rhel,7,7Server,14,14.1) + touch build/rhel_7_14 + +# RHEL 8 +build/rhel_8_9.5: + $(call build_rpm,rhel,8,8Server,9.5,9.5.25) + touch build/rhel_8_9.5 + +build/rhel_8_9.6: + $(call build_rpm,rhel,8,8Server,9.6,9.6.24) + touch build/rhel_8_9.6 + +build/rhel_8_10: + $(call build_rpm,rhel,8,8Server,10,10.19) + touch build/rhel_8_10 + +build/rhel_8_11: + $(call build_rpm,rhel,8,8Server,11,11.14) + touch build/rhel_8_11 + +build/rhel_8_12: + $(call build_rpm,rhel,8,8Server,12,12.9) + touch build/rhel_8_12 + +build/rhel_8_13: + $(call build_rpm,rhel,8,8Server,13,13.5) + touch build/rhel_8_13 + +build/rhel_8_14: + $(call build_rpm,rhel,8,8Server,14,14.1) + touch build/rhel_8_14 diff --git a/packaging/pkg/Makefile.suse b/packaging/pkg/Makefile.suse new file mode 100644 index 000000000..c71ebd389 --- /dev/null +++ b/packaging/pkg/Makefile.suse @@ -0,0 +1,57 @@ +# Suse 15.1 +build/suse_15.1_9.5: + $(call build_suse,suse,15.1,,9.5,9.5.25) + touch build/suse_15.1_9.5 + +build/suse_15.1_9.6: + $(call build_suse,suse,15.1,,9.6,9.6.24) + touch build/suse_15.1_9.6 + +build/suse_15.1_10: + $(call build_suse,suse,15.1,,10,10.19) + touch build/suse_15.1_10 + +build/suse_15.1_11: + $(call build_suse,suse,15.1,,11,11.14) + touch build/suse_15.1_11 + +build/suse_15.1_12: + $(call build_suse,suse,15.1,,12,12.9) + touch build/suse_15.1_12 + +build/suse_15.1_13: + $(call build_suse,suse,15.1,,13,13.5) + touch build/suse_15.1_13 + +build/suse_15.1_14: + $(call build_suse,suse,15.1,,14,14.1) + touch build/suse_15.1_14 + +# Suse 15.2 +build/suse_15.2_9.5: + $(call build_suse,suse,15.2,,9.5,9.5.25) + touch build/suse_15.2_9.5 + +build/suse_15.2_9.6: + $(call build_suse,suse,15.2,,9.6,9.6.24) + touch build/suse_15.2_9.6 + +build/suse_15.2_10: + $(call build_suse,suse,15.2,,10,10.19) + touch build/suse_15.2_10 + +build/suse_15.2_11: + $(call build_suse,suse,15.2,,11,11.14) + touch build/suse_15.2_11 + +build/suse_15.2_12: + $(call build_suse,suse,15.2,,12,12.9) + touch build/suse_15.2_12 + +build/suse_15.2_13: + $(call build_suse,suse,15.2,,13,13.5) + touch build/suse_15.2_13 + +build/suse_15.2_14: + $(call build_suse,suse,15.2,,14,14.1) + touch build/suse_15.2_14 diff --git a/packaging/pkg/Makefile.ubuntu b/packaging/pkg/Makefile.ubuntu new file mode 100644 index 000000000..02acd6c67 --- /dev/null +++ b/packaging/pkg/Makefile.ubuntu @@ -0,0 +1,58 @@ +# UBUNTU 20.04 +build/ubuntu_20.04_9.5: + $(call build_deb,ubuntu,20.04,focal,9.5,9.5.25) + touch build/ubuntu_20.04_9.5 + +build/ubuntu_20.04_9.6: + $(call build_deb,ubuntu,20.04,focal,9.6,9.6.24) + touch build/ubuntu_20.04_9.6 + +build/ubuntu_20.04_10: + $(call build_deb,ubuntu,20.04,focal,10,10.19) + touch build/ubuntu_20.04_10 + +build/ubuntu_20.04_11: + $(call build_deb,ubuntu,20.04,focal,11,11.14) + touch build/ubuntu_20.04_11 + +build/ubuntu_20.04_12: + $(call build_deb,ubuntu,20.04,focal,12,12.9) + touch build/ubuntu_20.04_12 + +build/ubuntu_20.04_13: + $(call build_deb,ubuntu,20.04,focal,13,13.5) + touch build/ubuntu_20.04_13 + +build/ubuntu_20.04_14: + $(call build_deb,ubuntu,20.04,focal,14,14.1) + touch build/ubuntu_20.04_14 + +# UBUNTU 18.04 +build/ubuntu_18.04_9.5: + $(call build_deb,ubuntu,18.04,bionic,9.5,9.5.25) + touch build/ubuntu_18.04_9.5 + +build/ubuntu_18.04_9.6: + $(call build_deb,ubuntu,18.04,bionic,9.6,9.6.24) + touch build/ubuntu_18.04_9.6 + +build/ubuntu_18.04_10: + $(call build_deb,ubuntu,18.04,bionic,10,10.19) + touch build/ubuntu_18.04_10 + +build/ubuntu_18.04_11: + $(call build_deb,ubuntu,18.04,bionic,11,11.14) + touch build/ubuntu_18.04_11 + +build/ubuntu_18.04_12: + $(call build_deb,ubuntu,18.04,bionic,12,12.9) + touch build/ubuntu_18.04_12 + +build/ubuntu_18.04_13: + $(call build_deb,ubuntu,18.04,bionic,13,13.5) + touch build/ubuntu_18.04_13 + +build/ubuntu_18.04_14: + $(call build_deb,ubuntu,18.04,bionic,14,14.1) + touch build/ubuntu_18.04_14 + diff --git a/packaging/pkg/scripts/alt.sh b/packaging/pkg/scripts/alt.sh new file mode 100755 index 000000000..7c3971d6a --- /dev/null +++ b/packaging/pkg/scripts/alt.sh @@ -0,0 +1,122 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -xe +set -o pipefail + +# THere is no std/ent packages for PG 9.5 +if [[ ${PG_VERSION} == '9.5' ]] && [[ ${PBK_EDITION} != '' ]] ; then + exit 0 +fi + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 +apt-get update -y + +mkdir /root/build +cd /root/build + +# Copy rpmbuild +cp -rv /app/in/specs/rpm/rpmbuild /root/ + +# download pbk +git clone $PKG_URL pg_probackup-${PKG_VERSION} +cd pg_probackup-${PKG_VERSION} +git checkout ${PKG_HASH} +cd .. + +# tarball it +if [[ ${PBK_EDITION} == '' ]] ; then + tar -cjf pg_probackup-${PKG_VERSION}.tar.bz2 pg_probackup-${PKG_VERSION} + mv pg_probackup-${PKG_VERSION}.tar.bz2 /root/rpmbuild/SOURCES + rm -rf pg_probackup-${PKG_VERSION} +else + mv pg_probackup-${PKG_VERSION} /root/rpmbuild/SOURCES +fi + + +if [[ ${PBK_EDITION} == '' ]] ; then + # Download PostgreSQL source + wget -q http://ftp.postgresql.org/pub/source/v${PG_FULL_VERSION}/postgresql-${PG_FULL_VERSION}.tar.bz2 -O postgresql-${PG_VERSION}.tar.bz2 + mv postgresql-${PG_VERSION}.tar.bz2 /root/rpmbuild/SOURCES/ + +else + tar -xf /app/in/tarballs/pgpro.tar.bz2 -C /root/rpmbuild/SOURCES/ + cd /root/rpmbuild/SOURCES/pgpro + + PGPRO_TOC=$(echo ${PG_FULL_VERSION} | sed 's|\.|_|g') + if [[ ${PBK_EDITION} == 'std' ]] ; then + git checkout "PGPRO${PGPRO_TOC}_1" + else + git checkout "PGPROEE${PGPRO_TOC}_1" + fi + rm -rf .git + + cd /root/rpmbuild/SOURCES/ + mv pgpro postgrespro-${PBK_EDITION}-${PG_FULL_VERSION} + chown -R root:root postgrespro-${PBK_EDITION}-${PG_FULL_VERSION} +fi + + +#cd /root/rpmbuild/SOURCES +#sed -i "s/@PG_VERSION@/${PKG_VERSION}/" pg_probackup.repo + +# build postgresql +echo '%_allow_root_build yes' > /root/.rpmmacros +echo '%_topdir %{getenv:HOME}/rpmbuild' >> /root/.rpmmacros + +cd /root/rpmbuild/SPECS +if [[ ${PBK_EDITION} == '' ]] ; then + sed -i "s/@PKG_VERSION@/${PKG_VERSION}/" pg_probackup.alt.spec + sed -i "s/@PKG_RELEASE@/${PKG_RELEASE}/" pg_probackup.alt.spec + sed -i "s/@PKG_HASH@/${PKG_HASH}/" pg_probackup.alt.spec + sed -i "s/@PG_VERSION@/${PG_VERSION}/" pg_probackup.alt.spec + sed -i "s/@PG_FULL_VERSION@/${PG_FULL_VERSION}/" pg_probackup.alt.spec +else + sed -i "s/@EDITION@/${PBK_EDITION}/" pg_probackup.alt.forks.spec + sed -i "s/@EDITION_FULL@/${PBK_EDITION_FULL}/" pg_probackup.alt.forks.spec + sed -i "s/@PKG_VERSION@/${PKG_VERSION}/" pg_probackup.alt.forks.spec + sed -i "s/@PKG_RELEASE@/${PKG_RELEASE}/" pg_probackup.alt.forks.spec + sed -i "s/@PKG_HASH@/${PKG_HASH}/" pg_probackup.alt.forks.spec + sed -i "s/@PG_VERSION@/${PG_VERSION}/" pg_probackup.alt.forks.spec + sed -i "s/@PG_FULL_VERSION@/${PG_FULL_VERSION}/" pg_probackup.alt.forks.spec + + if [ ${PG_VERSION} != '9.6' ]; then + sed -i "s|@PREFIX@|/opt/pgpro/${EDITION}-${PG_VERSION}|g" pg_probackup.alt.forks.spec + fi +fi + +# ALT Linux suck as detecting dependecies, so the manual hint is required +if [ ${DISTRIB_VERSION} == '7' ]; then + apt-get install libpq5.10 + +elif [ ${DISTRIB_VERSION} == '8' ]; then + apt-get install libpq5.12 + +else + apt-get install libpq5 +fi + +# install dependencies +#stolen from postgrespro +apt-get install -y flex libldap-devel libpam-devel libreadline-devel libssl-devel + +if [[ ${PBK_EDITION} == '' ]] ; then + # build pg_probackup + rpmbuild -bs pg_probackup.alt.spec + rpmbuild -ba pg_probackup.alt.spec #2>&1 | tee -ai /app/out/build.log + + # write artefacts to out directory + rm -rf /app/out/* + cp -arv /root/rpmbuild/{RPMS,SRPMS} /app/out +else + rpmbuild -ba pg_probackup.alt.forks.spec #2>&1 | tee -ai /app/out/build.log + # write artefacts to out directory + rm -rf /app/out/* + # cp -arv /root/rpmbuild/{RPMS,SRPMS} /app/out + cp -arv /root/rpmbuild/RPMS /app/out +fi diff --git a/packaging/pkg/scripts/deb.sh b/packaging/pkg/scripts/deb.sh new file mode 100755 index 000000000..6e134635a --- /dev/null +++ b/packaging/pkg/scripts/deb.sh @@ -0,0 +1,148 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -xe +set -o pipefail + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 + +# There is no std/ent packages for PG 9.5 +if [[ ${PG_VERSION} == '9.5' ]] && [[ ${PBK_EDITION} != '' ]] ; then + exit 0 +fi + +# PACKAGES NEEDED +apt-get --allow-releaseinfo-change update -y && apt-get install -y git wget bzip2 devscripts equivs + +# Prepare +export DEBIAN_FRONTEND=noninteractive +echo 'debconf debconf/frontend select Noninteractive' | debconf-set-selections + +if [ ${CODENAME} == 'jessie' ]; then + printf "deb http://archive.debian.org/debian/ jessie main\ndeb-src http://archive.debian.org/debian/ jessie main\ndeb http://security.debian.org jessie/updates main\ndeb-src http://security.debian.org jessie/updates main" > /etc/apt/sources.list +fi + +apt-get -qq update -y + +# download PKG_URL if PKG_HASH is omitted +mkdir /root/build +cd /root/build + +# clone pbk repo +git clone $PKG_URL ${PKG_NAME}_${PKG_VERSION} +cd ${PKG_NAME}_${PKG_VERSION} +git checkout ${PKG_HASH} +cd .. + +PG_TOC=$(echo ${PG_VERSION} | sed 's|\.||g') +# Download PostgreSQL source if building for vanilla +if [[ ${PBK_EDITION} == '' ]] ; then + wget -q http://ftp.postgresql.org/pub/source/v${PG_FULL_VERSION}/postgresql-${PG_FULL_VERSION}.tar.bz2 +fi + +cd /root/build/${PKG_NAME}_${PKG_VERSION} +cp -av /app/in/specs/deb/pg_probackup/debian ./ +if [[ ${PBK_EDITION} == '' ]] ; then + sed -i "s/@PKG_NAME@/${PKG_NAME}/g" debian/changelog + sed -i "s/@PKG_VERSION@/${PKG_VERSION}/g" debian/changelog + sed -i "s/@PKG_RELEASE@/${PKG_RELEASE}/g" debian/changelog + sed -i "s/@PKG_HASH@/${PKG_HASH}/g" debian/changelog + sed -i "s/@CODENAME@/${CODENAME}/g" debian/changelog + + sed -i "s/@PKG_NAME@/${PKG_NAME}/g" debian/control + sed -i "s/@PG_VERSION@/${PG_VERSION}/g" debian/control + + sed -i "s/@PG_VERSION@/${PG_VERSION}/" debian/pg_probackup.install + mv debian/pg_probackup.install debian/${PKG_NAME}.install + + sed -i "s/@PKG_NAME@/${PKG_NAME}/g" debian/rules + sed -i "s/@PG_TOC@/${PG_TOC}/g" debian/rules + sed -i "s/@PG_VERSION@/${PG_VERSION}/g" debian/rules + sed -i "s/@PG_FULL_VERSION@/${PG_FULL_VERSION}/g" debian/rules + sed -i "s|@PREFIX@|/stump|g" debian/rules +else + sed -i "s/@PKG_NAME@/pg-probackup-${PBK_EDITION}-${PG_VERSION}/g" debian/changelog + sed -i "s/@PKG_VERSION@/${PKG_VERSION}/g" debian/changelog + sed -i "s/@PKG_RELEASE@/${PKG_RELEASE}/g" debian/changelog + sed -i "s/@PKG_HASH@/${PKG_HASH}/g" debian/changelog + sed -i "s/@CODENAME@/${CODENAME}/g" debian/changelog + + sed -i "s/@PKG_NAME@/pg-probackup-${PBK_EDITION}-${PG_VERSION}/g" debian/control + sed -i "s/pg-probackup-@PG_VERSION@/pg-probackup-${PBK_EDITION}-${PG_VERSION}/g" debian/control + sed -i "s/@PG_VERSION@/${PG_VERSION}/g" debian/control + sed -i "s/PostgreSQL/PostgresPro ${PBK_EDITION_FULL}/g" debian/control + + sed -i "s/pg_probackup-@PG_VERSION@/pg_probackup-${PBK_EDITION}-${PG_VERSION}/" debian/pg_probackup.install + mv debian/pg_probackup.install debian/pg-probackup-${PBK_EDITION}-${PG_VERSION}.install + + sed -i "s/@PKG_NAME@/pg-probackup-${PBK_EDITION}-${PG_VERSION}/g" debian/rules + sed -i "s/@PG_TOC@/${PG_TOC}/g" debian/rules + sed -i "s/pg_probackup-@PG_VERSION@/pg_probackup-${PBK_EDITION}-${PG_VERSION}/g" debian/rules + sed -i "s/postgresql-@PG_FULL_VERSION@/postgrespro-${PBK_EDITION}-${PG_FULL_VERSION}/g" debian/rules + + if [ ${PG_VERSION} == '9.6' ]; then + sed -i "s|@PREFIX@|/stump|g" debian/rules + else + sed -i "s|@PREFIX@|/opt/pgpro/${PBK_EDITION}-${PG_VERSION}|g" debian/rules + fi +fi + +# Build dependencies +mk-build-deps --install --remove --tool 'apt-get --no-install-recommends --yes' debian/control +rm -rf ./*.deb + +# Pack source to orig.tar.gz +mkdir -p /root/build/dsc +if [[ ${PBK_EDITION} == '' ]] ; then + mv /root/build/postgresql-${PG_FULL_VERSION}.tar.bz2 \ + /root/build/dsc/${PKG_NAME}_${PKG_VERSION}.orig-postgresql${PG_TOC}.tar.bz2 + + cd /root/build/${PKG_NAME}_${PKG_VERSION} + tar -xf /root/build/dsc/${PKG_NAME}_${PKG_VERSION}.orig-postgresql${PG_TOC}.tar.bz2 + cd /root/build + + tar -czf ${PKG_NAME}_${PKG_VERSION}.orig.tar.gz \ + ${PKG_NAME}_${PKG_VERSION} + + mv /root/build/${PKG_NAME}_${PKG_VERSION}.orig.tar.gz /root/build/dsc + + cd /root/build/${PKG_NAME}_${PKG_VERSION} + tar -xf /root/build/dsc/${PKG_NAME}_${PKG_VERSION}.orig-postgresql${PG_TOC}.tar.bz2 +else + tar -xf /app/in/tarballs/pgpro.tar.bz2 -C /root/build/dsc/ + cd /root/build/dsc/pgpro + + PGPRO_TOC=$(echo ${PG_FULL_VERSION} | sed 's|\.|_|g') + if [[ ${PBK_EDITION} == 'std' ]] ; then + git checkout "PGPRO${PGPRO_TOC}_1" + else + git checkout "PGPROEE${PGPRO_TOC}_1" + fi + + mv /root/build/dsc/pgpro /root/build/${PKG_NAME}_${PKG_VERSION}/postgrespro-${PBK_EDITION}-${PG_FULL_VERSION} +fi + +# BUILD: SOURCE PKG +if [[ ${PBK_EDITION} == '' ]] ; then + cd /root/build/dsc + dpkg-source -b /root/build/${PKG_NAME}_${PKG_VERSION} +fi + +# BUILD: DEB PKG +cd /root/build/${PKG_NAME}_${PKG_VERSION} +dpkg-buildpackage -b #&> /app/out/build.log + +# COPY ARTEFACTS +rm -rf /app/out/* +cd /root/build +cp -v *.deb /app/out +cp -v *.changes /app/out + +if [[ ${PBK_EDITION} == '' ]] ; then + cp -arv dsc /app/out +fi diff --git a/packaging/pkg/scripts/rpm.sh b/packaging/pkg/scripts/rpm.sh new file mode 100755 index 000000000..2fec4a700 --- /dev/null +++ b/packaging/pkg/scripts/rpm.sh @@ -0,0 +1,159 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2021 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + + +#yum upgrade -y || echo "some packages in docker fail to install" +#if [ -f /etc/rosa-release ]; then +# # Avoids old yum bugs on rosa-6 +# yum upgrade -y || echo "some packages in docker fail to install" +#fi + +set -xe +set -o pipefail + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 + +if [ ${DISTRIB} = 'centos' ] ; then + sed -i 's|^baseurl=http://|baseurl=https://|g' /etc/yum.repos.d/*.repo + if [ ${DISTRIB_VERSION} = '8' ]; then + sed -i 's|mirrorlist|#mirrorlist|g' /etc/yum.repos.d/CentOS-*.repo + sed -i 's|#baseurl=http://mirror.centos.org|baseurl=http://vault.centos.org|g' /etc/yum.repos.d/CentOS-*.repo + fi + yum update -y + if [ ${DISTRIB_VERSION} = '8' ]; then + sed -i 's|mirrorlist|#mirrorlist|g' /etc/yum.repos.d/CentOS-*.repo + sed -i 's|#baseurl=http://mirror.centos.org|baseurl=http://vault.centos.org|g' /etc/yum.repos.d/CentOS-*.repo + fi +fi + +# PACKAGES NEEDED +yum install -y git wget bzip2 rpm-build +if [ ${DISTRIB} = 'oraclelinux' -a ${DISTRIB_VERSION} = '8' -a -n ${PBK_EDITION} ] ; then + yum install -y bison flex +fi + +mkdir /root/build +cd /root/build +rpm --rebuilddb && yum clean all + +# Copy rpmbuild +cp -rv /app/in/specs/rpm/rpmbuild /root/ + +# download pbk +git clone $PKG_URL pg_probackup-${PKG_VERSION} +cd pg_probackup-${PKG_VERSION} +git checkout ${PKG_HASH} + +# move it to source +cd /root/build +if [[ ${PBK_EDITION} == '' ]] ; then + tar -cjf pg_probackup-${PKG_VERSION}.tar.bz2 pg_probackup-${PKG_VERSION} + mv pg_probackup-${PKG_VERSION}.tar.bz2 /root/rpmbuild/SOURCES + rm -rf pg_probackup-${PKG_VERSION} +else + mv pg_probackup-${PKG_VERSION} /root/rpmbuild/SOURCES +fi + +if [[ ${PBK_EDITION} == '' ]] ; then + + # Download PostgreSQL source + wget -q http://ftp.postgresql.org/pub/source/v${PG_FULL_VERSION}/postgresql-${PG_FULL_VERSION}.tar.bz2 -O /root/rpmbuild/SOURCES/postgresql-${PG_VERSION}.tar.bz2 + + cd /root/rpmbuild/SOURCES/ + sed -i "s/@DISTRIB@/${DISTRIB}/" pg_probackup.repo + if [ $DISTRIB == 'centos' ] + then sed -i "s/@SHORT_CODENAME@/Centos/" pg_probackup.repo + elif [ $DISTRIB == 'rhel' ] + then sed -i "s/@SHORT_CODENAME@/RedHat/" pg_probackup.repo + elif [ $DISTRIB == 'oraclelinux' ] + then sed -i "s/@SHORT_CODENAME@/Oracle/" pg_probackup.repo + fi +else + tar -xf /app/in/tarballs/pgpro.tar.bz2 -C /root/rpmbuild/SOURCES/ + cd /root/rpmbuild/SOURCES/pgpro + + PGPRO_TOC=$(echo ${PG_FULL_VERSION} | sed 's|\.|_|g') + if [[ ${PBK_EDITION} == 'std' ]] ; then + git checkout "PGPRO${PGPRO_TOC}_1" + else + git checkout "PGPROEE${PGPRO_TOC}_1" + fi + rm -rf .git + + cd /root/rpmbuild/SOURCES/ + sed -i "s/@DISTRIB@/${DISTRIB}/" pg_probackup-forks.repo + if [ $DISTRIB == 'centos' ] + then sed -i "s/@SHORT_CODENAME@/Centos/" pg_probackup-forks.repo + elif [ $DISTRIB == 'rhel' ] + then sed -i "s/@SHORT_CODENAME@/RedHat/" pg_probackup-forks.repo + elif [ $DISTRIB == 'oraclelinux' ] + then sed -i "s/@SHORT_CODENAME@/Oracle/" pg_probackup-forks.repo + fi + + mv pgpro postgrespro-${PBK_EDITION}-${PG_FULL_VERSION} + chown -R root:root postgrespro-${PBK_EDITION}-${PG_FULL_VERSION} + +# tar -cjf postgrespro-${PBK_EDITION}-${PG_FULL_VERSION}.tar.bz2 postgrespro-${PBK_EDITION}-${PG_FULL_VERSION} +fi + +cd /root/rpmbuild/SPECS +if [[ ${PBK_EDITION} == '' ]] ; then + sed -i "s/@PKG_VERSION@/${PKG_VERSION}/" pg_probackup.spec + sed -i "s/@PKG_RELEASE@/${PKG_RELEASE}/" pg_probackup.spec + sed -i "s/@PKG_HASH@/${PKG_HASH}/" pg_probackup.spec + sed -i "s/@PG_VERSION@/${PG_VERSION}/" pg_probackup.spec + sed -i "s/@PG_FULL_VERSION@/${PG_FULL_VERSION}/" pg_probackup.spec + + sed -i "s/@PKG_VERSION@/${PKG_VERSION}/" pg_probackup-repo.spec + sed -i "s/@PKG_RELEASE@/${PKG_RELEASE}/" pg_probackup-repo.spec +else + sed -i "s/@EDITION@/${PBK_EDITION}/" pg_probackup-pgpro.spec + sed -i "s/@EDITION_FULL@/${PBK_EDITION_FULL}/" pg_probackup-pgpro.spec + sed -i "s/@PKG_VERSION@/${PKG_VERSION}/" pg_probackup-pgpro.spec + sed -i "s/@PKG_RELEASE@/${PKG_RELEASE}/" pg_probackup-pgpro.spec + sed -i "s/@PKG_HASH@/${PKG_HASH}/" pg_probackup-pgpro.spec + sed -i "s/@PG_VERSION@/${PG_VERSION}/" pg_probackup-pgpro.spec + sed -i "s/@PG_FULL_VERSION@/${PG_FULL_VERSION}/" pg_probackup-pgpro.spec + + if [ ${PG_VERSION} != '9.6' ]; then + sed -i "s|@PREFIX@|/opt/pgpro/${EDITION}-${PG_VERSION}|g" pg_probackup-pgpro.spec + fi + + sed -i "s/@PKG_VERSION@/${PKG_VERSION}/" pg_probackup-repo-forks.spec + sed -i "s/@PKG_RELEASE@/${PKG_RELEASE}/" pg_probackup-repo-forks.spec +fi + +if [[ ${PBK_EDITION} == '' ]] ; then + + # install dependencies + yum-builddep -y pg_probackup.spec + + # build pg_probackup + rpmbuild -bs pg_probackup.spec + rpmbuild -ba pg_probackup.spec + + # build repo files + rpmbuild -bs pg_probackup-repo.spec + rpmbuild -ba pg_probackup-repo.spec + + # write artefacts to out directory + rm -rf /app/out/* + cp -arv /root/rpmbuild/{RPMS,SRPMS} /app/out +else + # install dependencies + yum-builddep -y pg_probackup-pgpro.spec + # build pg_probackup + rpmbuild -ba pg_probackup-pgpro.spec + + # build repo files + rpmbuild -ba pg_probackup-repo-forks.spec + + # write artefacts to out directory + rm -rf /app/out/* + cp -arv /root/rpmbuild/RPMS /app/out +fi diff --git a/packaging/pkg/scripts/suse.sh b/packaging/pkg/scripts/suse.sh new file mode 100755 index 000000000..76b444b5b --- /dev/null +++ b/packaging/pkg/scripts/suse.sh @@ -0,0 +1,95 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + + +#yum upgrade -y || echo "some packages in docker fail to install" +#if [ -f /etc/rosa-release ]; then +# # Avoids old yum bugs on rosa-6 +# yum upgrade -y || echo "some packages in docker fail to install" +#fi + +set -xe +set -o pipefail + +# currenctly we do not build std|ent packages for Suse +if [[ ${PBK_EDITION} != '' ]] ; then + exit 0 +fi + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 +zypper clean + +# PACKAGES NEEDED +zypper install -y git wget bzip2 rpm-build + +mkdir /root/build +cd /root/build + +# Copy rpmbuild +cp -rv /app/in/specs/rpm/rpmbuild /root/ + +# download pbk +git clone $PKG_URL pg_probackup-${PKG_VERSION} +cd pg_probackup-${PKG_VERSION} +git checkout ${PKG_HASH} +cd .. + +# tarball it +tar -cjf pg_probackup-${PKG_VERSION}.tar.bz2 pg_probackup-${PKG_VERSION} +mv pg_probackup-${PKG_VERSION}.tar.bz2 /root/rpmbuild/SOURCES +rm -rf pg_probackup-${PKG_VERSION} + +# Download PostgreSQL source +wget -q http://ftp.postgresql.org/pub/source/v${PG_FULL_VERSION}/postgresql-${PG_FULL_VERSION}.tar.bz2 -O /root/rpmbuild/SOURCES/postgresql-${PG_VERSION}.tar.bz2 + +rm -rf /usr/src/packages +ln -s /root/rpmbuild /usr/src/packages + +cd /root/rpmbuild/SOURCES +sed -i "s/@PG_VERSION@/${PKG_VERSION}/" pg_probackup.repo + + +# change to build dir +cd /root/rpmbuild/SOURCES +sed -i "s/@DISTRIB@/${DISTRIB}/" pg_probackup.repo +if [ $DISTRIB == 'centos' ] + then sed -i "s/@SHORT_CODENAME@/Centos/" pg_probackup.repo +elif [ $DISTRIB == 'rhel' ] + then sed -i "s/@SHORT_CODENAME@/RedHat/" pg_probackup.repo +elif [ $DISTRIB == 'oraclelinux' ] + then sed -i "s/@SHORT_CODENAME@/Oracle/" pg_probackup.repo +elif [ $DISTRIB == 'suse' ] + then sed -i "s/@SHORT_CODENAME@/SUSE/" pg_probackup.repo +fi + +cd /root/rpmbuild/SPECS +sed -i "s/@PKG_VERSION@/${PKG_VERSION}/" pg_probackup.spec +sed -i "s/@PKG_RELEASE@/${PKG_RELEASE}/" pg_probackup.spec +sed -i "s/@PKG_HASH@/${PKG_HASH}/" pg_probackup.spec +sed -i "s/@PG_VERSION@/${PG_VERSION}/" pg_probackup.spec +sed -i "s/@PG_FULL_VERSION@/${PG_FULL_VERSION}/" pg_probackup.spec + +sed -i "s/@PG_VERSION@/${PG_VERSION}/" pg_probackup-repo.spec +sed -i "s/@PKG_VERSION@/${PKG_VERSION}/" pg_probackup-repo.spec +sed -i "s/@PKG_RELEASE@/${PKG_RELEASE}/" pg_probackup-repo.spec + +# install dependencies +zypper -n install \ + $(rpmspec --parse pg_probackup.spec | grep BuildRequires | cut -d':' -f2 | xargs) + +# build pg_probackup +rpmbuild -bs pg_probackup.spec +rpmbuild -ba pg_probackup.spec #2>&1 | tee -ai /app/out/build.log + +# build repo files, TODO: move to separate repo +rpmbuild -ba pg_probackup-repo.spec + +# write artefacts to out directory +rm -rf /app/out/* + +cp -arv /root/rpmbuild/{RPMS,SRPMS} /app/out diff --git a/packaging/pkg/specs/deb/pg_probackup/debian/changelog b/packaging/pkg/specs/deb/pg_probackup/debian/changelog new file mode 100644 index 000000000..5b9160220 --- /dev/null +++ b/packaging/pkg/specs/deb/pg_probackup/debian/changelog @@ -0,0 +1,11 @@ +@PKG_NAME@ (@PKG_VERSION@-@PKG_RELEASE@.@PKG_HASH@.@CODENAME@) @CODENAME@; urgency=medium + + * @PKG_VERSION@ + + -- Grigory Smolkin Wed, 9 Feb 2018 10:22:08 +0300 + +@PKG_NAME@ (2.0.14-1.@CODENAME@) @CODENAME@; urgency=medium + + * Initial package + + -- Grigory Smolkin Fri, 29 Jan 2018 10:22:08 +0300 diff --git a/packaging/pkg/specs/deb/pg_probackup/debian/compat b/packaging/pkg/specs/deb/pg_probackup/debian/compat new file mode 100644 index 000000000..ec635144f --- /dev/null +++ b/packaging/pkg/specs/deb/pg_probackup/debian/compat @@ -0,0 +1 @@ +9 diff --git a/packaging/pkg/specs/deb/pg_probackup/debian/control b/packaging/pkg/specs/deb/pg_probackup/debian/control new file mode 100644 index 000000000..8f1d42007 --- /dev/null +++ b/packaging/pkg/specs/deb/pg_probackup/debian/control @@ -0,0 +1,29 @@ +Source: @PKG_NAME@ +Section: database +Priority: optional +Maintainer: PostgresPro DBA +Uploaders: Grigory Smolkin +Build-Depends: + debhelper (>= 9), + bison, + dpkg-dev, + flex, + gettext, + zlib1g-dev | libz-dev, + libpq5 +Standards-Version: 3.9.6 +Homepage: https://github.com/postgrespro/pg_probackup + +Package: @PKG_NAME@ +Architecture: any +Depends: ${misc:Depends}, ${shlibs:Depends} +Description: Backup tool for PostgreSQL. + . + This package provides pg_probackup binary for PostgreSQL @PG_VERSION@. + +Package: @PKG_NAME@-dbg +Depends: @PKG_NAME@ +Architecture: any +Description: Backup tool for PostgreSQL. + . + This package provides detached debugging symbols for pg_probackup diff --git a/packaging/pkg/specs/deb/pg_probackup/debian/pg_probackup.install b/packaging/pkg/specs/deb/pg_probackup/debian/pg_probackup.install new file mode 100644 index 000000000..ed904ca40 --- /dev/null +++ b/packaging/pkg/specs/deb/pg_probackup/debian/pg_probackup.install @@ -0,0 +1 @@ +pg_probackup-@PG_VERSION@ /usr/bin/ \ No newline at end of file diff --git a/packaging/pkg/specs/deb/pg_probackup/debian/rules b/packaging/pkg/specs/deb/pg_probackup/debian/rules new file mode 100644 index 000000000..309a9a1d4 --- /dev/null +++ b/packaging/pkg/specs/deb/pg_probackup/debian/rules @@ -0,0 +1,29 @@ +#!/usr/bin/make -f + +# Uncomment this to turn on verbose mode. +export DH_VERBOSE=1 + +%: + dh $@ + +override_dh_auto_clean: + # skip + +override_dh_auto_build: + cd postgresql-@PG_FULL_VERSION@ && ./configure --enable-debug --without-readline --prefix=@PREFIX@ &&\ + make MAKELEVEL=0 install DESTDIR=$(CURDIR)/debian/tmp && cd .. &&\ + make USE_PGXS=1 top_srcdir=$(CURDIR)/postgresql-@PG_FULL_VERSION@ PG_CONFIG=$(CURDIR)/debian/tmp/@PREFIX@/bin/pg_config &&\ + mv pg_probackup pg_probackup-@PG_VERSION@ + +override_dh_auto_test: + # skip + +override_dh_auto_install: + # skip + +override_dh_strip: + dh_strip --dbg-package=@PKG_NAME@-dbg + +override_dh_auto_clean: + # skip + #make clean top_srcdir=$(CURDIR)/pg@PG_TOC@-source PG_CONFIG=$(CURDIR)/debian/tmp/stump/bin/pg_config diff --git a/packaging/pkg/specs/deb/pg_probackup/debian/source/format b/packaging/pkg/specs/deb/pg_probackup/debian/source/format new file mode 100644 index 000000000..163aaf8d8 --- /dev/null +++ b/packaging/pkg/specs/deb/pg_probackup/debian/source/format @@ -0,0 +1 @@ +3.0 (quilt) diff --git a/packaging/pkg/specs/rpm/rpmbuild/SOURCES/GPG-KEY-PG_PROBACKUP b/packaging/pkg/specs/rpm/rpmbuild/SOURCES/GPG-KEY-PG_PROBACKUP new file mode 100644 index 000000000..c11d9c015 --- /dev/null +++ b/packaging/pkg/specs/rpm/rpmbuild/SOURCES/GPG-KEY-PG_PROBACKUP @@ -0,0 +1,52 @@ +-----BEGIN PGP PUBLIC KEY BLOCK----- +Version: GnuPG v2.0.22 (GNU/Linux) + +mQINBFpy9DABEADd44hR3o4i4DrUephrr7iHPHcRH0Zego3A36NdOf0ymP94H8Bi +U8C6YyKFbltShh18IC3QZJK04hLRQEs6sPKC2XHwlz+Tndi49Z45pfV54xEVKmBS +IZ5AM9y1FxwQAOzu6pZGu32DWDXZzhI7nLuY8rqAMMuzKeRcGm3sQ6ZcAwYOLT+e +ZAxkUL05MBGDaLc91HtKiurRlHuMySiVdkNu9ebTGV4zZv+ocBK8iC5rJjTJCv78 +eLkrRgjp7/MuLQ7mmiwfZx5lUIO9S87HDeH940mcYWRGUsdCbj0791wHY0PXlqhH +6lCLGur9/5yM88pGU79uahCblxsYdue6fdgIZR0hQBUxnLphI2SCshP89VDEwaP2 +dlC/qESJ3xyULkdJz67wlhOrIPC9T1d2pa5MUurOK0yTFH7j4JLWjBgU59h31ZEF +NMHde+Fwv+lL/yRht2Xz7HG5Rt8ogn4/rPBloXr1v83iN34aZnnqanyhSbE9xUhP +RNK3fBxXmX9IjFsBhRelPcv5NWNnxnnMkEfhoZvrAy+ykUGLP+J+Rj+d5v/8nAUc +taxqAXlUz1VabR0BVISBsRY+ket4O2dJ1WbZ8KXG6q/F9UMpS0v9aRdb1JyzrWCw +wT/l3q9x89i27SgDZgAfEFhvbMN6hUmFyVoMBgk8kqvi4b3lZZGCeuLX5wARAQAB +tCxQb3N0Z3JlU1FMIFByb2Zlc3Npb25hbCA8ZGJhQHBvc3RncmVzcHJvLnJ1PokC +OQQTAQIAIwUCWnL0MAIbAwcLCQgHAwIBBhUIAgkKCwQWAgMBAh4BAheAAAoJEKeJ +efZjbXF+zDUP/RfYxlq3erzP/cG6/LghZlJy6hGuUgyDFj2zUVAbpoFhqCAmaNLc ++bBYMCyNRhS8/oXushCSxUV8D7LRIRIRdtbNAnd4MNl6U4ORF6JcdPPNLROzwMik +3TmIVACMdjb9IRF5+8jVrIgDPI/FVtf5qp0Ot6OBtpD5oWQ7ubZ31RPR3pacdujK +jlbzL5Y6HsonhMbSJU/d0d9DylMvX4Gcxdw7M2Pfe3E6mjPJmcHiKuCKln2eLOsg +53HA/RWKy+uYDo+vdefUsQCIdnC5VghnXG8FTuvVqeqiSeU2XdyuzjndqxKZNrMw +YK1POK7R55R1aKJaOKEwnfd5oN02p77U+R/vb/mDcfZWbXI8JrHwPKVOQiEl0S+G +ePPW57EmX9yFuWAzcOPp9yCt/+roVry1ICifrFaLOhtT+/vle0j3+rbn31BMPsjf +QbREVetHfWB0N78k/hKC8SivDdrXsdqovcGgSAjFzPEdznvx9vKopwz2CQ6DK25Q +3M4j79Akcaa08k5Wphyx48PbhzSeE/d4xVzey7ge0BwYMdNGXKeyBjT6h9e+iySE +UTZ3/3c7O1D8p2EfPUMT/aI5fWlLBXlT5fDp2yX0HMTt/NUIXAiTHb5BDnZ+4ld3 +KXjHw4WzaOfHBfGDjJDtHPgdTEJTsQbH8//D+wwU3ueNS1ho4DpLqc+YuQINBFpy +9DABEADJMkgQ2m4g4LX7FNnmQbRgDcuhL8Y0VRGST+5x6xvb2em1boQHUaTt7/3z +DnaIRrZqrFP09O6xblSjEu9FZE+JuQGNyC4TH9fjvKnkRlqTF6X87nRVGByRmrdL +lPp9XPJY2Mc7c0PisncI/j7d9PmUHOSmaWeLG/WqMbzZA+s1IWjC0tqIN2k5ivTN +PfRm+9ebEHMUN+D7yZQMBlCmFexwy6h5pAioyj4tAOHqxfNDE33qezaeBn/E1BpW +NyegKwNtPUL0t2kXTO5tspKKCcny4HJ7K60gak0fWp42qVygwSPR54ztFM+6XjCh +0MmZ/mAdzLd6OJiP8RfMCfXbXpK4793+Cw0AK3Mu+mnJ26kz1KEZ9DKiAEhBhK3r +Z3/isUc8LcVYLHIduH9b/K50FjgR0T1Lm4r6Hpf6nTROlfiFSMXJU0HepAzMPHRq +EWqTJ49UgI7Llf+aBP7fGLqRPvWJpAJaQkMiUxfP5JYYCb+45d7I54iXQCD6ToK1 +bDnh+zZIrwyUIxPfFQh1xPYyFWRELJpeOFzm+espqiVFPXpBoimVlytwNrGdbxbY +SO0eEVlE41AjD8cgk+ibAvt/moT2+Mps/t083LR+J92kj+iX/D4NHVy4CjJTrhwO +rI3FrxtdU+NFXULyj0KslOKuyG5WuHLQvfL5P3JGuTkP4iJOTQARAQABiQIfBBgB +AgAJBQJacvQwAhsMAAoJEKeJefZjbXF+8JgQAJqlO1ftIsJvZ/+4ZVVOTPx5ZmYs +ABp4/2gaiLdhajN8ynbZqtCyjtQwSCLJFf2CcDL8XUooJzdQECkqdiI7ouYSFBzO +ui3jjCuFz5oHv88OtX2cIRxHqlZQmXEHvk0gH61xDV5CWBJmjxdRcsC7n1I8DSVg +Qmuq06S+xIX6rHf2CRxYKahBip71u7OIH4BRV44y26xf1a8an+8BkqF9+mYt7zqO +vyMCJ1UftXcuE5SxY54jnNAavF7Kq/2Yp7v3aYqFREngxtbWudyo7QW5EuToSvY2 +qY6tpInahWjuXxeARsFzp4fB0Eo/yH+iqG30zkQCuxLyxzbMMcNQP4if3yV6uO14 +LqapZLrMp6IMTfHDKmbbtDQ2RpRRut3K4khXRQ1LjGKziOU4ZCEazrXEijm2AlKw +7JS3POGvM+VAiaGNBqfdHpTwXXT7zkxJjfJC3Et/6fHy1xuCtUtMs41PjHS/HWi8 +w70T8XpKub+ewVElxq2D83rx07w3HuBtVUyqG0XgcACwqQA1vMLJaR3VoX1024ho +sf2PtZqQ7SCgt0hkZAT72j05nz4bIxUIcDkAGtd9FDPQ4Ixi6fRfTJpZ7lIEV5as +Zs9C0hrxmWgJwSGgQa2Waylvw47fMwfMn+gUNRqwanyOjVYfpSJafLc6Ol43bQN/ +jCKs4enncezhjcAh +=TVZj +-----END PGP PUBLIC KEY BLOCK----- \ No newline at end of file diff --git a/packaging/pkg/specs/rpm/rpmbuild/SOURCES/GPG-KEY-PG_PROBACKUP-FORKS b/packaging/pkg/specs/rpm/rpmbuild/SOURCES/GPG-KEY-PG_PROBACKUP-FORKS new file mode 100644 index 000000000..c11d9c015 --- /dev/null +++ b/packaging/pkg/specs/rpm/rpmbuild/SOURCES/GPG-KEY-PG_PROBACKUP-FORKS @@ -0,0 +1,52 @@ +-----BEGIN PGP PUBLIC KEY BLOCK----- +Version: GnuPG v2.0.22 (GNU/Linux) + +mQINBFpy9DABEADd44hR3o4i4DrUephrr7iHPHcRH0Zego3A36NdOf0ymP94H8Bi +U8C6YyKFbltShh18IC3QZJK04hLRQEs6sPKC2XHwlz+Tndi49Z45pfV54xEVKmBS +IZ5AM9y1FxwQAOzu6pZGu32DWDXZzhI7nLuY8rqAMMuzKeRcGm3sQ6ZcAwYOLT+e +ZAxkUL05MBGDaLc91HtKiurRlHuMySiVdkNu9ebTGV4zZv+ocBK8iC5rJjTJCv78 +eLkrRgjp7/MuLQ7mmiwfZx5lUIO9S87HDeH940mcYWRGUsdCbj0791wHY0PXlqhH +6lCLGur9/5yM88pGU79uahCblxsYdue6fdgIZR0hQBUxnLphI2SCshP89VDEwaP2 +dlC/qESJ3xyULkdJz67wlhOrIPC9T1d2pa5MUurOK0yTFH7j4JLWjBgU59h31ZEF +NMHde+Fwv+lL/yRht2Xz7HG5Rt8ogn4/rPBloXr1v83iN34aZnnqanyhSbE9xUhP +RNK3fBxXmX9IjFsBhRelPcv5NWNnxnnMkEfhoZvrAy+ykUGLP+J+Rj+d5v/8nAUc +taxqAXlUz1VabR0BVISBsRY+ket4O2dJ1WbZ8KXG6q/F9UMpS0v9aRdb1JyzrWCw +wT/l3q9x89i27SgDZgAfEFhvbMN6hUmFyVoMBgk8kqvi4b3lZZGCeuLX5wARAQAB +tCxQb3N0Z3JlU1FMIFByb2Zlc3Npb25hbCA8ZGJhQHBvc3RncmVzcHJvLnJ1PokC +OQQTAQIAIwUCWnL0MAIbAwcLCQgHAwIBBhUIAgkKCwQWAgMBAh4BAheAAAoJEKeJ +efZjbXF+zDUP/RfYxlq3erzP/cG6/LghZlJy6hGuUgyDFj2zUVAbpoFhqCAmaNLc ++bBYMCyNRhS8/oXushCSxUV8D7LRIRIRdtbNAnd4MNl6U4ORF6JcdPPNLROzwMik +3TmIVACMdjb9IRF5+8jVrIgDPI/FVtf5qp0Ot6OBtpD5oWQ7ubZ31RPR3pacdujK +jlbzL5Y6HsonhMbSJU/d0d9DylMvX4Gcxdw7M2Pfe3E6mjPJmcHiKuCKln2eLOsg +53HA/RWKy+uYDo+vdefUsQCIdnC5VghnXG8FTuvVqeqiSeU2XdyuzjndqxKZNrMw +YK1POK7R55R1aKJaOKEwnfd5oN02p77U+R/vb/mDcfZWbXI8JrHwPKVOQiEl0S+G +ePPW57EmX9yFuWAzcOPp9yCt/+roVry1ICifrFaLOhtT+/vle0j3+rbn31BMPsjf +QbREVetHfWB0N78k/hKC8SivDdrXsdqovcGgSAjFzPEdznvx9vKopwz2CQ6DK25Q +3M4j79Akcaa08k5Wphyx48PbhzSeE/d4xVzey7ge0BwYMdNGXKeyBjT6h9e+iySE +UTZ3/3c7O1D8p2EfPUMT/aI5fWlLBXlT5fDp2yX0HMTt/NUIXAiTHb5BDnZ+4ld3 +KXjHw4WzaOfHBfGDjJDtHPgdTEJTsQbH8//D+wwU3ueNS1ho4DpLqc+YuQINBFpy +9DABEADJMkgQ2m4g4LX7FNnmQbRgDcuhL8Y0VRGST+5x6xvb2em1boQHUaTt7/3z +DnaIRrZqrFP09O6xblSjEu9FZE+JuQGNyC4TH9fjvKnkRlqTF6X87nRVGByRmrdL +lPp9XPJY2Mc7c0PisncI/j7d9PmUHOSmaWeLG/WqMbzZA+s1IWjC0tqIN2k5ivTN +PfRm+9ebEHMUN+D7yZQMBlCmFexwy6h5pAioyj4tAOHqxfNDE33qezaeBn/E1BpW +NyegKwNtPUL0t2kXTO5tspKKCcny4HJ7K60gak0fWp42qVygwSPR54ztFM+6XjCh +0MmZ/mAdzLd6OJiP8RfMCfXbXpK4793+Cw0AK3Mu+mnJ26kz1KEZ9DKiAEhBhK3r +Z3/isUc8LcVYLHIduH9b/K50FjgR0T1Lm4r6Hpf6nTROlfiFSMXJU0HepAzMPHRq +EWqTJ49UgI7Llf+aBP7fGLqRPvWJpAJaQkMiUxfP5JYYCb+45d7I54iXQCD6ToK1 +bDnh+zZIrwyUIxPfFQh1xPYyFWRELJpeOFzm+espqiVFPXpBoimVlytwNrGdbxbY +SO0eEVlE41AjD8cgk+ibAvt/moT2+Mps/t083LR+J92kj+iX/D4NHVy4CjJTrhwO +rI3FrxtdU+NFXULyj0KslOKuyG5WuHLQvfL5P3JGuTkP4iJOTQARAQABiQIfBBgB +AgAJBQJacvQwAhsMAAoJEKeJefZjbXF+8JgQAJqlO1ftIsJvZ/+4ZVVOTPx5ZmYs +ABp4/2gaiLdhajN8ynbZqtCyjtQwSCLJFf2CcDL8XUooJzdQECkqdiI7ouYSFBzO +ui3jjCuFz5oHv88OtX2cIRxHqlZQmXEHvk0gH61xDV5CWBJmjxdRcsC7n1I8DSVg +Qmuq06S+xIX6rHf2CRxYKahBip71u7OIH4BRV44y26xf1a8an+8BkqF9+mYt7zqO +vyMCJ1UftXcuE5SxY54jnNAavF7Kq/2Yp7v3aYqFREngxtbWudyo7QW5EuToSvY2 +qY6tpInahWjuXxeARsFzp4fB0Eo/yH+iqG30zkQCuxLyxzbMMcNQP4if3yV6uO14 +LqapZLrMp6IMTfHDKmbbtDQ2RpRRut3K4khXRQ1LjGKziOU4ZCEazrXEijm2AlKw +7JS3POGvM+VAiaGNBqfdHpTwXXT7zkxJjfJC3Et/6fHy1xuCtUtMs41PjHS/HWi8 +w70T8XpKub+ewVElxq2D83rx07w3HuBtVUyqG0XgcACwqQA1vMLJaR3VoX1024ho +sf2PtZqQ7SCgt0hkZAT72j05nz4bIxUIcDkAGtd9FDPQ4Ixi6fRfTJpZ7lIEV5as +Zs9C0hrxmWgJwSGgQa2Waylvw47fMwfMn+gUNRqwanyOjVYfpSJafLc6Ol43bQN/ +jCKs4enncezhjcAh +=TVZj +-----END PGP PUBLIC KEY BLOCK----- \ No newline at end of file diff --git a/packaging/pkg/specs/rpm/rpmbuild/SOURCES/pg_probackup-forks.repo b/packaging/pkg/specs/rpm/rpmbuild/SOURCES/pg_probackup-forks.repo new file mode 100644 index 000000000..d26b058cd --- /dev/null +++ b/packaging/pkg/specs/rpm/rpmbuild/SOURCES/pg_probackup-forks.repo @@ -0,0 +1,6 @@ +[pg_probackup-forks] +name=PG_PROBACKUP @SHORT_CODENAME@ packages for PostgresPro Standard and Enterprise - $basearch +baseurl=https://repo.postgrespro.ru/pg_probackup-forks/rpm/latest/@DISTRIB@-$releasever-$basearch +enabled=1 +gpgcheck=1 +gpgkey=file:///etc/pki/rpm-gpg/GPG-KEY-PG_PROBACKUP diff --git a/packaging/pkg/specs/rpm/rpmbuild/SOURCES/pg_probackup.repo b/packaging/pkg/specs/rpm/rpmbuild/SOURCES/pg_probackup.repo new file mode 100644 index 000000000..33dc31a24 --- /dev/null +++ b/packaging/pkg/specs/rpm/rpmbuild/SOURCES/pg_probackup.repo @@ -0,0 +1,13 @@ +[pg_probackup] +name=PG_PROBACKUP Packages for @SHORT_CODENAME@ Linux - $basearch +baseurl=https://repo.postgrespro.ru/pg_probackup/rpm/latest/@DISTRIB@-$releasever-$basearch +enabled=1 +gpgcheck=1 +gpgkey=file:///etc/pki/rpm-gpg/GPG-KEY-PG_PROBACKUP + +[pg_probackup-sources] +name=PG_PROBACKUP Source Packages for @SHORT_CODENAME@ Linux - $basearch +baseurl=https://repo.postgrespro.ru/pg_probackup/srpm/latest/@DISTRIB@-$releasever-$basearch +enabled=1 +gpgcheck=1 +gpgkey=file:///etc/pki/rpm-gpg/GPG-KEY-PG_PROBACKUP diff --git a/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup-pgpro.spec b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup-pgpro.spec new file mode 100644 index 000000000..8955b3fa7 --- /dev/null +++ b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup-pgpro.spec @@ -0,0 +1,71 @@ +%global version @PKG_VERSION@ +%global release @PKG_RELEASE@ +%global hash @PKG_HASH@ +%global pgsql_major @PG_VERSION@ +%global pgsql_full @PG_FULL_VERSION@ +%global edition @EDITION@ +%global edition_full @EDITION_FULL@ +%global prefix @PREFIX@ + +Name: pg_probackup-%{edition}-%{pgsql_major} +Version: %{version} +Release: %{release}.%{hash} +Summary: Backup utility for PostgresPro %{edition_full} +Group: Applications/Databases +License: BSD +Url: http://postgrespro.ru/ +#Source0: postgrespro-%{edition}-%{pgsql_full}.tar.bz2 +#Source1: pg_probackup-%{version}.tar.bz2 +Source0: postgrespro-%{edition}-%{pgsql_full} +Source1: pg_probackup-%{version} +BuildRequires: gcc make perl glibc-devel +BuildRequires: openssl-devel gettext zlib-devel + + +%description +Backup tool for PostgresPro %{edition_full}. + +%prep +#%setup -q -b1 -n pg_probackup-%{version}.tar.bz2 +mv %{_topdir}/SOURCES/postgrespro-%{edition}-%{pgsql_full} %{_topdir}/BUILD +cd %{_topdir}/BUILD/postgrespro-%{edition}-%{pgsql_full} +mv %{_topdir}/SOURCES/pg_probackup-%{version} contrib/pg_probackup + +mkdir %{_topdir}/SOURCES/postgrespro-%{edition}-%{pgsql_full} +mkdir %{_topdir}/SOURCES/pg_probackup-%{version} + +%build +#cd %{_topdir}/SOURCES/postgrespro-%{edition}-%{pgsql_full} +#mv %{_topdir}/SOURCES/postgrespro-%{edition}-%{pgsql_full} ./ +#cd postgrespro-%{edition}-%{pgsql_full} +#mv %{_topdir}/SOURCES/pg_probackup-%{version} contrib/pg_probackup +cd %{_topdir}/BUILD/postgrespro-%{edition}-%{pgsql_full} + +%if "%{pgsql_major}" == "9.6" +./configure --enable-debug --without-readline +%else +./configure --enable-debug --without-readline --prefix=%{prefix} +%endif +make -C 'src/common' +make -C 'src/port' +make -C 'src/interfaces' +cd contrib/pg_probackup && make + +%install +cd %{_topdir}/BUILD/postgrespro-%{edition}-%{pgsql_full} +%{__mkdir} -p %{buildroot}%{_bindir} +%{__install} -p -m 755 contrib/pg_probackup/pg_probackup %{buildroot}%{_bindir}/%{name} + +%files +%{_bindir}/%{name} + +%clean +rm -rf $RPM_BUILD_ROOT + + +%changelog +* Wed Feb 9 2018 Grigory Smolkin - %{version}-%{release}.%{hash} +- @PKG_VERSION@ + +* Fri Jan 29 2018 Grigory Smolkin - 2.0.14-1 +- Initial release. diff --git a/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup-repo-forks.spec b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup-repo-forks.spec new file mode 100644 index 000000000..47adb250f --- /dev/null +++ b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup-repo-forks.spec @@ -0,0 +1,49 @@ +%global version @PKG_VERSION@ +%global release @PKG_RELEASE@ + +Summary: pg_probackup repo RPM +Name: pg_probackup-repo-forks +Version: %{version} +Release: %{release} +Group: Applications/Databases +License: BSD +Url: http://postgrespro.ru/ + +Source0: http://repo.postgrespro.ru/pg_probackup-forks/keys/GPG-KEY-PG_PROBACKUP +Source1: pg_probackup-forks.repo + +BuildArch: noarch + +%description +This package contains yum configuration for @SHORT_CODENAME@, and also the GPG key +for pg_probackup RPMs for PostgresPro Standard and Enterprise. + +%prep +%setup -q -c -T +install -pm 644 %{SOURCE0} . +install -pm 644 %{SOURCE1} . + +%build + +%install +rm -rf $RPM_BUILD_ROOT + +#GPG Key +install -Dpm 644 %{SOURCE0} \ + $RPM_BUILD_ROOT%{_sysconfdir}/pki/rpm-gpg/GPG-KEY-PG_PROBACKUP + +# yum +install -dm 755 $RPM_BUILD_ROOT%{_sysconfdir}/yum.repos.d +install -pm 644 %{SOURCE1} $RPM_BUILD_ROOT%{_sysconfdir}/yum.repos.d + +%clean +rm -rf $RPM_BUILD_ROOT + +%files +%defattr(-,root,root,-) +%config(noreplace) /etc/yum.repos.d/* +/etc/pki/rpm-gpg/* + +%changelog +* Fri Oct 26 2019 Grigory Smolkin +- Initial package diff --git a/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup-repo.spec b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup-repo.spec new file mode 100644 index 000000000..da54bc7b1 --- /dev/null +++ b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup-repo.spec @@ -0,0 +1,58 @@ +%global version @PKG_VERSION@ +%global release @PKG_RELEASE@ + +Summary: PG_PROBACKUP RPMs +Name: pg_probackup-repo +Version: %{version} +Release: %{release} +Group: Applications/Databases +License: BSD +Url: http://postgrespro.ru/ + +Source0: http://repo.postgrespro.ru/pg_probackup/keys/GPG-KEY-PG_PROBACKUP +Source1: pg_probackup.repo + +BuildArch: noarch + +%description +This package contains yum configuration for Centos, and also the GPG key for PG_PROBACKUP RPMs. + +%prep +%setup -q -c -T +install -pm 644 %{SOURCE0} . +install -pm 644 %{SOURCE1} . + +%build + +%install +rm -rf $RPM_BUILD_ROOT + +#GPG Key +install -Dpm 644 %{SOURCE0} \ + $RPM_BUILD_ROOT%{_sysconfdir}/pki/rpm-gpg/GPG-KEY-PG_PROBACKUP + +# yum /etc/zypp/repos.d/repo-update.repo + +%if 0%{?suse_version} + install -dm 755 $RPM_BUILD_ROOT%{_sysconfdir}/zypp/repos.d + install -pm 644 %{SOURCE1} $RPM_BUILD_ROOT%{_sysconfdir}/zypp/repos.d +%else + install -dm 755 $RPM_BUILD_ROOT%{_sysconfdir}/yum.repos.d + install -pm 644 %{SOURCE1} $RPM_BUILD_ROOT%{_sysconfdir}/yum.repos.d +%endif + +%clean +rm -rf $RPM_BUILD_ROOT + +%files +%defattr(-,root,root,-) +%if 0%{?suse_version} + %config(noreplace) /etc/zypp/repos.d/* +%else + %config(noreplace) /etc/yum.repos.d/* +%endif +/etc/pki/rpm-gpg/* + +%changelog +* Mon Jun 29 2020 Grigory Smolkin +- release update diff --git a/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup.alt.forks.spec b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup.alt.forks.spec new file mode 100644 index 000000000..cbb57e42a --- /dev/null +++ b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup.alt.forks.spec @@ -0,0 +1,67 @@ +%global version @PKG_VERSION@ +%global release @PKG_RELEASE@ +%global hash @PKG_HASH@ +%global pgsql_major @PG_VERSION@ +%global pgsql_full @PG_FULL_VERSION@ +%global edition @EDITION@ +%global edition_full @EDITION_FULL@ +%global prefix @PREFIX@ + +#%set_verify_elf_method unresolved=relaxed, rpath=relaxed +%set_verify_elf_method rpath=relaxed,unresolved=relaxed + +Name: pg_probackup-%{edition}-%{pgsql_major} +Version: %{version} +Release: %{release}.%{hash} +Summary: Backup utility for PostgresPro %{edition_full} +Group: Applications/Databases +License: BSD +Url: http://postgrespro.ru/ +#Source0: postgrespro-%{edition}-%{pgsql_full}.tar.bz2 +#Source1: pg_probackup-%{edition}-%{version}.tar.bz2 +Source0: postgrespro-%{edition}-%{pgsql_full} +Source1: pg_probackup-%{version} +BuildRequires: gcc make perl glibc-devel bison flex +BuildRequires: readline-devel openssl-devel gettext zlib-devel + + +%description +Backup tool for PostgresPro %{edition_full}. + +%prep +#%setup -q -b1 -n postgrespro-%{edition}-%{pgsql_full} +mv %{_topdir}/SOURCES/postgrespro-%{edition}-%{pgsql_full} %{_topdir}/BUILD +cd %{_topdir}/BUILD/postgrespro-%{edition}-%{pgsql_full} +mv %{_topdir}/SOURCES/pg_probackup-%{version} contrib/pg_probackup + +mkdir %{_topdir}/SOURCES/postgrespro-%{edition}-%{pgsql_full} +mkdir %{_topdir}/SOURCES/pg_probackup-%{edition}-%{version} +mkdir %{_topdir}/SOURCES/pg_probackup-%{version} + +%build +cd %{_topdir}/BUILD/postgrespro-%{edition}-%{pgsql_full} +%if "%{pgsql_major}" == "9.6" +./configure --enable-debug +%else +./configure --enable-debug --disable-online-upgrade --prefix=%{prefix} +%endif +make -C 'src/common' +make -C 'src/port' +make -C 'src/interfaces' +cd contrib/pg_probackup && make + +%install +cd %{_topdir}/BUILD/postgrespro-%{edition}-%{pgsql_full} +%{__mkdir} -p %{buildroot}%{_bindir} +%{__install} -p -m 755 contrib/pg_probackup/pg_probackup %{buildroot}%{_bindir}/%{name} + +%files +%{_bindir}/%{name} + +%clean +rm -rf $RPM_BUILD_ROOT + + +%changelog +* Mon Nov 17 2019 Grigory Smolkin - 2.2.6-1 +- Initial release. diff --git a/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup.alt.spec b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup.alt.spec new file mode 100644 index 000000000..3105ffa67 --- /dev/null +++ b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup.alt.spec @@ -0,0 +1,48 @@ +%global version @PKG_VERSION@ +%global release @PKG_RELEASE@ +%global hash @PKG_HASH@ +%global pgsql_major @PG_VERSION@ +%global pgsql_full @PG_FULL_VERSION@ +%set_verify_elf_method rpath=relaxed + +Name: pg_probackup-%{pgsql_major} +Version: %{version} +Release: %{release}.%{hash} +Summary: Backup utility for PostgreSQL +Group: Applications/Databases +License: BSD +Url: http://postgrespro.ru/ +Source0: http://ftp.postgresql.org/pub/source/v%{pgsql_full}/postgresql-%{pgsql_major}.tar.bz2 +Source1: pg_probackup-%{version}.tar.bz2 +BuildRequires: gcc make perl glibc-devel bison flex +BuildRequires: readline-devel openssl-devel gettext zlib-devel + + +%description +Backup tool for PostgreSQL. + +%prep +%setup -q -b1 -n postgresql-%{pgsql_full} + +%build +mv %{_builddir}/pg_probackup-%{version} contrib/pg_probackup +./configure --enable-debug --without-readline +make -C 'src/common' +make -C 'src/port' +make -C 'src/interfaces' +cd contrib/pg_probackup && make + +%install +%{__mkdir} -p %{buildroot}%{_bindir} +%{__install} -p -m 755 contrib/pg_probackup/pg_probackup %{buildroot}%{_bindir}/%{name} + +%files +%{_bindir}/%{name} + +%clean +rm -rf $RPM_BUILD_ROOT + + +%changelog +* Mon Nov 17 2019 Grigory Smolkin - 2.2.6-1 +- Initial release. diff --git a/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup.spec b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup.spec new file mode 100644 index 000000000..e5fb5ad48 --- /dev/null +++ b/packaging/pkg/specs/rpm/rpmbuild/SPECS/pg_probackup.spec @@ -0,0 +1,48 @@ +%global version @PKG_VERSION@ +%global release @PKG_RELEASE@ +%global hash @PKG_HASH@ +%global pgsql_major @PG_VERSION@ +%global pgsql_full @PG_FULL_VERSION@ + +Name: pg_probackup-%{pgsql_major} +Version: %{version} +Release: %{release}.%{hash} +Summary: Backup utility for PostgreSQL +Group: Applications/Databases +License: BSD +Url: http://postgrespro.ru/ +Source0: http://ftp.postgresql.org/pub/source/v%{pgsql_full}/postgresql-%{pgsql_major}.tar.bz2 +Source1: pg_probackup-%{version}.tar.bz2 +BuildRequires: gcc make perl glibc-devel openssl-devel gettext zlib-devel + +%description +Backup tool for PostgreSQL. + +%prep +%setup -q -b1 -n postgresql-%{pgsql_full} + +%build +mv %{_builddir}/pg_probackup-%{version} contrib/pg_probackup +./configure --enable-debug --without-readline +make -C 'src/common' +make -C 'src/port' +make -C 'src/interfaces' +cd contrib/pg_probackup && make + +%install +%{__mkdir} -p %{buildroot}%{_bindir} +%{__install} -p -m 755 contrib/pg_probackup/pg_probackup %{buildroot}%{_bindir}/%{name} + +%files +%{_bindir}/%{name} + +%clean +rm -rf $RPM_BUILD_ROOT + + +%changelog +* Wed Feb 9 2018 Grigory Smolkin - %{version}-%{release}.%{hash} +- @PKG_VERSION@ + +* Fri Jan 29 2018 Grigory Smolkin - 2.0.14-1 +- Initial release. diff --git a/packaging/pkg/tarballs/.gitkeep b/packaging/pkg/tarballs/.gitkeep new file mode 100644 index 000000000..e69de29bb diff --git a/packaging/repo/reprepro-conf/changelog.script b/packaging/repo/reprepro-conf/changelog.script new file mode 100755 index 000000000..4ff1f1787 --- /dev/null +++ b/packaging/repo/reprepro-conf/changelog.script @@ -0,0 +1,246 @@ +#!/bin/sh +# This is an example script that can be hooked into reprepro +# to either generate a hierachy like packages.debian.org/changelogs/ +# or to generate changelog files in the "third party sites" +# location apt-get changelogs looks if it is not found in +# Apt::Changelogs::Server. +# +# All you have to do is to: +# - copy it into you conf/ directory, +# - if you want "third party site" style changelogs, edit the +# CHANGELOGDIR variable below, +# and +# - add the following to any distribution in conf/distributions +# you want to have changelogs and copyright files extracted: +#Log: +# --type=dsc changelogs.example +# (note the space at the beginning of the second line). +# This will cause this script to extract changelogs for all +# newly added source packages. (To generate them for already +# existing packages, call "reprepro rerunnotifiers"). + +# DEPENDENCIES: dpkg >= 1.13.9 + +if test "x${REPREPRO_OUT_DIR:+set}" = xset ; then + # Note: due to cd, REPREPRO_*_DIR will no longer + # be usable. And only things relative to outdir will work... + cd "${REPREPRO_OUT_DIR}" || exit 1 +else + # this will also trigger if reprepro < 3.5.1 is used, + # in that case replace this with a manual cd to the + # correct directory... + cat "changelog.example needs to be run by reprepro!" >&2 + exit 1 +fi + +# CHANGELOGDIR set means generate full hierachy +# (clients need to set Apt::Changelogs::Server to use that) +#CHANGELOGDIR=changelogs + +# CHANGELOGDIR empty means generate changelog (and only changelog) files +# in the new "third party site" place apt-get changelog is using as fallback: +#CHANGELOGDIR= + +# Set to avoid using some predefined TMPDIR or even /tmp as +# tempdir: + +# TMPDIR=/var/cache/whateveryoucreated + +if test -z "$CHANGELOGDIR" ; then +addsource() { + DSCFILE="$1" + CANONDSCFILE="$(readlink --canonicalize "$DSCFILE")" + CHANGELOGFILE="${DSCFILE%.dsc}.changelog" + BASEDIR="$(dirname "$CHANGELOGFILE")" + if ! [ -f "$CHANGELOGFILE" ] ; then + EXTRACTDIR="$(mktemp -d)" + (cd -- "$EXTRACTDIR" && dpkg-source --no-copy -x "$CANONDSCFILE" > /dev/null) + install --mode=644 -- "$EXTRACTDIR"/*/debian/changelog "$CHANGELOGFILE" + chmod -R u+rwX -- "$EXTRACTDIR" + rm -r -- "$EXTRACTDIR" + fi + if [ -L "$BASEDIR"/current."$CODENAME" ] ; then + # should not be there, just to be sure + rm -f -- "$BASEDIR"/current."$CODENAME" + fi + # mark this as needed by this distribution + ln -s -- "$(basename "$CHANGELOGFILE")" "$BASEDIR/current.$CODENAME" + JUSTADDED="$CHANGELOGFILE" +} +delsource() { + DSCFILE="$1" + CHANGELOGFILE="${DSCFILE%.dsc}.changelog" + BASEDIR="$(dirname "$CHANGELOGFILE")" + BASENAME="$(basename "$CHANGELOGFILE")" + if [ "x$JUSTADDED" = "x$CHANGELOGFILE" ] ; then + exit 0 + fi +# echo "delete, basedir=$BASEDIR changelog=$CHANGELOGFILE, dscfile=$DSCFILE, " + if [ "x$(readlink "$BASEDIR/current.$CODENAME")" = "x$BASENAME" ] ; then + rm -- "$BASEDIR/current.$CODENAME" + fi + NEEDED=0 + for c in "$BASEDIR"/current.* ; do + if [ "x$(readlink -- "$c")" = "x$BASENAME" ] ; then + NEEDED=1 + fi + done + if [ "$NEEDED" -eq 0 -a -f "$CHANGELOGFILE" ] ; then + rm -r -- "$CHANGELOGFILE" + # to remove the directory if now empty + rmdir --ignore-fail-on-non-empty -- "$BASEDIR" + fi +} + +else # "$CHANGELOGDIR" set: + +addsource() { + DSCFILE="$1" + CANONDSCFILE="$(readlink --canonicalize "$DSCFILE")" + TARGETDIR="${CHANGELOGDIR}/${DSCFILE%.dsc}" + SUBDIR="$(basename $TARGETDIR)" + BASEDIR="$(dirname $TARGETDIR)" + if ! [ -d "$TARGETDIR" ] ; then + echo "extract $CANONDSCFILE information to $TARGETDIR" + mkdir -p -- "$TARGETDIR" + EXTRACTDIR="$(mktemp -d)" + (cd -- "$EXTRACTDIR" && dpkg-source --no-copy -x "$CANONDSCFILE" > /dev/null) + install --mode=644 -- "$EXTRACTDIR"/*/debian/copyright "$TARGETDIR/copyright" + install --mode=644 -- "$EXTRACTDIR"/*/debian/changelog "$TARGETDIR/changelog" + chmod -R u+rwX -- "$EXTRACTDIR" + rm -r -- "$EXTRACTDIR" + fi + if [ -L "$BASEDIR"/current."$CODENAME" ] ; then + # should not be there, just to be sure + rm -f -- "$BASEDIR"/current."$CODENAME" + fi + # mark this as needed by this distribution + ln -s -- "$SUBDIR" "$BASEDIR/current.$CODENAME" + JUSTADDED="$TARGETDIR" +} +delsource() { + DSCFILE="$1" + TARGETDIR="${CHANGELOGDIR}/${DSCFILE%.dsc}" + SUBDIR="$(basename $TARGETDIR)" + BASEDIR="$(dirname $TARGETDIR)" + if [ "x$JUSTADDED" = "x$TARGETDIR" ] ; then + exit 0 + fi +# echo "delete, basedir=$BASEDIR targetdir=$TARGETDIR, dscfile=$DSCFILE, " + if [ "x$(readlink "$BASEDIR/current.$CODENAME")" = "x$SUBDIR" ] ; then + rm -- "$BASEDIR/current.$CODENAME" + fi + NEEDED=0 + for c in "$BASEDIR"/current.* ; do + if [ "x$(readlink -- "$c")" = "x$SUBDIR" ] ; then + NEEDED=1 + fi + done + if [ "$NEEDED" -eq 0 -a -d "$TARGETDIR" ] ; then + rm -r -- "$TARGETDIR" + # to remove the directory if now empty + rmdir --ignore-fail-on-non-empty -- "$BASEDIR" + fi +} +fi # CHANGELOGDIR + +ACTION="$1" +CODENAME="$2" +PACKAGETYPE="$3" +if [ "x$PACKAGETYPE" != "xdsc" ] ; then +# the --type=dsc should cause this to never happen, but better safe than sorry. + exit 1 +fi +COMPONENT="$4" +ARCHITECTURE="$5" +if [ "x$ARCHITECTURE" != "xsource" ] ; then + exit 1 +fi +NAME="$6" +shift 6 +JUSTADDED="" +if [ "x$ACTION" = "xadd" -o "x$ACTION" = "xinfo" ] ; then + VERSION="$1" + shift + if [ "x$1" != "x--" ] ; then + exit 2 + fi + shift + while [ "$#" -gt 0 ] ; do + case "$1" in + *.dsc) + addsource "$1" + ;; + --) + exit 2 + ;; + esac + shift + done +elif [ "x$ACTION" = "xremove" ] ; then + OLDVERSION="$1" + shift + if [ "x$1" != "x--" ] ; then + exit 2 + fi + shift + while [ "$#" -gt 0 ] ; do + case "$1" in + *.dsc) + delsource "$1" + ;; + --) + exit 2 + ;; + esac + shift + done +elif [ "x$ACTION" = "xreplace" ] ; then + VERSION="$1" + shift + OLDVERSION="$1" + shift + if [ "x$1" != "x--" ] ; then + exit 2 + fi + shift + while [ "$#" -gt 0 -a "x$1" != "x--" ] ; do + case "$1" in + *.dsc) + addsource "$1" + ;; + esac + shift + done + if [ "x$1" != "x--" ] ; then + exit 2 + fi + shift + while [ "$#" -gt 0 ] ; do + case "$1" in + *.dsc) + delsource "$1" + ;; + --) + exit 2 + ;; + esac + shift + done +fi + +exit 0 +# Copyright 2007,2008,2012 Bernhard R. Link +# +# This program is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License version 2 as +# published by the Free Software Foundation. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program; if not, write to the Free Software +# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02111-1301 USA \ No newline at end of file diff --git a/packaging/repo/reprepro-conf/distributions b/packaging/repo/reprepro-conf/distributions new file mode 100644 index 000000000..7dce7e6d0 --- /dev/null +++ b/packaging/repo/reprepro-conf/distributions @@ -0,0 +1,179 @@ +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: squeeze +Architectures: amd64 i386 source +Components: main-squeeze +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: wheezy +Architectures: amd64 i386 source +Components: main-wheezy +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: jessie +Architectures: amd64 i386 source +Components: main-jessie +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: bullseye +Architectures: amd64 i386 source +Components: main-bullseye +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: wily +Architectures: amd64 i386 source +Components: main-wily +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: precise +Architectures: amd64 i386 source +Components: main-precise +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: vivid +Architectures: amd64 i386 source +Components: main-vivid +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: trusty +Architectures: amd64 i386 source +Components: main-trusty +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: lucid +Architectures: amd64 i386 source +Components: main-lucid +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: cosmic +Architectures: amd64 i386 source +Components: main-cosmic +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: xenial +Architectures: amd64 i386 source +Components: main-xenial +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: yakkety +Architectures: amd64 i386 source +Components: main-yakkety +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: zesty +Architectures: amd64 i386 source +Components: main-zesty +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: stretch +Architectures: amd64 i386 source +Components: main-stretch +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: buster +Architectures: amd64 i386 source +Components: main-buster +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: artful +Architectures: amd64 i386 source +Components: main-artful +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: bionic +Architectures: amd64 i386 source +Components: main-bionic +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script + +Origin: repo.postgrespro.ru +Label: PostgreSQL backup utility pg_probackup +Codename: focal +Architectures: amd64 i386 source +Components: main-focal +Description: PostgresPro pg_probackup repo +SignWith: yes +Log: + --type=dsc changelog.script diff --git a/packaging/repo/rpm-conf/rpmmacros b/packaging/repo/rpm-conf/rpmmacros new file mode 100644 index 000000000..e00a76d68 --- /dev/null +++ b/packaging/repo/rpm-conf/rpmmacros @@ -0,0 +1,5 @@ +%_signature gpg +%_gpg_path /root/.gnupg +%_gpg_name PostgreSQL Professional +%_gpgbin /usr/bin/gpg +%_gpg_check_password_cmd /bin/true diff --git a/packaging/repo/scripts/alt.sh b/packaging/repo/scripts/alt.sh new file mode 100755 index 000000000..4cda313ef --- /dev/null +++ b/packaging/repo/scripts/alt.sh @@ -0,0 +1,62 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -exu +set -o errexit +set -o pipefail + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 + +export INPUT_DIR=/app/in #dir with builded rpm +export OUT_DIR=/app/www/${PBK_PKG_REPO} + +apt-get update -y +apt-get install -qq -y apt-repo-tools gnupg rsync perl less wget + +if [[ ${PBK_EDITION} == '' ]] ; then + REPO_SUFFIX='vanilla' + FORK='PostgreSQL' +else + REPO_SUFFIX='forks' + FORK='PostgresPro' +fi + +cd $INPUT_DIR + +cp -arv /app/repo/$PBK_PKG_REPO/gnupg /root/.gnupg +chmod -R 0600 /root/.gnupg +for pkg in $(ls); do + for pkg_full_version in $(ls ./$pkg); do + + # THere is no std/ent packages for PG 9.5 + if [[ ${pkg} == 'pg_probackup-std-9.5' ]] || [[ ${pkg} == 'pg_probackup-ent-9.5' ]] ; then + continue; + fi + + RPM_DIR=${OUT_DIR}/rpm/${pkg_full_version}/altlinux-p${DISTRIB_VERSION}/x86_64/RPMS.${REPO_SUFFIX} + mkdir -p "$RPM_DIR" + cp -arv $INPUT_DIR/$pkg/$pkg_full_version/RPMS/x86_64/* $RPM_DIR/ + + genbasedir --architecture=x86_64 --architectures=x86_64 --origin=repo.postgrespro.ru \ + --label="${FORK} backup utility pg_probackup" --description "${FORK} pg_probackup repo" \ + --version=$pkg_full_version --bloat --progress --create \ + --topdir=${OUT_DIR}/rpm/${pkg_full_version}/altlinux-p${DISTRIB_VERSION} x86_64 ${REPO_SUFFIX} + + # SRPM is available only for vanilla + if [[ ${PBK_EDITION} == '' ]] ; then + SRPM_DIR=${OUT_DIR}/srpm/${pkg_full_version}/altlinux-p${DISTRIB_VERSION}/x86_64/SRPMS.${REPO_SUFFIX} + mkdir -p "$SRPM_DIR" + cp -arv $INPUT_DIR/$pkg/$pkg_full_version/SRPMS/* $SRPM_DIR/ + + genbasedir --architecture=x86_64 --architectures=x86_64 --origin=repo.postgrespro.ru \ + --label="${FORK} backup utility pg_probackup sources" --description "${FORK} pg_probackup repo" \ + --version=$pkg_full_version --bloat --progress --create \ + --topdir=${OUT_DIR}/srpm/${pkg_full_version}/altlinux-p${DISTRIB_VERSION} x86_64 ${REPO_SUFFIX} + fi + done +done diff --git a/packaging/repo/scripts/deb.sh b/packaging/repo/scripts/deb.sh new file mode 100755 index 000000000..31416972d --- /dev/null +++ b/packaging/repo/scripts/deb.sh @@ -0,0 +1,51 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -exu +set -o errexit +set -o pipefail + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 + +export INPUT_DIR=/app/in # dir with builded deb +export OUT_DIR=/app/www/${PBK_PKG_REPO} +#export REPO_DIR=/app/repo + +cd $INPUT_DIR + +export DEB_DIR=$OUT_DIR/deb +export KEYS_DIR=$OUT_DIR/keys +export CONF=/app/repo/reprepro-conf +mkdir -p "$KEYS_DIR" +cp -av /app/repo/${PBK_PKG_REPO}/gnupg /root/.gnupg + +rsync /app/repo/${PBK_PKG_REPO}/gnupg/key.public $KEYS_DIR/GPG-KEY-PG_PROBACKUP +echo -e 'User-agent: *\nDisallow: /' > $OUT_DIR/robots.txt + +mkdir -p $DEB_DIR +cd $DEB_DIR +cp -av $CONF ./conf + +# make remove-debpkg tool +echo -n "#!" > remove-debpkg +echo "/bin/sh" >> remove-debpkg +echo "CODENAME=\$1" >> remove-debpkg +echo "DEBFILE=\$2" >> remove-debpkg +echo "DEBNAME=\`basename \$DEBFILE | sed -e 's/_.*//g'\`" >> remove-debpkg +echo "reprepro --waitforlock 5 remove \$CODENAME \$DEBNAME" >> remove-debpkg +chmod +x remove-debpkg + +#find $INPUT_DIR/ -name '*.changes' -exec reprepro -P optional -Vb . include ${CODENAME} {} \; +find $INPUT_DIR -name "*${CODENAME}*.deb" -exec ./remove-debpkg $CODENAME {} \; +find $INPUT_DIR -name "*${CODENAME}*.dsc" -exec reprepro --waitforlock 5 -i undefinedtarget --ignore=missingfile -P optional -S main -Vb . includedsc $CODENAME {} \; +find $INPUT_DIR -name "*${CODENAME}*.deb" -exec reprepro --waitforlock 5 -i undefinedtarget --ignore=missingfile -P optional -Vb . includedeb $CODENAME {} \; +reprepro export $CODENAME + +rm -f remove-debpkg +rm -rf ./conf +rm -rf /root/.gnupg diff --git a/packaging/repo/scripts/rpm.sh b/packaging/repo/scripts/rpm.sh new file mode 100755 index 000000000..524789cbf --- /dev/null +++ b/packaging/repo/scripts/rpm.sh @@ -0,0 +1,58 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -ex +set -o errexit +set -o pipefail + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 + +export INPUT_DIR=/app/in #dir with builded rpm +export OUT_DIR=/app/www/${PBK_PKG_REPO} +export KEYS_DIR=$OUT_DIR/keys + +# deploy keys +mkdir -p "$KEYS_DIR" +rsync /app/repo/$PBK_PKG_REPO/gnupg/key.public $KEYS_DIR/GPG-KEY-PG_PROBACKUP +chmod 755 $KEYS_DIR +echo -e 'User-agent: *\nDisallow: /' > $OUT_DIR/robots.txt + +cd $INPUT_DIR + +cp -arv /app/repo/rpm-conf/rpmmacros /root/.rpmmacros +cp -arv /app/repo/$PBK_PKG_REPO/gnupg /root/.gnupg +chmod -R 0600 /root/.gnupg +chown -R root:root /root/.gnupg + +for pkg in $(ls ${INPUT_DIR}); do + for pkg_full_version in $(ls ${INPUT_DIR}/$pkg); do + if [[ ${PBK_EDITION} == '' ]] ; then + cp $INPUT_DIR/$pkg/$pkg_full_version/RPMS/noarch/pg_probackup-repo-*.noarch.rpm \ + $KEYS_DIR/pg_probackup-repo-$DISTRIB.noarch.rpm + else + cp $INPUT_DIR/$pkg/$pkg_full_version/RPMS/noarch/pg_probackup-repo-*.noarch.rpm \ + $KEYS_DIR/pg_probackup-repo-forks-$DISTRIB.noarch.rpm + fi + + [ ! -z "$CODENAME" ] && export DISTRIB_VERSION=$CODENAME + RPM_DIR=$OUT_DIR/rpm/$pkg_full_version/${DISTRIB}-${DISTRIB_VERSION}-x86_64 + mkdir -p "$RPM_DIR" + cp -arv $INPUT_DIR/$pkg/$pkg_full_version/RPMS/x86_64/* $RPM_DIR/ + for f in $(ls $RPM_DIR/*.rpm); do rpm --addsign $f || exit 1; done + createrepo $RPM_DIR/ + + if [[ ${PBK_EDITION} == '' ]] ; then + SRPM_DIR=$OUT_DIR/srpm/$pkg_full_version/${DISTRIB}-${DISTRIB_VERSION}-x86_64 + mkdir -p "$SRPM_DIR" + cp -arv $INPUT_DIR/$pkg/$pkg_full_version/SRPMS/* $SRPM_DIR/ + for f in $(ls $SRPM_DIR/*.rpm); do rpm --addsign $f || exit 1; done + createrepo $SRPM_DIR/ + fi + + done +done diff --git a/packaging/repo/scripts/suse.sh b/packaging/repo/scripts/suse.sh new file mode 100755 index 000000000..c85e0ff10 --- /dev/null +++ b/packaging/repo/scripts/suse.sh @@ -0,0 +1,72 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -ex +set -o errexit +set -o pipefail + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 + +# currenctly we do not build std|ent packages for Suse +if [[ ${PBK_EDITION} != '' ]] ; then + exit 0 +fi + +export INPUT_DIR=/app/in #dir with builded rpm +export OUT_DIR=/app/www/${PBK_PKG_REPO} +export KEYS_DIR=$OUT_DIR/keys +# deploy keys + +zypper install -y createrepo +rm -rf /root/.gnupg + +cd $INPUT_DIR + +mkdir -p $KEYS_DIR +chmod 755 $KEYS_DIR +rsync /app/repo/$PBK_PKG_REPO/gnupg/key.public $KEYS_DIR/GPG-KEY-PG_PROBACKUP + +echo -e 'User-agent: *\nDisallow: /' > $OUT_DIR/robots.txt + +cp -arv /app/repo/rpm-conf/rpmmacros /root/.rpmmacros +cp -arv /app/repo/$PBK_PKG_REPO/gnupg /root/.gnupg +chmod -R 0600 /root/.gnupg + +for pkg in $(ls); do + for pkg_full_version in $(ls ./$pkg); do + + cp $INPUT_DIR/$pkg/$pkg_full_version/RPMS/noarch/pg_probackup-repo-*.noarch.rpm \ + $KEYS_DIR/pg_probackup-repo-$DISTRIB.noarch.rpm + [ ! -z "$CODENAME" ] && export DISTRIB_VERSION=$CODENAME + RPM_DIR=$OUT_DIR/rpm/$pkg_full_version/${DISTRIB}-${DISTRIB_VERSION}-x86_64 + SRPM_DIR=$OUT_DIR/srpm/$pkg_full_version/${DISTRIB}-${DISTRIB_VERSION}-x86_64 + + # rm -rf "$RPM_DIR" && mkdir -p "$RPM_DIR" + # rm -rf "$SRPM_DIR" && mkdir -p "$SRPM_DIR" + mkdir -p "$RPM_DIR" + mkdir -p "$SRPM_DIR" + + cp -arv $INPUT_DIR/$pkg/$pkg_full_version/RPMS/x86_64/* $RPM_DIR/ + cp -arv $INPUT_DIR/$pkg/$pkg_full_version/SRPMS/* $SRPM_DIR/ + + for f in $(ls $RPM_DIR/*.rpm); do rpm --addsign $f || exit 1; done + for f in $(ls $SRPM_DIR/*.rpm); do rpm --addsign $f || exit 1; done + + createrepo $RPM_DIR/ + createrepo $SRPM_DIR/ + + # rpm --addsign $RPM_DIR/repodata/repomd.xml + # rpm --addsign $SRPM_DIR/repodata/repomd.xml + + gpg --batch --yes -a --detach-sign $RPM_DIR/repodata/repomd.xml + gpg --batch --yes -a --detach-sign $SRPM_DIR/repodata/repomd.xml + + cp -a /root/.gnupg/key.public $RPM_DIR/repodata/repomd.xml.key + cp -a /root/.gnupg/key.public $SRPM_DIR/repodata/repomd.xml.key + done +done diff --git a/packaging/test/Makefile.alt b/packaging/test/Makefile.alt new file mode 100644 index 000000000..bd00cef7f --- /dev/null +++ b/packaging/test/Makefile.alt @@ -0,0 +1,50 @@ +# ALT 9 +build/test_alt_9_9.6: + $(call test_alt,alt,9,,9.6,9.6.24) + touch build/test_alt_9_9.6 + +build/test_alt_9_10: + $(call test_alt,alt,9,,10,10.19) + touch build/test_alt_9_10 + +build/test_alt_9_11: + $(call test_alt,alt,9,,11,11.14) + touch build/test_alt_9_11 + +build/test_alt_9_12: + $(call test_alt,alt,9,,12,12.9) + touch build/test_alt_9_12 + +build/test_alt_9_13: + $(call test_alt,alt,9,,13,13.5) + touch build/test_alt_9_13 + +build/test_alt_9_14: + $(call test_alt,alt,9,,14,14.1) + touch build/test_alt_9_14 + +# ALT 8 +build/test_alt_8_9.6: + $(call test_alt,alt,8,,9.6,9.6.24) + touch build/test_alt_8_9.6 + +build/test_alt_8_10: + $(call test_alt,alt,8,,10,10.19) + touch build/test_alt_8_10 + +build/test_alt_8_11: + $(call test_alt,alt,8,,11,11.14) + touch build/test_alt_8_11 + +build/test_alt_8_12: + $(call test_alt,alt,8,,12,12.9) + touch build/test_alt_8_12 + +build/test_alt_8_13: + $(call test_alt,alt,8,,13,13.5) + touch build/test_alt_8_13 + +build/test_alt_8_14: + $(call test_alt,alt,8,,14,14.1) + touch build/test_alt_8_14 + diff --git a/packaging/test/Makefile.centos b/packaging/test/Makefile.centos new file mode 100644 index 000000000..9d30a324b --- /dev/null +++ b/packaging/test/Makefile.centos @@ -0,0 +1,49 @@ +# CENTOS 7 +build/test_centos_7_9.6: + $(call test_rpm,centos,7,,9.6,9.6.24) + touch build/test_centos_7_9.6 + +build/test_centos_7_10: + $(call test_rpm,centos,7,,10,10.19) + touch build/test_centos_7_10 + +build/test_centos_7_11: + $(call test_rpm,centos,7,,11,11.14) + touch build/test_centos_7_11 + +build/test_centos_7_12: + $(call test_rpm,centos,7,,12,12.9) + touch build/test_centos_7_12 + +build/test_centos_7_13: + $(call test_rpm,centos,7,,13,13.5) + touch build/test_centos_7_13 + +build/test_centos_7_14: + $(call test_rpm,centos,7,,14,14.1) + touch build/test_centos_7_14 + +# CENTOS 8 +build/test_centos_8_9.6: + $(call test_rpm,centos,8,,9.6,9.6.24) + touch build/test_centos_8_9.6 + +build/test_centos_8_10: + $(call test_rpm,centos,8,,10,10.19) + touch build/test_centos_8_10 + +build/test_centos_8_11: + $(call test_rpm,centos,8,,11,11.14) + touch build/test_centos_8_11 + +build/test_centos_8_12: + $(call test_rpm,centos,8,,12,12.9) + touch build/test_centos_8_12 + +build/test_centos_8_13: + $(call test_rpm,centos,8,,13,13.5) + touch build/test_centos_8_13 + +build/test_centos_8_14: + $(call test_rpm,centos,8,,14,14.1) + touch build/test_centos_8_14 diff --git a/packaging/test/Makefile.debian b/packaging/test/Makefile.debian new file mode 100644 index 000000000..e4d904f62 --- /dev/null +++ b/packaging/test/Makefile.debian @@ -0,0 +1,74 @@ +# DEBIAN 9 +build/test_debian_9_9.6: + $(call test_deb,debian,9,stretch,9.6,9.6.24) + touch build/test_debian_9_9.6 + +build/test_debian_9_10: + $(call test_deb,debian,9,stretch,10,10.19) + touch build/test_debian_9_10 + +build/test_debian_9_11: + $(call test_deb,debian,9,stretch,11,11.14) + touch build/test_debian_9_11 + +build/test_debian_9_12: + $(call test_deb,debian,9,stretch,12,12.9) + touch build/test_debian_9_12 + +build/test_debian_9_13: + $(call test_deb,debian,9,stretch,13,13.5) + touch build/test_debian_9_13 + +build/test_debian_9_14: + $(call test_deb,debian,9,stretch,14,14.1) + touch build/test_debian_9_14 + +# DEBIAN 10 +build/test_debian_10_9.6: + $(call test_deb,debian,10,buster,9.6,9.6.24) + touch build/test_debian_10_9.6 + +build/test_debian_10_10: + $(call test_deb,debian,10,buster,10,10.19) + touch build/test_debian_10_10 + +build/test_debian_10_11: + $(call test_deb,debian,10,buster,11,11.14) + touch build/test_debian_10_11 + +build/test_debian_10_12: + $(call test_deb,debian,10,buster,12,12.9) + touch build/test_debian_10_12 + +build/test_debian_10_13: + $(call test_deb,debian,10,buster,13,13.5) + touch build/test_debian_10_13 + +build/test_debian_10_14: + $(call test_deb,debian,10,buster,14,14.1) + touch build/test_debian_10_14 + +# DEBIAN 11 +build/test_debian_11_9.6: + $(call test_deb,debian,11,bullseye,9.6,9.6.24) + touch build/test_debian_11_9.6 + +build/test_debian_11_10: + $(call test_deb,debian,11,bullseye,10,10.19) + touch build/test_debian_11_10 + +build/test_debian_11_11: + $(call test_deb,debian,11,bullseye,11,11.14) + touch build/test_debian_11_11 + +build/test_debian_11_12: + $(call test_deb,debian,11,bullseye,12,12.9) + touch build/test_debian_11_12 + +build/test_debian_11_13: + $(call test_deb,debian,11,bullseye,13,13.5) + touch build/test_debian_11_13 + +build/test_debian_11_14: + $(call test_deb,debian,11,bullseye,14,14.1) + touch build/test_debian_11_14 diff --git a/packaging/test/Makefile.oraclelinux b/packaging/test/Makefile.oraclelinux new file mode 100644 index 000000000..0efe6574d --- /dev/null +++ b/packaging/test/Makefile.oraclelinux @@ -0,0 +1,49 @@ +# ORACLE LINUX 7 +build/test_oraclelinux_7_9.6: + $(call test_rpm,oraclelinux,7,,9.6,9.6.24) + touch build/test_oraclelinux_7_9.6 + +build/test_oraclelinux_7_10: + $(call test_rpm,oraclelinux,7,,10,10.19) + touch build/test_oraclelinux_7_10 + +build/test_oraclelinux_7_11: + $(call test_rpm,oraclelinux,7,,11,11.14) + touch build/test_oraclelinux_7_11 + +build/test_oraclelinux_7_12: + $(call test_rpm,oraclelinux,7,,12,12.9) + touch build/test_oraclelinux_7_12 + +build/test_oraclelinux_7_13: + $(call test_rpm,oraclelinux,7,,13,13.5) + touch build/test_oraclelinux_7_13 + +build/test_oraclelinux_7_14: + $(call test_rpm,oraclelinux,7,,14,14.1) + touch build/test_oraclelinux_7_14 + +# ORACLE LINUX 8 +build/test_oraclelinux_8_9.6: + $(call test_rpm,oraclelinux,8,,9.6,9.6.24) + touch build/test_oraclelinux_8_9.6 + +build/test_oraclelinux_8_10: + $(call test_rpm,oraclelinux,8,,10,10.19) + touch build/test_oraclelinux_8_10 + +build/test_oraclelinux_8_11: + $(call test_rpm,oraclelinux,8,,11,11.14) + touch build/test_oraclelinux_8_11 + +build/test_oraclelinux_8_12: + $(call test_rpm,oraclelinux,8,,12,12.9) + touch build/test_oraclelinux_8_12 + +build/test_oraclelinux_8_13: + $(call test_rpm,oraclelinux,8,,13,13.5) + touch build/test_oraclelinux_8_13 + +build/test_oraclelinux_8_14: + $(call test_rpm,oraclelinux,8,,14,14.1) + touch build/test_oraclelinux_8_14 diff --git a/packaging/test/Makefile.rhel b/packaging/test/Makefile.rhel new file mode 100644 index 000000000..3b26c8942 --- /dev/null +++ b/packaging/test/Makefile.rhel @@ -0,0 +1,49 @@ +# RHEL 7 +build/test_rhel_7_9.6: + $(call test_rpm,rhel,7,7Server,9.6,9.6.24) + touch build/test_rhel_7_9.6 + +build/test_rhel_7_10: + $(call test_rpm,rhel,7,7Server,10,10.19) + touch build/test_rhel_7_10 + +build/test_rhel_7_11: + $(call test_rpm,rhel,7,7Server,11,11.14) + touch build/test_rhel_7_11 + +build/test_rhel_7_12: + $(call test_rpm,rhel,7,7Server,12,12.9) + touch build/test_rhel_7_12 + +build/test_rhel_7_13: + $(call test_rpm,rhel,7,7Server,13,13.5) + touch build/test_rhel_7_13 + +build/test_rhel_7_14: + $(call test_rpm,rhel,7,7Server,14,14.1) + touch build/test_rhel_7_14 + +# RHEL 8 +build/test_rhel_8_9.6: + $(call test_rpm,rhel,8,8Server,9.6,9.6.24) + touch build/test_rhel_8_9.6 + +build/test_rhel_8_10: + $(call test_rpm,rhel,8,8Server,10,10.19) + touch build/test_rhel_8_10 + +build/test_rhel_8_11: + $(call test_rpm,rhel,8,8Server,11,11.14) + touch build/test_rhel_8_11 + +build/test_rhel_8_12: + $(call test_rpm,rhel,8,8Server,12,12.9) + touch build/test_rhel_8_12 + +build/test_rhel_8_13: + $(call test_rpm,rhel,8,8Server,13,13.5) + touch build/test_rhel_8_13 + +build/test_rhel_8_14: + $(call test_rpm,rhel,8,8Server,14,14.1) + touch build/test_rhel_8_14 diff --git a/packaging/test/Makefile.suse b/packaging/test/Makefile.suse new file mode 100644 index 000000000..19e8d52d8 --- /dev/null +++ b/packaging/test/Makefile.suse @@ -0,0 +1,49 @@ +# Suse 15.1 +build/test_suse_15.1_9.6: + $(call test_suse,suse,15.1,,9.6,9.6.24) + touch build/test_suse_15.1_9.6 + +build/test_suse_15.1_10: + $(call test_suse,suse,15.1,,10,10.19) + touch build/test_suse_15.1_10 + +build/test_suse_15.1_11: + $(call test_suse,suse,15.1,,11,11.14) + touch build/test_suse_15.1_11 + +build/test_suse_15.1_12: + $(call test_suse,suse,15.1,,12,12.9) + touch build/test_suse_15.1_12 + +build/test_suse_15.1_13: + $(call test_suse,suse,15.1,,13,13.5) + touch build/test_suse_15.1_13 + +build/test_suse_15.1_14: + $(call test_suse,suse,15.1,,14,14.1) + touch build/test_suse_15.1_14 + +# Suse 15.2 +build/test_suse_15.2_9.6: + $(call test_suse,suse,15.2,,9.6,9.6.24) + touch build/test_suse_15.2_9.6 + +build/test_suse_15.2_10: + $(call test_suse,suse,15.2,,10,10.19) + touch build/test_suse_15.2_10 + +build/test_suse_15.2_11: + $(call test_suse,suse,15.2,,11,11.14) + touch build/test_suse_15.2_11 + +build/test_suse_15.2_12: + $(call test_suse,suse,15.2,,12,12.9) + touch build/test_suse_15.2_12 + +build/test_suse_15.2_13: + $(call test_suse,suse,15.2,,13,13.5) + touch build/test_suse_15.2_13 + +build/test_suse_15.2_14: + $(call test_suse,suse,15.2,,14,14.1) + touch build/test_suse_15.2_14 diff --git a/packaging/test/Makefile.ubuntu b/packaging/test/Makefile.ubuntu new file mode 100644 index 000000000..86a257b91 --- /dev/null +++ b/packaging/test/Makefile.ubuntu @@ -0,0 +1,49 @@ +# UBUNTU 18.04 +build/test_ubuntu_18.04_9.6: + $(call test_deb,ubuntu,18.04,bionic,9.6,9.6.24) + touch build/test_ubuntu_18.04_9.6 + +build/test_ubuntu_18.04_10: + $(call test_deb,ubuntu,18.04,bionic,10,10.19) + touch build/test_ubuntu_18.04_10 + +build/test_ubuntu_18.04_11: + $(call test_deb,ubuntu,18.04,bionic,11,11.14) + touch build/test_ubuntu_18.04_11 + +build/test_ubuntu_18.04_12: + $(call test_deb,ubuntu,18.04,bionic,12,12.9) + touch build/test_ubuntu_18.04_12 + +build/test_ubuntu_18.04_13: + $(call test_deb,ubuntu,18.04,bionic,13,13.5) + touch build/test_ubuntu_18.04_13 + +build/test_ubuntu_18.04_14: + $(call test_deb,ubuntu,18.04,bionic,14,14.1) + touch build/test_ubuntu_18.04_14 + +# UBUNTU 20.04 +build/test_ubuntu_20.04_9.6: + $(call test_deb,ubuntu,20.04,focal,9.6,9.6.24) + touch build/test_ubuntu_20.04_9.6 + +build/test_ubuntu_20.04_10: + $(call test_deb,ubuntu,20.04,focal,10,10.19) + touch build/test_ubuntu_20.04_10 + +build/test_ubuntu_20.04_11: + $(call test_deb,ubuntu,20.04,focal,11,11.14) + touch build/test_ubuntu_20.04_11 + +build/test_ubuntu_20.04_12: + $(call test_deb,ubuntu,20.04,focal,12,12.9) + touch build/test_ubuntu_20.04_12 + +build/test_ubuntu_20.04_13: + $(call test_deb,ubuntu,20.04,focal,13,13.5) + touch build/test_ubuntu_20.04_13 + +build/test_ubuntu_20.04_14: + $(call test_deb,ubuntu,20.04,focal,14,14.1) + touch build/test_ubuntu_20.04_14 diff --git a/packaging/test/scripts/alt.sh b/packaging/test/scripts/alt.sh new file mode 100755 index 000000000..262864474 --- /dev/null +++ b/packaging/test/scripts/alt.sh @@ -0,0 +1,72 @@ +#!/usr/bin/env bash + +set -xe +set -o pipefail + +ulimit -n 1024 + +apt-get clean -y +apt-get update -y +apt-get install nginx su -y + +adduser nginx + +cat < /etc/nginx/nginx.conf +user nginx; +worker_processes 1; +error_log /var/log/nginx/error.log; +events { + worker_connections 1024; +} +http { + server { + listen 80 default; + root /app/www; + } +} +EOF + +/etc/init.d/nginx start + +# install POSTGRESQL + +export PGDATA=/var/lib/pgsql/${PG_VERSION}/data + +# install old packages +echo "rpm http://repo.postgrespro.ru/pg_probackup/rpm/latest/altlinux-p7 x86_64 vanilla" > /etc/apt/sources.list.d/pg_probackup.list +apt-get update +apt-get install ${PKG_NAME} -y +${PKG_NAME} --help +${PKG_NAME} --version + +# install new packages +echo "127.0.0.1 repo.postgrespro.ru" >> /etc/hosts +echo "rpm http://repo.postgrespro.ru/pg_probackup/rpm/latest/altlinux-p${DISTRIB_VERSION} x86_64 vanilla" > /etc/apt/sources.list.d/pg_probackup.list +echo "rpm [p${DISTRIB_VERSION}] http://mirror.yandex.ru/altlinux p${DISTRIB_VERSION}/branch/x86_64 debuginfo" > /etc/apt/sources.list.d/debug.list + +apt-get update -y +apt-get install ${PKG_NAME} -y +${PKG_NAME} --help +${PKG_NAME} --version + +exit 0 + +# TODO: run init, add-instance, backup and restore +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pgbench --no-vacuum -t 1000 -c 1" +su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA} --no-sync --compress" +su postgres -c "${PKG_NAME} show --instance=node -B /tmp/backup -D ${PGDATA}" + +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl stop -D ${PGDATA}" +rm -rf ${PGDATA} + +su postgres -c "${PKG_NAME} restore --instance=node -B /tmp/backup -D ${PGDATA} --no-sync" +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl start -w -D ${PGDATA}" + +sleep 5 + +echo "select count(*) from pgbench_accounts;" | su postgres -c "/usr/pgsql-${PG_VERSION}/bin/psql" || exit 1 + +exit 0 # while PG12 is not working + +# SRC PACKAGE +cd /mnt diff --git a/packaging/test/scripts/alt_forks.sh b/packaging/test/scripts/alt_forks.sh new file mode 100755 index 000000000..c406e5358 --- /dev/null +++ b/packaging/test/scripts/alt_forks.sh @@ -0,0 +1,75 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -xe +set -o pipefail + +ulimit -n 1024 + +if [ ${PBK_EDITION} == 'ent' ]; then + exit 0 +fi + +apt-get clean -y +apt-get update -y +apt-get install nginx su -y + +adduser nginx + +cat < /etc/nginx/nginx.conf +user nginx; +worker_processes 1; +error_log /var/log/nginx/error.log; +events { + worker_connections 1024; +} +http { + server { + listen 80 default; + root /app/www; + } +} +EOF + +/etc/init.d/nginx start + +# install POSTGRESQL + +export PGDATA=/var/lib/pgsql/${PG_VERSION}/data + +# install old packages + +# install new packages +echo "127.0.0.1 repo.postgrespro.ru" >> /etc/hosts +echo "rpm http://repo.postgrespro.ru/pg_probackup-forks/rpm/latest/altlinux-p${DISTRIB_VERSION} x86_64 forks" > /etc/apt/sources.list.d/pg_probackup.list +echo "rpm [p${DISTRIB_VERSION}] http://mirror.yandex.ru/altlinux p${DISTRIB_VERSION}/branch/x86_64 debuginfo" > /etc/apt/sources.list.d/debug.list + +apt-get update -y +apt-get install ${PKG_NAME} ${PKG_NAME}-debuginfo -y +${PKG_NAME} --help +${PKG_NAME} --version + +exit 0 + +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pgbench --no-vacuum -t 1000 -c 1" +su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA}" +su postgres -c "${PKG_NAME} show --instance=node -B /tmp/backup -D ${PGDATA}" + +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl stop -D ${PGDATA}" +rm -rf ${PGDATA} + +su postgres -c "${PKG_NAME} restore --instance=node -B /tmp/backup -D ${PGDATA}" +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl start -w -D ${PGDATA}" + +sleep 5 + +echo "select count(*) from pgbench_accounts;" | su postgres -c "/usr/pgsql-${PG_VERSION}/bin/psql" || exit 1 + +exit 0 # while PG12 is not working + +# SRC PACKAGE +cd /mnt diff --git a/packaging/test/scripts/deb.sh b/packaging/test/scripts/deb.sh new file mode 100755 index 000000000..fca9a23d8 --- /dev/null +++ b/packaging/test/scripts/deb.sh @@ -0,0 +1,143 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2021 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -xe +set -o pipefail + +ulimit -n 1024 + +PG_TOG=$(echo $PG_VERSION | sed 's|\.||g') + +# upgrade and utils +# export parameters +export DEBIAN_FRONTEND=noninteractive +echo 'debconf debconf/frontend select Noninteractive' | debconf-set-selections + +if [ ${DISTRIB} = 'ubuntu' -a ${CODENAME} = 'xenial' ] ; then + apt-get -qq update +elif [ ${DISTRIB} = 'debian' -a ${CODENAME} = 'stretch' ] ; then + apt-get -qq update +else + apt-get -qq --allow-releaseinfo-change update +fi + +apt-get -qq install -y wget nginx gnupg lsb-release +#apt-get -qq install -y libterm-readline-gnu-perl dialog gnupg procps + +# echo -e 'Package: *\nPin: origin test.postgrespro.ru\nPin-Priority: 800' >\ +# /etc/apt/preferences.d/pgpro-800 + +# install nginx +echo "127.0.0.1 test.postgrespro.ru" >> /etc/hosts +cat < /etc/nginx/nginx.conf +user www-data; +worker_processes 1; +error_log /var/log/nginx/error.log; +events { + worker_connections 1024; +} +http { + server { + listen 80 default; + root /app/www; + } +} +EOF +nginx -s reload || (pkill -9 nginx || nginx -c /etc/nginx/nginx.conf &) + +# install POSTGRESQL +#if [ ${CODENAME} == 'precise' ] && [ ${PG_VERSION} != '10' ] && [ ${PG_VERSION} != '11' ]; then + sh -c 'echo "deb http://apt.postgresql.org/pub/repos/apt/ $(lsb_release -cs)-pgdg main" > /etc/apt/sources.list.d/pgdg.list' + wget --no-check-certificate -O - https://www.postgresql.org/media/keys/ACCC4CF8.asc | apt-key add - + apt-get update -y + apt-get install -y postgresql-${PG_VERSION} +#fi + +# install pg_probackup from current public repo +echo "deb [arch=amd64] http://repo.postgrespro.ru/pg_probackup/deb/ $(lsb_release -cs) main-$(lsb_release -cs)" >\ + /etc/apt/sources.list.d/pg_probackup-old.list +wget -O - http://repo.postgrespro.ru/pg_probackup/keys/GPG-KEY-PG_PROBACKUP | apt-key add - && apt-get update + +apt-get install -y pg-probackup-${PG_VERSION} +pg_probackup-${PG_VERSION} --help +pg_probackup-${PG_VERSION} --version + +# Artful do no have PostgreSQL packages at all, Precise do not have PostgreSQL 10 +#if [ ${CODENAME} == 'precise' ] && [ ${PG_VERSION} != '10' ] && [ ${PG_VERSION} != '11' ]; then + export PGDATA=/var/lib/postgresql/${PG_VERSION}/data + su postgres -c "/usr/lib/postgresql/${PG_VERSION}/bin/initdb -k -D ${PGDATA}" + su postgres -c "pg_probackup-${PG_VERSION} init -B /tmp/backup" + su postgres -c "pg_probackup-${PG_VERSION} add-instance --instance=node -B /tmp/backup -D ${PGDATA}" + + echo "wal_level=hot_standby" >> ${PGDATA}/postgresql.auto.conf + echo "fsync=off" >> ${PGDATA}/postgresql.auto.conf + echo "archive_mode=on" >> ${PGDATA}/postgresql.auto.conf + echo "archive_command='pg_probackup-${PG_VERSION} archive-push --no-sync -B /tmp/backup compress --instance=node --wal-file-path %p --wal-file-name %f'" >> ${PGDATA}/postgresql.auto.conf + + su postgres -c "/usr/lib/postgresql/${PG_VERSION}/bin/pg_ctl start -D ${PGDATA}" + sleep 5 + su postgres -c "pg_probackup-${PG_VERSION} backup --instance=node -b full -B /tmp/backup -D ${PGDATA} --no-sync" + su postgres -c "pg_probackup-${PG_VERSION} show --instance=node -B /tmp/backup -D ${PGDATA}" + su postgres -c "pg_probackup-${PG_VERSION} show --instance=node -B /tmp/backup --archive -D ${PGDATA}" + + su postgres -c "/usr/lib/postgresql/${PG_VERSION}/bin/pgbench --no-vacuum -i -s 5" + su postgres -c "pg_probackup-${PG_VERSION} backup --instance=node -b page -B /tmp/backup -D ${PGDATA} --no-sync" +#fi + +# install new packages +echo "deb [arch=amd64] http://test.postgrespro.ru/pg_probackup/deb/ $(lsb_release -cs) main-$(lsb_release -cs)" >\ + /etc/apt/sources.list.d/pg_probackup-new.list +wget -O - http://test.postgrespro.ru/pg_probackup/keys/GPG-KEY-PG_PROBACKUP | apt-key add - +apt-get update +apt-get install -y pg-probackup-${PG_VERSION} +pg_probackup-${PG_VERSION} --help +pg_probackup-${PG_VERSION} --version + +#if [ ${CODENAME} == 'precise' ] && [ ${PG_VERSION} != '10' ] && [ ${PG_VERSION} != '11' ]; then +# echo "wal_level=hot_standby" >> ${PGDATA}/postgresql.auto.conf +# echo "archive_mode=on" >> ${PGDATA}/postgresql.auto.conf +# echo "archive_command='${PKG_NAME} archive-push -B /tmp/backup --compress --instance=node --wal-file-path %p --wal-file-name %f'" >> ${PGDATA}/postgresql.auto.conf +# su postgres -c "/usr/lib/postgresql/${PG_VERSION}/bin/pg_ctl restart -D ${PGDATA}" +# sleep 5 +# su postgres -c "${PKG_NAME} init -B /tmp/backup" +# su postgres -c "${PKG_NAME} add-instance --instance=node -B /tmp/backup -D ${PGDATA}" + +# su postgres -c "pg_probackup-${PG_VERSION} init -B /tmp/backup" +# su postgres -c "pg_probackup-${PG_VERSION} add-instance --instance=node -B /tmp/backup -D ${PGDATA}" + su postgres -c "pg_probackup-${PG_VERSION} backup --instance=node --compress -b delta -B /tmp/backup -D ${PGDATA} --no-sync" + su postgres -c "/usr/lib/postgresql/${PG_VERSION}/bin/pgbench --no-vacuum -t 1000 -c 1" + su postgres -c "pg_probackup-${PG_VERSION} backup --instance=node -b page -B /tmp/backup -D ${PGDATA} --no-sync" + su postgres -c "pg_probackup-${PG_VERSION} show --instance=node -B /tmp/backup -D ${PGDATA}" + + su postgres -c "/usr/lib/postgresql/${PG_VERSION}/bin/pg_ctl stop -D ${PGDATA}" + rm -rf ${PGDATA} + + su postgres -c "pg_probackup-${PG_VERSION} restore --instance=node -B /tmp/backup -D ${PGDATA} --no-sync" + su postgres -c "/usr/lib/postgresql/${PG_VERSION}/bin/pg_ctl start -w -t 60 -D ${PGDATA}" + +sleep 5 +echo "select count(*) from pgbench_accounts;" | su postgres -c "/usr/lib/postgresql/${PG_VERSION}/bin/psql" || exit 1 +#fi + +# CHECK SRC package +apt-get install -y dpkg-dev +echo "deb-src [arch=amd64] http://test.postgrespro.ru/pg_probackup/deb/ $(lsb_release -cs) main-$(lsb_release -cs)" >>\ + /etc/apt/sources.list.d/pg_probackup.list + +wget -O - http://test.postgrespro.ru/pg_probackup/keys/GPG-KEY-PG_PROBACKUP | apt-key add - && apt-get update + +apt-get update -y + +cd /mnt +apt-get source pg-probackup-${PG_VERSION} +exit 0 + +cd pg-probackup-${PG_VERSION}-${PKG_VERSION} +#mk-build-deps --install --remove --tool 'apt-get --no-install-recommends --yes' debian/control +#rm -rf ./*.deb +apt-get install -y debhelper bison flex gettext zlib1g-dev +dpkg-buildpackage -us -uc diff --git a/packaging/test/scripts/deb_forks.sh b/packaging/test/scripts/deb_forks.sh new file mode 100755 index 000000000..e05695608 --- /dev/null +++ b/packaging/test/scripts/deb_forks.sh @@ -0,0 +1,153 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -xe +set -o pipefail + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 + +# TODO: remove after release +exit 0 + +if [ ${PBK_EDITION} == 'ent' ]; then + exit 0 +fi + +if [ ${PBK_EDITION} == 'std' ] && [ ${PG_VERSION} == '9.6' ]; then + exit 0 +fi + +# upgrade and utils +# export parameters +export DEBIAN_FRONTEND=noninteractive +echo 'debconf debconf/frontend select Noninteractive' | debconf-set-selections + +#if [ ${CODENAME} == 'jessie' ]; then +#printf "deb http://archive.debian.org/debian/ jessie main\ndeb-src http://archive.debian.org/debian/ jessie main\ndeb http://security.debian.org jessie/updates main\ndeb-src http://security.debian.org jessie/updates main" > /etc/apt/sources.list +#fi + +if [ ${DISTRIB} = 'debian' -a ${CODENAME} = 'stretch' ] ; then + apt-get -qq update +else + apt-get -qq --allow-releaseinfo-change update +fi +apt-get -qq install -y wget nginx gnupg lsb-release apt-transport-https +#apt-get -qq install -y libterm-readline-gnu-perl dialog gnupg procps + +# echo -e 'Package: *\nPin: origin test.postgrespro.ru\nPin-Priority: 800' >\ +# /etc/apt/preferences.d/pgpro-800 + +# install nginx +echo "127.0.0.1 test.postgrespro.ru" >> /etc/hosts +cat < /etc/nginx/nginx.conf +user www-data; +worker_processes 1; +error_log /var/log/nginx/error.log; +events { + worker_connections 1024; +} +http { + server { + listen 80 default; + root /app/www; + } +} +EOF +nginx -s reload || (pkill -9 nginx || nginx -c /etc/nginx/nginx.conf &) + +# install POSTGRESPRO +if [ ${PBK_EDITION} == 'std' ]; then + sh -c 'echo "deb https://repo.postgrespro.ru/pgpro-${PG_VERSION}/${DISTRIB}/ $(lsb_release -cs) main" > /etc/apt/sources.list.d/pgpro.list' + wget --quiet -O - https://repo.postgrespro.ru/pgpro-${PG_VERSION}/keys/GPG-KEY-POSTGRESPRO | apt-key add - + apt-get update -y + + apt-get install -y postgrespro-std-${PG_VERSION} + BINDIR="/opt/pgpro/std-${PG_VERSION}/bin" + export LD_LIBRARY_PATH=/opt/pgpro/std-${PG_VERSION}/lib/ +fi + +# install pg_probackup from current public repo +echo "deb [arch=amd64] http://repo.postgrespro.ru/pg_probackup-forks/deb/ $(lsb_release -cs) main-$(lsb_release -cs)" >\ + /etc/apt/sources.list.d/pg_probackup-old.list +wget -O - http://repo.postgrespro.ru/pg_probackup-forks/keys/GPG-KEY-PG_PROBACKUP | apt-key add - && apt-get update + +apt-get install -y pg-probackup-${PBK_EDITION}-${PG_VERSION} +pg_probackup-${PBK_EDITION}-${PG_VERSION} --help +pg_probackup-${PBK_EDITION}-${PG_VERSION} --version + + +if [ ${PBK_EDITION} == 'std' ]; then + export PGDATA=/tmp/data + su postgres -c "${BINDIR}/initdb -k -D ${PGDATA}" + su postgres -c "pg_probackup-${PBK_EDITION}-${PG_VERSION} init -B /tmp/backup" + su postgres -c "pg_probackup-${PBK_EDITION}-${PG_VERSION} add-instance --instance=node -B /tmp/backup -D ${PGDATA}" + + echo "wal_level=hot_standby" >> ${PGDATA}/postgresql.auto.conf + echo "fsync=off" >> ${PGDATA}/postgresql.auto.conf + echo "archive_mode=on" >> ${PGDATA}/postgresql.auto.conf + echo "archive_command='pg_probackup-${PBK_EDITION}-${PG_VERSION} archive-push --no-sync -B /tmp/backup compress --instance=node --wal-file-path %p --wal-file-name %f'" >> ${PGDATA}/postgresql.auto.conf + + su postgres -c "${BINDIR}/pg_ctl stop -w -t 60 -D /var/lib/pgpro/std-${PG_VERSION}/data" || echo "it is all good" + su postgres -c "${BINDIR}/pg_ctl start -D ${PGDATA}" + sleep 5 + su postgres -c "pg_probackup-${PBK_EDITION}-${PG_VERSION} backup --instance=node -b full -B /tmp/backup -D ${PGDATA} --no-sync" + su postgres -c "pg_probackup-${PBK_EDITION}-${PG_VERSION} show --instance=node -B /tmp/backup -D ${PGDATA}" + su postgres -c "pg_probackup-${PBK_EDITION}-${PG_VERSION} show --instance=node -B /tmp/backup -D ${PGDATA} --archive" + + su postgres -c "${BINDIR}/pgbench --no-vacuum -i -s 5" + su postgres -c "pg_probackup-${PBK_EDITION}-${PG_VERSION} backup --instance=node -b page -B /tmp/backup -D ${PGDATA} --no-sync" +fi + +# install new packages +echo "deb [arch=amd64] http://test.postgrespro.ru/pg_probackup-forks/deb/ $(lsb_release -cs) main-$(lsb_release -cs)" >\ + /etc/apt/sources.list.d/pg_probackup-new.list +wget -O - http://test.postgrespro.ru/pg_probackup-forks/keys/GPG-KEY-PG_PROBACKUP | apt-key add - +apt-get update -y + +#if [ ${PBK_EDITION} == 'std' ] && [ ${PG_VERSION} == '9.6' ]; then +# apt-get install -y libpq5 pg-probackup-${PBK_EDITION}-${PG_VERSION} +#else +# apt-get install -y pg-probackup-${PBK_EDITION}-${PG_VERSION} +#fi + +apt-get install -y pg-probackup-${PBK_EDITION}-${PG_VERSION} + +# in Ent 11 and 10 because of PQselect vanilla libpq5 is incompatible with Ent pg_probackup +if [ ${PBK_EDITION} == 'ent' ]; then + if [ ${PG_VERSION} == '11' ] || [ ${PG_VERSION} == '10' ] || [ ${PG_VERSION} == '9.6' ]; then + exit 0 + fi +fi + +pg_probackup-${PBK_EDITION}-${PG_VERSION} --help +pg_probackup-${PBK_EDITION}-${PG_VERSION} --version + +if [ ${PBK_EDITION} == 'ent' ]; then + exit 0 +fi + +if [ ${PBK_EDITION} == 'std' ] && [ ${PG_VERSION} == '9.6' ]; then + exit 0 +fi + + +#if [ ${CODENAME} == 'precise' ] && [ ${PG_VERSION} != '10' ] && [ ${PG_VERSION} != '11' ]; then + su postgres -c "${BINDIR}/pgbench --no-vacuum -t 1000 -c 1" + su postgres -c "pg_probackup-${PBK_EDITION}-${PG_VERSION} backup --instance=node -b page -B /tmp/backup -D ${PGDATA}" + su postgres -c "pg_probackup-${PBK_EDITION}-${PG_VERSION} show --instance=node -B /tmp/backup -D ${PGDATA}" + + su postgres -c "${BINDIR}/pg_ctl stop -w -t 60 -D ${PGDATA}" + rm -rf ${PGDATA} + + su postgres -c "pg_probackup-${PBK_EDITION}-${PG_VERSION} restore --instance=node -B /tmp/backup -D ${PGDATA}" + su postgres -c "${BINDIR}/pg_ctl start -w -t 60 -D ${PGDATA}" + +sleep 5 +echo "select count(*) from pgbench_accounts;" | su postgres -c "${BINDIR}/psql" || exit 1 + +exit 0 diff --git a/packaging/test/scripts/rpm.sh b/packaging/test/scripts/rpm.sh new file mode 100755 index 000000000..87d430ef8 --- /dev/null +++ b/packaging/test/scripts/rpm.sh @@ -0,0 +1,178 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2021 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -xe +set -o pipefail + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 + +PG_TOG=$(echo $PG_VERSION | sed 's|\.||g') + +if [ ${DISTRIB} != 'rhel' -o ${DISTRIB_VERSION} != '7' ]; then + # update of rpm package is broken in rhel-7 (26/12/2022) + #yum update -y + if [ ${DISTRIB} = 'centos' -a ${DISTRIB_VERSION} = '8' ]; then + sed -i 's|mirrorlist|#mirrorlist|g' /etc/yum.repos.d/CentOS-*.repo + sed -i 's|#baseurl=http://mirror.centos.org|baseurl=http://vault.centos.org|g' /etc/yum.repos.d/CentOS-*.repo + fi + yum update -y + if [ ${DISTRIB} = 'centos' -a ${DISTRIB_VERSION} = '8' ]; then + sed -i 's|mirrorlist|#mirrorlist|g' /etc/yum.repos.d/CentOS-*.repo + sed -i 's|#baseurl=http://mirror.centos.org|baseurl=http://vault.centos.org|g' /etc/yum.repos.d/CentOS-*.repo + fi +fi +# yum upgrade -y || echo 'some packages in docker failed to upgrade' +# yum install -y sudo + +if [ ${DISTRIB_VERSION} = '6' ]; then + yum install -y https://nginx.org/packages/rhel/6/x86_64/RPMS/nginx-1.8.1-1.el6.ngx.x86_64.rpm +elif [ ${DISTRIB_VERSION} = '7' ]; then + yum install -y https://nginx.org/packages/rhel/7/x86_64/RPMS/nginx-1.8.1-1.el7.ngx.x86_64.rpm +elif [ ${DISTRIB_VERSION} = '8' -a \( ${DISTRIB} = 'rhel' -o ${DISTRIB} = 'oraclelinux' \) ]; then + yum install -y nginx +else + yum install epel-release -y + yum install -y nginx +fi + +if ! getent group nginx > /dev/null 2>&1 ; then + addgroup --system --quiet nginx +fi +if ! getent passwd nginx > /dev/null 2>&1 ; then + adduser --quiet \ + --system --disabled-login --ingroup nginx \ + --home /var/run/nginx/ --no-create-home \ + nginx +fi + +cat < /etc/nginx/nginx.conf +user nginx; +worker_processes 1; +error_log /var/log/nginx/error.log; +events { + worker_connections 1024; +} +http { + server { + listen 80 default; + root /app/www; + } +} +EOF +nginx -s reload || (pkill -9 nginx || nginx -c /etc/nginx/nginx.conf &) + +# install POSTGRESQL +rpm -ivh https://download.postgresql.org/pub/repos/yum/reporpms/EL-${DISTRIB_VERSION}-x86_64/pgdg-redhat-repo-latest.noarch.rpm + +if [ ${DISTRIB} == 'oraclelinux' ] && [ ${DISTRIB_VERSION} == '8' ]; then + dnf -qy module disable postgresql +fi + +if [ ${DISTRIB} == 'centos' ] && [ ${DISTRIB_VERSION} == '8' ]; then + dnf -qy module disable postgresql +fi + +# PGDG doesn't support install of PG-9.6 from repo package anymore +if [ ${PG_VERSION} == '9.6' ] && [ ${DISTRIB_VERSION} == '7' ]; then + # ugly hack: use repo settings from PG10 + sed -i 's/10/9.6/' /etc/yum.repos.d/pgdg-redhat-all.repo +fi + +yum install -y postgresql${PG_TOG}-server.x86_64 +export PGDATA=/var/lib/pgsql/${PG_VERSION}/data + +# install old packages +yum install -y http://repo.postgrespro.ru/pg_probackup/keys/pg_probackup-repo-${DISTRIB}.noarch.rpm +yum install -y ${PKG_NAME} +${PKG_NAME} --help +${PKG_NAME} --version + +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/initdb -k -D ${PGDATA}" +echo "fsync=off" >> ${PGDATA}/postgresql.auto.conf +echo "wal_level=hot_standby" >> ${PGDATA}/postgresql.auto.conf +echo "archive_mode=on" >> ${PGDATA}/postgresql.auto.conf +echo "archive_command='${PKG_NAME} archive-push --no-sync -B /tmp/backup --instance=node --wal-file-path %p --wal-file-name %f'" >> ${PGDATA}/postgresql.auto.conf +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl start -D ${PGDATA}" +sleep 5 + +su postgres -c "${PKG_NAME} init -B /tmp/backup" +su postgres -c "${PKG_NAME} add-instance --instance=node -B /tmp/backup -D ${PGDATA}" +su postgres -c "${PKG_NAME} backup --instance=node --compress -b full -B /tmp/backup -D ${PGDATA} --no-sync" +su postgres -c "${PKG_NAME} show --instance=node -B /tmp/backup -D ${PGDATA} --archive" + +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pgbench --no-vacuum -i -s 5" +su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA} --no-sync" + +# install new packages +echo "127.0.0.1 repo.postgrespro.ru" >> /etc/hosts + +# yum remove -y pg_probackup-repo +#yum install -y http://repo.postgrespro.ru/pg_probackup/keys/pg_probackup-repo-${DISTRIB}.noarch.rpm +yum clean all -y + +sed -i "s/https/http/g" /etc/yum.repos.d/pg_probackup.repo + +yum update -y ${PKG_NAME} +${PKG_NAME} --help +${PKG_NAME} --version + +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pgbench --no-vacuum -t 1000 -c 1" +su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA} --no-sync" +su postgres -c "${PKG_NAME} show --instance=node -B /tmp/backup -D ${PGDATA}" + +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl stop -D ${PGDATA}" +rm -rf ${PGDATA} + +su postgres -c "${PKG_NAME} restore --instance=node -B /tmp/backup -D ${PGDATA} --no-sync" +su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl start -w -D ${PGDATA}" + +sleep 5 + +echo "select count(*) from pgbench_accounts;" | su postgres -c "/usr/pgsql-${PG_VERSION}/bin/psql" || exit 1 + +#else +# echo "127.0.0.1 repo.postgrespro.ru" >> /etc/hosts +# rpm -ivh http://repo.postgrespro.ru/pg_probackup/keys/pg_probackup-repo-${DISTRIB}.noarch.rpm +# yum install -y ${PKG_NAME} +# ${PKG_NAME} --help +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/initdb -k -D ${PGDATA}" +# su postgres -c "${PKG_NAME} init -B /tmp/backup" +# su postgres -c "${PKG_NAME} add-instance --instance=node -B /tmp/backup -D ${PGDATA}" +# echo "wal_level=hot_standby" >> ${PGDATA}/postgresql.auto.conf +# echo "archive_mode=on" >> ${PGDATA}/postgresql.auto.conf +# echo "archive_command='${PKG_NAME} archive-push -B /tmp/backup --instance=node --wal-file-path %p --wal-file-name %f'" >> ${PGDATA}/postgresql.auto.conf +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl start -D ${PGDATA}" +# sleep 5 +# su postgres -c "${PKG_NAME} backup --instance=node --compress -b full -B /tmp/backup -D ${PGDATA}" +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pgbench --no-vacuum -i -s 10" +# su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA}" +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pgbench --no-vacuum -t 1000 -c 1" +# su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA}" +# su postgres -c "${PKG_NAME} show --instance=node -B /tmp/backup -D ${PGDATA}" +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl stop -D ${PGDATA}" +# rm -rf ${PGDATA} +# su postgres -c "${PKG_NAME} restore --instance=node -B /tmp/backup -D ${PGDATA}" +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl start -D ${PGDATA}" +# sleep 10 +# echo "select count(*) from pgbench_accounts;" | su postgres -c "/usr/pgsql-${PG_VERSION}/bin/psql" || exit 1 +#fi + +exit 0 # while PG12 is not working + +# SRC PACKAGE +cd /mnt +yum install yum-utils rpm-build -y +yumdownloader --source ${PKG_NAME} +rpm -ivh ./*.rpm +cd /root/rpmbuild/SPECS +exit 0 + +# build pg_probackup +yum-builddep -y pg_probackup.spec +rpmbuild -bs pg_probackup.spec +rpmbuild -ba pg_probackup.spec #2>&1 | tee -ai /app/out/build.log diff --git a/packaging/test/scripts/rpm_forks.sh b/packaging/test/scripts/rpm_forks.sh new file mode 100755 index 000000000..d57711697 --- /dev/null +++ b/packaging/test/scripts/rpm_forks.sh @@ -0,0 +1,188 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -xe +set -o pipefail + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 + +PG_TOG=$(echo $PG_VERSION | sed 's|\.||g') + +if [ ${DISTRIB} != 'rhel' -o ${DISTRIB_VERSION} != '7' ]; then + # update of rpm package is broken in rhel-7 (26/12/2022) + if [ ${DISTRIB} = 'centos' -a ${DISTRIB_VERSION} = '8' ]; then + sed -i 's|mirrorlist|#mirrorlist|g' /etc/yum.repos.d/CentOS-*.repo + sed -i 's|#baseurl=http://mirror.centos.org|baseurl=http://vault.centos.org|g' /etc/yum.repos.d/CentOS-*.repo + fi + yum update -y + if [ ${DISTRIB} = 'centos' -a ${DISTRIB_VERSION} = '8' ]; then + sed -i 's|mirrorlist|#mirrorlist|g' /etc/yum.repos.d/CentOS-*.repo + sed -i 's|#baseurl=http://mirror.centos.org|baseurl=http://vault.centos.org|g' /etc/yum.repos.d/CentOS-*.repo + fi +fi + +if [ ${PBK_EDITION} == 'ent' ]; then + exit 0 +fi + +# yum upgrade -y || echo 'some packages in docker failed to upgrade' +# yum install -y sudo + +if [ ${DISTRIB} == 'rhel' ] && [ ${DISTRIB_VERSION} == '6' ]; then + exit 0; +elif [ ${DISTRIB} == 'oraclelinux' ] && [ ${DISTRIB_VERSION} == '6' ]; then + exit 0; +elif [ ${DISTRIB} == 'oraclelinux' ] && [ ${DISTRIB_VERSION} == '8' ]; then + yum install -y nginx +elif [ ${DISTRIB_VERSION} == '7' ]; then + yum install -y https://nginx.org/packages/rhel/7/x86_64/RPMS/nginx-1.8.1-1.el7.ngx.x86_64.rpm +elif [ ${DISTRIB} == 'oraclelinux' ] && [ ${DISTRIB_VERSION} == '6' ]; then + yum install -y https://nginx.org/packages/rhel/6/x86_64/RPMS/nginx-1.8.1-1.el6.ngx.x86_64.rpm +else + yum install epel-release -y + yum install -y nginx +fi + +if ! getent group nginx > /dev/null 2>&1 ; then + addgroup --system --quiet nginx +fi +if ! getent passwd nginx > /dev/null 2>&1 ; then + adduser --quiet \ + --system --disabled-login --ingroup nginx \ + --home /var/run/nginx/ --no-create-home \ + nginx +fi + +cat < /etc/nginx/nginx.conf +user nginx; +worker_processes 1; +error_log /var/log/nginx/error.log; +events { + worker_connections 1024; +} +http { + server { + listen 80 default; + root /app/www; + } +} +EOF +nginx -s reload || (pkill -9 nginx || nginx -c /etc/nginx/nginx.conf &) + +# if [ ${DISTRIB} == 'centos' ]; then + +# install old packages +yum install -y http://repo.postgrespro.ru/pg_probackup-forks/keys/pg_probackup-repo-forks-${DISTRIB}.noarch.rpm +sed -i "s/https/http/g" /etc/yum.repos.d/pg_probackup-forks.repo + +yum install -y ${PKG_NAME} +${PKG_NAME} --help +${PKG_NAME} --version + +if [ $PBK_EDITION == 'std' ] ; then + + # install POSTGRESQL + # rpm -ivh https://download.postgresql.org/pub/repos/yum/reporpms/EL-${DISTRIB_VERSION}-x86_64/pgdg-redhat-repo-latest.noarch.rpm + #if [[ ${PG_VERSION} == '11' ]] || [[ ${PG_VERSION} == '12' ]]; then + # rpm -ivh https://repo.postgrespro.ru/pgpro-${PG_VERSION}/keys/postgrespro-std-${PG_VERSION}.${DISTRIB}.yum-${PG_VERSION}-0.3.noarch.rpm + #else + # rpm -ivh https://repo.postgrespro.ru/pgpro-${PG_VERSION}/keys/postgrespro-std-${PG_VERSION}.${DISTRIB}.yum-${PG_VERSION}-0.3.noarch.rpm + #fi + curl -o pgpro-repo-add.sh https://repo.postgrespro.ru/pgpro-${PG_VERSION}/keys/pgpro-repo-add.sh + sh pgpro-repo-add.sh + + if [[ ${PG_VERSION} == '9.6' ]]; then + yum install -y postgrespro${PG_TOG}-server.x86_64 + BINDIR="/usr/pgpro-${PG_VERSION}/bin" + else + yum install -y postgrespro-std-${PG_TOG}-server.x86_64 + BINDIR="/opt/pgpro/std-${PG_VERSION}/bin" + export LD_LIBRARY_PATH=/opt/pgpro/std-${PG_VERSION}/lib/ + fi + + export PGDATA=/tmp/data + + su postgres -c "${BINDIR}/initdb -k -D ${PGDATA}" + echo "wal_level=hot_standby" >> ${PGDATA}/postgresql.auto.conf + echo "archive_mode=on" >> ${PGDATA}/postgresql.auto.conf + echo "fsync=off" >> ${PGDATA}/postgresql.auto.conf + echo "archive_command='${PKG_NAME} archive-push -B /tmp/backup --instance=node --wal-file-path %p --wal-file-name %f'" >> ${PGDATA}/postgresql.auto.conf + su postgres -c "${BINDIR}/pg_ctl start -D ${PGDATA}" + sleep 5 + + su postgres -c "${PKG_NAME} init -B /tmp/backup" + su postgres -c "${PKG_NAME} add-instance --instance=node -B /tmp/backup -D ${PGDATA}" + su postgres -c "${PKG_NAME} backup --instance=node --compress -b full -B /tmp/backup -D ${PGDATA}" + su postgres -c "${PKG_NAME} show --instance=node -B /tmp/backup -D ${PGDATA} --archive" + + su postgres -c "${BINDIR}/pgbench --no-vacuum -i -s 5" + su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA}" +fi + +# install new packages +echo "127.0.0.1 repo.postgrespro.ru" >> /etc/hosts + +# yum remove -y pg_probackup-repo +#yum install -y http://repo.postgrespro.ru/pg_probackup-forks/keys/pg_probackup-repo-forks-${DISTRIB}.noarch.rpm +#yum clean all -y + +sed -i "s/https/http/g" /etc/yum.repos.d/pg_probackup-forks.repo + +# yum update -y ${PKG_NAME} +yum install -y ${PKG_NAME} + +${PKG_NAME} --help +${PKG_NAME} --version + +if [ $PBK_EDITION == 'ent' ]; then + exit 0 +fi + +# su postgres -c "${BINDIR}/pgbench --no-vacuum -t 1000 -c 1" +su postgres -c "${BINDIR}/pgbench --no-vacuum -i -s 5" +su postgres -c "${PKG_NAME} backup --instance=node -b full -B /tmp/backup -D ${PGDATA}" +su postgres -c "${PKG_NAME} show --instance=node -B /tmp/backup -D ${PGDATA}" + +su postgres -c "${BINDIR}/pg_ctl stop -D ${PGDATA}" +rm -rf ${PGDATA} + +su postgres -c "${PKG_NAME} restore --instance=node -B /tmp/backup -D ${PGDATA}" +su postgres -c "${BINDIR}/pg_ctl start -w -D ${PGDATA}" + +sleep 5 + +echo "select count(*) from pgbench_accounts;" | su postgres -c "${BINDIR}/psql" || exit 1 + +#else +# echo "127.0.0.1 repo.postgrespro.ru" >> /etc/hosts +# rpm -ivh http://repo.postgrespro.ru/pg_probackup/keys/pg_probackup-repo-${DISTRIB}.noarch.rpm +# yum install -y ${PKG_NAME} +# ${PKG_NAME} --help +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/initdb -k -D ${PGDATA}" +# su postgres -c "${PKG_NAME} init -B /tmp/backup" +# su postgres -c "${PKG_NAME} add-instance --instance=node -B /tmp/backup -D ${PGDATA}" +# echo "wal_level=hot_standby" >> ${PGDATA}/postgresql.auto.conf +# echo "archive_mode=on" >> ${PGDATA}/postgresql.auto.conf +# echo "archive_command='${PKG_NAME} archive-push -B /tmp/backup --instance=node --wal-file-path %p --wal-file-name %f'" >> ${PGDATA}/postgresql.auto.conf +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl start -D ${PGDATA}" +# sleep 5 +# su postgres -c "${PKG_NAME} backup --instance=node --compress -b full -B /tmp/backup -D ${PGDATA}" +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pgbench --no-vacuum -i -s 10" +# su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA}" +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pgbench --no-vacuum -t 1000 -c 1" +# su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA}" +# su postgres -c "${PKG_NAME} show --instance=node -B /tmp/backup -D ${PGDATA}" +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl stop -D ${PGDATA}" +# rm -rf ${PGDATA} +# su postgres -c "${PKG_NAME} restore --instance=node -B /tmp/backup -D ${PGDATA}" +# su postgres -c "/usr/pgsql-${PG_VERSION}/bin/pg_ctl start -D ${PGDATA}" +# sleep 10 +# echo "select count(*) from pgbench_accounts;" | su postgres -c "/usr/pgsql-${PG_VERSION}/bin/psql" || exit 1 +#fi + +exit 0 diff --git a/packaging/test/scripts/suse.sh b/packaging/test/scripts/suse.sh new file mode 100755 index 000000000..ff630b479 --- /dev/null +++ b/packaging/test/scripts/suse.sh @@ -0,0 +1,128 @@ +#!/usr/bin/env bash + +# Copyright Notice: +# © (C) Postgres Professional 2015-2016 http://www.postgrespro.ru/ +# Distributed under Apache License 2.0 +# Распространяется по лицензии Apache 2.0 + +set -xe +set -o pipefail + +# fix https://github.com/moby/moby/issues/23137 +ulimit -n 1024 + +# currenctly we do not build std|ent packages for Suse +if [[ ${PBK_EDITION} != '' ]] ; then + exit 0 +fi + +PG_TOG=$(echo $PG_VERSION | sed 's|\.||g') + +if [ ${PG_TOG} == '13' ]; then # no packages for PG13 + exit 0 +fi + +if [ ${PG_TOG} == '11' ]; then # no packages for PG11 + exit 0 +fi + +if [ ${PG_TOG} == '95' ]; then # no packages for PG95 + exit 0 +fi + +zypper install -y nginx +if ! getent group nginx > /dev/null 2>&1 ; then + addgroup --system --quiet nginx +fi +if ! getent passwd nginx > /dev/null 2>&1 ; then + adduser --quiet \ + --system --disabled-login --ingroup nginx \ + --home /var/run/nginx/ --no-create-home \ + nginx +fi + +useradd postgres + +cat < /etc/nginx/nginx.conf +user nginx; +worker_processes 1; +error_log /var/log/nginx/error.log; +events { + worker_connections 1024; +} +http { + server { + listen 80 default; + root /app/www; + } +} +EOF +nginx -s reload || (pkill -9 nginx || nginx -c /etc/nginx/nginx.conf &) + +# install POSTGRESQL +zypper install -y postgresql${PG_TOG} postgresql${PG_TOG}-server postgresql${PG_TOG}-contrib +export PGDATA=/tmp/data + +# install old packages +zypper install --allow-unsigned-rpm -y http://repo.postgrespro.ru/pg_probackup/keys/pg_probackup-repo-${DISTRIB}.noarch.rpm +zypper --gpg-auto-import-keys install -y ${PKG_NAME} +${PKG_NAME} --help +${PKG_NAME} --version + +su postgres -c "/usr/lib/postgresql${PG_TOG}/bin/initdb -k -D ${PGDATA}" +echo "fsync=off" >> ${PGDATA}/postgresql.auto.conf +echo "wal_level=hot_standby" >> ${PGDATA}/postgresql.auto.conf +echo "archive_mode=on" >> ${PGDATA}/postgresql.auto.conf +echo "archive_command='${PKG_NAME} archive-push --no-sync -B /tmp/backup --instance=node --wal-file-path %p --wal-file-name %f'" >> ${PGDATA}/postgresql.auto.conf +su postgres -c "/usr/lib/postgresql${PG_TOG}/bin/pg_ctl start -D ${PGDATA}" +sleep 5 + +su postgres -c "${PKG_NAME} init -B /tmp/backup" +su postgres -c "${PKG_NAME} add-instance --instance=node -B /tmp/backup -D ${PGDATA}" +su postgres -c "${PKG_NAME} backup --instance=node --compress -b full -B /tmp/backup -D ${PGDATA} --no-sync" +su postgres -c "${PKG_NAME} show --instance=node -B /tmp/backup -D ${PGDATA} --archive" + +su postgres -c "/usr/lib/postgresql${PG_TOG}/bin/pgbench --no-vacuum -i -s 5" +su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA} --no-sync" + +# install new packages +echo "127.0.0.1 repo.postgrespro.ru" >> /etc/hosts +zypper clean all -y + +sed -i "s/https/http/g" /etc/zypp/repos.d/pg_probackup.repo + +zypper update -y ${PKG_NAME} +${PKG_NAME} --help +${PKG_NAME} --version + +su postgres -c "/usr/lib/postgresql${PG_TOG}/bin/pgbench --no-vacuum -t 1000 -c 1" +su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA} --no-sync" + +su postgres -c "/usr/lib/postgresql${PG_TOG}/bin/pgbench --no-vacuum -t 1000 -c 1" +su postgres -c "${PKG_NAME} backup --instance=node -b page -B /tmp/backup -D ${PGDATA} --no-sync" +su postgres -c "${PKG_NAME} show --instance=node -B /tmp/backup -D ${PGDATA}" + +su postgres -c "/usr/lib/postgresql${PG_TOG}/bin/pg_ctl stop -D ${PGDATA}" +rm -rf ${PGDATA} + +su postgres -c "${PKG_NAME} restore --instance=node -B /tmp/backup -D ${PGDATA} --no-sync" +su postgres -c "/usr/lib/postgresql${PG_TOG}/bin/pg_ctl start -w -D ${PGDATA}" + +sleep 5 + +echo "select count(*) from pgbench_accounts;" | su postgres -c "/usr/lib/postgresql${PG_TOG}/bin/psql" || exit 1 + +exit 0 + +# SRC PACKAGE +cd /mnt +yum install yum-utils rpm-build -y +yumdownloader --source ${PKG_NAME} +rpm -ivh ./*.rpm +cd /root/rpmbuild/SPECS +exit 0 + +# build pg_probackup +yum-builddep -y pg_probackup.spec +rpmbuild -bs pg_probackup.spec +rpmbuild -ba pg_probackup.spec #2>&1 | tee -ai /app/out/build.log diff --git a/packaging/test/scripts/suse_forks.sh b/packaging/test/scripts/suse_forks.sh new file mode 100755 index 000000000..b83f1ddd9 --- /dev/null +++ b/packaging/test/scripts/suse_forks.sh @@ -0,0 +1,5 @@ +#!/usr/bin/env bash + +set -xe +set -o pipefail +exit 0 diff --git a/po/LINGUAS b/po/LINGUAS new file mode 100644 index 000000000..562ba4cf0 --- /dev/null +++ b/po/LINGUAS @@ -0,0 +1 @@ +ru diff --git a/po/ru.po b/po/ru.po new file mode 100644 index 000000000..1263675c2 --- /dev/null +++ b/po/ru.po @@ -0,0 +1,1880 @@ +# Russian message translation file for pg_probackup +# Copyright (C) 2022 PostgreSQL Global Development Group +# This file is distributed under the same license as the pg_probackup (PostgreSQL) package. +# Vyacheslav Makarov , 2022. +msgid "" +msgstr "" +"Project-Id-Version: pg_probackup (PostgreSQL)\n" +"Report-Msgid-Bugs-To: bugs@postgrespro.ru\n" +"POT-Creation-Date: 2022-04-08 11:33+0300\n" +"PO-Revision-Date: 2022-MO-DA HO:MI+ZONE\n" +"Last-Translator: Vyacheslav Makarov \n" +"Language-Team: Russian \n" +"Language: ru\n" +"MIME-Version: 1.0\n" +"Content-Type: text/plain; charset=UTF-8\n" +"Content-Transfer-Encoding: 8bit\n" +"Plural-Forms: nplurals=3; plural=(n%10==1 && n%100!=11 ? 0 : n%10>=2 && n" +"%10<=4 && (n%100<10 || n%100>=20) ? 1 : 2);\n" + +#: src/help.c:84 +#, c-format +msgid "" +"\n" +"%s - utility to manage backup/recovery of PostgreSQL database.\n" +msgstr "" +"\n" +"%s - утилита для управления резервным копированием/восстановлением базы данных PostgreSQL.\n" + +#: src/help.c:86 +#, c-format +msgid "" +"\n" +" %s help [COMMAND]\n" +msgstr "" + +#: src/help.c:88 +#, c-format +msgid "" +"\n" +" %s version\n" +msgstr "" + +#: src/help.c:90 +#, c-format +msgid "" +"\n" +" %s init -B backup-path\n" +msgstr "" + +#: src/help.c:92 +#, c-format +msgid "" +"\n" +" %s set-config -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:93 src/help.c:791 +#, c-format +msgid " [-D pgdata-path]\n" +msgstr "" + +#: src/help.c:94 src/help.c:130 src/help.c:218 +#, c-format +msgid " [--external-dirs=external-directories-paths]\n" +msgstr "" + +#: src/help.c:95 src/help.c:132 src/help.c:305 src/help.c:731 src/help.c:794 +#, c-format +msgid " [--log-level-console=log-level-console]\n" +msgstr "" + +#: src/help.c:96 src/help.c:133 src/help.c:306 src/help.c:732 src/help.c:795 +#, c-format +msgid " [--log-level-file=log-level-file]\n" +msgstr "" + +#: src/help.c:97 src/help.c:134 src/help.c:307 src/help.c:733 src/help.c:796 +#, c-format +msgid " [--log-filename=log-filename]\n" +msgstr "" + +#: src/help.c:98 src/help.c:135 src/help.c:308 src/help.c:734 src/help.c:797 +#, c-format +msgid " [--error-log-filename=error-log-filename]\n" +msgstr "" + +#: src/help.c:99 src/help.c:136 src/help.c:309 src/help.c:735 src/help.c:798 +#, c-format +msgid " [--log-directory=log-directory]\n" +msgstr "" + +#: src/help.c:100 src/help.c:137 src/help.c:310 src/help.c:736 src/help.c:799 +#, c-format +msgid " [--log-rotation-size=log-rotation-size]\n" +msgstr "" + +#: src/help.c:101 src/help.c:800 +#, c-format +msgid " [--log-rotation-age=log-rotation-age]\n" +msgstr "" + +#: src/help.c:102 src/help.c:140 src/help.c:203 src/help.c:313 src/help.c:674 +#: src/help.c:801 +#, c-format +msgid " [--retention-redundancy=retention-redundancy]\n" +msgstr "" + +#: src/help.c:103 src/help.c:141 src/help.c:204 src/help.c:314 src/help.c:675 +#: src/help.c:802 +#, c-format +msgid " [--retention-window=retention-window]\n" +msgstr "" + +#: src/help.c:104 src/help.c:142 src/help.c:205 src/help.c:315 src/help.c:676 +#: src/help.c:803 +#, c-format +msgid " [--wal-depth=wal-depth]\n" +msgstr "" + +#: src/help.c:105 src/help.c:144 src/help.c:235 src/help.c:317 src/help.c:804 +#: src/help.c:948 +#, c-format +msgid " [--compress-algorithm=compress-algorithm]\n" +msgstr "" + +#: src/help.c:106 src/help.c:145 src/help.c:236 src/help.c:318 src/help.c:805 +#: src/help.c:949 +#, c-format +msgid " [--compress-level=compress-level]\n" +msgstr "" + +#: src/help.c:107 src/help.c:232 src/help.c:806 src/help.c:945 +#, c-format +msgid " [--archive-timeout=timeout]\n" +msgstr "" + +#: src/help.c:108 src/help.c:147 src/help.c:259 src/help.c:320 src/help.c:807 +#: src/help.c:1045 +#, c-format +msgid " [-d dbname] [-h host] [-p port] [-U username]\n" +msgstr "" + +#: src/help.c:109 src/help.c:149 src/help.c:174 src/help.c:219 src/help.c:237 +#: src/help.c:247 src/help.c:261 src/help.c:322 src/help.c:449 src/help.c:808 +#: src/help.c:906 src/help.c:950 src/help.c:994 src/help.c:1047 +#, c-format +msgid " [--remote-proto] [--remote-host]\n" +msgstr "" + +#: src/help.c:110 src/help.c:150 src/help.c:175 src/help.c:220 src/help.c:238 +#: src/help.c:248 src/help.c:262 src/help.c:323 src/help.c:450 src/help.c:809 +#: src/help.c:907 src/help.c:951 src/help.c:995 src/help.c:1048 +#, c-format +msgid " [--remote-port] [--remote-path] [--remote-user]\n" +msgstr "" + +#: src/help.c:111 src/help.c:151 src/help.c:176 src/help.c:221 src/help.c:239 +#: src/help.c:249 src/help.c:263 src/help.c:324 src/help.c:451 src/help.c:1049 +#, c-format +msgid " [--ssh-options]\n" +msgstr "" + +#: src/help.c:112 +#, c-format +msgid " [--restore-command=cmdline] [--archive-host=destination]\n" +msgstr "" + +#: src/help.c:113 src/help.c:178 +#, c-format +msgid " [--archive-port=port] [--archive-user=username]\n" +msgstr "" + +#: src/help.c:114 src/help.c:119 src/help.c:123 src/help.c:153 src/help.c:179 +#: src/help.c:188 src/help.c:194 src/help.c:209 src/help.c:214 src/help.c:222 +#: src/help.c:226 src/help.c:240 src/help.c:250 src/help.c:264 +#, c-format +msgid " [--help]\n" +msgstr "" + +#: src/help.c:116 +#, c-format +msgid "" +"\n" +" %s set-backup -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:117 +#, c-format +msgid " -i backup-id [--ttl=interval] [--expire-time=timestamp]\n" +msgstr "" + +#: src/help.c:118 +#, c-format +msgid " [--note=text]\n" +msgstr "" + +#: src/help.c:121 +#, c-format +msgid "" +"\n" +" %s show-config -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:122 +#, c-format +msgid " [--format=format]\n" +msgstr "" + +#: src/help.c:125 +#, c-format +msgid "" +"\n" +" %s backup -B backup-path -b backup-mode --instance=instance_name\n" +msgstr "" + +#: src/help.c:126 src/help.c:299 +#, c-format +msgid " [-D pgdata-path] [-C]\n" +msgstr "" + +#: src/help.c:127 src/help.c:300 +#, c-format +msgid " [--stream [-S slot-name] [--temp-slot]]\n" +msgstr "" + +#: src/help.c:128 src/help.c:301 +#, c-format +msgid " [--backup-pg-log] [-j num-threads] [--progress]\n" +msgstr "" + +#: src/help.c:129 src/help.c:168 src/help.c:302 src/help.c:433 +#, c-format +msgid " [--no-validate] [--skip-block-validation]\n" +msgstr "" + +#: src/help.c:131 src/help.c:304 +#, c-format +msgid " [--no-sync]\n" +msgstr "" + +#: src/help.c:138 src/help.c:311 +#, c-format +msgid " [--log-rotation-age=log-rotation-age] [--no-color]\n" +msgstr "" + +#: src/help.c:139 src/help.c:312 +#, c-format +msgid " [--delete-expired] [--delete-wal] [--merge-expired]\n" +msgstr "" + +#: src/help.c:143 src/help.c:316 +#, c-format +msgid " [--compress]\n" +msgstr "" + +#: src/help.c:146 src/help.c:319 +#, c-format +msgid " [--archive-timeout=archive-timeout]\n" +msgstr "" + +#: src/help.c:148 src/help.c:260 src/help.c:321 src/help.c:1046 +#, c-format +msgid " [-w --no-password] [-W --password]\n" +msgstr "" + +#: src/help.c:152 +#, c-format +msgid " [--ttl=interval] [--expire-time=timestamp] [--note=text]\n" +msgstr "" + +#: src/help.c:156 +#, c-format +msgid "" +"\n" +" %s restore -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:157 src/help.c:431 +#, c-format +msgid " [-D pgdata-path] [-i backup-id] [-j num-threads]\n" +msgstr "" + +#: src/help.c:158 src/help.c:183 src/help.c:439 src/help.c:552 +#, c-format +msgid " [--recovery-target-time=time|--recovery-target-xid=xid\n" +msgstr "" + +#: src/help.c:159 src/help.c:184 src/help.c:440 src/help.c:553 +#, c-format +msgid " |--recovery-target-lsn=lsn [--recovery-target-inclusive=boolean]]\n" +msgstr "" + +#: src/help.c:160 src/help.c:185 src/help.c:441 src/help.c:554 +#, c-format +msgid " [--recovery-target-timeline=timeline]\n" +msgstr "" + +#: src/help.c:161 src/help.c:442 +#, c-format +msgid " [--recovery-target=immediate|latest]\n" +msgstr "" + +#: src/help.c:162 src/help.c:186 src/help.c:443 src/help.c:555 +#, c-format +msgid " [--recovery-target-name=target-name]\n" +msgstr "" + +#: src/help.c:163 src/help.c:444 +#, c-format +msgid " [--recovery-target-action=pause|promote|shutdown]\n" +msgstr "" + +#: src/help.c:164 src/help.c:445 src/help.c:793 +#, c-format +msgid " [--restore-command=cmdline]\n" +msgstr "" + +#: src/help.c:165 +#, c-format +msgid " [-R | --restore-as-replica] [--force]\n" +msgstr "" + +#: src/help.c:166 src/help.c:447 +#, c-format +msgid " [--primary-conninfo=primary_conninfo]\n" +msgstr "" + +#: src/help.c:167 src/help.c:448 +#, c-format +msgid " [-S | --primary-slot-name=slotname]\n" +msgstr "" + +#: src/help.c:169 +#, c-format +msgid " [-T OLDDIR=NEWDIR] [--progress]\n" +msgstr "" + +#: src/help.c:170 src/help.c:435 +#, c-format +msgid " [--external-mapping=OLDDIR=NEWDIR]\n" +msgstr "" + +#: src/help.c:171 +#, c-format +msgid " [--skip-external-dirs] [--no-sync]\n" +msgstr "" + +#: src/help.c:172 src/help.c:437 +#, c-format +msgid " [-I | --incremental-mode=none|checksum|lsn]\n" +msgstr "" + +#: src/help.c:173 +#, c-format +msgid " [--db-include | --db-exclude]\n" +msgstr "" + +#: src/help.c:177 +#, c-format +msgid " [--archive-host=hostname]\n" +msgstr "" + +#: src/help.c:181 +#, c-format +msgid "" +"\n" +" %s validate -B backup-path [--instance=instance_name]\n" +msgstr "" + +#: src/help.c:182 src/help.c:551 +#, c-format +msgid " [-i backup-id] [--progress] [-j num-threads]\n" +msgstr "" + +#: src/help.c:187 +#, c-format +msgid " [--skip-block-validation]\n" +msgstr "" + +#: src/help.c:190 +#, c-format +msgid "" +"\n" +" %s checkdb [-B backup-path] [--instance=instance_name]\n" +msgstr "" + +#: src/help.c:191 +#, c-format +msgid " [-D pgdata-path] [--progress] [-j num-threads]\n" +msgstr "" + +#: src/help.c:192 src/help.c:603 +#, c-format +msgid " [--amcheck] [--skip-block-validation]\n" +msgstr "" + +#: src/help.c:193 +#, c-format +msgid " [--heapallindexed] [--checkunique]\n" +msgstr "" + +#: src/help.c:196 +#, c-format +msgid "" +"\n" +" %s show -B backup-path\n" +msgstr "" + +#: src/help.c:197 src/help.c:657 +#, c-format +msgid " [--instance=instance_name [-i backup-id]]\n" +msgstr "" + +#: src/help.c:198 +#, c-format +msgid " [--format=format] [--archive]\n" +msgstr "" + +#: src/help.c:199 +#, c-format +msgid " [--no-color] [--help]\n" +msgstr "" + +#: src/help.c:201 +#, c-format +msgid "" +"\n" +" %s delete -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:202 src/help.c:673 +#, c-format +msgid " [-j num-threads] [--progress]\n" +msgstr "" + +#: src/help.c:206 +#, c-format +msgid " [-i backup-id | --delete-expired | --merge-expired | --status=backup_status]\n" +msgstr "" + +#: src/help.c:207 +#, c-format +msgid " [--delete-wal]\n" +msgstr "" + +#: src/help.c:208 +#, c-format +msgid " [--dry-run] [--no-validate] [--no-sync]\n" +msgstr "" + +#: src/help.c:211 +#, c-format +msgid "" +"\n" +" %s merge -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:212 +#, c-format +msgid " -i backup-id [--progress] [-j num-threads]\n" +msgstr "" + +#: src/help.c:213 src/help.c:730 +#, c-format +msgid " [--no-validate] [--no-sync]\n" +msgstr "" + +#: src/help.c:216 +#, c-format +msgid "" +"\n" +" %s add-instance -B backup-path -D pgdata-path\n" +msgstr "" + +#: src/help.c:217 src/help.c:225 src/help.c:904 +#, c-format +msgid " --instance=instance_name\n" +msgstr "" + +#: src/help.c:224 +#, c-format +msgid "" +"\n" +" %s del-instance -B backup-path\n" +msgstr "" + +#: src/help.c:228 +#, c-format +msgid "" +"\n" +" %s archive-push -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:229 src/help.c:244 src/help.c:942 src/help.c:990 +#, c-format +msgid " --wal-file-name=wal-file-name\n" +msgstr "" + +#: src/help.c:230 src/help.c:943 src/help.c:991 +#, c-format +msgid " [--wal-file-path=wal-file-path]\n" +msgstr "" + +#: src/help.c:231 src/help.c:245 src/help.c:944 src/help.c:992 +#, c-format +msgid " [-j num-threads] [--batch-size=batch_size]\n" +msgstr "" + +#: src/help.c:233 src/help.c:946 +#, c-format +msgid " [--no-ready-rename] [--no-sync]\n" +msgstr "" + +#: src/help.c:234 src/help.c:947 +#, c-format +msgid " [--overwrite] [--compress]\n" +msgstr "" + +#: src/help.c:242 +#, c-format +msgid "" +"\n" +" %s archive-get -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:243 +#, c-format +msgid " --wal-file-path=wal-file-path\n" +msgstr "" + +#: src/help.c:246 src/help.c:993 +#, c-format +msgid " [--no-validate-wal]\n" +msgstr "" + +#: src/help.c:252 +#, c-format +msgid "" +"\n" +" %s catchup -b catchup-mode\n" +msgstr "" + +#: src/help.c:253 src/help.c:1039 +#, c-format +msgid " --source-pgdata=path_to_pgdata_on_remote_server\n" +msgstr "" + +#: src/help.c:254 src/help.c:1040 +#, c-format +msgid " --destination-pgdata=path_to_local_dir\n" +msgstr "" + +#: src/help.c:255 +#, c-format +msgid " [--stream [-S slot-name] [--temp-slot | --perm-slot]]\n" +msgstr "" + +#: src/help.c:256 src/help.c:1042 +#, c-format +msgid " [-j num-threads]\n" +msgstr "" + +#: src/help.c:257 src/help.c:434 src/help.c:1043 +#, c-format +msgid " [-T OLDDIR=NEWDIR]\n" +msgstr "" + +#: src/help.c:258 src/help.c:1044 +#, c-format +msgid " [--exclude-path=path_prefix]\n" +msgstr "" + +#: src/help.c:270 +#, c-format +msgid "Read the website for details <%s>.\n" +msgstr "Подробнее читайте на сайте <%s>.\n" + +#: src/help.c:272 +#, c-format +msgid "Report bugs to <%s>.\n" +msgstr "Сообщайте об ошибках в <%s>.\n" + +#: src/help.c:279 +#, c-format +msgid "" +"\n" +"Unknown command. Try pg_probackup help\n" +"\n" +msgstr "" +"\n" +"Неизвестная команда. Попробуйте pg_probackup help\n" +"\n" + +#: src/help.c:285 +#, c-format +msgid "" +"\n" +"This command is intended for internal use\n" +"\n" +msgstr "" + +#: src/help.c:291 +#, c-format +msgid "" +"\n" +"%s init -B backup-path\n" +"\n" +msgstr "" + +#: src/help.c:292 +#, c-format +msgid "" +" -B, --backup-path=backup-path location of the backup storage area\n" +"\n" +msgstr "" + +#: src/help.c:298 +#, c-format +msgid "" +"\n" +"%s backup -B backup-path -b backup-mode --instance=instance_name\n" +msgstr "" + +#: src/help.c:303 src/help.c:792 +#, c-format +msgid " [-E external-directories-paths]\n" +msgstr "" + +#: src/help.c:325 +#, c-format +msgid "" +" [--ttl=interval] [--expire-time=timestamp] [--note=text]\n" +"\n" +msgstr "" + +#: src/help.c:327 src/help.c:455 src/help.c:558 src/help.c:606 src/help.c:660 +#: src/help.c:679 src/help.c:739 src/help.c:812 src/help.c:895 src/help.c:910 +#: src/help.c:934 src/help.c:954 src/help.c:998 +#, c-format +msgid " -B, --backup-path=backup-path location of the backup storage area\n" +msgstr "" + +#: src/help.c:328 +#, c-format +msgid " -b, --backup-mode=backup-mode backup mode=FULL|PAGE|DELTA|PTRACK\n" +msgstr "" + +#: src/help.c:329 src/help.c:456 src/help.c:559 src/help.c:607 src/help.c:680 +#: src/help.c:740 src/help.c:813 src/help.c:896 +#, c-format +msgid " --instance=instance_name name of the instance\n" +msgstr "" + +#: src/help.c:330 src/help.c:458 src/help.c:608 src/help.c:814 src/help.c:911 +#, c-format +msgid " -D, --pgdata=pgdata-path location of the database storage area\n" +msgstr "" + +#: src/help.c:331 +#, c-format +msgid " -C, --smooth-checkpoint do smooth checkpoint before backup\n" +msgstr "" + +#: src/help.c:332 +#, c-format +msgid " --stream stream the transaction log and include it in the backup\n" +msgstr "" + +#: src/help.c:333 src/help.c:1054 +#, c-format +msgid " -S, --slot=SLOTNAME replication slot to use\n" +msgstr "" + +#: src/help.c:334 src/help.c:1055 +#, c-format +msgid " --temp-slot use temporary replication slot\n" +msgstr "" + +#: src/help.c:335 +#, c-format +msgid " --backup-pg-log backup of '%s' directory\n" +msgstr "" + +#: src/help.c:336 src/help.c:460 src/help.c:563 src/help.c:611 src/help.c:682 +#: src/help.c:743 src/help.c:960 src/help.c:1004 src/help.c:1058 +#, c-format +msgid " -j, --threads=NUM number of parallel threads\n" +msgstr "" + +#: src/help.c:337 src/help.c:462 src/help.c:562 src/help.c:610 src/help.c:683 +#: src/help.c:744 +#, c-format +msgid " --progress show progress\n" +msgstr "" + +#: src/help.c:338 +#, c-format +msgid " --no-validate disable validation after backup\n" +msgstr "" + +#: src/help.c:339 src/help.c:466 src/help.c:573 +#, c-format +msgid " --skip-block-validation set to validate only file-level checksum\n" +msgstr "" + +#: src/help.c:340 src/help.c:815 src/help.c:914 +#, c-format +msgid " -E --external-dirs=external-directories-paths\n" +msgstr "" + +#: src/help.c:341 src/help.c:816 src/help.c:915 +#, c-format +msgid " backup some directories not from pgdata \n" +msgstr "" + +#: src/help.c:342 src/help.c:817 src/help.c:916 +#, c-format +msgid " (example: --external-dirs=/tmp/dir1:/tmp/dir2)\n" +msgstr "" + +#: src/help.c:343 +#, c-format +msgid " --no-sync do not sync backed up files to disk\n" +msgstr "" + +#: src/help.c:344 +#, c-format +msgid " --note=text add note to backup\n" +msgstr "" + +#: src/help.c:345 src/help.c:784 +#, c-format +msgid " (example: --note='backup before app update to v13.1')\n" +msgstr "" + +#: src/help.c:347 src/help.c:508 src/help.c:575 src/help.c:622 src/help.c:702 +#: src/help.c:748 src/help.c:820 +#, c-format +msgid "" +"\n" +" Logging options:\n" +msgstr "" + +#: src/help.c:348 src/help.c:509 src/help.c:576 src/help.c:623 src/help.c:703 +#: src/help.c:749 src/help.c:821 +#, c-format +msgid " --log-level-console=log-level-console\n" +msgstr "" + +#: src/help.c:349 src/help.c:510 src/help.c:577 src/help.c:624 src/help.c:704 +#: src/help.c:750 src/help.c:822 +#, c-format +msgid " level for console logging (default: info)\n" +msgstr "" + +#: src/help.c:350 src/help.c:353 src/help.c:511 src/help.c:514 src/help.c:578 +#: src/help.c:581 src/help.c:625 src/help.c:628 src/help.c:705 src/help.c:708 +#: src/help.c:751 src/help.c:754 src/help.c:823 src/help.c:826 +#, c-format +msgid " available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n" +msgstr "" + +#: src/help.c:351 src/help.c:512 src/help.c:579 src/help.c:626 src/help.c:706 +#: src/help.c:752 src/help.c:824 +#, c-format +msgid " --log-level-file=log-level-file\n" +msgstr "" + +#: src/help.c:352 src/help.c:513 src/help.c:580 src/help.c:627 src/help.c:707 +#: src/help.c:753 src/help.c:825 +#, c-format +msgid " level for file logging (default: off)\n" +msgstr "" + +#: src/help.c:354 src/help.c:515 src/help.c:582 src/help.c:629 src/help.c:709 +#: src/help.c:755 src/help.c:827 +#, c-format +msgid " --log-filename=log-filename\n" +msgstr "" + +#: src/help.c:355 src/help.c:516 src/help.c:583 src/help.c:630 src/help.c:710 +#: src/help.c:756 src/help.c:828 +#, c-format +msgid " filename for file logging (default: 'pg_probackup.log')\n" +msgstr "" + +#: src/help.c:356 src/help.c:517 src/help.c:584 src/help.c:711 src/help.c:757 +#: src/help.c:829 +#, c-format +msgid " support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n" +msgstr "" + +#: src/help.c:357 src/help.c:518 src/help.c:585 src/help.c:632 src/help.c:712 +#: src/help.c:758 src/help.c:830 +#, c-format +msgid " --error-log-filename=error-log-filename\n" +msgstr "" + +#: src/help.c:358 src/help.c:519 src/help.c:586 src/help.c:633 src/help.c:713 +#: src/help.c:759 src/help.c:831 +#, c-format +msgid " filename for error logging (default: none)\n" +msgstr "" + +#: src/help.c:359 src/help.c:520 src/help.c:587 src/help.c:634 src/help.c:714 +#: src/help.c:760 src/help.c:832 +#, c-format +msgid " --log-directory=log-directory\n" +msgstr "" + +#: src/help.c:360 src/help.c:521 src/help.c:588 src/help.c:635 src/help.c:715 +#: src/help.c:761 src/help.c:833 +#, c-format +msgid " directory for file logging (default: BACKUP_PATH/log)\n" +msgstr "" + +#: src/help.c:361 src/help.c:522 src/help.c:589 src/help.c:636 src/help.c:716 +#: src/help.c:762 src/help.c:834 +#, c-format +msgid " --log-rotation-size=log-rotation-size\n" +msgstr "" + +#: src/help.c:362 src/help.c:523 src/help.c:590 src/help.c:637 src/help.c:717 +#: src/help.c:763 src/help.c:835 +#, c-format +msgid " rotate logfile if its size exceeds this value; 0 disables; (default: 0)\n" +msgstr "" + +#: src/help.c:363 src/help.c:524 src/help.c:591 src/help.c:638 src/help.c:718 +#: src/help.c:764 src/help.c:836 +#, c-format +msgid " available units: 'kB', 'MB', 'GB', 'TB' (default: kB)\n" +msgstr "" + +#: src/help.c:364 src/help.c:525 src/help.c:592 src/help.c:639 src/help.c:719 +#: src/help.c:765 src/help.c:837 +#, c-format +msgid " --log-rotation-age=log-rotation-age\n" +msgstr "" + +#: src/help.c:365 src/help.c:526 src/help.c:593 src/help.c:640 src/help.c:720 +#: src/help.c:766 src/help.c:838 +#, c-format +msgid " rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n" +msgstr "" + +#: src/help.c:366 src/help.c:527 src/help.c:594 src/help.c:641 src/help.c:721 +#: src/help.c:767 src/help.c:839 +#, c-format +msgid " available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n" +msgstr "" + +#: src/help.c:367 src/help.c:528 src/help.c:642 +#, c-format +msgid " --no-color disable the coloring of error and warning console messages\n" +msgstr "" + +#: src/help.c:369 src/help.c:687 src/help.c:841 +#, c-format +msgid "" +"\n" +" Retention options:\n" +msgstr "" + +#: src/help.c:370 src/help.c:688 +#, c-format +msgid " --delete-expired delete backups expired according to current\n" +msgstr "" + +#: src/help.c:371 src/help.c:373 +#, c-format +msgid " retention policy after successful backup completion\n" +msgstr "" + +#: src/help.c:372 src/help.c:690 +#, c-format +msgid " --merge-expired merge backups expired according to current\n" +msgstr "" + +#: src/help.c:374 src/help.c:692 +#, c-format +msgid " --delete-wal remove redundant files in WAL archive\n" +msgstr "" + +#: src/help.c:375 src/help.c:693 src/help.c:842 +#, c-format +msgid " --retention-redundancy=retention-redundancy\n" +msgstr "" + +#: src/help.c:376 src/help.c:694 src/help.c:843 +#, c-format +msgid " number of full backups to keep; 0 disables; (default: 0)\n" +msgstr "" + +#: src/help.c:377 src/help.c:695 src/help.c:844 +#, c-format +msgid " --retention-window=retention-window\n" +msgstr "" + +#: src/help.c:378 src/help.c:696 src/help.c:845 +#, c-format +msgid " number of days of recoverability; 0 disables; (default: 0)\n" +msgstr "" + +#: src/help.c:379 src/help.c:697 +#, c-format +msgid " --wal-depth=wal-depth number of latest valid backups per timeline that must\n" +msgstr "" + +#: src/help.c:380 src/help.c:698 +#, c-format +msgid " retain the ability to perform PITR; 0 disables; (default: 0)\n" +msgstr "" + +#: src/help.c:381 src/help.c:699 +#, c-format +msgid " --dry-run perform a trial run without any changes\n" +msgstr "" + +#: src/help.c:383 +#, c-format +msgid "" +"\n" +" Pinning options:\n" +msgstr "" + +#: src/help.c:384 src/help.c:778 +#, c-format +msgid " --ttl=interval pin backup for specified amount of time; 0 unpin\n" +msgstr "" + +#: src/help.c:385 src/help.c:779 +#, c-format +msgid " available units: 'ms', 's', 'min', 'h', 'd' (default: s)\n" +msgstr "" + +#: src/help.c:386 src/help.c:780 +#, c-format +msgid " (example: --ttl=20d)\n" +msgstr "" + +#: src/help.c:387 src/help.c:781 +#, c-format +msgid " --expire-time=time pin backup until specified time stamp\n" +msgstr "" + +#: src/help.c:388 src/help.c:782 +#, c-format +msgid " (example: --expire-time='2024-01-01 00:00:00+03')\n" +msgstr "" + +#: src/help.c:390 src/help.c:849 src/help.c:967 +#, c-format +msgid "" +"\n" +" Compression options:\n" +msgstr "" + +#: src/help.c:391 src/help.c:850 src/help.c:968 +#, c-format +msgid " --compress alias for --compress-algorithm='zlib' and --compress-level=1\n" +msgstr "" + +#: src/help.c:392 src/help.c:851 src/help.c:969 +#, c-format +msgid " --compress-algorithm=compress-algorithm\n" +msgstr "" + +#: src/help.c:393 +#, c-format +msgid " available options: 'zlib', 'pglz', 'none' (default: none)\n" +msgstr "" + +#: src/help.c:394 src/help.c:853 src/help.c:971 +#, c-format +msgid " --compress-level=compress-level\n" +msgstr "" + +#: src/help.c:395 src/help.c:854 src/help.c:972 +#, c-format +msgid " level of compression [0-9] (default: 1)\n" +msgstr "" + +#: src/help.c:397 src/help.c:856 +#, c-format +msgid "" +"\n" +" Archive options:\n" +msgstr "" + +#: src/help.c:398 src/help.c:857 +#, c-format +msgid " --archive-timeout=timeout wait timeout for WAL segment archiving (default: 5min)\n" +msgstr "" + +#: src/help.c:400 src/help.c:644 src/help.c:859 src/help.c:1066 +#, c-format +msgid "" +"\n" +" Connection options:\n" +msgstr "" + +#: src/help.c:401 src/help.c:645 src/help.c:860 src/help.c:1067 +#, c-format +msgid " -U, --pguser=USERNAME user name to connect as (default: current local user)\n" +msgstr "" + +#: src/help.c:402 src/help.c:646 src/help.c:861 src/help.c:1068 +#, c-format +msgid " -d, --pgdatabase=DBNAME database to connect (default: username)\n" +msgstr "" + +#: src/help.c:403 src/help.c:647 src/help.c:862 src/help.c:1069 +#, c-format +msgid " -h, --pghost=HOSTNAME database server host or socket directory(default: 'local socket')\n" +msgstr "" + +#: src/help.c:404 src/help.c:648 src/help.c:863 src/help.c:1070 +#, c-format +msgid " -p, --pgport=PORT database server port (default: 5432)\n" +msgstr "" + +#: src/help.c:405 src/help.c:649 src/help.c:1071 +#, c-format +msgid " -w, --no-password never prompt for password\n" +msgstr "" + +#: src/help.c:406 +#, c-format +msgid " -W, --password force password prompt\n" +msgstr "" + +#: src/help.c:408 src/help.c:530 src/help.c:865 src/help.c:917 src/help.c:974 +#: src/help.c:1009 src/help.c:1074 +#, c-format +msgid "" +"\n" +" Remote options:\n" +msgstr "" + +#: src/help.c:409 src/help.c:531 src/help.c:866 src/help.c:918 src/help.c:975 +#: src/help.c:1010 src/help.c:1075 +#, c-format +msgid " --remote-proto=protocol remote protocol to use\n" +msgstr "" + +#: src/help.c:410 src/help.c:532 src/help.c:867 src/help.c:919 src/help.c:976 +#: src/help.c:1011 src/help.c:1076 +#, c-format +msgid " available options: 'ssh', 'none' (default: ssh)\n" +msgstr "" + +#: src/help.c:411 src/help.c:533 src/help.c:868 src/help.c:920 +#, c-format +msgid " --remote-host=destination remote host address or hostname\n" +msgstr "" + +#: src/help.c:412 src/help.c:534 src/help.c:869 src/help.c:921 src/help.c:978 +#: src/help.c:1013 src/help.c:1078 +#, c-format +msgid " --remote-port=port remote host port (default: 22)\n" +msgstr "" + +#: src/help.c:413 src/help.c:535 src/help.c:870 src/help.c:922 src/help.c:979 +#: src/help.c:1014 src/help.c:1079 +#, c-format +msgid " --remote-path=path path to directory with pg_probackup binary on remote host\n" +msgstr "" + +#: src/help.c:414 src/help.c:536 src/help.c:871 src/help.c:923 src/help.c:980 +#: src/help.c:1015 src/help.c:1080 +#, c-format +msgid " (default: current binary path)\n" +msgstr "" + +#: src/help.c:415 src/help.c:537 src/help.c:872 src/help.c:924 src/help.c:981 +#: src/help.c:1016 src/help.c:1081 +#, c-format +msgid " --remote-user=username user name for ssh connection (default: current user)\n" +msgstr "" + +#: src/help.c:416 src/help.c:538 src/help.c:873 src/help.c:925 src/help.c:982 +#: src/help.c:1017 src/help.c:1082 +#, c-format +msgid " --ssh-options=ssh_options additional ssh options (default: none)\n" +msgstr "" + +#: src/help.c:417 src/help.c:539 src/help.c:874 +#, c-format +msgid " (example: --ssh-options='-c cipher_spec -F configfile')\n" +msgstr "" + +#: src/help.c:419 src/help.c:881 +#, c-format +msgid "" +"\n" +" Replica options:\n" +msgstr "" + +#: src/help.c:420 src/help.c:882 +#, c-format +msgid " --master-user=user_name user name to connect to master (deprecated)\n" +msgstr "" + +#: src/help.c:421 src/help.c:883 +#, c-format +msgid " --master-db=db_name database to connect to master (deprecated)\n" +msgstr "" + +#: src/help.c:422 src/help.c:884 +#, c-format +msgid " --master-host=host_name database server host of master (deprecated)\n" +msgstr "" + +#: src/help.c:423 src/help.c:885 +#, c-format +msgid " --master-port=port database server port of master (deprecated)\n" +msgstr "" + +#: src/help.c:424 src/help.c:886 +#, c-format +msgid "" +" --replica-timeout=timeout wait timeout for WAL segment streaming through replication (deprecated)\n" +"\n" +msgstr "" + +#: src/help.c:430 +#, c-format +msgid "" +"\n" +"%s restore -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:432 +#, c-format +msgid " [--progress] [--force] [--no-sync]\n" +msgstr "" + +#: src/help.c:436 +#, c-format +msgid " [--skip-external-dirs]\n" +msgstr "" + +#: src/help.c:438 +#, c-format +msgid " [--db-include dbname | --db-exclude dbname]\n" +msgstr "" + +#: src/help.c:446 +#, c-format +msgid " [-R | --restore-as-replica]\n" +msgstr "" + +#: src/help.c:452 +#, c-format +msgid " [--archive-host=hostname] [--archive-port=port]\n" +msgstr "" + +#: src/help.c:453 +#, c-format +msgid "" +" [--archive-user=username]\n" +"\n" +msgstr "" + +#: src/help.c:459 +#, c-format +msgid " -i, --backup-id=backup-id backup to restore\n" +msgstr "" + +#: src/help.c:463 +#, c-format +msgid " --force ignore invalid status of the restored backup\n" +msgstr "" + +#: src/help.c:464 +#, c-format +msgid " --no-sync do not sync restored files to disk\n" +msgstr "" + +#: src/help.c:465 +#, c-format +msgid " --no-validate disable backup validation during restore\n" +msgstr "" + +#: src/help.c:468 src/help.c:1060 +#, c-format +msgid " -T, --tablespace-mapping=OLDDIR=NEWDIR\n" +msgstr "" + +#: src/help.c:469 src/help.c:1061 +#, c-format +msgid " relocate the tablespace from directory OLDDIR to NEWDIR\n" +msgstr "" + +#: src/help.c:470 +#, c-format +msgid " --external-mapping=OLDDIR=NEWDIR\n" +msgstr "" + +#: src/help.c:471 +#, c-format +msgid " relocate the external directory from OLDDIR to NEWDIR\n" +msgstr "" + +#: src/help.c:472 +#, c-format +msgid " --skip-external-dirs do not restore all external directories\n" +msgstr "" + +#: src/help.c:474 +#, c-format +msgid "" +"\n" +" Incremental restore options:\n" +msgstr "" + +#: src/help.c:475 +#, c-format +msgid " -I, --incremental-mode=none|checksum|lsn\n" +msgstr "" + +#: src/help.c:476 +#, c-format +msgid " reuse valid pages available in PGDATA if they have not changed\n" +msgstr "" + +#: src/help.c:477 +#, c-format +msgid " (default: none)\n" +msgstr "" + +#: src/help.c:479 +#, c-format +msgid "" +"\n" +" Partial restore options:\n" +msgstr "" + +#: src/help.c:480 +#, c-format +msgid " --db-include dbname restore only specified databases\n" +msgstr "" + +#: src/help.c:481 +#, c-format +msgid " --db-exclude dbname do not restore specified databases\n" +msgstr "" + +#: src/help.c:483 +#, c-format +msgid "" +"\n" +" Recovery options:\n" +msgstr "" + +#: src/help.c:484 src/help.c:564 +#, c-format +msgid " --recovery-target-time=time time stamp up to which recovery will proceed\n" +msgstr "" + +#: src/help.c:485 src/help.c:565 +#, c-format +msgid " --recovery-target-xid=xid transaction ID up to which recovery will proceed\n" +msgstr "" + +#: src/help.c:486 src/help.c:566 +#, c-format +msgid " --recovery-target-lsn=lsn LSN of the write-ahead log location up to which recovery will proceed\n" +msgstr "" + +#: src/help.c:487 src/help.c:567 +#, c-format +msgid " --recovery-target-inclusive=boolean\n" +msgstr "" + +#: src/help.c:488 src/help.c:568 +#, c-format +msgid " whether we stop just after the recovery target\n" +msgstr "" + +#: src/help.c:489 src/help.c:569 +#, c-format +msgid " --recovery-target-timeline=timeline\n" +msgstr "" + +#: src/help.c:490 src/help.c:570 +#, c-format +msgid " recovering into a particular timeline\n" +msgstr "" + +#: src/help.c:491 +#, c-format +msgid " --recovery-target=immediate|latest\n" +msgstr "" + +#: src/help.c:492 +#, c-format +msgid " end recovery as soon as a consistent state is reached or as late as possible\n" +msgstr "" + +#: src/help.c:493 src/help.c:571 +#, c-format +msgid " --recovery-target-name=target-name\n" +msgstr "" + +#: src/help.c:494 src/help.c:572 +#, c-format +msgid " the named restore point to which recovery will proceed\n" +msgstr "" + +#: src/help.c:495 +#, c-format +msgid " --recovery-target-action=pause|promote|shutdown\n" +msgstr "" + +#: src/help.c:496 +#, c-format +msgid " action the server should take once the recovery target is reached\n" +msgstr "" + +#: src/help.c:497 +#, c-format +msgid " (default: pause)\n" +msgstr "" + +#: src/help.c:498 src/help.c:818 +#, c-format +msgid " --restore-command=cmdline command to use as 'restore_command' in recovery.conf; 'none' disables\n" +msgstr "" + +#: src/help.c:500 +#, c-format +msgid "" +"\n" +" Standby options:\n" +msgstr "" + +#: src/help.c:501 +#, c-format +msgid " -R, --restore-as-replica write a minimal recovery.conf in the output directory\n" +msgstr "" + +#: src/help.c:502 +#, c-format +msgid " to ease setting up a standby server\n" +msgstr "" + +#: src/help.c:503 +#, c-format +msgid " --primary-conninfo=primary_conninfo\n" +msgstr "" + +#: src/help.c:504 +#, c-format +msgid " connection string to be used for establishing connection\n" +msgstr "" + +#: src/help.c:505 +#, c-format +msgid " with the primary server\n" +msgstr "" + +#: src/help.c:506 +#, c-format +msgid " -S, --primary-slot-name=slotname replication slot to be used for WAL streaming from the primary server\n" +msgstr "" + +#: src/help.c:541 src/help.c:876 +#, c-format +msgid "" +"\n" +" Remote WAL archive options:\n" +msgstr "" + +#: src/help.c:542 src/help.c:877 +#, c-format +msgid " --archive-host=destination address or hostname for ssh connection to archive host\n" +msgstr "" + +#: src/help.c:543 src/help.c:878 +#, c-format +msgid " --archive-port=port port for ssh connection to archive host (default: 22)\n" +msgstr "" + +#: src/help.c:544 +#, c-format +msgid "" +" --archive-user=username user name for ssh connection to archive host (default: PostgreSQL user)\n" +"\n" +msgstr "" + +#: src/help.c:550 +#, c-format +msgid "" +"\n" +"%s validate -B backup-path [--instance=instance_name]\n" +msgstr "" + +#: src/help.c:556 +#, c-format +msgid "" +" [--skip-block-validation]\n" +"\n" +msgstr "" + +#: src/help.c:560 +#, c-format +msgid " -i, --backup-id=backup-id backup to validate\n" +msgstr "" + +#: src/help.c:595 src/help.c:722 src/help.c:768 +#, c-format +msgid "" +" --no-color disable the coloring of error and warning console messages\n" +"\n" +msgstr "" + +#: src/help.c:601 +#, c-format +msgid "" +"\n" +"%s checkdb [-B backup-path] [--instance=instance_name]\n" +msgstr "" + +#: src/help.c:602 +#, c-format +msgid " [-D pgdata-path] [-j num-threads] [--progress]\n" +msgstr "" + +#: src/help.c:604 +#, c-format +msgid "" +" [--heapallindexed] [--checkunique]\n" +"\n" +msgstr "" + +#: src/help.c:612 +#, c-format +msgid " --skip-block-validation skip file-level checking\n" +msgstr "" + +#: src/help.c:613 src/help.c:618 src/help.c:620 +#, c-format +msgid " can be used only with '--amcheck' option\n" +msgstr "" + +#: src/help.c:614 +#, c-format +msgid " --amcheck in addition to file-level block checking\n" +msgstr "" + +#: src/help.c:615 +#, c-format +msgid " check btree indexes via function 'bt_index_check()'\n" +msgstr "" + +#: src/help.c:616 +#, c-format +msgid " using 'amcheck' or 'amcheck_next' extensions\n" +msgstr "" + +#: src/help.c:617 +#, c-format +msgid " --heapallindexed also check that heap is indexed\n" +msgstr "" + +#: src/help.c:619 +#, c-format +msgid " --checkunique also check unique constraints\n" +msgstr "" + +#: src/help.c:631 +#, c-format +msgid " support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log\n" +msgstr "" + +#: src/help.c:650 src/help.c:1072 +#, c-format +msgid "" +" -W, --password force password prompt\n" +"\n" +msgstr "" + +#: src/help.c:656 +#, c-format +msgid "" +"\n" +"%s show -B backup-path\n" +msgstr "" + +#: src/help.c:658 +#, c-format +msgid "" +" [--format=format] [--archive]\n" +"\n" +msgstr "" + +#: src/help.c:661 +#, c-format +msgid " --instance=instance_name show info about specific instance\n" +msgstr "" + +#: src/help.c:662 +#, c-format +msgid " -i, --backup-id=backup-id show info about specific backups\n" +msgstr "" + +#: src/help.c:663 +#, c-format +msgid " --archive show WAL archive information\n" +msgstr "" + +#: src/help.c:664 +#, c-format +msgid " --format=format show format=PLAIN|JSON\n" +msgstr "" + +#: src/help.c:665 +#, c-format +msgid "" +" --no-color disable the coloring for plain format\n" +"\n" +msgstr "" + +#: src/help.c:671 +#, c-format +msgid "" +"\n" +"%s delete -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:672 +#, c-format +msgid " [-i backup-id | --delete-expired | --merge-expired] [--delete-wal]\n" +msgstr "" + +#: src/help.c:677 +#, c-format +msgid "" +" [--no-validate] [--no-sync]\n" +"\n" +msgstr "" + +#: src/help.c:681 +#, c-format +msgid " -i, --backup-id=backup-id backup to delete\n" +msgstr "" + +#: src/help.c:684 src/help.c:745 +#, c-format +msgid " --no-validate disable validation during retention merge\n" +msgstr "" + +#: src/help.c:685 src/help.c:746 +#, c-format +msgid " --no-sync do not sync merged files to disk\n" +msgstr "" + +#: src/help.c:689 src/help.c:691 +#, c-format +msgid " retention policy\n" +msgstr "" + +#: src/help.c:700 +#, c-format +msgid " --status=backup_status delete all backups with specified status\n" +msgstr "" + +#: src/help.c:728 +#, c-format +msgid "" +"\n" +"%s merge -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:729 +#, c-format +msgid " -i backup-id [-j num-threads] [--progress]\n" +msgstr "" + +#: src/help.c:737 +#, c-format +msgid "" +" [--log-rotation-age=log-rotation-age]\n" +"\n" +msgstr "" + +#: src/help.c:741 +#, c-format +msgid " -i, --backup-id=backup-id backup to merge\n" +msgstr "" + +#: src/help.c:774 +#, c-format +msgid "" +"\n" +"%s set-backup -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:775 +#, c-format +msgid " -i backup-id\n" +msgstr "" + +#: src/help.c:776 +#, c-format +msgid "" +" [--ttl=interval] [--expire-time=time] [--note=text]\n" +"\n" +msgstr "" + +#: src/help.c:783 +#, c-format +msgid " --note=text add note to backup; 'none' to remove note\n" +msgstr "" + +#: src/help.c:790 +#, c-format +msgid "" +"\n" +"%s set-config -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:810 src/help.c:908 src/help.c:952 src/help.c:996 +#, c-format +msgid "" +" [--ssh-options]\n" +"\n" +msgstr "" + +#: src/help.c:846 +#, c-format +msgid " --wal-depth=wal-depth number of latest valid backups with ability to perform\n" +msgstr "" + +#: src/help.c:847 +#, c-format +msgid " the point in time recovery; disables; (default: 0)\n" +msgstr "" + +#: src/help.c:852 src/help.c:970 +#, c-format +msgid " available options: 'zlib','pglz','none' (default: 'none')\n" +msgstr "" + +#: src/help.c:879 +#, c-format +msgid " --archive-user=username user name for ssh connection to archive host (default: PostgreSQL user)\n" +msgstr "" + +#: src/help.c:892 +#, c-format +msgid "" +"\n" +"%s show-config -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:893 +#, c-format +msgid "" +" [--format=format]\n" +"\n" +msgstr "" + +#: src/help.c:897 +#, c-format +msgid "" +" --format=format show format=PLAIN|JSON\n" +"\n" +msgstr "" + +#: src/help.c:903 +#, c-format +msgid "" +"\n" +"%s add-instance -B backup-path -D pgdata-path\n" +msgstr "" + +#: src/help.c:905 +#, c-format +msgid " [-E external-directory-path]\n" +msgstr "" + +#: src/help.c:912 +#, c-format +msgid " --instance=instance_name name of the new instance\n" +msgstr "" + +#: src/help.c:926 src/help.c:983 src/help.c:1018 src/help.c:1083 +#, c-format +msgid "" +" (example: --ssh-options='-c cipher_spec -F configfile')\n" +"\n" +msgstr "" + +#: src/help.c:932 +#, c-format +msgid "" +"\n" +"%s del-instance -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:935 +#, c-format +msgid "" +" --instance=instance_name name of the instance to delete\n" +"\n" +msgstr "" + +#: src/help.c:941 +#, c-format +msgid "" +"\n" +"%s archive-push -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:955 src/help.c:999 +#, c-format +msgid " --instance=instance_name name of the instance to delete\n" +msgstr "" + +#: src/help.c:956 src/help.c:1002 +#, c-format +msgid " --wal-file-name=wal-file-name\n" +msgstr "" + +#: src/help.c:957 +#, c-format +msgid " name of the file to copy into WAL archive\n" +msgstr "" + +#: src/help.c:958 src/help.c:1000 +#, c-format +msgid " --wal-file-path=wal-file-path\n" +msgstr "" + +#: src/help.c:959 +#, c-format +msgid " relative destination path of the WAL archive\n" +msgstr "" + +#: src/help.c:961 +#, c-format +msgid " --batch-size=NUM number of files to be copied\n" +msgstr "" + +#: src/help.c:962 +#, c-format +msgid " --archive-timeout=timeout wait timeout before discarding stale temp file(default: 5min)\n" +msgstr "" + +#: src/help.c:963 +#, c-format +msgid " --no-ready-rename do not rename '.ready' files in 'archive_status' directory\n" +msgstr "" + +#: src/help.c:964 +#, c-format +msgid " --no-sync do not sync WAL file to disk\n" +msgstr "" + +#: src/help.c:965 +#, c-format +msgid " --overwrite overwrite archived WAL file\n" +msgstr "" + +#: src/help.c:977 src/help.c:1012 src/help.c:1077 +#, c-format +msgid " --remote-host=hostname remote host address or hostname\n" +msgstr "" + +#: src/help.c:989 +#, c-format +msgid "" +"\n" +"%s archive-get -B backup-path --instance=instance_name\n" +msgstr "" + +#: src/help.c:1001 +#, c-format +msgid " relative destination path name of the WAL file on the server\n" +msgstr "" + +#: src/help.c:1003 +#, c-format +msgid " name of the WAL file to retrieve from the archive\n" +msgstr "" + +#: src/help.c:1005 +#, c-format +msgid " --batch-size=NUM number of files to be prefetched\n" +msgstr "" + +#: src/help.c:1006 +#, c-format +msgid " --prefetch-dir=path location of the store area for prefetched WAL files\n" +msgstr "" + +#: src/help.c:1007 +#, c-format +msgid " --no-validate-wal skip validation of prefetched WAL file before using it\n" +msgstr "" + +#: src/help.c:1024 +#, c-format +msgid "" +"\n" +"%s help [command]\n" +msgstr "" + +#: src/help.c:1025 +#, c-format +msgid "" +"%s command --help\n" +"\n" +msgstr "" + +#: src/help.c:1031 +#, c-format +msgid "" +"\n" +"%s version\n" +msgstr "" + +#: src/help.c:1032 +#, c-format +msgid "" +"%s --version\n" +"\n" +msgstr "" + +#: src/help.c:1038 +#, c-format +msgid "" +"\n" +"%s catchup -b catchup-mode\n" +msgstr "" + +#: src/help.c:1041 +#, c-format +msgid " [--stream [-S slot-name]] [--temp-slot | --perm-slot]\n" +msgstr "" + +#: src/help.c:1050 +#, c-format +msgid "" +" [--help]\n" +"\n" +msgstr "" + +#: src/help.c:1052 +#, c-format +msgid " -b, --backup-mode=catchup-mode catchup mode=FULL|DELTA|PTRACK\n" +msgstr "" + +#: src/help.c:1053 +#, c-format +msgid " --stream stream the transaction log (only supported mode)\n" +msgstr "" + +#: src/help.c:1056 +#, c-format +msgid " -P --perm-slot create permanent replication slot\n" +msgstr "" + +#: src/help.c:1062 +#, c-format +msgid " -x, --exclude-path=path_prefix files with path_prefix (relative to pgdata) will be\n" +msgstr "" + +#: src/help.c:1063 +#, c-format +msgid " excluded from catchup (can be used multiple times)\n" +msgstr "" + +#: src/help.c:1064 +#, c-format +msgid " Dangerous option! Use at your own risk!\n" +msgstr "" diff --git a/src/archive.c b/src/archive.c index 5a1053cd1..7d753c8b3 100644 --- a/src/archive.c +++ b/src/archive.c @@ -3,462 +3,951 @@ * archive.c: - pg_probackup specific archive commands for archive backups. * * - * Portions Copyright (c) 2018-2019, Postgres Professional + * Portions Copyright (c) 2018-2022, Postgres Professional * *------------------------------------------------------------------------- */ +#include #include "pg_probackup.h" +#include "utils/thread.h" +#include "instr_time.h" + +static void *push_files(void *arg); +static void *get_files(void *arg); +static bool get_wal_file(const char *filename, const char *from_path, const char *to_path, + bool prefetch_mode); +static int get_wal_file_internal(const char *from_path, const char *to_path, FILE *out, + bool is_decompress); +#ifdef HAVE_LIBZ +static const char *get_gz_error(gzFile gzf, int errnum); +#endif +//static void copy_file_attributes(const char *from_path, +// fio_location from_location, +// const char *to_path, fio_location to_location, +// bool unlink_on_error); -#include +static bool next_wal_segment_exists(TimeLineID tli, XLogSegNo segno, const char *prefetch_dir, uint32 wal_seg_size); +static uint32 run_wal_prefetch(const char *prefetch_dir, const char *archive_dir, TimeLineID tli, + XLogSegNo first_segno, int num_threads, bool inclusive, int batch_size, + uint32 wal_seg_size); +static bool wal_satisfy_from_prefetch(TimeLineID tli, XLogSegNo segno, const char *wal_file_name, + const char *prefetch_dir, const char *absolute_wal_file_path, + uint32 wal_seg_size, bool parse_wal); + +static uint32 maintain_prefetch(const char *prefetch_dir, XLogSegNo first_segno, uint32 wal_seg_size); -static void push_wal_file(const char *from_path, const char *to_path, - bool is_compress, bool overwrite); -static void get_wal_file(const char *from_path, const char *to_path); +static bool prefetch_stop = false; +static uint32 xlog_seg_size; + +typedef struct +{ + const char *first_filename; + const char *pg_xlog_dir; + const char *archive_dir; + const char *archive_status_dir; + bool overwrite; + bool compress; + bool no_sync; + bool no_ready_rename; + uint32 archive_timeout; + + CompressAlg compress_alg; + int compress_level; + int thread_num; + + parray *files; + + uint32 n_pushed; + uint32 n_skipped; + + /* + * Return value from the thread. + * 0 means there is no error, + * 1 - there is an error. + * 2 - no error, but nothing to push + */ + int ret; +} archive_push_arg; + +typedef struct +{ + const char *prefetch_dir; + const char *archive_dir; + int thread_num; + parray *files; + uint32 n_fetched; +} archive_get_arg; + +typedef struct WALSegno +{ + char name[MAXFNAMELEN]; + volatile pg_atomic_flag lock; + volatile pg_atomic_uint32 done; + struct WALSegno* prev; +} WALSegno; + +static int push_file_internal_uncompressed(WALSegno *wal_file_name, const char *pg_xlog_dir, + const char *archive_dir, bool overwrite, bool no_sync, + uint32 archive_timeout); #ifdef HAVE_LIBZ -static const char *get_gz_error(gzFile gzf, int errnum); +static int push_file_internal_gz(WALSegno *wal_file_name, const char *pg_xlog_dir, + const char *archive_dir, bool overwrite, bool no_sync, + int compress_level, uint32 archive_timeout); #endif -static bool fileEqualCRC(const char *path1, const char *path2, - bool path2_is_compressed); -static void copy_file_attributes(const char *from_path, - fio_location from_location, - const char *to_path, fio_location to_location, - bool unlink_on_error); + +static int push_file(WALSegno *xlogfile, const char *archive_status_dir, + const char *pg_xlog_dir, const char *archive_dir, + bool overwrite, bool no_sync, uint32 archive_timeout, + bool no_ready_rename, bool is_compress, + int compress_level); + +static parray *setup_push_filelist(const char *archive_status_dir, + const char *first_file, int batch_size); /* + * At this point, we already done one roundtrip to archive server + * to get instance config. + * * pg_probackup specific archive command for archive backups - * set archive_command = 'pg_probackup archive-push -B /home/anastasia/backup - * --wal-file-path %p --wal-file-name %f', to move backups into arclog_path. - * Where archlog_path is $BACKUP_PATH/wal/system_id. - * Currently it just copies wal files to the new location. - * TODO: Planned options: list the arclog content, - * compute and validate checksums. + * set archive_command to + * 'pg_probackup archive-push -B /home/anastasia/backup --wal-file-name %f', + * to move backups into arclog_path. + * Where archlog_path is $BACKUP_PATH/wal/instance_name */ -int -do_archive_push(char *wal_file_path, char *wal_file_name, bool overwrite) +void +do_archive_push(InstanceState *instanceState, InstanceConfig *instance, char *pg_xlog_dir, + char *wal_file_name, int batch_size, bool overwrite, + bool no_sync, bool no_ready_rename) { - char backup_wal_file_path[MAXPGPATH]; - char absolute_wal_file_path[MAXPGPATH]; - char current_dir[MAXPGPATH]; - uint64 system_id; + uint64 i; + /* usually instance pgdata/pg_wal/archive_status, empty if no_ready_rename or batch_size == 1 */ + char archive_status_dir[MAXPGPATH] = ""; bool is_compress = false; - if (wal_file_name == NULL && wal_file_path == NULL) - elog(ERROR, "required parameters are not specified: --wal-file-name %%f --wal-file-path %%p"); + /* arrays with meta info for multi threaded backup */ + pthread_t *threads; + archive_push_arg *threads_args; + bool push_isok = true; - if (wal_file_name == NULL) - elog(ERROR, "required parameter not specified: --wal-file-name %%f"); - - if (wal_file_path == NULL) - elog(ERROR, "required parameter not specified: --wal-file-path %%p"); - - canonicalize_path(wal_file_path); + /* reporting */ + uint32 n_total_pushed = 0; + uint32 n_total_skipped = 0; + uint32 n_total_failed = 0; + instr_time start_time, end_time; + double push_time; + char pretty_time_str[20]; - if (!getcwd(current_dir, sizeof(current_dir))) - elog(ERROR, "getcwd() error"); + /* files to push in multi-thread mode */ + parray *batch_files = NULL; + int n_threads; - /* verify that archive-push --instance parameter is valid */ - system_id = get_system_identifier(current_dir); + if (!no_ready_rename || batch_size > 1) + join_path_components(archive_status_dir, pg_xlog_dir, "archive_status"); - if (instance_config.pgdata == NULL) - elog(ERROR, "cannot read pg_probackup.conf for this instance"); +#ifdef HAVE_LIBZ + if (instance->compress_alg == ZLIB_COMPRESS) + is_compress = true; +#endif - if(system_id != instance_config.system_identifier) - elog(ERROR, "Refuse to push WAL segment %s into archive. Instance parameters mismatch." - "Instance '%s' should have SYSTEM_ID = " UINT64_FORMAT " instead of " UINT64_FORMAT, - wal_file_name, instance_name, instance_config.system_identifier, - system_id); + /* Setup filelist and locks */ + batch_files = setup_push_filelist(archive_status_dir, wal_file_name, batch_size); + + n_threads = num_threads; + if (num_threads > parray_num(batch_files)) + n_threads = parray_num(batch_files); + + elog(INFO, "pg_probackup archive-push WAL file: %s, " + "threads: %i/%i, batch: %lu/%i, compression: %s", + wal_file_name, n_threads, num_threads, + parray_num(batch_files), batch_size, + is_compress ? "zlib" : "none"); + + num_threads = n_threads; + + /* Single-thread push + * We don`t want to start multi-thread push, if number of threads in equal to 1, + * or the number of files ready to push is small. + * Multithreading in remote mode isn`t cheap, + * establishing ssh connection can take 100-200ms, so running and terminating + * one thread using generic multithread approach can take + * almost as much time as copying itself. + * TODO: maybe we should be more conservative and force single thread + * push if batch_files array is small. + */ + if (num_threads == 1 || (parray_num(batch_files) == 1)) + { + INSTR_TIME_SET_CURRENT(start_time); + for (i = 0; i < parray_num(batch_files); i++) + { + int rc; + WALSegno *xlogfile = (WALSegno *) parray_get(batch_files, i); + bool first_wal = strcmp(xlogfile->name, wal_file_name) == 0; + + rc = push_file(xlogfile, first_wal ? NULL : archive_status_dir, + pg_xlog_dir, instanceState->instance_wal_subdir_path, + overwrite, no_sync, + instance->archive_timeout, + no_ready_rename || first_wal, + is_compress && IsXLogFileName(xlogfile->name) ? true : false, + instance->compress_level); + if (rc == 0) + n_total_pushed++; + else + n_total_skipped++; + } - /* Create 'archlog_path' directory. Do nothing if it already exists. */ - fio_mkdir(arclog_path, DIR_PERMISSION, FIO_BACKUP_HOST); + push_isok = true; + goto push_done; + } - join_path_components(absolute_wal_file_path, current_dir, wal_file_path); - join_path_components(backup_wal_file_path, arclog_path, wal_file_name); + /* init thread args with its own segno */ + threads = (pthread_t *) palloc(sizeof(pthread_t) * num_threads); + threads_args = (archive_push_arg *) palloc(sizeof(archive_push_arg) * num_threads); - elog(INFO, "pg_probackup archive-push from %s to %s", absolute_wal_file_path, backup_wal_file_path); + for (i = 0; i < num_threads; i++) + { + archive_push_arg *arg = &(threads_args[i]); + + arg->first_filename = wal_file_name; + arg->archive_dir = instanceState->instance_wal_subdir_path; + arg->pg_xlog_dir = pg_xlog_dir; + arg->archive_status_dir = (!no_ready_rename || batch_size > 1) ? archive_status_dir : NULL; + arg->overwrite = overwrite; + arg->compress = is_compress; + arg->no_sync = no_sync; + arg->no_ready_rename = no_ready_rename; + arg->archive_timeout = instance->archive_timeout; + + arg->compress_alg = instance->compress_alg; + arg->compress_level = instance->compress_level; + + arg->files = batch_files; + arg->n_pushed = 0; + arg->n_skipped = 0; + + arg->thread_num = i+1; + /* By default there are some error */ + arg->ret = 1; + } - if (instance_config.compress_alg == PGLZ_COMPRESS) - elog(ERROR, "pglz compression is not supported"); + /* Run threads */ + INSTR_TIME_SET_CURRENT(start_time); + for (i = 0; i < num_threads; i++) + { + archive_push_arg *arg = &(threads_args[i]); + pthread_create(&threads[i], NULL, push_files, arg); + } -#ifdef HAVE_LIBZ - if (instance_config.compress_alg == ZLIB_COMPRESS) - is_compress = IsXLogFileName(wal_file_name); -#endif + /* Wait threads */ + for (i = 0; i < num_threads; i++) + { + pthread_join(threads[i], NULL); + if (threads_args[i].ret == 1) + { + push_isok = false; + n_total_failed++; + } - push_wal_file(absolute_wal_file_path, backup_wal_file_path, is_compress, - overwrite); - elog(INFO, "pg_probackup archive-push completed successfully"); + n_total_pushed += threads_args[i].n_pushed; + n_total_skipped += threads_args[i].n_skipped; + } - return 0; + /* Note, that we are leaking memory here, + * because pushing into archive is a very + * time-sensitive operation, so we skip freeing stuff. + */ + +push_done: + fio_disconnect(); + /* calculate elapsed time */ + INSTR_TIME_SET_CURRENT(end_time); + INSTR_TIME_SUBTRACT(end_time, start_time); + push_time = INSTR_TIME_GET_DOUBLE(end_time); + pretty_time_interval(push_time, pretty_time_str, 20); + + if (push_isok) + /* report number of files pushed into archive */ + elog(INFO, "pg_probackup archive-push completed successfully, " + "pushed: %u, skipped: %u, time elapsed: %s", + n_total_pushed, n_total_skipped, pretty_time_str); + else + elog(ERROR, "pg_probackup archive-push failed, " + "pushed: %i, skipped: %u, failed: %u, time elapsed: %s", + n_total_pushed, n_total_skipped, n_total_failed, + pretty_time_str); } +/* ------------- INTERNAL FUNCTIONS ---------- */ /* - * pg_probackup specific restore command. - * Move files from arclog_path to pgdata/wal_file_path. + * Copy files from pg_wal to archive catalog with possible compression. */ -int -do_archive_get(char *wal_file_path, char *wal_file_name) +static void * +push_files(void *arg) { - char backup_wal_file_path[MAXPGPATH]; - char absolute_wal_file_path[MAXPGPATH]; - char current_dir[MAXPGPATH]; + int i; + int rc; + archive_push_arg *args = (archive_push_arg *) arg; - if (wal_file_name == NULL && wal_file_path == NULL) - elog(ERROR, "required parameters are not specified: --wal-file-name %%f --wal-file-path %%p"); + my_thread_num = args->thread_num; - if (wal_file_name == NULL) - elog(ERROR, "required parameter not specified: --wal-file-name %%f"); + for (i = 0; i < parray_num(args->files); i++) + { + bool no_ready_rename = args->no_ready_rename; + WALSegno *xlogfile = (WALSegno *) parray_get(args->files, i); - if (wal_file_path == NULL) - elog(ERROR, "required parameter not specified: --wal-file-path %%p"); + if (!pg_atomic_test_set_flag(&xlogfile->lock)) + continue; - canonicalize_path(wal_file_path); + /* Do not rename ready file of the first file, + * we do this to avoid flooding PostgreSQL log with + * warnings about ready file been missing. + */ + if (strcmp(args->first_filename, xlogfile->name) == 0) + no_ready_rename = true; + + rc = push_file(xlogfile, args->archive_status_dir, + args->pg_xlog_dir, args->archive_dir, + args->overwrite, args->no_sync, + args->archive_timeout, no_ready_rename, + /* do not compress .backup, .partial and .history files */ + args->compress && IsXLogFileName(xlogfile->name) ? true : false, + args->compress_level); + + if (rc == 0) + args->n_pushed++; + else + args->n_skipped++; + } - if (!getcwd(current_dir, sizeof(current_dir))) - elog(ERROR, "getcwd() error"); + /* close ssh connection */ + fio_disconnect(); - join_path_components(absolute_wal_file_path, current_dir, wal_file_path); - join_path_components(backup_wal_file_path, arclog_path, wal_file_name); + args->ret = 0; + return NULL; +} - elog(INFO, "pg_probackup archive-get from %s to %s", - backup_wal_file_path, absolute_wal_file_path); - get_wal_file(backup_wal_file_path, absolute_wal_file_path); - elog(INFO, "pg_probackup archive-get completed successfully"); +int +push_file(WALSegno *xlogfile, const char *archive_status_dir, + const char *pg_xlog_dir, const char *archive_dir, + bool overwrite, bool no_sync, uint32 archive_timeout, + bool no_ready_rename, bool is_compress, + int compress_level) +{ + int rc; - return 0; + elog(LOG, "pushing file \"%s\"", xlogfile->name); + + /* If compression is not required, then just copy it as is */ + if (!is_compress) + rc = push_file_internal_uncompressed(xlogfile, pg_xlog_dir, + archive_dir, overwrite, no_sync, + archive_timeout); +#ifdef HAVE_LIBZ + else + rc = push_file_internal_gz(xlogfile, pg_xlog_dir, archive_dir, + overwrite, no_sync, compress_level, + archive_timeout); +#endif + + pg_atomic_write_u32(&xlogfile->done, 1); + + /* take '--no-ready-rename' flag into account */ + if (!no_ready_rename && archive_status_dir != NULL) + { + char wal_file_dummy[MAXPGPATH]; + char wal_file_ready[MAXPGPATH]; + char wal_file_done[MAXPGPATH]; + + join_path_components(wal_file_dummy, archive_status_dir, xlogfile->name); + snprintf(wal_file_ready, MAXPGPATH, "%s.%s", wal_file_dummy, "ready"); + snprintf(wal_file_done, MAXPGPATH, "%s.%s", wal_file_dummy, "done"); + + canonicalize_path(wal_file_ready); + canonicalize_path(wal_file_done); + /* It is ok to rename status file in archive_status directory */ + elog(LOG, "Rename \"%s\" to \"%s\"", wal_file_ready, wal_file_done); + + /* do not error out, if rename failed */ + if (fio_rename(wal_file_ready, wal_file_done, FIO_DB_HOST) < 0) + elog(WARNING, "Cannot rename ready file \"%s\" to \"%s\": %s", + wal_file_ready, wal_file_done, strerror(errno)); + } + + return rc; } -/* ------------- INTERNAL FUNCTIONS ---------- */ /* - * Copy WAL segment from pgdata to archive catalog with possible compression. + * Copy non WAL file, such as .backup or .history file, into WAL archive. + * Such files are not compressed. + * Returns: + * 0 - file was successfully pushed + * 1 - push was skipped because file already exists in the archive and + * has the same checksum */ -void -push_wal_file(const char *from_path, const char *to_path, bool is_compress, - bool overwrite) +int +push_file_internal_uncompressed(WALSegno *wal_file, const char *pg_xlog_dir, + const char *archive_dir, bool overwrite, bool no_sync, + uint32 archive_timeout) { FILE *in = NULL; int out = -1; - char buf[XLOG_BLCKSZ]; - const char *to_path_p; - char to_path_temp[MAXPGPATH]; - int errno_temp; + char *buf = pgut_malloc(OUT_BUF_SIZE); /* 1MB buffer */ + const char *wal_file_name = wal_file->name; + char from_fullpath[MAXPGPATH]; + char to_fullpath[MAXPGPATH]; + /* partial handling */ + struct stat st; + char to_fullpath_part[MAXPGPATH]; + int partial_try_count = 0; + int partial_file_size = 0; + bool partial_is_stale = true; + /* remote agent error message */ + char *errmsg = NULL; + + /* from path */ + join_path_components(from_fullpath, pg_xlog_dir, wal_file_name); + canonicalize_path(from_fullpath); + /* to path */ + join_path_components(to_fullpath, archive_dir, wal_file_name); + canonicalize_path(to_fullpath); + + /* Open source file for read */ + in = fopen(from_fullpath, PG_BINARY_R); + if (in == NULL) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot open source file \"%s\": %s", from_fullpath, strerror(errno)); + } -#ifdef HAVE_LIBZ - char gz_to_path[MAXPGPATH]; - gzFile gz_out = NULL; + /* disable stdio buffering for input file */ + setvbuf(in, NULL, _IONBF, BUFSIZ); + + /* open destination partial file for write */ + snprintf(to_fullpath_part, sizeof(to_fullpath_part), "%s.part", to_fullpath); - if (is_compress) + /* Grab lock by creating temp file in exclusive mode */ + out = fio_open(to_fullpath_part, O_RDWR | O_CREAT | O_EXCL | PG_BINARY, FIO_BACKUP_HOST); + if (out < 0) { - snprintf(gz_to_path, sizeof(gz_to_path), "%s.gz", to_path); - to_path_p = gz_to_path; + if (errno != EEXIST) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Failed to open temp WAL file \"%s\": %s", + to_fullpath_part, strerror(errno)); + } + /* Already existing destination temp file is not an error condition */ } else -#endif - to_path_p = to_path; + goto part_opened; + + /* + * Partial file already exists, it could have happened due to: + * 1. failed archive-push + * 2. concurrent archiving + * + * For ARCHIVE_TIMEOUT period we will try to create partial file + * and look for the size of already existing partial file, to + * determine if it is changing or not. + * If after ARCHIVE_TIMEOUT we still failed to create partial + * file, we will make a decision about discarding + * already existing partial file. + */ + + while (partial_try_count < archive_timeout) + { + if (fio_stat(to_fullpath_part, &st, false, FIO_BACKUP_HOST) < 0) + { + if (errno == ENOENT) + { + //part file is gone, lets try to grab it + out = fio_open(to_fullpath_part, O_RDWR | O_CREAT | O_EXCL | PG_BINARY, FIO_BACKUP_HOST); + if (out < 0) + { + if (errno != EEXIST) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Failed to open temp WAL file \"%s\": %s", + to_fullpath_part, strerror(errno)); + } + } + else + /* Successfully created partial file */ + break; + } + else + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot stat temp WAL file \"%s\": %s", to_fullpath_part, strerror(errno)); + } + } - /* open file for read */ - in = fio_fopen(from_path, PG_BINARY_R, FIO_DB_HOST); - if (in == NULL) - elog(ERROR, "Cannot open source WAL file \"%s\": %s", from_path, - strerror(errno)); + /* first round */ + if (!partial_try_count) + { + elog(LOG, "Temp WAL file already exists, waiting on it %u seconds: \"%s\"", + archive_timeout, to_fullpath_part); + partial_file_size = st.st_size; + } - /* Check if possible to skip copying */ - if (fileExists(to_path_p, FIO_BACKUP_HOST)) - { - if (fileEqualCRC(from_path, to_path_p, is_compress)) - return; - /* Do not copy and do not rise error. Just quit as normal. */ - else if (!overwrite) - elog(ERROR, "WAL segment \"%s\" already exists.", to_path_p); - } + /* file size is changing */ + if (st.st_size > partial_file_size) + partial_is_stale = false; - /* open backup file for write */ -#ifdef HAVE_LIBZ - if (is_compress) + sleep(1); + partial_try_count++; + } + /* The possible exit conditions: + * 1. File is grabbed + * 2. File is not grabbed, and it is not stale + * 2. File is not grabbed, and it is stale. + */ + + /* + * If temp file was not grabbed for ARCHIVE_TIMEOUT and temp file is not stale, + * then exit with error. + */ + if (out < 0) { - snprintf(to_path_temp, sizeof(to_path_temp), "%s.partial", gz_to_path); + if (!partial_is_stale) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Failed to open temp WAL file \"%s\" in %i seconds", + to_fullpath_part, archive_timeout); + } + + /* Partial segment is considered stale, so reuse it */ + elog(LOG, "Reusing stale temp WAL file \"%s\"", to_fullpath_part); + fio_unlink(to_fullpath_part, FIO_BACKUP_HOST); - gz_out = fio_gzopen(to_path_temp, PG_BINARY_W, instance_config.compress_level, FIO_BACKUP_HOST); - if (gz_out == NULL) - elog(ERROR, "Cannot open destination temporary WAL file \"%s\": %s", - to_path_temp, strerror(errno)); + out = fio_open(to_fullpath_part, O_RDWR | O_CREAT | O_EXCL | PG_BINARY, FIO_BACKUP_HOST); + if (out < 0) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot open temp WAL file \"%s\": %s", to_fullpath_part, strerror(errno)); + } } - else -#endif + +part_opened: + elog(LOG, "Temp WAL file successfully created: \"%s\"", to_fullpath_part); + /* Check if possible to skip copying */ + if (fileExists(to_fullpath, FIO_BACKUP_HOST)) { - snprintf(to_path_temp, sizeof(to_path_temp), "%s.partial", to_path); + pg_crc32 crc32_src; + pg_crc32 crc32_dst; - out = fio_open(to_path_temp, O_RDWR | O_CREAT | O_EXCL | PG_BINARY, FIO_BACKUP_HOST); - if (out < 0) - elog(ERROR, "Cannot open destination temporary WAL file \"%s\": %s", - to_path_temp, strerror(errno)); + crc32_src = fio_get_crc32(from_fullpath, FIO_DB_HOST, false, false); + crc32_dst = fio_get_crc32(to_fullpath, FIO_BACKUP_HOST, false, false); + + if (crc32_src == crc32_dst) + { + elog(LOG, "WAL file already exists in archive with the same " + "checksum, skip pushing: \"%s\"", from_fullpath); + /* cleanup */ + fclose(in); + fio_close(out); + fio_unlink(to_fullpath_part, FIO_BACKUP_HOST); + return 1; + } + else + { + if (overwrite) + elog(LOG, "WAL file already exists in archive with " + "different checksum, overwriting: \"%s\"", to_fullpath); + else + { + /* Overwriting is forbidden, + * so we must unlink partial file and exit with error. + */ + fio_unlink(to_fullpath_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "WAL file already exists in archive with " + "different checksum: \"%s\"", to_fullpath); + } + } } /* copy content */ + errno = 0; for (;;) { - ssize_t read_len = 0; + size_t read_len = 0; - read_len = fio_fread(in, buf, sizeof(buf)); + read_len = fread(buf, 1, OUT_BUF_SIZE, in); - if (read_len < 0) + if (ferror(in)) { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_BACKUP_HOST); - elog(ERROR, - "Cannot read source WAL file \"%s\": %s", - from_path, strerror(errno_temp)); + fio_unlink(to_fullpath_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot read source file \"%s\": %s", + from_fullpath, strerror(errno)); } - if (read_len > 0) + if (read_len > 0 && fio_write_async(out, buf, read_len) != read_len) { -#ifdef HAVE_LIBZ - if (is_compress) - { - if (fio_gzwrite(gz_out, buf, read_len) != read_len) - { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot write to compressed WAL file \"%s\": %s", - to_path_temp, get_gz_error(gz_out, errno_temp)); - } - } - else -#endif - { - if (fio_write(out, buf, read_len) != read_len) - { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot write to WAL file \"%s\": %s", - to_path_temp, strerror(errno_temp)); - } - } + fio_unlink(to_fullpath_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot write to destination temp file \"%s\": %s", + to_fullpath_part, strerror(errno)); } - if (read_len == 0) + if (feof(in)) break; } -#ifdef HAVE_LIBZ - if (is_compress) + /* close source file */ + fclose(in); + + /* Writing is asynchronous in case of push in remote mode, so check agent status */ + if (fio_check_error_fd(out, &errmsg)) { - if (fio_gzclose(gz_out) != 0) - { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot close compressed WAL file \"%s\": %s", - to_path_temp, get_gz_error(gz_out, errno_temp)); - } + fio_unlink(to_fullpath_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot write to the remote file \"%s\": %s", + to_fullpath_part, errmsg); } - else -#endif + + if (wal_file->prev != NULL) { - if (fio_flush(out) != 0 || fio_close(out) != 0) + while (!pg_atomic_read_u32(&wal_file->prev->done)) { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot write WAL file \"%s\": %s", - to_path_temp, strerror(errno_temp)); + if (thread_interrupted || interrupted) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Terminated while waiting for prev file"); + } + usleep(250); } } - if (fio_fclose(in)) + /* close temp file */ + if (fio_close(out) != 0) { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot close source WAL file \"%s\": %s", - from_path, strerror(errno_temp)); + fio_unlink(to_fullpath_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot close temp WAL file \"%s\": %s", + to_fullpath_part, strerror(errno)); + } + + /* sync temp file to disk */ + if (!no_sync) + { + if (fio_sync(to_fullpath_part, FIO_BACKUP_HOST) != 0) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Failed to sync file \"%s\": %s", + to_fullpath_part, strerror(errno)); + } } - /* update file permission. */ - copy_file_attributes(from_path, FIO_DB_HOST, to_path_temp, FIO_BACKUP_HOST, true); + elog(LOG, "Rename \"%s\" to \"%s\"", to_fullpath_part, to_fullpath); + + //copy_file_attributes(from_path, FIO_DB_HOST, to_path_temp, FIO_BACKUP_HOST, true); - if (fio_rename(to_path_temp, to_path_p, FIO_BACKUP_HOST) < 0) + /* Rename temp file to destination file */ + if (fio_rename(to_fullpath_part, to_fullpath, FIO_BACKUP_HOST) < 0) { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot rename WAL file \"%s\" to \"%s\": %s", - to_path_temp, to_path_p, strerror(errno_temp)); + fio_unlink(to_fullpath_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot rename file \"%s\" to \"%s\": %s", + to_fullpath_part, to_fullpath, strerror(errno)); } -#ifdef HAVE_LIBZ - if (is_compress) - elog(INFO, "WAL file compressed to \"%s\"", gz_to_path); -#endif + pg_free(buf); + return 0; } +#ifdef HAVE_LIBZ /* - * Copy WAL segment from archive catalog to pgdata with possible decompression. + * Push WAL segment into archive and apply streaming compression to it. + * Returns: + * 0 - file was successfully pushed + * 1 - push was skipped because file already exists in the archive and + * has the same checksum */ -void -get_wal_file(const char *from_path, const char *to_path) +int +push_file_internal_gz(WALSegno *wal_file, const char *pg_xlog_dir, + const char *archive_dir, bool overwrite, bool no_sync, + int compress_level, uint32 archive_timeout) { FILE *in = NULL; - int out; - char buf[XLOG_BLCKSZ]; - const char *from_path_p = from_path; - char to_path_temp[MAXPGPATH]; - int errno_temp; - bool is_decompress = false; + gzFile out = NULL; + char *buf = pgut_malloc(OUT_BUF_SIZE); + const char *wal_file_name = wal_file->name; + char from_fullpath[MAXPGPATH]; + char to_fullpath[MAXPGPATH]; + char to_fullpath_gz[MAXPGPATH]; + + /* partial handling */ + struct stat st; + + char to_fullpath_gz_part[MAXPGPATH]; + int partial_try_count = 0; + int partial_file_size = 0; + bool partial_is_stale = true; + /* remote agent errormsg */ + char *errmsg = NULL; + + /* from path */ + join_path_components(from_fullpath, pg_xlog_dir, wal_file_name); + canonicalize_path(from_fullpath); + /* to path */ + join_path_components(to_fullpath, archive_dir, wal_file_name); + canonicalize_path(to_fullpath); + + /* destination file with .gz suffix */ + snprintf(to_fullpath_gz, sizeof(to_fullpath_gz), "%s.gz", to_fullpath); + /* destination temp file */ + snprintf(to_fullpath_gz_part, sizeof(to_fullpath_gz_part), "%s.part", to_fullpath_gz); + + /* Open source file for read */ + in = fopen(from_fullpath, PG_BINARY_R); + if (in == NULL) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot open source WAL file \"%s\": %s", + from_fullpath, strerror(errno)); + } -#ifdef HAVE_LIBZ - char gz_from_path[MAXPGPATH]; - gzFile gz_in = NULL; -#endif + /* disable stdio buffering for input file */ + setvbuf(in, NULL, _IONBF, BUFSIZ); - /* First check source file for existance */ - if (fio_access(from_path, F_OK, FIO_BACKUP_HOST) != 0) + /* Grab lock by creating temp file in exclusive mode */ + out = fio_gzopen(to_fullpath_gz_part, PG_BINARY_W, compress_level, FIO_BACKUP_HOST); + if (out == NULL) { -#ifdef HAVE_LIBZ - /* - * Maybe we need to decompress the file. Check it with .gz - * extension. - */ - snprintf(gz_from_path, sizeof(gz_from_path), "%s.gz", from_path); - if (fio_access(gz_from_path, F_OK, FIO_BACKUP_HOST) == 0) + if (errno != EEXIST) { - /* Found compressed file */ - is_decompress = true; - from_path_p = gz_from_path; + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot open temp WAL file \"%s\": %s", + to_fullpath_gz_part, strerror(errno)); } -#endif - /* Didn't find compressed file */ - if (!is_decompress) - elog(ERROR, "Source WAL file \"%s\" doesn't exist", - from_path); - } - - /* open file for read */ - if (!is_decompress) - { - in = fio_fopen(from_path, PG_BINARY_R, FIO_BACKUP_HOST); - if (in == NULL) - elog(ERROR, "Cannot open source WAL file \"%s\": %s", - from_path, strerror(errno)); + /* Already existing destination temp file is not an error condition */ } -#ifdef HAVE_LIBZ else + goto part_opened; + + /* + * Partial file already exists, it could have happened due to: + * 1. failed archive-push + * 2. concurrent archiving + * + * For ARCHIVE_TIMEOUT period we will try to create partial file + * and look for the size of already existing partial file, to + * determine if it is changing or not. + * If after ARCHIVE_TIMEOUT we still failed to create partial + * file, we will make a decision about discarding + * already existing partial file. + */ + + while (partial_try_count < archive_timeout) { - gz_in = fio_gzopen(gz_from_path, PG_BINARY_R, Z_DEFAULT_COMPRESSION, - FIO_BACKUP_HOST); - if (gz_in == NULL) - elog(ERROR, "Cannot open compressed WAL file \"%s\": %s", - gz_from_path, strerror(errno)); + if (fio_stat(to_fullpath_gz_part, &st, false, FIO_BACKUP_HOST) < 0) + { + if (errno == ENOENT) + { + //part file is gone, lets try to grab it + out = fio_gzopen(to_fullpath_gz_part, PG_BINARY_W, compress_level, FIO_BACKUP_HOST); + if (out == NULL) + { + if (errno != EEXIST) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Failed to open temp WAL file \"%s\": %s", + to_fullpath_gz_part, strerror(errno)); + } + } + else + /* Successfully created partial file */ + break; + } + else + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot stat temp WAL file \"%s\": %s", + to_fullpath_gz_part, strerror(errno)); + } + } + + /* first round */ + if (!partial_try_count) + { + elog(LOG, "Temp WAL file already exists, waiting on it %u seconds: \"%s\"", + archive_timeout, to_fullpath_gz_part); + partial_file_size = st.st_size; + } + + /* file size is changing */ + if (st.st_size > partial_file_size) + partial_is_stale = false; + + sleep(1); + partial_try_count++; } -#endif + /* The possible exit conditions: + * 1. File is grabbed + * 2. File is not grabbed, and it is not stale + * 2. File is not grabbed, and it is stale. + */ + + /* + * If temp file was not grabbed for ARCHIVE_TIMEOUT and temp file is not stale, + * then exit with error. + */ + if (out == NULL) + { + if (!partial_is_stale) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Failed to open temp WAL file \"%s\" in %i seconds", + to_fullpath_gz_part, archive_timeout); + } - /* open backup file for write */ - snprintf(to_path_temp, sizeof(to_path_temp), "%s.partial", to_path); + /* Partial segment is considered stale, so reuse it */ + elog(LOG, "Reusing stale temp WAL file \"%s\"", to_fullpath_gz_part); + fio_unlink(to_fullpath_gz_part, FIO_BACKUP_HOST); - out = fio_open(to_path_temp, O_RDWR | O_CREAT | O_EXCL | PG_BINARY, FIO_DB_HOST); - if (out < 0) - elog(ERROR, "Cannot open destination temporary WAL file \"%s\": %s", - to_path_temp, strerror(errno)); + out = fio_gzopen(to_fullpath_gz_part, PG_BINARY_W, compress_level, FIO_BACKUP_HOST); + if (out == NULL) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot open temp WAL file \"%s\": %s", + to_fullpath_gz_part, strerror(errno)); + } + } - /* copy content */ - for (;;) +part_opened: + elog(LOG, "Temp WAL file successfully created: \"%s\"", to_fullpath_gz_part); + /* Check if possible to skip copying, + */ + if (fileExists(to_fullpath_gz, FIO_BACKUP_HOST)) { - int read_len = 0; + pg_crc32 crc32_src; + pg_crc32 crc32_dst; -#ifdef HAVE_LIBZ - if (is_decompress) + crc32_src = fio_get_crc32(from_fullpath, FIO_DB_HOST, false, false); + crc32_dst = fio_get_crc32(to_fullpath_gz, FIO_BACKUP_HOST, true, false); + + if (crc32_src == crc32_dst) { - read_len = fio_gzread(gz_in, buf, sizeof(buf)); - if (read_len <= 0 && !fio_gzeof(gz_in)) - { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_DB_HOST); - elog(ERROR, "Cannot read compressed WAL file \"%s\": %s", - gz_from_path, get_gz_error(gz_in, errno_temp)); - } + elog(LOG, "WAL file already exists in archive with the same " + "checksum, skip pushing: \"%s\"", from_fullpath); + /* cleanup */ + fclose(in); + fio_gzclose(out); + fio_unlink(to_fullpath_gz_part, FIO_BACKUP_HOST); + return 1; } else -#endif { - read_len = fio_fread(in, buf, sizeof(buf)); - if (read_len < 0) + if (overwrite) + elog(LOG, "WAL file already exists in archive with " + "different checksum, overwriting: \"%s\"", to_fullpath_gz); + else { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_DB_HOST); - elog(ERROR, "Cannot read source WAL file \"%s\": %s", - from_path, strerror(errno_temp)); + /* Overwriting is forbidden, + * so we must unlink partial file and exit with error. + */ + fio_unlink(to_fullpath_gz_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "WAL file already exists in archive with " + "different checksum: \"%s\"", to_fullpath_gz); } } + } - if (read_len > 0) - { - if (fio_write(out, buf, read_len) != read_len) - { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_DB_HOST); - elog(ERROR, "Cannot write to WAL file \"%s\": %s", to_path_temp, - strerror(errno_temp)); - } - } + /* copy content */ + /* TODO: move to separate function */ + for (;;) + { + size_t read_len = 0; - /* Check for EOF */ -#ifdef HAVE_LIBZ - if (is_decompress) + read_len = fread(buf, 1, OUT_BUF_SIZE, in); + + if (ferror(in)) { - if (fio_gzeof(gz_in) || read_len == 0) - break; + fio_unlink(to_fullpath_gz_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot read from source file \"%s\": %s", + from_fullpath, strerror(errno)); } - else -#endif + + if (read_len > 0 && fio_gzwrite(out, buf, read_len) != read_len) { - if (/* feof(in) || */ read_len == 0) - break; + fio_unlink(to_fullpath_gz_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot write to compressed temp WAL file \"%s\": %s", + to_fullpath_gz_part, get_gz_error(out, errno)); } + + if (feof(in)) + break; } - if (fio_flush(out) != 0 || fio_close(out) != 0) + /* close source file */ + fclose(in); + + /* Writing is asynchronous in case of push in remote mode, so check agent status */ + if (fio_check_error_fd_gz(out, &errmsg)) { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_DB_HOST); - elog(ERROR, "Cannot write WAL file \"%s\": %s", - to_path_temp, strerror(errno_temp)); + fio_unlink(to_fullpath_gz_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot write to the remote compressed file \"%s\": %s", + to_fullpath_gz_part, errmsg); } -#ifdef HAVE_LIBZ - if (is_decompress) + if (wal_file->prev != NULL) { - if (fio_gzclose(gz_in) != 0) + while (!pg_atomic_read_u32(&wal_file->prev->done)) { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_DB_HOST); - elog(ERROR, "Cannot close compressed WAL file \"%s\": %s", - gz_from_path, get_gz_error(gz_in, errno_temp)); + if (thread_interrupted || interrupted) + { + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Terminated while waiting for prev file"); + } + usleep(250); } } - else -#endif + + /* close temp file, TODO: make it synchronous */ + if (fio_gzclose(out) != 0) + { + fio_unlink(to_fullpath_gz_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot close compressed temp WAL file \"%s\": %s", + to_fullpath_gz_part, strerror(errno)); + } + + /* sync temp file to disk */ + if (!no_sync) { - if (fio_fclose(in)) + if (fio_sync(to_fullpath_gz_part, FIO_BACKUP_HOST) != 0) { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_DB_HOST); - elog(ERROR, "Cannot close source WAL file \"%s\": %s", - from_path, strerror(errno_temp)); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Failed to sync file \"%s\": %s", + to_fullpath_gz_part, strerror(errno)); } } - /* update file permission. */ - copy_file_attributes(from_path_p, FIO_BACKUP_HOST, to_path_temp, FIO_DB_HOST, true); + elog(LOG, "Rename \"%s\" to \"%s\"", + to_fullpath_gz_part, to_fullpath_gz); - if (fio_rename(to_path_temp, to_path, FIO_DB_HOST) < 0) + //copy_file_attributes(from_path, FIO_DB_HOST, to_path_temp, FIO_BACKUP_HOST, true); + + /* Rename temp file to destination file */ + if (fio_rename(to_fullpath_gz_part, to_fullpath_gz, FIO_BACKUP_HOST) < 0) { - errno_temp = errno; - fio_unlink(to_path_temp, FIO_DB_HOST); - elog(ERROR, "Cannot rename WAL file \"%s\" to \"%s\": %s", - to_path_temp, to_path, strerror(errno_temp)); + fio_unlink(to_fullpath_gz_part, FIO_BACKUP_HOST); + pg_atomic_write_u32(&wal_file->done, 1); + elog(ERROR, "Cannot rename file \"%s\" to \"%s\": %s", + to_fullpath_gz_part, to_fullpath_gz, strerror(errno)); } -#ifdef HAVE_LIBZ - if (is_decompress) - elog(INFO, "WAL file decompressed from \"%s\"", gz_from_path); -#endif + pg_free(buf); + + return 0; } +#endif #ifdef HAVE_LIBZ /* @@ -478,85 +967,871 @@ get_gz_error(gzFile gzf, int errnum) } #endif -/* - * compare CRC of two WAL files. - * If necessary, decompress WAL file from path2 - */ -static bool -fileEqualCRC(const char *path1, const char *path2, bool path2_is_compressed) +/* Copy file attributes */ +//static void +//copy_file_attributes(const char *from_path, fio_location from_location, +// const char *to_path, fio_location to_location, +// bool unlink_on_error) +//{ +// struct stat st; +// +// if (fio_stat(from_path, &st, true, from_location) == -1) +// { +// if (unlink_on_error) +// fio_unlink(to_path, to_location); +// elog(ERROR, "Cannot stat file \"%s\": %s", +// from_path, strerror(errno)); +// } +// +// if (fio_chmod(to_path, st.st_mode, to_location) == -1) +// { +// if (unlink_on_error) +// fio_unlink(to_path, to_location); +// elog(ERROR, "Cannot change mode of file \"%s\": %s", +// to_path, strerror(errno)); +// } +//} + +static int +walSegnoCompareName(const void *f1, const void *f2) { - pg_crc32 crc1; - pg_crc32 crc2; + WALSegno *w1 = *(WALSegno**)f1; + WALSegno *w2 = *(WALSegno**)f2; - /* Get checksum of backup file */ -#ifdef HAVE_LIBZ - if (path2_is_compressed) + return strcmp(w1->name, w2->name); +} + +/* Look for files with '.ready' suffix in archive_status directory + * and pack such files into batch sized array. + */ +parray * +setup_push_filelist(const char *archive_status_dir, const char *first_file, + int batch_size) +{ + WALSegno *xlogfile = NULL; + parray *status_files = NULL; + parray *batch_files = parray_new(); + size_t i; + + /* guarantee that first filename is in batch list */ + xlogfile = palloc0(sizeof(WALSegno)); + pg_atomic_init_flag(&xlogfile->lock); + pg_atomic_init_u32(&xlogfile->done, 0); + snprintf(xlogfile->name, MAXFNAMELEN, "%s", first_file); + parray_append(batch_files, xlogfile); + + if (batch_size < 2) + return batch_files; + + /* get list of files from archive_status */ + status_files = parray_new(); + dir_list_file(status_files, archive_status_dir, false, false, false, false, true, 0, FIO_DB_HOST); + parray_qsort(status_files, pgFileCompareName); + + for (i = 0; i < parray_num(status_files); i++) { - char buf [1024]; - gzFile gz_in = NULL; + int result = 0; + char filename[MAXFNAMELEN]; + char suffix[MAXFNAMELEN]; + pgFile *file = (pgFile *) parray_get(status_files, i); - INIT_FILE_CRC32(true, crc2); - gz_in = fio_gzopen(path2, PG_BINARY_R, Z_DEFAULT_COMPRESSION, FIO_BACKUP_HOST); - if (gz_in == NULL) - /* File cannot be read */ - elog(ERROR, - "Cannot compare WAL file \"%s\" with compressed \"%s\"", - path1, path2); + result = sscanf(file->name, "%[^.]%s", (char *) &filename, (char *) &suffix); - for (;;) - { - int read_len = fio_gzread(gz_in, buf, sizeof(buf)); - if (read_len <= 0 && !fio_gzeof(gz_in)) - { - /* An error occurred while reading the file */ - elog(WARNING, - "Cannot compare WAL file \"%s\" with compressed \"%s\": %d", - path1, path2, read_len); - return false; - } - COMP_FILE_CRC32(true, crc2, buf, read_len); - if (fio_gzeof(gz_in) || read_len == 0) - break; - } - FIN_FILE_CRC32(true, crc2); + if (result != 2) + continue; - if (fio_gzclose(gz_in) != 0) - elog(ERROR, "Cannot close compressed WAL file \"%s\": %s", - path2, get_gz_error(gz_in, errno)); - } - else -#endif - { - crc2 = pgFileGetCRC(path2, true, true, NULL, FIO_BACKUP_HOST); - } + if (strcmp(suffix, ".ready") != 0) + continue; - /* Get checksum of original file */ - crc1 = pgFileGetCRC(path1, true, true, NULL, FIO_DB_HOST); + /* first filename already in batch list */ + if (strcmp(filename, first_file) == 0) + continue; - return EQ_CRC32C(crc1, crc2); -} + xlogfile = palloc0(sizeof(WALSegno)); + pg_atomic_init_flag(&xlogfile->lock); + pg_atomic_init_u32(&xlogfile->done, 0); -/* Copy file attributes */ -static void -copy_file_attributes(const char *from_path, fio_location from_location, - const char *to_path, fio_location to_location, - bool unlink_on_error) + snprintf(xlogfile->name, MAXFNAMELEN, "%s", filename); + parray_append(batch_files, xlogfile); + + if (parray_num(batch_files) >= batch_size) + break; + } + + parray_qsort(batch_files, walSegnoCompareName); + for (i = 1; i < parray_num(batch_files); i++) + { + xlogfile = (WALSegno*) parray_get(batch_files, i); + xlogfile->prev = (WALSegno*) parray_get(batch_files, i-1); + } + + /* cleanup */ + parray_walk(status_files, pgFileFree); + parray_free(status_files); + + return batch_files; +} + +/* + * pg_probackup specific restore command. + * Move files from arclog_path to pgdata/wal_file_path. + * + * The problem with archive-get: we must be very careful about + * erroring out, because postgres will interpretent our negative exit code + * as the fact, that requested file is missing and may take irreversible actions. + * So if file copying has failed we must retry several times before bailing out. + * + * TODO: add support of -D option. + * TOTHINK: what can be done about ssh connection been broken? + * TOTHINk: do we need our own rmtree function ? + * TOTHINk: so sort of async prefetch ? + + */ +void +do_archive_get(InstanceState *instanceState, InstanceConfig *instance, const char *prefetch_dir_arg, + char *wal_file_path, char *wal_file_name, int batch_size, + bool validate_wal) +{ + int fail_count = 0; + char backup_wal_file_path[MAXPGPATH]; + char absolute_wal_file_path[MAXPGPATH]; + char current_dir[MAXPGPATH]; + char prefetch_dir[MAXPGPATH]; + char pg_xlog_dir[MAXPGPATH]; + char prefetched_file[MAXPGPATH]; + + /* reporting */ + uint32 n_fetched = 0; + int n_actual_threads = num_threads; + uint32 n_files_in_prefetch = 0; + + /* time reporting */ + instr_time start_time, end_time; + double get_time; + char pretty_time_str[20]; + + if (wal_file_name == NULL) + elog(ERROR, "Required parameter not specified: --wal-file-name %%f"); + + if (wal_file_path == NULL) + elog(ERROR, "Required parameter not specified: --wal_file_path %%p"); + + if (!getcwd(current_dir, sizeof(current_dir))) + elog(ERROR, "getcwd() error"); + + /* path to PGDATA/pg_wal directory */ + join_path_components(pg_xlog_dir, current_dir, XLOGDIR); + + /* destination full filepath, usually it is PGDATA/pg_wal/RECOVERYXLOG */ + join_path_components(absolute_wal_file_path, current_dir, wal_file_path); + + /* full filepath to WAL file in archive directory. + * $BACKUP_PATH/wal/instance_name/000000010000000000000001 */ + join_path_components(backup_wal_file_path, instanceState->instance_wal_subdir_path, wal_file_name); + + INSTR_TIME_SET_CURRENT(start_time); + if (num_threads > batch_size) + n_actual_threads = batch_size; + elog(INFO, "pg_probackup archive-get WAL file: %s, remote: %s, threads: %i/%i, batch: %i", + wal_file_name, IsSshProtocol() ? "ssh" : "none", n_actual_threads, num_threads, batch_size); + + num_threads = n_actual_threads; + + elog(VERBOSE, "Obtaining XLOG_SEG_SIZE from pg_control file"); + instance->xlog_seg_size = get_xlog_seg_size(current_dir); + + /* Prefetch optimization kicks in only if simple XLOG segments is requested + * and batching is enabled. + * + * We check that file do exists in prefetch directory, then we validate it and + * rename to destination path. + * If file do not exists, then we run prefetch and rename it. + */ + if (IsXLogFileName(wal_file_name) && batch_size > 1) + { + XLogSegNo segno; + TimeLineID tli; + + GetXLogFromFileName(wal_file_name, &tli, &segno, instance->xlog_seg_size); + + if (prefetch_dir_arg) + /* use provided prefetch directory */ + snprintf(prefetch_dir, sizeof(prefetch_dir), "%s", prefetch_dir_arg); + else + /* use default path */ + join_path_components(prefetch_dir, pg_xlog_dir, "pbk_prefetch"); + + /* Construct path to WAL file in prefetch directory. + * current_dir/pg_wal/pbk_prefech/000000010000000000000001 + */ + join_path_components(prefetched_file, prefetch_dir, wal_file_name); + + /* check if file is available in prefetch directory */ + if (access(prefetched_file, F_OK) == 0) + { + /* Prefetched WAL segment is available, before using it, we must validate it. + * But for validation to work properly(because of contrecord), we must be sure + * that next WAL segment is also available in prefetch directory. + * If next segment do not exists in prefetch directory, we must provide it from + * archive. If it is NOT available in the archive, then file in prefetch directory + * cannot be trusted. In this case we discard all prefetched files and + * copy requested file directly from archive. + */ + if (!next_wal_segment_exists(tli, segno, prefetch_dir, instance->xlog_seg_size)) + n_fetched = run_wal_prefetch(prefetch_dir, instanceState->instance_wal_subdir_path, + tli, segno, num_threads, false, batch_size, + instance->xlog_seg_size); + + n_files_in_prefetch = maintain_prefetch(prefetch_dir, segno, instance->xlog_seg_size); + + if (wal_satisfy_from_prefetch(tli, segno, wal_file_name, prefetch_dir, + absolute_wal_file_path, instance->xlog_seg_size, + validate_wal)) + { + n_files_in_prefetch--; + elog(INFO, "pg_probackup archive-get used prefetched WAL segment %s, prefetch state: %u/%u", + wal_file_name, n_files_in_prefetch, batch_size); + goto get_done; + } + else + { + /* discard prefetch */ +// n_fetched = 0; + pgut_rmtree(prefetch_dir, false, false); + } + } + else + { + /* Do prefetch maintenance here */ + + mkdir(prefetch_dir, DIR_PERMISSION); /* In case prefetch directory do not exists yet */ + + /* We`ve failed to satisfy current request from prefetch directory, + * therefore we can discard its content, since it may be corrupted or + * contain stale files. + * + * UPDATE: we should not discard prefetch easily, because failing to satisfy + * request for WAL may come from this recovery behavior: + * https://www.postgresql.org/message-id/flat/16159-f5a34a3a04dc67e0%40postgresql.org + */ +// rmtree(prefetch_dir, false); + + /* prefetch files */ + n_fetched = run_wal_prefetch(prefetch_dir, instanceState->instance_wal_subdir_path, + tli, segno, num_threads, true, batch_size, + instance->xlog_seg_size); + + n_files_in_prefetch = maintain_prefetch(prefetch_dir, segno, instance->xlog_seg_size); + + if (wal_satisfy_from_prefetch(tli, segno, wal_file_name, prefetch_dir, absolute_wal_file_path, + instance->xlog_seg_size, validate_wal)) + { + n_files_in_prefetch--; + elog(INFO, "pg_probackup archive-get copied WAL file %s, prefetch state: %u/%u", + wal_file_name, n_files_in_prefetch, batch_size); + goto get_done; + } +// else +// { +// /* yet again failed to satisfy request from prefetch */ +// n_fetched = 0; +// rmtree(prefetch_dir, false); +// } + } + } + + /* we use it to extend partial file later */ + xlog_seg_size = instance->xlog_seg_size; + + /* Either prefetch didn`t cut it, or batch mode is disabled or + * the requested file is not WAL segment. + * Copy file from the archive directly. + * Retry several times before bailing out. + * + * TODO: + * files copied from archive directly are not validated, which is not ok. + * TOTHINK: + * Current WAL validation cannot be applied to partial files. + */ + + while (fail_count < 3) + { + if (get_wal_file(wal_file_name, backup_wal_file_path, absolute_wal_file_path, false)) + { + fail_count = 0; + elog(LOG, "pg_probackup archive-get copied WAL file %s", wal_file_name); + n_fetched++; + break; + } + else + fail_count++; + + elog(LOG, "Failed to get WAL file %s, retry %i/3", wal_file_name, fail_count); + } + + /* TODO/TOTHINK: + * If requested file is corrupted, we have no way to warn PostgreSQL about it. + * We either can: + * 1. feed to recovery and let PostgreSQL sort it out. Currently we do this. + * 2. error out. + * + * Also note, that we can detect corruption only if prefetch mode is used. + * TODO: if corruption or network problem encountered, kill yourself + * with SIGTERN to prevent recovery from starting up database. + */ + +get_done: + INSTR_TIME_SET_CURRENT(end_time); + INSTR_TIME_SUBTRACT(end_time, start_time); + get_time = INSTR_TIME_GET_DOUBLE(end_time); + pretty_time_interval(get_time, pretty_time_str, 20); + + if (fail_count == 0) + elog(INFO, "pg_probackup archive-get completed successfully, fetched: %i/%i, time elapsed: %s", + n_fetched, batch_size, pretty_time_str); + else + elog(ERROR, "pg_probackup archive-get failed to deliver WAL file: %s, time elapsed: %s", + wal_file_name, pretty_time_str); +} + +/* + * Copy batch_size of regular WAL segments into prefetch directory, + * starting with first_file. + * + * inclusive - should we copy first_file or not. + */ +uint32 run_wal_prefetch(const char *prefetch_dir, const char *archive_dir, + TimeLineID tli, XLogSegNo first_segno, int num_threads, + bool inclusive, int batch_size, uint32 wal_seg_size) +{ + int i; + XLogSegNo segno; + parray *batch_files = parray_new(); + int n_total_fetched = 0; + + if (!inclusive) + first_segno++; + + for (segno = first_segno; segno < (first_segno + batch_size); segno++) + { + WALSegno *xlogfile = palloc(sizeof(WALSegno)); + pg_atomic_init_flag(&xlogfile->lock); + + /* construct filename for WAL segment */ + GetXLogFileName(xlogfile->name, tli, segno, wal_seg_size); + + parray_append(batch_files, xlogfile); + + } + + /* copy segments */ + if (num_threads == 1) + { + for (i = 0; i < parray_num(batch_files); i++) + { + char to_fullpath[MAXPGPATH]; + char from_fullpath[MAXPGPATH]; + WALSegno *xlogfile = (WALSegno *) parray_get(batch_files, i); + + join_path_components(to_fullpath, prefetch_dir, xlogfile->name); + join_path_components(from_fullpath, archive_dir, xlogfile->name); + + /* It is ok, maybe requested batch is greater than the number of available + * files in the archive + */ + if (!get_wal_file(xlogfile->name, from_fullpath, to_fullpath, true)) + { + elog(LOG, "Thread [%d]: Failed to prefetch WAL segment %s", 0, xlogfile->name); + break; + } + + n_total_fetched++; + } + } + else + { + /* arrays with meta info for multi threaded archive-get */ + pthread_t *threads; + archive_get_arg *threads_args; + + /* init thread args */ + threads = (pthread_t *) palloc(sizeof(pthread_t) * num_threads); + threads_args = (archive_get_arg *) palloc(sizeof(archive_get_arg) * num_threads); + + for (i = 0; i < num_threads; i++) + { + archive_get_arg *arg = &(threads_args[i]); + + arg->prefetch_dir = prefetch_dir; + arg->archive_dir = archive_dir; + + arg->thread_num = i+1; + arg->files = batch_files; + arg->n_fetched = 0; + } + + /* Run threads */ + for (i = 0; i < num_threads; i++) + { + archive_get_arg *arg = &(threads_args[i]); + pthread_create(&threads[i], NULL, get_files, arg); + } + + /* Wait threads */ + for (i = 0; i < num_threads; i++) + { + pthread_join(threads[i], NULL); + n_total_fetched += threads_args[i].n_fetched; + } + } + /* TODO: free batch_files */ + return n_total_fetched; +} + +/* + * Copy files from archive catalog to pg_wal. + */ +static void * +get_files(void *arg) +{ + int i; + char to_fullpath[MAXPGPATH]; + char from_fullpath[MAXPGPATH]; + archive_get_arg *args = (archive_get_arg *) arg; + + my_thread_num = args->thread_num; + + for (i = 0; i < parray_num(args->files); i++) + { + WALSegno *xlogfile = (WALSegno *) parray_get(args->files, i); + + if (prefetch_stop) + break; + + if (!pg_atomic_test_set_flag(&xlogfile->lock)) + continue; + + join_path_components(from_fullpath, args->archive_dir, xlogfile->name); + join_path_components(to_fullpath, args->prefetch_dir, xlogfile->name); + + if (!get_wal_file(xlogfile->name, from_fullpath, to_fullpath, true)) + { + /* It is ok, maybe requested batch is greater than the number of available + * files in the archive + */ + elog(LOG, "Failed to prefetch WAL segment %s", xlogfile->name); + prefetch_stop = true; + break; + } + + args->n_fetched++; + } + + /* close ssh connection */ + fio_disconnect(); + + return NULL; +} + +/* + * Copy WAL segment from archive catalog to pgdata with possible decompression. + * When running in prefetch mode, we should not error out. + */ +bool +get_wal_file(const char *filename, const char *from_fullpath, + const char *to_fullpath, bool prefetch_mode) +{ + int rc = FILE_MISSING; + FILE *out; + char from_fullpath_gz[MAXPGPATH]; + bool src_partial = false; + + snprintf(from_fullpath_gz, sizeof(from_fullpath_gz), "%s.gz", from_fullpath); + + /* open destination file */ + out = fopen(to_fullpath, PG_BINARY_W); + if (!out) + { + elog(WARNING, "Failed to open file '%s': %s", + to_fullpath, strerror(errno)); + return false; + } + + if (chmod(to_fullpath, FILE_PERMISSION) == -1) + { + elog(WARNING, "Cannot change mode of file '%s': %s", + to_fullpath, strerror(errno)); + fclose(out); + unlink(to_fullpath); + return false; + } + + /* disable buffering for output file */ + setvbuf(out, NULL, _IONBF, BUFSIZ); + + /* In prefetch mode, we do look only for full WAL segments + * In non-prefetch mode, do look up '.partial' and '.gz.partial' + * segments. + */ + if (fio_is_remote(FIO_BACKUP_HOST)) + { + char *errmsg = NULL; + /* get file via ssh */ +#ifdef HAVE_LIBZ + /* If requested file is regular WAL segment, then try to open it with '.gz' suffix... */ + if (IsXLogFileName(filename)) + rc = fio_send_file_gz(from_fullpath_gz, out, &errmsg); + if (rc == FILE_MISSING) +#endif + /* ... failing that, use uncompressed */ + rc = fio_send_file(from_fullpath, out, false, NULL, &errmsg); + + /* When not in prefetch mode, try to use partial file */ + if (rc == FILE_MISSING && !prefetch_mode && IsXLogFileName(filename)) + { + char from_partial[MAXPGPATH]; + +#ifdef HAVE_LIBZ + /* '.gz.partial' goes first ... */ + snprintf(from_partial, sizeof(from_partial), "%s.gz.partial", from_fullpath); + rc = fio_send_file_gz(from_partial, out, &errmsg); + if (rc == FILE_MISSING) +#endif + { + /* ... failing that, use '.partial' */ + snprintf(from_partial, sizeof(from_partial), "%s.partial", from_fullpath); + rc = fio_send_file(from_partial, out, false, NULL, &errmsg); + } + + if (rc == SEND_OK) + src_partial = true; + } + + if (rc == WRITE_FAILED) + elog(WARNING, "Cannot write to file '%s': %s", + to_fullpath, strerror(errno)); + + if (errmsg) + elog(WARNING, "%s", errmsg); + + pg_free(errmsg); + } + else + { + /* get file locally */ +#ifdef HAVE_LIBZ + /* If requested file is regular WAL segment, then try to open it with '.gz' suffix... */ + if (IsXLogFileName(filename)) + rc = get_wal_file_internal(from_fullpath_gz, to_fullpath, out, true); + if (rc == FILE_MISSING) +#endif + /* ... failing that, use uncompressed */ + rc = get_wal_file_internal(from_fullpath, to_fullpath, out, false); + + /* When not in prefetch mode, try to use partial file */ + if (rc == FILE_MISSING && !prefetch_mode && IsXLogFileName(filename)) + { + char from_partial[MAXPGPATH]; + +#ifdef HAVE_LIBZ + /* '.gz.partial' goes first ... */ + snprintf(from_partial, sizeof(from_partial), "%s.gz.partial", from_fullpath); + rc = get_wal_file_internal(from_partial, to_fullpath, out, true); + if (rc == FILE_MISSING) +#endif + { + /* ... failing that, use '.partial' */ + snprintf(from_partial, sizeof(from_partial), "%s.partial", from_fullpath); + rc = get_wal_file_internal(from_partial, to_fullpath, out, false); + } + + if (rc == SEND_OK) + src_partial = true; + } + } + + if (!prefetch_mode && (rc == FILE_MISSING)) + elog(LOG, "Target WAL file is missing: %s", filename); + + if (rc < 0) + { + fclose(out); + unlink(to_fullpath); + return false; + } + + /* If partial file was used as source, then it is very likely that destination + * file is not equal to XLOG_SEG_SIZE - that is the way pg_receivexlog works. + * We must manually extent it up to XLOG_SEG_SIZE. + */ + if (src_partial) + { + + if (fflush(out) != 0) + { + elog(WARNING, "Cannot flush file \"%s\": %s", to_fullpath, strerror(errno)); + fclose(out); + unlink(to_fullpath); + return false; + } + + if (ftruncate(fileno(out), xlog_seg_size) != 0) + { + elog(WARNING, "Cannot extend file \"%s\": %s", to_fullpath, strerror(errno)); + fclose(out); + unlink(to_fullpath); + return false; + } + } + + if (fclose(out) != 0) + { + elog(WARNING, "Cannot close file '%s': %s", to_fullpath, strerror(errno)); + unlink(to_fullpath); + return false; + } + + elog(LOG, "WAL file successfully %s: %s", + prefetch_mode ? "prefetched" : "copied", filename); + return true; +} + +/* + * Copy WAL segment with possible decompression from local archive. + * Return codes: + * FILE_MISSING (-1) + * OPEN_FAILED (-2) + * READ_FAILED (-3) + * WRITE_FAILED (-4) + * ZLIB_ERROR (-5) + */ +int +get_wal_file_internal(const char *from_path, const char *to_path, FILE *out, + bool is_decompress) +{ +#ifdef HAVE_LIBZ + gzFile gz_in = NULL; +#endif + FILE *in = NULL; + char *buf = pgut_malloc(OUT_BUF_SIZE); /* 1MB buffer */ + int exit_code = 0; + + elog(LOG, "Attempting to %s WAL file '%s'", + is_decompress ? "open compressed" : "open", from_path); + + /* open source file for read */ + if (!is_decompress) + { + in = fopen(from_path, PG_BINARY_R); + if (in == NULL) + { + if (errno == ENOENT) + exit_code = FILE_MISSING; + else + { + elog(WARNING, "Cannot open source WAL file \"%s\": %s", + from_path, strerror(errno)); + exit_code = OPEN_FAILED; + } + goto cleanup; + } + + /* disable stdio buffering */ + setvbuf(out, NULL, _IONBF, BUFSIZ); + } +#ifdef HAVE_LIBZ + else + { + gz_in = gzopen(from_path, PG_BINARY_R); + if (gz_in == NULL) + { + if (errno == ENOENT) + exit_code = FILE_MISSING; + else + { + elog(WARNING, "Cannot open compressed WAL file \"%s\": %s", + from_path, strerror(errno)); + exit_code = OPEN_FAILED; + } + + goto cleanup; + } + } +#endif + + /* copy content */ + for (;;) + { + int read_len = 0; + +#ifdef HAVE_LIBZ + if (is_decompress) + { + read_len = gzread(gz_in, buf, OUT_BUF_SIZE); + + if (read_len <= 0) + { + if (gzeof(gz_in)) + break; + else + { + elog(WARNING, "Cannot read compressed WAL file \"%s\": %s", + from_path, get_gz_error(gz_in, errno)); + exit_code = READ_FAILED; + break; + } + } + } + else +#endif + { + read_len = fread(buf, 1, OUT_BUF_SIZE, in); + + if (ferror(in)) + { + elog(WARNING, "Cannot read source WAL file \"%s\": %s", + from_path, strerror(errno)); + exit_code = READ_FAILED; + break; + } + + if (read_len == 0 && feof(in)) + break; + } + + if (read_len > 0) + { + if (fwrite(buf, 1, read_len, out) != read_len) + { + elog(WARNING, "Cannot write to WAL file '%s': %s", + to_path, strerror(errno)); + exit_code = WRITE_FAILED; + break; + } + } + } + +cleanup: +#ifdef HAVE_LIBZ + if (gz_in) + gzclose(gz_in); +#endif + if (in) + fclose(in); + + pg_free(buf); + return exit_code; +} + +bool next_wal_segment_exists(TimeLineID tli, XLogSegNo segno, const char *prefetch_dir, uint32 wal_seg_size) { - struct stat st; + char next_wal_filename[MAXFNAMELEN]; + char next_wal_fullpath[MAXPGPATH]; - if (fio_stat(from_path, &st, true, from_location) == -1) + GetXLogFileName(next_wal_filename, tli, segno + 1, wal_seg_size); + + join_path_components(next_wal_fullpath, prefetch_dir, next_wal_filename); + + if (access(next_wal_fullpath, F_OK) == 0) + return true; + + return false; +} + +/* Try to use content of prefetch directory to satisfy request for WAL segment + * If file is found, then validate it and rename. + * If requested file do not exists or validation has failed, then + * caller must copy WAL file directly from archive. + */ +bool wal_satisfy_from_prefetch(TimeLineID tli, XLogSegNo segno, const char *wal_file_name, + const char *prefetch_dir, const char *absolute_wal_file_path, + uint32 wal_seg_size, bool parse_wal) +{ + char prefetched_file[MAXPGPATH]; + + join_path_components(prefetched_file, prefetch_dir, wal_file_name); + + /* If prefetched file do not exists, then nothing can be done */ + if (access(prefetched_file, F_OK) != 0) + return false; + + /* If the next WAL segment do not exists in prefetch directory, + * then current segment cannot be validated, therefore cannot be used + * to satisfy recovery request. + */ + if (parse_wal && !next_wal_segment_exists(tli, segno, prefetch_dir, wal_seg_size)) + return false; + + if (parse_wal && !validate_wal_segment(tli, segno, prefetch_dir, wal_seg_size)) { - if (unlink_on_error) - fio_unlink(to_path, to_location); - elog(ERROR, "Cannot stat file \"%s\": %s", - from_path, strerror(errno)); + /* prefetched WAL segment is not looking good */ + elog(LOG, "Prefetched WAL segment %s is invalid, cannot use it", wal_file_name); + unlink(prefetched_file); + return false; } - if (fio_chmod(to_path, st.st_mode, to_location) == -1) + /* file is available in prefetch directory */ + if (rename(prefetched_file, absolute_wal_file_path) == 0) + return true; + else { - if (unlink_on_error) - fio_unlink(to_path, to_location); - elog(ERROR, "Cannot change mode of file \"%s\": %s", - to_path, strerror(errno)); + elog(WARNING, "Cannot rename file '%s' to '%s': %s", + prefetched_file, absolute_wal_file_path, strerror(errno)); + unlink(prefetched_file); } + + return false; +} + +/* + * Maintain prefetch directory: drop redundant files + * Return number of files in prefetch directory. + */ +uint32 maintain_prefetch(const char *prefetch_dir, XLogSegNo first_segno, uint32 wal_seg_size) +{ + DIR *dir; + struct dirent *dir_ent; + uint32 n_files = 0; + + XLogSegNo segno; + TimeLineID tli; + + char fullpath[MAXPGPATH]; + + dir = opendir(prefetch_dir); + if (dir == NULL) + { + if (errno != ENOENT) + elog(WARNING, "Cannot open directory \"%s\": %s", prefetch_dir, strerror(errno)); + + return n_files; + } + + while ((dir_ent = readdir(dir))) + { + /* Skip entries point current dir or parent dir */ + if (strcmp(dir_ent->d_name, ".") == 0 || + strcmp(dir_ent->d_name, "..") == 0) + continue; + + if (IsXLogFileName(dir_ent->d_name)) + { + + GetXLogFromFileName(dir_ent->d_name, &tli, &segno, wal_seg_size); + + /* potentially useful segment, keep it */ + if (segno >= first_segno) + { + n_files++; + continue; + } + } + + join_path_components(fullpath, prefetch_dir, dir_ent->d_name); + unlink(fullpath); + } + + closedir(dir); + + return n_files; } diff --git a/src/backup.c b/src/backup.c index 29d6d34e6..78c3512e9 100644 --- a/src/backup.c +++ b/src/backup.c @@ -3,7 +3,7 @@ * backup.c: backup DB cluster, archived WAL * * Portions Copyright (c) 2009-2013, NIPPON TELEGRAPH AND TELEPHONE CORPORATION - * Portions Copyright (c) 2015-2019, Postgres Professional + * Portions Copyright (c) 2015-2022, Postgres Professional * *------------------------------------------------------------------------- */ @@ -13,9 +13,11 @@ #if PG_VERSION_NUM < 110000 #include "catalog/catalog.h" #endif +#if PG_VERSION_NUM < 120000 +#include "access/transam.h" +#endif #include "catalog/pg_tablespace.h" #include "pgtar.h" -#include "receivelog.h" #include "streamutil.h" #include @@ -25,63 +27,19 @@ #include "utils/thread.h" #include "utils/file.h" - -/* - * Macro needed to parse ptrack. - * NOTE Keep those values synchronized with definitions in ptrack.h - */ -#define PTRACK_BITS_PER_HEAPBLOCK 1 -#define HEAPBLOCKS_PER_BYTE (BITS_PER_BYTE / PTRACK_BITS_PER_HEAPBLOCK) - -static int standby_message_timeout = 10 * 1000; /* 10 sec = default */ -static XLogRecPtr stop_backup_lsn = InvalidXLogRecPtr; -static XLogRecPtr stop_stream_lsn = InvalidXLogRecPtr; - -/* - * How long we should wait for streaming end in seconds. - * Retrieved as checkpoint_timeout + checkpoint_timeout * 0.1 - */ -static uint32 stream_stop_timeout = 0; -/* Time in which we started to wait for streaming end */ -static time_t stream_stop_begin = 0; - -const char *progname = "pg_probackup"; +//const char *progname = "pg_probackup"; /* list of files contained in backup */ -static parray *backup_files_list = NULL; +parray *backup_files_list = NULL; /* We need critical section for datapagemap_add() in case of using threads */ static pthread_mutex_t backup_pagemap_mutex = PTHREAD_MUTEX_INITIALIZER; -/* - * We need to wait end of WAL streaming before execute pg_stop_backup(). - */ -typedef struct -{ - const char *basedir; - PGconn *conn; - - /* - * Return value from the thread. - * 0 means there is no error, 1 - there is an error. - */ - int ret; - - XLogRecPtr startpos; - TimeLineID starttli; -} StreamThreadArg; - -static pthread_t stream_thread; -static StreamThreadArg stream_thread_arg = {"", NULL, 1}; - -static int is_ptrack_enable = false; -bool is_ptrack_support = false; +// TODO: move to PGnodeInfo bool exclusive_backup = false; /* Is pg_start_backup() was executed */ -static bool backup_in_progress = false; -/* Is pg_stop_backup() was sent */ -static bool pg_stop_backup_is_sent = false; +bool backup_in_progress = false; /* * Backup routines @@ -90,70 +48,58 @@ static void backup_cleanup(bool fatal, void *userdata); static void *backup_files(void *arg); -static void do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo); +static void do_backup_pg(InstanceState *instanceState, PGconn *backup_conn, + PGNodeInfo *nodeInfo, bool no_sync, bool backup_logs); -static void pg_start_backup(const char *label, bool smooth, pgBackup *backup, - PGNodeInfo *nodeInfo, PGconn *backup_conn, PGconn *master_conn); static void pg_switch_wal(PGconn *conn); -static void pg_stop_backup(pgBackup *backup, PGconn *pg_startbackup_conn, PGNodeInfo *nodeInfo); -static int checkpoint_timeout(PGconn *backup_conn); -//static void backup_list_file(parray *files, const char *root, ) -static XLogRecPtr wait_wal_lsn(XLogRecPtr lsn, bool is_start_lsn, - bool wait_prev_segment); -static void make_pagemap_from_ptrack(parray* files, PGconn* backup_conn); -static void *StreamLog(void *arg); -static void IdentifySystem(StreamThreadArg *stream_thread_arg); +static void pg_stop_backup(InstanceState *instanceState, pgBackup *backup, PGconn *pg_startbackup_conn, PGNodeInfo *nodeInfo); static void check_external_for_tablespaces(parray *external_list, PGconn *backup_conn); +static parray *get_database_map(PGconn *pg_startbackup_conn); -/* Ptrack functions */ -static void pg_ptrack_clear(PGconn *backup_conn); -static bool pg_ptrack_support(PGconn *backup_conn); -static bool pg_ptrack_enable(PGconn *backup_conn); -static bool pg_ptrack_get_and_clear_db(Oid dbOid, Oid tblspcOid, - PGconn *backup_conn); -static char *pg_ptrack_get_and_clear(Oid tablespace_oid, - Oid db_oid, - Oid rel_oid, - size_t *result_size, - PGconn *backup_conn); -static XLogRecPtr get_last_ptrack_lsn(PGconn *backup_conn); +/* pgpro specific functions */ +static bool pgpro_support(PGconn *conn); /* Check functions */ -static bool pg_checksum_enable(PGconn *conn); +static bool pg_is_checksum_enabled(PGconn *conn); static bool pg_is_in_recovery(PGconn *conn); static bool pg_is_superuser(PGconn *conn); static void check_server_version(PGconn *conn, PGNodeInfo *nodeInfo); static void confirm_block_size(PGconn *conn, const char *name, int blcksz); -static void set_cfs_datafiles(parray *files, const char *root, char *relative, size_t i); +static void rewind_and_mark_cfs_datafiles(parray *files, const char *root, char *relative, size_t i); +static bool remove_excluded_files_criterion(void *value, void *exclude_args); +static void backup_cfs_segment(int i, pgFile *file, backup_files_arg *arguments); +static void process_file(int i, pgFile *file, backup_files_arg *arguments); + +static StopBackupCallbackParams stop_callback_params; static void backup_stopbackup_callback(bool fatal, void *userdata) { - PGconn *pg_startbackup_conn = (PGconn *) userdata; + StopBackupCallbackParams *st = (StopBackupCallbackParams *) userdata; /* * If backup is in progress, notify stop of backup to PostgreSQL */ if (backup_in_progress) { - elog(WARNING, "backup in progress, stop backup"); - pg_stop_backup(NULL, pg_startbackup_conn, NULL); /* don't care about stop_lsn in case of error */ + elog(WARNING, "A backup is in progress, stopping it."); + /* don't care about stop_lsn in case of error */ + pg_stop_backup_send(st->conn, st->server_version, current.from_replica, exclusive_backup, NULL); } } /* * Take a backup of a single postgresql instance. - * Move files from 'pgdata' to a subdirectory in 'backup_path'. + * Move files from 'pgdata' to a subdirectory in backup catalog. */ static void -do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) +do_backup_pg(InstanceState *instanceState, PGconn *backup_conn, + PGNodeInfo *nodeInfo, bool no_sync, bool backup_logs) { int i; - char database_path[MAXPGPATH]; char external_prefix[MAXPGPATH]; /* Temp value. Used as template */ - char dst_backup_path[MAXPGPATH]; char label[1024]; XLogRecPtr prev_backup_start_lsn = InvalidXLogRecPtr; @@ -166,12 +112,19 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) parray *prev_backup_filelist = NULL; parray *backup_list = NULL; parray *external_dirs = NULL; + parray *database_map = NULL; + + /* used for multitimeline incremental backup */ + parray *tli_list = NULL; - pgFile *pg_control = NULL; - PGconn *master_conn = NULL; - PGconn *pg_startbackup_conn = NULL; + /* for fancy reporting */ + time_t start_time, end_time; + char pretty_time[20]; + char pretty_bytes[20]; - elog(LOG, "Database backup start"); + pgFile *src_pg_control_file = NULL; + + elog(INFO, "Database backup start"); if(current.external_dir_str) { external_dirs = make_external_directory_list(current.external_dir_str, @@ -179,8 +132,20 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) check_external_for_tablespaces(external_dirs, backup_conn); } + /* notify start of backup to PostgreSQL server */ + time2iso(label, lengthof(label), current.start_time, false); + strncat(label, " with pg_probackup", lengthof(label) - + strlen(" with pg_probackup")); + + /* Call pg_start_backup function in PostgreSQL connect */ + pg_start_backup(label, smooth_checkpoint, ¤t, nodeInfo, backup_conn); + /* Obtain current timeline */ - current.tli = get_current_timeline(false); +#if PG_VERSION_NUM >= 90600 + current.tli = get_current_timeline(backup_conn); +#else + current.tli = get_current_timeline_from_control(instance_config.pgdata, FIO_DB_HOST, false); +#endif /* * In incremental backup mode ensure that already-validated @@ -190,136 +155,163 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) current.backup_mode == BACKUP_MODE_DIFF_PTRACK || current.backup_mode == BACKUP_MODE_DIFF_DELTA) { - char prev_backup_filelist_path[MAXPGPATH]; - /* get list of backups already taken */ - backup_list = catalog_get_backup_list(INVALID_BACKUP_ID); + backup_list = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); prev_backup = catalog_get_last_data_backup(backup_list, current.tli, current.start_time); if (prev_backup == NULL) - elog(ERROR, "Valid backup on current timeline %X is not found. " - "Create new FULL backup before an incremental one.", + { + /* try to setup multi-timeline backup chain */ + elog(WARNING, "Valid full backup on current timeline %u is not found, " + "trying to look up on previous timelines", current.tli); - pgBackupGetPath(prev_backup, prev_backup_filelist_path, - lengthof(prev_backup_filelist_path), DATABASE_FILE_LIST); + tli_list = get_history_streaming(&instance_config.conn_opt, current.tli, backup_list); + if (!tli_list) + { + elog(WARNING, "Failed to obtain current timeline history file via replication protocol"); + /* fallback to using archive */ + tli_list = catalog_get_timelines(instanceState, &instance_config); + } + + if (parray_num(tli_list) == 0) + elog(WARNING, "Cannot find valid backup on previous timelines, " + "WAL archive is not available"); + else + { + prev_backup = get_multi_timeline_parent(backup_list, tli_list, current.tli, + current.start_time, &instance_config); + + if (prev_backup == NULL) + elog(WARNING, "Cannot find valid backup on previous timelines"); + } + + /* failed to find suitable parent, error out */ + if (!prev_backup) + elog(ERROR, "Create new full backup before an incremental one"); + } + } + + if (prev_backup) + { + if (parse_program_version(prev_backup->program_version) > parse_program_version(PROGRAM_VERSION)) + elog(ERROR, "pg_probackup binary version is %s, but backup %s version is %s. " + "pg_probackup do not guarantee to be forward compatible. " + "Please upgrade pg_probackup binary.", + PROGRAM_VERSION, backup_id_of(prev_backup), prev_backup->program_version); + + elog(INFO, "Parent backup: %s", backup_id_of(prev_backup)); + /* Files of previous backup needed by DELTA backup */ - prev_backup_filelist = dir_read_file_list(NULL, NULL, prev_backup_filelist_path, FIO_BACKUP_HOST); + prev_backup_filelist = get_backup_filelist(prev_backup, true); /* If lsn is not NULL, only pages with higher lsn will be copied. */ prev_backup_start_lsn = prev_backup->start_lsn; current.parent_backup = prev_backup->start_time; - write_backup(¤t); + write_backup(¤t, true); } /* * It`s illegal to take PTRACK backup if LSN from ptrack_control() is not - * equal to stop_lsn of previous backup. + * equal to start_lsn of previous backup. */ if (current.backup_mode == BACKUP_MODE_DIFF_PTRACK) { - XLogRecPtr ptrack_lsn = get_last_ptrack_lsn(backup_conn); + XLogRecPtr ptrack_lsn = get_last_ptrack_lsn(backup_conn, nodeInfo); - if (ptrack_lsn > prev_backup->stop_lsn || ptrack_lsn == InvalidXLogRecPtr) + // new ptrack (>=2.0) is more robust and checks Start LSN + if (ptrack_lsn > prev_backup->start_lsn || ptrack_lsn == InvalidXLogRecPtr) { - elog(ERROR, "LSN from ptrack_control %X/%X differs from STOP LSN of previous backup %X/%X.\n" + elog(ERROR, "LSN from ptrack_control %X/%X is greater than Start LSN of previous backup %X/%X.\n" "Create new full backup before an incremental one.", (uint32) (ptrack_lsn >> 32), (uint32) (ptrack_lsn), - (uint32) (prev_backup->stop_lsn >> 32), - (uint32) (prev_backup->stop_lsn)); + (uint32) (prev_backup->start_lsn >> 32), + (uint32) (prev_backup->start_lsn)); } } - /* Clear ptrack files for FULL and PAGE backup */ - if (current.backup_mode != BACKUP_MODE_DIFF_PTRACK && is_ptrack_enable) - pg_ptrack_clear(backup_conn); - - /* notify start of backup to PostgreSQL server */ - time2iso(label, lengthof(label), current.start_time); - strncat(label, " with pg_probackup", lengthof(label) - - strlen(" with pg_probackup")); - - /* Create connection to master server needed to call pg_start_backup */ - if (current.from_replica && exclusive_backup) - { - master_conn = pgut_connect(instance_config.master_conn_opt.pghost, - instance_config.master_conn_opt.pgport, - instance_config.master_conn_opt.pgdatabase, - instance_config.master_conn_opt.pguser); - pg_startbackup_conn = master_conn; - } - else - pg_startbackup_conn = backup_conn; - - pg_start_backup(label, smooth_checkpoint, ¤t, nodeInfo, backup_conn, pg_startbackup_conn); - - /* For incremental backup check that start_lsn is not from the past */ + /* For incremental backup check that start_lsn is not from the past + * Though it will not save us if PostgreSQL instance is actually + * restored STREAM backup. + */ if (current.backup_mode != BACKUP_MODE_FULL && prev_backup->start_lsn > current.start_lsn) elog(ERROR, "Current START LSN %X/%X is lower than START LSN %X/%X of previous backup %s. " "It may indicate that we are trying to backup PostgreSQL instance from the past.", (uint32) (current.start_lsn >> 32), (uint32) (current.start_lsn), (uint32) (prev_backup->start_lsn >> 32), (uint32) (prev_backup->start_lsn), - base36enc(prev_backup->start_time)); + backup_id_of(prev_backup)); /* Update running backup meta with START LSN */ - write_backup(¤t); + write_backup(¤t, true); + + /* In PAGE mode or in ARCHIVE wal-mode wait for current segment */ + if (current.backup_mode == BACKUP_MODE_DIFF_PAGE || !current.stream) + { + /* Check that archive_dir can be reached */ + if (fio_access(instanceState->instance_wal_subdir_path, F_OK, FIO_BACKUP_HOST) != 0) + elog(ERROR, "WAL archive directory is not accessible \"%s\": %s", + instanceState->instance_wal_subdir_path, strerror(errno)); - pgBackupGetPath(¤t, database_path, lengthof(database_path), - DATABASE_DIR); - pgBackupGetPath(¤t, external_prefix, lengthof(external_prefix), - EXTERNAL_DIR); + /* + * Do not wait start_lsn for stream backup. + * Because WAL streaming will start after pg_start_backup() in stream + * mode. + */ + wait_wal_lsn(instanceState->instance_wal_subdir_path, current.start_lsn, true, current.tli, false, true, ERROR, false); + } /* start stream replication */ - if (stream_wal) + if (current.stream) { - /* How long we should wait for streaming end after pg_stop_backup */ - stream_stop_timeout = checkpoint_timeout(backup_conn); - stream_stop_timeout = stream_stop_timeout + stream_stop_timeout * 0.1; + char stream_xlog_path[MAXPGPATH]; - join_path_components(dst_backup_path, database_path, PG_XLOG_DIR); - fio_mkdir(dst_backup_path, DIR_PERMISSION, FIO_BACKUP_HOST); + join_path_components(stream_xlog_path, current.database_dir, PG_XLOG_DIR); + fio_mkdir(stream_xlog_path, DIR_PERMISSION, FIO_BACKUP_HOST); - stream_thread_arg.basedir = dst_backup_path; + start_WAL_streaming(backup_conn, stream_xlog_path, &instance_config.conn_opt, + current.start_lsn, current.tli, true); - /* - * Connect in replication mode to the server. + /* Make sure that WAL streaming is working + * PAGE backup in stream mode is waited twice, first for + * segment in WAL archive and then for streamed segment */ - stream_thread_arg.conn = pgut_connect_replication(instance_config.conn_opt.pghost, - instance_config.conn_opt.pgport, - instance_config.conn_opt.pgdatabase, - instance_config.conn_opt.pguser); - /* sanity */ - IdentifySystem(&stream_thread_arg); - - /* By default there are some error */ - stream_thread_arg.ret = 1; - /* we must use startpos as start_lsn from start_backup */ - stream_thread_arg.startpos = current.start_lsn; - stream_thread_arg.starttli = current.tli; - - thread_interrupted = false; - pthread_create(&stream_thread, NULL, StreamLog, &stream_thread_arg); + wait_wal_lsn(stream_xlog_path, current.start_lsn, true, current.tli, false, true, ERROR, true); } - /* initialize backup list */ + /* initialize backup's file list */ backup_files_list = parray_new(); + join_path_components(external_prefix, current.root_dir, EXTERNAL_DIR); /* list files with the logical path. omit $PGDATA */ - dir_list_file(backup_files_list, instance_config.pgdata, - true, true, false, 0, FIO_DB_HOST); + fio_list_dir(backup_files_list, instance_config.pgdata, + true, true, false, backup_logs, true, 0); + + /* + * Get database_map (name to oid) for use in partial restore feature. + * It's possible that we fail and database_map will be NULL. + */ + database_map = get_database_map(backup_conn); /* * Append to backup list all files and directories * from external directory option */ if (external_dirs) + { for (i = 0; i < parray_num(external_dirs); i++) + { /* External dirs numeration starts with 1. * 0 value is not external dir */ - dir_list_file(backup_files_list, parray_get(external_dirs, i), - false, true, false, i+1, FIO_DB_HOST); + if (fio_is_remote(FIO_DB_HOST)) + fio_list_dir(backup_files_list, parray_get(external_dirs, i), + false, true, false, false, true, i+1); + else + dir_list_file(backup_files_list, parray_get(external_dirs, i), + false, true, false, false, true, i+1, FIO_LOCAL_HOST); + } + } /* close ssh session in main thread */ fio_disconnect(); @@ -332,6 +324,10 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) elog(ERROR, "PGDATA is almost empty. Either it was concurrently deleted or " "pg_probackup do not possess sufficient permissions to list PGDATA content"); + current.pgdata_bytes += calculate_datasize_of_filelist(backup_files_list); + pretty_size(current.pgdata_bytes, pretty_bytes, lengthof(pretty_bytes)); + elog(INFO, "PGDATA size: %s", pretty_bytes); + /* * Sort pathname ascending. It is necessary to create intermediate * directories sequentially. @@ -343,43 +339,67 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) * Sorted array is used at least in parse_filelist_filenames(), * extractPageMap(), make_pagemap_from_ptrack(). */ - parray_qsort(backup_files_list, pgFileComparePath); + parray_qsort(backup_files_list, pgFileCompareRelPathWithExternal); /* Extract information about files in backup_list parsing their names:*/ parse_filelist_filenames(backup_files_list, instance_config.pgdata); + elog(INFO, "Current Start LSN: %X/%X, TLI: %X", + (uint32) (current.start_lsn >> 32), (uint32) (current.start_lsn), + current.tli); if (current.backup_mode != BACKUP_MODE_FULL) - { - elog(LOG, "current_tli:%X", current.tli); - elog(LOG, "prev_backup->start_lsn: %X/%X", - (uint32) (prev_backup->start_lsn >> 32), (uint32) (prev_backup->start_lsn)); - elog(LOG, "current.start_lsn: %X/%X", - (uint32) (current.start_lsn >> 32), (uint32) (current.start_lsn)); - } + elog(INFO, "Parent Start LSN: %X/%X, TLI: %X", + (uint32) (prev_backup->start_lsn >> 32), (uint32) (prev_backup->start_lsn), + prev_backup->tli); /* * Build page mapping in incremental mode. */ - if (current.backup_mode == BACKUP_MODE_DIFF_PAGE) - { - /* - * Build the page map. Obtain information about changed pages - * reading WAL segments present in archives up to the point - * where this backup has started. - */ - extractPageMap(arclog_path, current.tli, instance_config.xlog_seg_size, - prev_backup->start_lsn, current.start_lsn); - } - else if (current.backup_mode == BACKUP_MODE_DIFF_PTRACK) + + if (current.backup_mode == BACKUP_MODE_DIFF_PAGE || + current.backup_mode == BACKUP_MODE_DIFF_PTRACK) { - /* - * Build the page map from ptrack information. - */ - make_pagemap_from_ptrack(backup_files_list, backup_conn); + bool pagemap_isok = true; + + time(&start_time); + elog(INFO, "Extracting pagemap of changed blocks"); + + if (current.backup_mode == BACKUP_MODE_DIFF_PAGE) + { + /* + * Build the page map. Obtain information about changed pages + * reading WAL segments present in archives up to the point + * where this backup has started. + */ + pagemap_isok = extractPageMap(instanceState->instance_wal_subdir_path, + instance_config.xlog_seg_size, + prev_backup->start_lsn, prev_backup->tli, + current.start_lsn, current.tli, tli_list); + } + else if (current.backup_mode == BACKUP_MODE_DIFF_PTRACK) + { + /* + * Build the page map from ptrack information. + */ + make_pagemap_from_ptrack_2(backup_files_list, backup_conn, + nodeInfo->ptrack_schema, + nodeInfo->ptrack_version_num, + prev_backup_start_lsn); + } + + time(&end_time); + + /* TODO: add ms precision */ + if (pagemap_isok) + elog(INFO, "Pagemap successfully extracted, time elapsed: %.0f sec", + difftime(end_time, start_time)); + else + elog(ERROR, "Pagemap extraction failed, time elasped: %.0f sec", + difftime(end_time, start_time)); } /* - * Make directories before backup and setup threads at the same time + * Make directories before backup */ for (i = 0; i < parray_num(backup_files_list); i++) { @@ -389,43 +409,57 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) if (S_ISDIR(file->mode)) { char dirpath[MAXPGPATH]; - char *dir_name; - - if (file->external_dir_num) - dir_name = GetRelativePath(file->path, - parray_get(external_dirs, - file->external_dir_num - 1)); - else - dir_name = GetRelativePath(file->path, instance_config.pgdata); - - elog(VERBOSE, "Create directory \"%s\"", dir_name); if (file->external_dir_num) { char temp[MAXPGPATH]; snprintf(temp, MAXPGPATH, "%s%d", external_prefix, file->external_dir_num); - join_path_components(dirpath, temp, dir_name); + join_path_components(dirpath, temp, file->rel_path); } else - join_path_components(dirpath, database_path, dir_name); + join_path_components(dirpath, current.database_dir, file->rel_path); + + elog(LOG, "Create directory '%s'", dirpath); fio_mkdir(dirpath, DIR_PERMISSION, FIO_BACKUP_HOST); } - /* setup threads */ - pg_atomic_clear_flag(&file->lock); } + /* + * find pg_control file + * We'll copy it last + */ + { + int control_file_elem_index; + pgFile search_key; + MemSet(&search_key, 0, sizeof(pgFile)); + /* pgFileCompareRelPathWithExternal uses only .rel_path and .external_dir_num for comparision */ + search_key.rel_path = XLOG_CONTROL_FILE; + search_key.external_dir_num = 0; + control_file_elem_index = parray_bsearch_index(backup_files_list, &search_key, pgFileCompareRelPathWithExternal); + + if (control_file_elem_index < 0) + elog(ERROR, "File \"%s\" not found in PGDATA %s", XLOG_CONTROL_FILE, current.database_dir); + src_pg_control_file = (pgFile *)parray_get(backup_files_list, control_file_elem_index); + } + + /* setup thread locks */ + pfilearray_clear_locks(backup_files_list); + /* Sort by size for load balancing */ parray_qsort(backup_files_list, pgFileCompareSize); /* Sort the array for binary search */ if (prev_backup_filelist) - parray_qsort(prev_backup_filelist, pgFileComparePathWithExternal); + parray_qsort(prev_backup_filelist, pgFileCompareRelPathWithExternal); /* write initial backup_content.control file and update backup.control */ write_backup_filelist(¤t, backup_files_list, - instance_config.pgdata, external_dirs); - write_backup(¤t); + instance_config.pgdata, external_dirs, true); + write_backup(¤t, true); + + /* Init backup page header map */ + init_header_map(¤t); /* init thread args with own file lists */ threads = (pthread_t *) palloc(sizeof(pthread_t) * num_threads); @@ -435,15 +469,15 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) { backup_files_arg *arg = &(threads_args[i]); + arg->nodeInfo = nodeInfo; arg->from_root = instance_config.pgdata; - arg->to_root = database_path; + arg->to_root = current.database_dir; arg->external_prefix = external_prefix; arg->external_dirs = external_dirs; arg->files_list = backup_files_list; arg->prev_filelist = prev_backup_filelist; arg->prev_start_lsn = prev_backup_start_lsn; - arg->conn_arg.conn = NULL; - arg->conn_arg.cancel_conn = NULL; + arg->hdr_map = &(current.hdr_map); arg->thread_num = i+1; /* By default there are some error */ arg->ret = 1; @@ -452,6 +486,7 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) /* Run threads */ thread_interrupted = false; elog(INFO, "Start transferring data files"); + time(&start_time); for (i = 0; i < num_threads; i++) { backup_files_arg *arg = &(threads_args[i]); @@ -467,24 +502,37 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) if (threads_args[i].ret == 1) backup_isok = false; } - if (backup_isok) - elog(INFO, "Data files are transferred"); - else - elog(ERROR, "Data files transferring failed"); - /* Remove disappeared during backup files from backup_list */ - for (i = 0; i < parray_num(backup_files_list); i++) + /* copy pg_control at very end */ + if (backup_isok) { - pgFile *tmp_file = (pgFile *) parray_get(backup_files_list, i); - if (tmp_file->write_size == FILE_NOT_FOUND) - { - pgFileFree(tmp_file); - parray_remove(backup_files_list, i); - i--; - } + elog(progress ? INFO : LOG, "Progress: Backup file \"%s\"", + src_pg_control_file->rel_path); + + char from_fullpath[MAXPGPATH]; + char to_fullpath[MAXPGPATH]; + join_path_components(from_fullpath, instance_config.pgdata, src_pg_control_file->rel_path); + join_path_components(to_fullpath, current.database_dir, src_pg_control_file->rel_path); + + backup_non_data_file(src_pg_control_file, NULL, + from_fullpath, to_fullpath, + current.backup_mode, current.parent_backup, + true); } + + + time(&end_time); + pretty_time_interval(difftime(end_time, start_time), + pretty_time, lengthof(pretty_time)); + if (backup_isok) + elog(INFO, "Data files are transferred, time elapsed: %s", + pretty_time); + else + elog(ERROR, "Data files transferring failed, time elapsed: %s", + pretty_time); + /* clean previous backup file list */ if (prev_backup_filelist) { @@ -493,79 +541,107 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) } /* Notify end of backup */ - pg_stop_backup(¤t, pg_startbackup_conn, nodeInfo); + pg_stop_backup(instanceState, ¤t, backup_conn, nodeInfo); /* In case of backup from replica >= 9.6 we must fix minRecPoint, * First we must find pg_control in backup_files_list. */ if (current.from_replica && !exclusive_backup) { - char pg_control_path[MAXPGPATH]; - - snprintf(pg_control_path, sizeof(pg_control_path), "%s/%s", - instance_config.pgdata, XLOG_CONTROL_FILE); + pgFile *pg_control = NULL; - for (i = 0; i < parray_num(backup_files_list); i++) - { - pgFile *tmp_file = (pgFile *) parray_get(backup_files_list, i); + pg_control = src_pg_control_file; - if (strcmp(tmp_file->path, pg_control_path) == 0) - { - pg_control = tmp_file; - break; - } - } if (!pg_control) elog(ERROR, "Failed to find file \"%s\" in backup filelist.", - pg_control_path); + XLOG_CONTROL_FILE); + + set_min_recovery_point(pg_control, current.database_dir, current.stop_lsn); + } - set_min_recovery_point(pg_control, database_path, current.stop_lsn); + /* close and sync page header map */ + if (current.hdr_map.fp) + { + cleanup_header_map(&(current.hdr_map)); + + if (fio_sync(current.hdr_map.path, FIO_BACKUP_HOST) != 0) + elog(ERROR, "Cannot sync file \"%s\": %s", current.hdr_map.path, strerror(errno)); } /* close ssh session in main thread */ fio_disconnect(); - /* Add archived xlog files into the list of files of this backup */ - if (stream_wal) + /* + * Add archived xlog files into the list of files of this backup + * NOTHING TO DO HERE + */ + + /* write database map to file and add it to control file */ + if (database_map) { - parray *xlog_files_list; - char pg_xlog_path[MAXPGPATH]; + write_database_map(¤t, database_map, backup_files_list); + /* cleanup */ + parray_walk(database_map, db_map_entry_free); + parray_free(database_map); + } + + /* Print the list of files to backup catalog */ + write_backup_filelist(¤t, backup_files_list, instance_config.pgdata, + external_dirs, true); + /* update backup control file to update size info */ + write_backup(¤t, true); - /* Scan backup PG_XLOG_DIR */ - xlog_files_list = parray_new(); - join_path_components(pg_xlog_path, database_path, PG_XLOG_DIR); - dir_list_file(xlog_files_list, pg_xlog_path, false, true, false, 0, - FIO_BACKUP_HOST); + /* Sync all copied files unless '--no-sync' flag is used */ + if (no_sync) + elog(WARNING, "Backup files are not synced to disk"); + else + { + elog(INFO, "Syncing backup files to disk"); + time(&start_time); - for (i = 0; i < parray_num(xlog_files_list); i++) + for (i = 0; i < parray_num(backup_files_list); i++) { - pgFile *file = (pgFile *) parray_get(xlog_files_list, i); - if (S_ISREG(file->mode)) - { - file->crc = pgFileGetCRC(file->path, true, false, - &file->read_size, FIO_BACKUP_HOST); - file->write_size = file->read_size; - } - /* Remove file path root prefix*/ - if (strstr(file->path, database_path) == file->path) + char to_fullpath[MAXPGPATH]; + pgFile *file = (pgFile *) parray_get(backup_files_list, i); + + /* TODO: sync directory ? */ + if (S_ISDIR(file->mode)) + continue; + + if (file->write_size <= 0) + continue; + + /* construct fullpath */ + if (file->external_dir_num == 0) + join_path_components(to_fullpath, current.database_dir, file->rel_path); + else { - char *ptr = file->path; + char external_dst[MAXPGPATH]; - file->path = pstrdup(GetRelativePath(ptr, database_path)); - free(ptr); + makeExternalDirPathByNum(external_dst, external_prefix, + file->external_dir_num); + join_path_components(to_fullpath, external_dst, file->rel_path); } + + if (fio_sync(to_fullpath, FIO_BACKUP_HOST) != 0) + elog(ERROR, "Cannot sync file \"%s\": %s", to_fullpath, strerror(errno)); } - /* Add xlog files into the list of backed up files */ - parray_concat(backup_files_list, xlog_files_list); - parray_free(xlog_files_list); + + time(&end_time); + pretty_time_interval(difftime(end_time, start_time), + pretty_time, lengthof(pretty_time)); + elog(INFO, "Backup files are synced, time elapsed: %s", pretty_time); } - /* Print the list of files to backup catalog */ - write_backup_filelist(¤t, backup_files_list, instance_config.pgdata, - external_dirs); - /* update backup control file to update size info */ - write_backup(¤t); + /* be paranoid about instance been from the past */ + if (current.backup_mode != BACKUP_MODE_FULL && + current.stop_lsn < prev_backup->stop_lsn) + elog(ERROR, "Current backup STOP LSN %X/%X is lower than STOP LSN %X/%X of previous backup %s. " + "It may indicate that we are trying to backup PostgreSQL instance from the past.", + (uint32) (current.stop_lsn >> 32), (uint32) (current.stop_lsn), + (uint32) (prev_backup->stop_lsn >> 32), (uint32) (prev_backup->stop_lsn), + backup_id_of(prev_backup)); /* clean external directories list */ if (external_dirs) @@ -578,6 +654,12 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) parray_free(backup_list); } + if (tli_list) + { + parray_walk(tli_list, timelineInfoFree); + parray_free(tli_list); + } + parray_walk(backup_files_list, pgFileFree); parray_free(backup_files_list); backup_files_list = NULL; @@ -592,7 +674,7 @@ do_backup_instance(PGconn *backup_conn, PGNodeInfo *nodeInfo) * check remote PostgreSQL instance. * Also checking system ID in this case serves no purpose, because * all work is done by server. - * + * * Returns established connection */ PGconn * @@ -613,13 +695,14 @@ pgdata_basic_setup(ConnectionOptions conn_opt, PGNodeInfo *nodeInfo) nodeInfo->block_size = BLCKSZ; nodeInfo->wal_block_size = XLOG_BLCKSZ; nodeInfo->is_superuser = pg_is_superuser(cur_conn); + nodeInfo->pgpro_support = pgpro_support(cur_conn); current.from_replica = pg_is_in_recovery(cur_conn); /* Confirm that this server version is supported */ check_server_version(cur_conn, nodeInfo); - if (pg_checksum_enable(cur_conn)) + if (pg_is_checksum_enabled(cur_conn)) current.checksum_version = 1; else current.checksum_version = 0; @@ -627,7 +710,7 @@ pgdata_basic_setup(ConnectionOptions conn_opt, PGNodeInfo *nodeInfo) nodeInfo->checksum_version = current.checksum_version; if (current.checksum_version) - elog(LOG, "This PostgreSQL instance was initialized with data block checksums. " + elog(INFO, "This PostgreSQL instance was initialized with data block checksums. " "Data block corruption will be detected"); else elog(WARNING, "This PostgreSQL instance was initialized without data block checksums. " @@ -636,9 +719,9 @@ pgdata_basic_setup(ConnectionOptions conn_opt, PGNodeInfo *nodeInfo) if (nodeInfo->is_superuser) elog(WARNING, "Current PostgreSQL role is superuser. " - "It is not recommended to run backup or checkdb as superuser."); + "It is not recommended to run pg_probackup under superuser."); - StrNCpy(current.server_version, nodeInfo->server_version_str, + strlcpy(current.server_version, nodeInfo->server_version_str, sizeof(current.server_version)); return cur_conn; @@ -646,29 +729,120 @@ pgdata_basic_setup(ConnectionOptions conn_opt, PGNodeInfo *nodeInfo) /* * Entry point of pg_probackup BACKUP subcommand. + * + * if start_time == INVALID_BACKUP_ID then we can generate backup_id */ int -do_backup(time_t start_time, bool no_validate) +do_backup(InstanceState *instanceState, pgSetBackupParams *set_backup_params, + bool no_validate, bool no_sync, bool backup_logs, time_t start_time) { - PGconn *backup_conn = NULL; - PGNodeInfo nodeInfo; - char pretty_data_bytes[20]; + PGconn *backup_conn = NULL; + PGNodeInfo nodeInfo; + time_t latest_backup_id = INVALID_BACKUP_ID; + char pretty_bytes[20]; + + if (!instance_config.pgdata) + elog(ERROR, "No postgres data directory specified.\n" + "Please specify it either using environment variable PGDATA or\n" + "command line option --pgdata (-D)"); /* Initialize PGInfonode */ pgNodeInit(&nodeInfo); - if (!instance_config.pgdata) - elog(ERROR, "required parameter not specified: PGDATA " - "(-D, --pgdata)"); + /* Save list of external directories */ + if (instance_config.external_dir_str && + (pg_strcasecmp(instance_config.external_dir_str, "none") != 0)) + current.external_dir_str = instance_config.external_dir_str; + + /* Find latest backup_id */ + { + parray *backup_list = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); + + if (parray_num(backup_list) > 0) + latest_backup_id = ((pgBackup *)parray_get(backup_list, 0))->backup_id; + + parray_walk(backup_list, pgBackupFree); + parray_free(backup_list); + } + + /* Try to pick backup_id and create backup directory with BACKUP_CONTROL_FILE */ + if (start_time != INVALID_BACKUP_ID) + { + /* If user already choosed backup_id for us, then try to use it. */ + if (start_time <= latest_backup_id) + /* don't care about freeing base36enc_dup memory, we exit anyway */ + elog(ERROR, "Can't assign backup_id from requested start_time (%s), " + "this time must be later that backup %s", + base36enc(start_time), base36enc(latest_backup_id)); + + current.backup_id = start_time; + pgBackupInitDir(¤t, instanceState->instance_backup_subdir_path); + } + else + { + /* We can generate our own unique backup_id + * Sometimes (when we try to backup twice in one second) + * backup_id will be duplicated -> try more times. + */ + int attempts = 10; + + if (time(NULL) < latest_backup_id) + elog(ERROR, "Can't assign backup_id, there is already a backup in future (%s)", + base36enc(latest_backup_id)); + + do + { + current.backup_id = time(NULL); + pgBackupInitDir(¤t, instanceState->instance_backup_subdir_path); + if (current.backup_id == INVALID_BACKUP_ID) + sleep(1); + } + while (current.backup_id == INVALID_BACKUP_ID && attempts-- > 0); + } + + /* If creation of backup dir was unsuccessful, there will be WARNINGS in logs already */ + if (current.backup_id == INVALID_BACKUP_ID) + elog(ERROR, "Can't create backup directory"); + + /* Update backup status and other metainfo. */ + current.status = BACKUP_STATUS_RUNNING; + /* XXX BACKUP_ID change it when backup_id wouldn't match start_time */ + current.start_time = current.backup_id; + + strlcpy(current.program_version, PROGRAM_VERSION, + sizeof(current.program_version)); current.compress_alg = instance_config.compress_alg; current.compress_level = instance_config.compress_level; + elog(INFO, "Backup start, pg_probackup version: %s, instance: %s, backup ID: %s, backup mode: %s, " + "wal mode: %s, remote: %s, compress-algorithm: %s, compress-level: %i", + PROGRAM_VERSION, instanceState->instance_name, backup_id_of(¤t), pgBackupGetBackupMode(¤t, false), + current.stream ? "STREAM" : "ARCHIVE", IsSshProtocol() ? "true" : "false", + deparse_compress_alg(current.compress_alg), current.compress_level); + + if (!lock_backup(¤t, true, true)) + elog(ERROR, "Cannot lock backup %s directory", + backup_id_of(¤t)); + write_backup(¤t, true); + + /* set the error processing function for the backup process */ + pgut_atexit_push(backup_cleanup, NULL); + + elog(LOG, "Backup destination is initialized"); + /* * setup backup_conn, do some compatibility checks and * fill basic info about instance */ backup_conn = pgdata_basic_setup(instance_config.conn_opt, &nodeInfo); + + if (current.from_replica) + elog(INFO, "Backup %s is going to be taken from standby", backup_id_of(¤t)); + + /* TODO, print PostgreSQL full version */ + //elog(INFO, "PostgreSQL version: %s", nodeInfo.server_version_str); + /* * Ensure that backup directory was initialized for the same PostgreSQL * instance we opened connection to. And that target backup database PGDATA @@ -676,32 +850,26 @@ do_backup(time_t start_time, bool no_validate) */ check_system_identifiers(backup_conn, instance_config.pgdata); - elog(INFO, "Backup start, pg_probackup version: %s, instance: %s, backup ID: %s, backup mode: %s, " - "wal-method: %s, remote: %s, replica: %s, compress-algorithm: %s, compress-level: %i", - PROGRAM_VERSION, instance_name, base36enc(start_time), pgBackupGetBackupMode(¤t), - current.stream ? "STREAM" : "ARCHIVE", IsSshProtocol() ? "true" : "false", - current.from_replica ? "true" : "false", deparse_compress_alg(current.compress_alg), - current.compress_level); - /* below perform checks specific for backup command */ #if PG_VERSION_NUM >= 110000 if (!RetrieveWalSegSize(backup_conn)) elog(ERROR, "Failed to retrieve wal_segment_size"); #endif - is_ptrack_support = pg_ptrack_support(backup_conn); - if (is_ptrack_support) - { - is_ptrack_enable = pg_ptrack_enable(backup_conn); - } + get_ptrack_version(backup_conn, &nodeInfo); + // elog(WARNING, "ptrack_version_num %d", ptrack_version_num); + + if (nodeInfo.ptrack_version_num > 0) + nodeInfo.is_ptrack_enabled = pg_is_ptrack_enabled(backup_conn, nodeInfo.ptrack_version_num); if (current.backup_mode == BACKUP_MODE_DIFF_PTRACK) { - if (!is_ptrack_support) + /* ptrack_version_num < 2.0 was already checked in get_ptrack_version() */ + if (nodeInfo.ptrack_version_num == 0) elog(ERROR, "This PostgreSQL instance does not support ptrack"); else { - if(!is_ptrack_enable) + if (!nodeInfo.is_ptrack_enabled) elog(ERROR, "Ptrack is disabled"); } } @@ -711,68 +879,67 @@ do_backup(time_t start_time, bool no_validate) if (instance_config.master_conn_opt.pghost == NULL) elog(ERROR, "Options for connection to master must be provided to perform backup from replica"); - /* Start backup. Update backup status. */ - current.status = BACKUP_STATUS_RUNNING; - current.start_time = start_time; - StrNCpy(current.program_version, PROGRAM_VERSION, - sizeof(current.program_version)); - - /* Save list of external directories */ - if (instance_config.external_dir_str && - pg_strcasecmp(instance_config.external_dir_str, "none") != 0) - { - current.external_dir_str = instance_config.external_dir_str; - } - - /* Create backup directory and BACKUP_CONTROL_FILE */ - if (pgBackupCreateDir(¤t)) - elog(ERROR, "Cannot create backup directory"); - if (!lock_backup(¤t)) - elog(ERROR, "Cannot lock backup %s directory", - base36enc(current.start_time)); - write_backup(¤t); - - elog(LOG, "Backup destination is initialized"); - - /* set the error processing function for the backup process */ - pgut_atexit_push(backup_cleanup, NULL); + /* add note to backup if requested */ + if (set_backup_params && set_backup_params->note) + add_note(¤t, set_backup_params->note); /* backup data */ - do_backup_instance(backup_conn, &nodeInfo); + do_backup_pg(instanceState, backup_conn, &nodeInfo, no_sync, backup_logs); pgut_atexit_pop(backup_cleanup, NULL); /* compute size of wal files of this backup stored in the archive */ if (!current.stream) { - current.wal_bytes = instance_config.xlog_seg_size * - (current.stop_lsn / instance_config.xlog_seg_size - - current.start_lsn / instance_config.xlog_seg_size + 1); + XLogSegNo start_segno; + XLogSegNo stop_segno; + + GetXLogSegNo(current.start_lsn, start_segno, instance_config.xlog_seg_size); + GetXLogSegNo(current.stop_lsn, stop_segno, instance_config.xlog_seg_size); + current.wal_bytes = (stop_segno - start_segno) * instance_config.xlog_seg_size; + + /* + * If start_lsn and stop_lsn are located in the same segment, then + * set wal_bytes to the size of 1 segment. + */ + if (current.wal_bytes <= 0) + current.wal_bytes = instance_config.xlog_seg_size; } /* Backup is done. Update backup status */ current.end_time = time(NULL); current.status = BACKUP_STATUS_DONE; - write_backup(¤t); + write_backup(¤t, true); + + /* Pin backup if requested */ + if (set_backup_params && + (set_backup_params->ttl > 0 || + set_backup_params->expire_time > 0)) + { + pin_backup(¤t, set_backup_params); + } if (!no_validate) - pgBackupValidate(¤t); + pgBackupValidate(¤t, NULL); /* Notify user about backup size */ - pretty_size(current.data_bytes, pretty_data_bytes, lengthof(pretty_data_bytes)); - elog(INFO, "Backup %s real size: %s", base36enc(current.start_time), pretty_data_bytes); + if (current.stream) + pretty_size(current.data_bytes + current.wal_bytes, pretty_bytes, lengthof(pretty_bytes)); + else + pretty_size(current.data_bytes, pretty_bytes, lengthof(pretty_bytes)); + elog(INFO, "Backup %s resident size: %s", backup_id_of(¤t), pretty_bytes); if (current.status == BACKUP_STATUS_OK || current.status == BACKUP_STATUS_DONE) - elog(INFO, "Backup %s completed", base36enc(current.start_time)); + elog(INFO, "Backup %s completed", backup_id_of(¤t)); else - elog(ERROR, "Backup %s failed", base36enc(current.start_time)); + elog(ERROR, "Backup %s failed", backup_id_of(¤t)); /* * After successful backup completion remove backups * which are expired according to retention policies */ if (delete_expired || merge_expired || delete_wal) - do_retention(); + do_retention(instanceState, no_validate, no_sync); return 0; } @@ -783,7 +950,7 @@ do_backup(time_t start_time, bool no_validate) static void check_server_version(PGconn *conn, PGNodeInfo *nodeInfo) { - PGresult *res; + PGresult *res = NULL; /* confirm server version */ nodeInfo->server_version = PQserverVersion(conn); @@ -801,51 +968,67 @@ check_server_version(PGconn *conn, PGNodeInfo *nodeInfo) if (nodeInfo->server_version < 90500) elog(ERROR, - "server version is %s, must be %s or higher", + "Server version is %s, must be %s or higher", nodeInfo->server_version_str, "9.5"); if (current.from_replica && nodeInfo->server_version < 90600) elog(ERROR, - "server version is %s, must be %s or higher for backup from replica", + "Server version is %s, must be %s or higher for backup from replica", nodeInfo->server_version_str, "9.6"); - /* TODO: search pg_proc for pgpro_edition before calling */ - res = pgut_execute_extended(conn, "SELECT pgpro_edition()", - 0, NULL, true, true); + if (nodeInfo->pgpro_support) + res = pgut_execute(conn, "SELECT pg_catalog.pgpro_edition()", 0, NULL); /* * Check major version of connected PostgreSQL and major version of * compiled PostgreSQL. */ #ifdef PGPRO_VERSION - if (PQresultStatus(res) == PGRES_FATAL_ERROR) + if (!res) + { /* It seems we connected to PostgreSQL (not Postgres Pro) */ - elog(ERROR, "%s was built with Postgres Pro %s %s, " - "but connection is made with PostgreSQL %s", - PROGRAM_NAME, PG_MAJORVERSION, PGPRO_EDITION, nodeInfo->server_version_str); - else if (strcmp(nodeInfo->server_version_str, PG_MAJORVERSION) != 0 && - strcmp(PQgetvalue(res, 0, 0), PGPRO_EDITION) != 0) - elog(ERROR, "%s was built with Postgres Pro %s %s, " - "but connection is made with Postgres Pro %s %s", - PROGRAM_NAME, PG_MAJORVERSION, PGPRO_EDITION, - nodeInfo->server_version_str, PQgetvalue(res, 0, 0)); + if(strcmp(PGPRO_EDITION, "1C") != 0) + { + elog(ERROR, "%s was built with Postgres Pro %s %s, " + "but connection is made with PostgreSQL %s", + PROGRAM_NAME, PG_MAJORVERSION, PGPRO_EDITION, nodeInfo->server_version_str); + } + /* We have PostgresPro for 1C and connect to PostgreSQL or PostgresPro for 1C + * Check the major version + */ + if (strcmp(nodeInfo->server_version_str, PG_MAJORVERSION) != 0) + elog(ERROR, "%s was built with PostgrePro %s %s, but connection is made with %s", + PROGRAM_NAME, PG_MAJORVERSION, PGPRO_EDITION, nodeInfo->server_version_str); + } + else + { + if (strcmp(nodeInfo->server_version_str, PG_MAJORVERSION) != 0 && + strcmp(PQgetvalue(res, 0, 0), PGPRO_EDITION) != 0) + elog(ERROR, "%s was built with Postgres Pro %s %s, " + "but connection is made with Postgres Pro %s %s", + PROGRAM_NAME, PG_MAJORVERSION, PGPRO_EDITION, + nodeInfo->server_version_str, PQgetvalue(res, 0, 0)); + } #else - if (PQresultStatus(res) != PGRES_FATAL_ERROR) + if (res) /* It seems we connected to Postgres Pro (not PostgreSQL) */ elog(ERROR, "%s was built with PostgreSQL %s, " "but connection is made with Postgres Pro %s %s", PROGRAM_NAME, PG_MAJORVERSION, nodeInfo->server_version_str, PQgetvalue(res, 0, 0)); - else if (strcmp(nodeInfo->server_version_str, PG_MAJORVERSION) != 0) - elog(ERROR, "%s was built with PostgreSQL %s, but connection is made with %s", - PROGRAM_NAME, PG_MAJORVERSION, nodeInfo->server_version_str); + else + { + if (strcmp(nodeInfo->server_version_str, PG_MAJORVERSION) != 0) + elog(ERROR, "%s was built with PostgreSQL %s, but connection is made with %s", + PROGRAM_NAME, PG_MAJORVERSION, nodeInfo->server_version_str); + } #endif - PQclear(res); + if (res) + PQclear(res); /* Do exclusive backup only for PostgreSQL 9.5 */ - exclusive_backup = nodeInfo->server_version < 90600 || - current.backup_mode == BACKUP_MODE_DIFF_PTRACK; + exclusive_backup = nodeInfo->server_version < 90600; } /* @@ -855,12 +1038,12 @@ check_server_version(PGconn *conn, PGNodeInfo *nodeInfo) * All system identifiers must be equal. */ void -check_system_identifiers(PGconn *conn, char *pgdata) +check_system_identifiers(PGconn *conn, const char *pgdata) { uint64 system_id_conn; uint64 system_id_pgdata; - system_id_pgdata = get_system_identifier(pgdata); + system_id_pgdata = get_system_identifier(pgdata, FIO_DB_HOST, false); system_id_conn = get_remote_system_identifier(conn); /* for checkdb check only system_id_pgdata and system_id_conn */ @@ -899,7 +1082,7 @@ confirm_block_size(PGconn *conn, const char *name, int blcksz) res = pgut_execute(conn, "SELECT pg_catalog.current_setting($1)", 1, &name); if (PQntuples(res) != 1 || PQnfields(res) != 1) - elog(ERROR, "cannot get %s: %s", name, PQerrorMessage(conn)); + elog(ERROR, "Cannot get %s: %s", name, PQerrorMessage(conn)); block_size = strtol(PQgetvalue(res, 0, 0), &endp, 10); if ((endp && *endp) || block_size != blcksz) @@ -913,40 +1096,41 @@ confirm_block_size(PGconn *conn, const char *name, int blcksz) /* * Notify start of backup to PostgreSQL server. */ -static void +void pg_start_backup(const char *label, bool smooth, pgBackup *backup, - PGNodeInfo *nodeInfo, PGconn *backup_conn, PGconn *pg_startbackup_conn) + PGNodeInfo *nodeInfo, PGconn *conn) { PGresult *res; const char *params[2]; uint32 lsn_hi; uint32 lsn_lo; - PGconn *conn; - params[0] = label; - /* For 9.5 replica we call pg_start_backup() on master */ - conn = pg_startbackup_conn; +#if PG_VERSION_NUM >= 150000 + elog(INFO, "wait for pg_backup_start()"); +#else + elog(INFO, "wait for pg_start_backup()"); +#endif /* 2nd argument is 'fast'*/ params[1] = smooth ? "false" : "true"; - if (!exclusive_backup) - res = pgut_execute(conn, - "SELECT pg_catalog.pg_start_backup($1, $2, false)", - 2, - params); - else - res = pgut_execute(conn, - "SELECT pg_catalog.pg_start_backup($1, $2)", - 2, - params); + res = pgut_execute(conn, +#if PG_VERSION_NUM >= 150000 + "SELECT pg_catalog.pg_backup_start($1, $2)", +#else + "SELECT pg_catalog.pg_start_backup($1, $2, false)", +#endif + 2, + params); /* * Set flag that pg_start_backup() was called. If an error will happen it * is necessary to call pg_stop_backup() in backup_cleanup(). */ backup_in_progress = true; - pgut_atexit_push(backup_stopbackup_callback, pg_startbackup_conn); + stop_callback_params.conn = conn; + stop_callback_params.server_version = nodeInfo->server_version; + pgut_atexit_push(backup_stopbackup_callback, &stop_callback_params); /* Extract timeline and LSN from results of pg_start_backup() */ XLogDataFromLSN(PQgetvalue(res, 0, 0), &lsn_hi, &lsn_lo); @@ -955,7 +1139,7 @@ pg_start_backup(const char *label, bool smooth, pgBackup *backup, PQclear(res); - if (current.backup_mode == BACKUP_MODE_DIFF_PAGE && + if ((!backup->stream || backup->backup_mode == BACKUP_MODE_DIFF_PAGE) && !backup->from_replica && !(nodeInfo->server_version < 90600 && !nodeInfo->is_superuser)) @@ -966,32 +1150,18 @@ pg_start_backup(const char *label, bool smooth, pgBackup *backup, * (because in 9.5 only superuser can switch WAL) */ pg_switch_wal(conn); - - if (current.backup_mode == BACKUP_MODE_DIFF_PAGE) - /* In PAGE mode wait for current segment... */ - wait_wal_lsn(backup->start_lsn, true, false); - /* - * Do not wait start_lsn for stream backup. - * Because WAL streaming will start after pg_start_backup() in stream - * mode. - */ - else if (!stream_wal) - /* ...for others wait for previous segment */ - wait_wal_lsn(backup->start_lsn, true, true); } /* * Switch to a new WAL segment. It should be called only for master. * For PG 9.5 it should be called only if pguser is superuser. */ -static void +void pg_switch_wal(PGconn *conn) { PGresult *res; - /* Remove annoying NOTICE messages generated by backend */ - res = pgut_execute(conn, "SET client_min_messages = warning;", 0, NULL); - PQclear(res); + pg_silent_client_messages(conn); #if PG_VERSION_NUM >= 100000 res = pgut_execute(conn, "SELECT pg_catalog.pg_switch_wal()", 0, NULL); @@ -1003,67 +1173,92 @@ pg_switch_wal(PGconn *conn) } /* - * Check if the instance supports ptrack - * TODO Maybe we should rather check ptrack_version()? + * Check if the instance is PostgresPro fork. */ static bool -pg_ptrack_support(PGconn *backup_conn) +pgpro_support(PGconn *conn) { - PGresult *res_db; + PGresult *res; - res_db = pgut_execute(backup_conn, - "SELECT proname FROM pg_proc WHERE proname='ptrack_version'", + res = pgut_execute(conn, + "SELECT proname FROM pg_catalog.pg_proc WHERE proname='pgpro_edition'::name AND pronamespace='pg_catalog'::regnamespace::oid", 0, NULL); - if (PQntuples(res_db) == 0) + + if (PQresultStatus(res) == PGRES_TUPLES_OK && + (PQntuples(res) == 1) && + (strcmp(PQgetvalue(res, 0, 0), "pgpro_edition") == 0)) { - PQclear(res_db); - return false; + PQclear(res); + return true; } - PQclear(res_db); - res_db = pgut_execute(backup_conn, - "SELECT pg_catalog.ptrack_version()", - 0, NULL); - if (PQntuples(res_db) == 0) + PQclear(res); + return false; +} + +/* + * Fill 'datname to Oid' map + * + * This function can fail to get the map for legal reasons, e.g. missing + * permissions on pg_database during `backup`. + * As long as user do not use partial restore feature it`s fine. + * + * To avoid breaking a backward compatibility don't throw an ERROR, + * throw a warning instead of an error and return NULL. + * Caller is responsible for checking the result. + */ +parray * +get_database_map(PGconn *conn) +{ + PGresult *res; + parray *database_map = NULL; + int i; + + /* + * Do not include template0 and template1 to the map + * as default databases that must always be restored. + */ + res = pgut_execute_extended(conn, + "SELECT oid, datname FROM pg_catalog.pg_database " + "WHERE datname NOT IN ('template1'::name, 'template0'::name)", + 0, NULL, true, true); + + /* Don't error out, simply return NULL. See comment above. */ + if (PQresultStatus(res) != PGRES_TUPLES_OK) { - PQclear(res_db); - return false; + PQclear(res); + elog(WARNING, "Failed to get database map: %s", + PQerrorMessage(conn)); + + return NULL; } - /* Now we support only ptrack versions upper than 1.5 */ - if (strcmp(PQgetvalue(res_db, 0, 0), "1.5") != 0 && - strcmp(PQgetvalue(res_db, 0, 0), "1.6") != 0 && - strcmp(PQgetvalue(res_db, 0, 0), "1.7") != 0) + /* Construct database map */ + for (i = 0; i < PQntuples(res); i++) { - elog(WARNING, "Update your ptrack to the version 1.5 or upper. Current version is %s", PQgetvalue(res_db, 0, 0)); - PQclear(res_db); - return false; - } + char *datname = NULL; + db_map_entry *db_entry = (db_map_entry *) pgut_malloc(sizeof(db_map_entry)); - PQclear(res_db); - return true; -} + /* get Oid */ + db_entry->dbOid = atoll(PQgetvalue(res, i, 0)); -/* Check if ptrack is enabled in target instance */ -static bool -pg_ptrack_enable(PGconn *backup_conn) -{ - PGresult *res_db; + /* get datname */ + datname = PQgetvalue(res, i, 1); + db_entry->datname = pgut_malloc(strlen(datname) + 1); + strcpy(db_entry->datname, datname); - res_db = pgut_execute(backup_conn, "SHOW ptrack_enable", 0, NULL); + if (database_map == NULL) + database_map = parray_new(); - if (strcmp(PQgetvalue(res_db, 0, 0), "on") != 0) - { - PQclear(res_db); - return false; + parray_append(database_map, db_entry); } - PQclear(res_db); - return true; + + return database_map; } /* Check if ptrack is enabled in target instance */ static bool -pg_checksum_enable(PGconn *conn) +pg_is_checksum_enabled(PGconn *conn) { PGresult *res_db; @@ -1113,277 +1308,75 @@ pg_is_superuser(PGconn *conn) return false; } -/* Clear ptrack files in all databases of the instance we connected to */ -static void -pg_ptrack_clear(PGconn *backup_conn) -{ - PGresult *res_db, - *res; - const char *dbname; - int i; - Oid dbOid, tblspcOid; - char *params[2]; - - params[0] = palloc(64); - params[1] = palloc(64); - res_db = pgut_execute(backup_conn, "SELECT datname, oid, dattablespace FROM pg_database", - 0, NULL); - - for(i = 0; i < PQntuples(res_db); i++) - { - PGconn *tmp_conn; - - dbname = PQgetvalue(res_db, i, 0); - if (strcmp(dbname, "template0") == 0) - continue; - - dbOid = atoi(PQgetvalue(res_db, i, 1)); - tblspcOid = atoi(PQgetvalue(res_db, i, 2)); - - tmp_conn = pgut_connect(instance_config.conn_opt.pghost, instance_config.conn_opt.pgport, - dbname, - instance_config.conn_opt.pguser); - - res = pgut_execute(tmp_conn, "SELECT pg_catalog.pg_ptrack_clear()", - 0, NULL); - PQclear(res); - - sprintf(params[0], "%i", dbOid); - sprintf(params[1], "%i", tblspcOid); - res = pgut_execute(tmp_conn, "SELECT pg_catalog.pg_ptrack_get_and_clear_db($1, $2)", - 2, (const char **)params); - PQclear(res); - - pgut_disconnect(tmp_conn); - } - - pfree(params[0]); - pfree(params[1]); - PQclear(res_db); -} - -static bool -pg_ptrack_get_and_clear_db(Oid dbOid, Oid tblspcOid, PGconn *backup_conn) -{ - char *params[2]; - char *dbname; - PGresult *res_db; - PGresult *res; - bool result; - - params[0] = palloc(64); - params[1] = palloc(64); - - sprintf(params[0], "%i", dbOid); - res_db = pgut_execute(backup_conn, - "SELECT datname FROM pg_database WHERE oid=$1", - 1, (const char **) params); - /* - * If database is not found, it's not an error. - * It could have been deleted since previous backup. - */ - if (PQntuples(res_db) != 1 || PQnfields(res_db) != 1) - return false; - - dbname = PQgetvalue(res_db, 0, 0); - - /* Always backup all files from template0 database */ - if (strcmp(dbname, "template0") == 0) - { - PQclear(res_db); - return true; - } - PQclear(res_db); - - sprintf(params[0], "%i", dbOid); - sprintf(params[1], "%i", tblspcOid); - res = pgut_execute(backup_conn, "SELECT pg_catalog.pg_ptrack_get_and_clear_db($1, $2)", - 2, (const char **)params); - - if (PQnfields(res) != 1) - elog(ERROR, "cannot perform pg_ptrack_get_and_clear_db()"); - - if (!parse_bool(PQgetvalue(res, 0, 0), &result)) - elog(ERROR, - "result of pg_ptrack_get_and_clear_db() is invalid: %s", - PQgetvalue(res, 0, 0)); - - PQclear(res); - pfree(params[0]); - pfree(params[1]); - - return result; -} - -/* Read and clear ptrack files of the target relation. - * Result is a bytea ptrack map of all segments of the target relation. - * case 1: we know a tablespace_oid, db_oid, and rel_filenode - * case 2: we know db_oid and rel_filenode (no tablespace_oid, because file in pg_default) - * case 3: we know only rel_filenode (because file in pg_global) - */ -static char * -pg_ptrack_get_and_clear(Oid tablespace_oid, Oid db_oid, Oid rel_filenode, - size_t *result_size, PGconn *backup_conn) -{ - PGconn *tmp_conn; - PGresult *res_db, - *res; - char *params[2]; - char *result; - char *val; - - params[0] = palloc(64); - params[1] = palloc(64); - - /* regular file (not in directory 'global') */ - if (db_oid != 0) - { - char *dbname; - - sprintf(params[0], "%i", db_oid); - res_db = pgut_execute(backup_conn, - "SELECT datname FROM pg_database WHERE oid=$1", - 1, (const char **) params); - /* - * If database is not found, it's not an error. - * It could have been deleted since previous backup. - */ - if (PQntuples(res_db) != 1 || PQnfields(res_db) != 1) - return NULL; - - dbname = PQgetvalue(res_db, 0, 0); - - if (strcmp(dbname, "template0") == 0) - { - PQclear(res_db); - return NULL; - } - - tmp_conn = pgut_connect(instance_config.conn_opt.pghost, instance_config.conn_opt.pgport, - dbname, - instance_config.conn_opt.pguser); - sprintf(params[0], "%i", tablespace_oid); - sprintf(params[1], "%i", rel_filenode); - res = pgut_execute(tmp_conn, "SELECT pg_catalog.pg_ptrack_get_and_clear($1, $2)", - 2, (const char **)params); - - if (PQnfields(res) != 1) - elog(ERROR, "cannot get ptrack file from database \"%s\" by tablespace oid %u and relation oid %u", - dbname, tablespace_oid, rel_filenode); - PQclear(res_db); - pgut_disconnect(tmp_conn); - } - /* file in directory 'global' */ - else - { - /* - * execute ptrack_get_and_clear for relation in pg_global - * Use backup_conn, cause we can do it from any database. - */ - sprintf(params[0], "%i", tablespace_oid); - sprintf(params[1], "%i", rel_filenode); - res = pgut_execute(backup_conn, "SELECT pg_catalog.pg_ptrack_get_and_clear($1, $2)", - 2, (const char **)params); - - if (PQnfields(res) != 1) - elog(ERROR, "cannot get ptrack file from pg_global tablespace and relation oid %u", - rel_filenode); - } - - val = PQgetvalue(res, 0, 0); - - /* TODO Now pg_ptrack_get_and_clear() returns bytea ending with \x. - * It should be fixed in future ptrack releases, but till then we - * can parse it. - */ - if (strcmp("x", val+1) == 0) - { - /* Ptrack file is missing */ - return NULL; - } - - result = (char *) PQunescapeBytea((unsigned char *) PQgetvalue(res, 0, 0), - result_size); - PQclear(res); - pfree(params[0]); - pfree(params[1]); - - return result; -} - /* - * Wait for target 'lsn'. + * Wait for target LSN or WAL segment, containing target LSN. * - * If current backup started in archive mode wait for 'lsn' to be archived in - * archive 'wal' directory with WAL segment file. - * If current backup started in stream mode wait for 'lsn' to be streamed in - * 'pg_wal' directory. + * Depending on value of flag in_stream_dir wait for target LSN to archived or + * streamed in 'archive_dir' or 'pg_wal' directory. * - * If 'is_start_lsn' is true and backup mode is PAGE then we wait for 'lsn' to - * be archived in archive 'wal' directory regardless stream mode. + * If flag 'is_start_lsn' is set then issue warning for first-time users. + * If flag 'in_prev_segment' is set, look for LSN in previous segment, + * with EndRecPtr >= Target LSN. It should be used only for solving + * invalid XRecOff problem. + * If flag 'segment_only' is set, then, instead of waiting for LSN, wait for segment, + * containing that LSN. + * If flags 'in_prev_segment' and 'segment_only' are both set, then wait for + * previous segment. * - * If 'wait_prev_segment' wait for previous segment. + * Flag 'in_stream_dir' determine whether we looking for WAL in 'pg_wal' directory or + * in archive. Do note, that we cannot rely sorely on global variable 'stream_wal' (current.stream) because, + * for example, PAGE backup must(!) look for start_lsn in archive regardless of wal_mode. * - * Returns LSN of last valid record if wait_prev_segment is not true, otherwise - * returns InvalidXLogRecPtr. + * 'timeout_elevel' determine the elevel for timeout elog message. If elevel lighter than + * ERROR is used, then return InvalidXLogRecPtr. TODO: return something more concrete, for example 1. + * + * Returns target LSN if such is found, failing that returns LSN of record prior to target LSN. + * Returns InvalidXLogRecPtr if 'segment_only' flag is used. */ -static XLogRecPtr -wait_wal_lsn(XLogRecPtr lsn, bool is_start_lsn, bool wait_prev_segment) +XLogRecPtr +wait_wal_lsn(const char *wal_segment_dir, XLogRecPtr target_lsn, bool is_start_lsn, TimeLineID tli, + bool in_prev_segment, bool segment_only, + int timeout_elevel, bool in_stream_dir) { - TimeLineID tli; XLogSegNo targetSegNo; - char pg_wal_dir[MAXPGPATH]; char wal_segment_path[MAXPGPATH], - *wal_segment_dir, wal_segment[MAXFNAMELEN]; bool file_exists = false; uint32 try_count = 0, timeout; + char *wal_delivery_str = in_stream_dir ? "streamed":"archived"; #ifdef HAVE_LIBZ char gz_wal_segment_path[MAXPGPATH]; #endif - tli = get_current_timeline(false); - /* Compute the name of the WAL file containing requested LSN */ - GetXLogSegNo(lsn, targetSegNo, instance_config.xlog_seg_size); - if (wait_prev_segment) + GetXLogSegNo(target_lsn, targetSegNo, instance_config.xlog_seg_size); + if (in_prev_segment) targetSegNo--; GetXLogFileName(wal_segment, tli, targetSegNo, instance_config.xlog_seg_size); + join_path_components(wal_segment_path, wal_segment_dir, wal_segment); /* - * In pg_start_backup we wait for 'lsn' in 'pg_wal' directory if it is + * In pg_start_backup we wait for 'target_lsn' in 'pg_wal' directory if it is * stream and non-page backup. Page backup needs archived WAL files, so we - * wait for 'lsn' in archive 'wal' directory for page backups. + * wait for 'target_lsn' in archive 'wal' directory for page backups. * * In pg_stop_backup it depends only on stream_wal. */ - if (stream_wal && - (current.backup_mode != BACKUP_MODE_DIFF_PAGE || !is_start_lsn)) - { - pgBackupGetPath2(¤t, pg_wal_dir, lengthof(pg_wal_dir), - DATABASE_DIR, PG_XLOG_DIR); - join_path_components(wal_segment_path, pg_wal_dir, wal_segment); - wal_segment_dir = pg_wal_dir; - } - else - { - join_path_components(wal_segment_path, arclog_path, wal_segment); - wal_segment_dir = arclog_path; - } + /* TODO: remove this in 3.0 (it is a cludge against some old bug with archive_timeout) */ if (instance_config.archive_timeout > 0) timeout = instance_config.archive_timeout; else timeout = ARCHIVE_TIMEOUT_DEFAULT; - if (wait_prev_segment) + if (segment_only) elog(LOG, "Looking for segment: %s", wal_segment); else elog(LOG, "Looking for LSN %X/%X in segment: %s", - (uint32) (lsn >> 32), (uint32) lsn, wal_segment); + (uint32) (target_lsn >> 32), (uint32) target_lsn, wal_segment); #ifdef HAVE_LIBZ snprintf(gz_wal_segment_path, sizeof(gz_wal_segment_path), "%s.gz", @@ -1412,38 +1405,48 @@ wait_wal_lsn(XLogRecPtr lsn, bool is_start_lsn, bool wait_prev_segment) if (file_exists) { - /* Do not check LSN for previous WAL segment */ - if (wait_prev_segment) + /* Do not check for target LSN */ + if (segment_only) return InvalidXLogRecPtr; /* - * A WAL segment found. Check LSN on it. + * A WAL segment found. Look for target LSN in it. */ - if (wal_contains_lsn(wal_segment_dir, lsn, tli, - instance_config.xlog_seg_size)) + if (!XRecOffIsNull(target_lsn) && + wal_contains_lsn(wal_segment_dir, target_lsn, tli, + instance_config.xlog_seg_size)) /* Target LSN was found */ { - elog(LOG, "Found LSN: %X/%X", (uint32) (lsn >> 32), (uint32) lsn); - return lsn; + elog(LOG, "Found LSN: %X/%X", (uint32) (target_lsn >> 32), (uint32) target_lsn); + return target_lsn; } /* - * If we failed to get LSN of valid record in a reasonable time, try + * If we failed to get target LSN in a reasonable time, try * to get LSN of last valid record prior to the target LSN. But only * in case of a backup from a replica. + * Note, that with NullXRecOff target_lsn we do not wait + * for 'timeout / 2' seconds before going for previous record, + * because such LSN cannot be delivered at all. + * + * There are two cases for this: + * 1. Replica returned readpoint LSN which just do not exists. We want to look + * for previous record in the same(!) WAL segment which endpoint points to this LSN. + * 2. Replica returened endpoint LSN with NullXRecOff. We want to look + * for previous record which endpoint points greater or equal LSN in previous WAL segment. */ - if (!exclusive_backup && current.from_replica && - (try_count > timeout / 4)) + if (current.from_replica && + (XRecOffIsNull(target_lsn) || try_count > timeout / 2)) { XLogRecPtr res; - res = get_last_wal_lsn(wal_segment_dir, current.start_lsn, - lsn, tli, false, - instance_config.xlog_seg_size); + res = get_prior_record_lsn(wal_segment_dir, current.start_lsn, target_lsn, tli, + in_prev_segment, instance_config.xlog_seg_size); + if (!XLogRecPtrIsInvalid(res)) { /* LSN of the prior record was found */ - elog(LOG, "Found prior LSN: %X/%X, it is used as stop LSN", + elog(LOG, "Found prior LSN: %X/%X", (uint32) (res >> 32), (uint32) res); return res; } @@ -1451,100 +1454,299 @@ wait_wal_lsn(XLogRecPtr lsn, bool is_start_lsn, bool wait_prev_segment) } sleep(1); - if (interrupted) - elog(ERROR, "Interrupted during waiting for WAL archiving"); + if (interrupted || thread_interrupted) + elog(ERROR, "Interrupted during waiting for WAL %s", in_stream_dir ? "streaming" : "archiving"); try_count++; /* Inform user if WAL segment is absent in first attempt */ if (try_count == 1) { - if (wait_prev_segment) - elog(INFO, "Wait for WAL segment %s to be archived", - wal_segment_path); + if (segment_only) + elog(INFO, "Wait for WAL segment %s to be %s", + wal_segment_path, wal_delivery_str); else - elog(INFO, "Wait for LSN %X/%X in archived WAL segment %s", - (uint32) (lsn >> 32), (uint32) lsn, wal_segment_path); + elog(INFO, "Wait for LSN %X/%X in %s WAL segment %s", + (uint32) (target_lsn >> 32), (uint32) target_lsn, + wal_delivery_str, wal_segment_path); } - if (!stream_wal && is_start_lsn && try_count == 30) - elog(WARNING, "By default pg_probackup assume WAL delivery method to be ARCHIVE. " - "If continius archiving is not set up, use '--stream' option to make autonomous backup. " - "Otherwise check that continius archiving works correctly."); + if (!current.stream && is_start_lsn && try_count == 30) + elog(WARNING, "By default pg_probackup assumes that WAL delivery method to be ARCHIVE. " + "If continuous archiving is not set up, use '--stream' option to make autonomous backup. " + "Otherwise check that continuous archiving works correctly."); if (timeout > 0 && try_count > timeout) { if (file_exists) - elog(ERROR, "WAL segment %s was archived, " - "but target LSN %X/%X could not be archived in %d seconds", - wal_segment, (uint32) (lsn >> 32), (uint32) lsn, timeout); + elog(timeout_elevel, "WAL segment %s was %s, " + "but target LSN %X/%X could not be %s in %d seconds", + wal_segment, wal_delivery_str, + (uint32) (target_lsn >> 32), (uint32) target_lsn, + wal_delivery_str, timeout); /* If WAL segment doesn't exist or we wait for previous segment */ else - elog(ERROR, - "Switched WAL segment %s could not be archived in %d seconds", - wal_segment, timeout); + elog(timeout_elevel, + "WAL segment %s could not be %s in %d seconds", + wal_segment, wal_delivery_str, timeout); + + return InvalidXLogRecPtr; } } } /* - * Notify end of backup to PostgreSQL server. + * Check stop_lsn (returned from pg_stop_backup()) and update backup->stop_lsn */ -static void -pg_stop_backup(pgBackup *backup, PGconn *pg_startbackup_conn, - PGNodeInfo *nodeInfo) +void +wait_wal_and_calculate_stop_lsn(const char *xlog_path, XLogRecPtr stop_lsn, pgBackup *backup) { - PGconn *conn; - PGresult *res; - PGresult *tablespace_map_content = NULL; - uint32 lsn_hi; - uint32 lsn_lo; - //XLogRecPtr restore_lsn = InvalidXLogRecPtr; - int pg_stop_backup_timeout = 0; - char path[MAXPGPATH]; - char backup_label[MAXPGPATH]; - FILE *fp; - pgFile *file; - size_t len; - char *val = NULL; - char *stop_backup_query = NULL; - bool stop_lsn_exists = false; + bool stop_lsn_exists = false; - /* - * We will use this values if there are no transactions between start_lsn - * and stop_lsn. + /* It is ok for replica to return invalid STOP LSN + * UPD: Apparently it is ok even for a master. */ - time_t recovery_time; - TransactionId recovery_xid; + if (!XRecOffIsValid(stop_lsn)) + { + XLogSegNo segno = 0; + XLogRecPtr lsn_tmp = InvalidXLogRecPtr; - if (!backup_in_progress) - elog(ERROR, "backup is not in progress"); + /* + * Even though the value is invalid, it's expected postgres behaviour + * and we're trying to fix it below. + */ + elog(LOG, "Invalid offset in stop_lsn value %X/%X, trying to fix", + (uint32) (stop_lsn >> 32), (uint32) (stop_lsn)); + + /* + * Note: even with gdb it is very hard to produce automated tests for + * contrecord + invalid LSN, so emulate it for manual testing. + */ + //lsn = lsn - XLOG_SEG_SIZE; + //elog(WARNING, "New Invalid stop_backup_lsn value %X/%X", + // (uint32) (stop_lsn >> 32), (uint32) (stop_lsn)); + + GetXLogSegNo(stop_lsn, segno, instance_config.xlog_seg_size); + + /* + * Note, that there is no guarantee that corresponding WAL file even exists. + * Replica may return LSN from future and keep staying in present. + * Or it can return invalid LSN. + * + * That's bad, since we want to get real LSN to save it in backup label file + * and to use it in WAL validation. + * + * So we try to do the following: + * 1. Wait 'archive_timeout' seconds for segment containing stop_lsn and + * look for the first valid record in it. + * It solves the problem of occasional invalid LSN on write-busy system. + * 2. Failing that, look for record in previous segment with endpoint + * equal or greater than stop_lsn. It may(!) solve the problem of invalid LSN + * on write-idle system. If that fails too, error out. + */ + + /* stop_lsn is pointing to a 0 byte of xlog segment */ + if (stop_lsn % instance_config.xlog_seg_size == 0) + { + /* Wait for segment with current stop_lsn, it is ok for it to never arrive */ + wait_wal_lsn(xlog_path, stop_lsn, false, backup->tli, + false, true, WARNING, backup->stream); + + /* Get the first record in segment with current stop_lsn */ + lsn_tmp = get_first_record_lsn(xlog_path, segno, backup->tli, + instance_config.xlog_seg_size, + instance_config.archive_timeout); + + /* Check that returned LSN is valid and greater than stop_lsn */ + if (XLogRecPtrIsInvalid(lsn_tmp) || + !XRecOffIsValid(lsn_tmp) || + lsn_tmp < stop_lsn) + { + /* Backup from master should error out here */ + if (!backup->from_replica) + elog(ERROR, "Failed to get next WAL record after %X/%X", + (uint32) (stop_lsn >> 32), + (uint32) (stop_lsn)); + + /* No luck, falling back to looking up for previous record */ + elog(WARNING, "Failed to get next WAL record after %X/%X, " + "looking for previous WAL record", + (uint32) (stop_lsn >> 32), + (uint32) (stop_lsn)); + + /* Despite looking for previous record there is not guarantee of success + * because previous record can be the contrecord. + */ + lsn_tmp = wait_wal_lsn(xlog_path, stop_lsn, false, backup->tli, + true, false, ERROR, backup->stream); + + /* sanity */ + if (!XRecOffIsValid(lsn_tmp) || XLogRecPtrIsInvalid(lsn_tmp)) + elog(ERROR, "Failed to get WAL record prior to %X/%X", + (uint32) (stop_lsn >> 32), + (uint32) (stop_lsn)); + } + } + /* stop lsn is aligned to xlog block size, just find next lsn */ + else if (stop_lsn % XLOG_BLCKSZ == 0) + { + /* Wait for segment with current stop_lsn */ + wait_wal_lsn(xlog_path, stop_lsn, false, backup->tli, + false, true, ERROR, backup->stream); + + /* Get the next closest record in segment with current stop_lsn */ + lsn_tmp = get_next_record_lsn(xlog_path, segno, backup->tli, + instance_config.xlog_seg_size, + instance_config.archive_timeout, + stop_lsn); + + /* sanity */ + if (!XRecOffIsValid(lsn_tmp) || XLogRecPtrIsInvalid(lsn_tmp)) + elog(ERROR, "Failed to get WAL record next to %X/%X", + (uint32) (stop_lsn >> 32), + (uint32) (stop_lsn)); + } + /* PostgreSQL returned something very illegal as STOP_LSN, error out */ + else + elog(ERROR, "Invalid stop_backup_lsn value %X/%X", + (uint32) (stop_lsn >> 32), (uint32) (stop_lsn)); + + /* Setting stop_backup_lsn will set stop point for streaming */ + stop_backup_lsn = lsn_tmp; + stop_lsn_exists = true; + } - conn = pg_startbackup_conn; + elog(INFO, "stop_lsn: %X/%X", + (uint32) (stop_lsn >> 32), (uint32) (stop_lsn)); - /* Remove annoying NOTICE messages generated by backend */ + /* + * Wait for stop_lsn to be archived or streamed. + * If replica returned valid STOP_LSN of not actually existing record, + * look for previous record with endpoint >= STOP_LSN. + */ + if (!stop_lsn_exists) + stop_backup_lsn = wait_wal_lsn(xlog_path, stop_lsn, false, backup->tli, + false, false, ERROR, backup->stream); + + backup->stop_lsn = stop_backup_lsn; +} + +/* Remove annoying NOTICE messages generated by backend */ +void +pg_silent_client_messages(PGconn *conn) +{ + PGresult *res; res = pgut_execute(conn, "SET client_min_messages = warning;", 0, NULL); PQclear(res); +} - /* Create restore point - * Only if backup is from master. - * For PG 9.5 create restore point only if pguser is superuser. - */ - if (backup != NULL && !current.from_replica && - !(nodeInfo->server_version < 90600 && - !nodeInfo->is_superuser)) - { - const char *params[1]; - char name[1024]; +void +pg_create_restore_point(PGconn *conn, time_t backup_start_time) +{ + PGresult *res; + const char *params[1]; + char name[1024]; - snprintf(name, lengthof(name), "pg_probackup, backup_id %s", - base36enc(backup->start_time)); - params[0] = name; + snprintf(name, lengthof(name), "pg_probackup, backup_id %s", + base36enc(backup_start_time)); + params[0] = name; - res = pgut_execute(conn, "SELECT pg_catalog.pg_create_restore_point($1)", - 1, params); - PQclear(res); - } + res = pgut_execute(conn, "SELECT pg_catalog.pg_create_restore_point($1)", + 1, params); + PQclear(res); +} + +void +pg_stop_backup_send(PGconn *conn, int server_version, bool is_started_on_replica, bool is_exclusive, char **query_text) +{ + static const char + stop_exlusive_backup_query[] = + /* + * Stop the non-exclusive backup. Besides stop_lsn it returns from + * pg_stop_backup(false) copy of the backup label and tablespace map + * so they can be written to disk by the caller. + * TODO, question: add NULLs as backup_label and tablespace_map? + */ + "SELECT" + " pg_catalog.txid_snapshot_xmax(pg_catalog.txid_current_snapshot())," + " current_timestamp(0)::timestamptz," + " pg_catalog.pg_stop_backup() as lsn", + stop_backup_on_master_query[] = + "SELECT" + " pg_catalog.txid_snapshot_xmax(pg_catalog.txid_current_snapshot())," + " current_timestamp(0)::timestamptz," + " lsn," + " labelfile," + " spcmapfile" + " FROM pg_catalog.pg_stop_backup(false, false)", + stop_backup_on_master_before10_query[] = + "SELECT" + " pg_catalog.txid_snapshot_xmax(pg_catalog.txid_current_snapshot())," + " current_timestamp(0)::timestamptz," + " lsn," + " labelfile," + " spcmapfile" + " FROM pg_catalog.pg_stop_backup(false)", + stop_backup_on_master_after15_query[] = + "SELECT" + " pg_catalog.txid_snapshot_xmax(pg_catalog.txid_current_snapshot())," + " current_timestamp(0)::timestamptz," + " lsn," + " labelfile," + " spcmapfile" + " FROM pg_catalog.pg_backup_stop(false)", + /* + * In case of backup from replica >= 9.6 we do not trust minRecPoint + * and stop_backup LSN, so we use latest replayed LSN as STOP LSN. + */ + stop_backup_on_replica_query[] = + "SELECT" + " pg_catalog.txid_snapshot_xmax(pg_catalog.txid_current_snapshot())," + " current_timestamp(0)::timestamptz," + " pg_catalog.pg_last_wal_replay_lsn()," + " labelfile," + " spcmapfile" + " FROM pg_catalog.pg_stop_backup(false, false)", + stop_backup_on_replica_before10_query[] = + "SELECT" + " pg_catalog.txid_snapshot_xmax(pg_catalog.txid_current_snapshot())," + " current_timestamp(0)::timestamptz," + " pg_catalog.pg_last_xlog_replay_location()," + " labelfile," + " spcmapfile" + " FROM pg_catalog.pg_stop_backup(false)", + stop_backup_on_replica_after15_query[] = + "SELECT" + " pg_catalog.txid_snapshot_xmax(pg_catalog.txid_current_snapshot())," + " current_timestamp(0)::timestamptz," + " pg_catalog.pg_last_wal_replay_lsn()," + " labelfile," + " spcmapfile" + " FROM pg_catalog.pg_backup_stop(false)"; + + const char * const stop_backup_query = + is_exclusive ? + stop_exlusive_backup_query : + server_version >= 150000 ? + (is_started_on_replica ? + stop_backup_on_replica_after15_query : + stop_backup_on_master_after15_query + ) : + (server_version >= 100000 ? + (is_started_on_replica ? + stop_backup_on_replica_query : + stop_backup_on_master_query + ) : + (is_started_on_replica ? + stop_backup_on_replica_before10_query : + stop_backup_on_master_before10_query + ) + ); + bool sent = false; + + /* Make proper timestamp format for parse_time(recovery_time) */ + pgut_execute(conn, "SET datestyle = 'ISO, DMY';", 0, NULL); + // TODO: check result /* * send pg_stop_backup asynchronously because we could came @@ -1552,342 +1754,308 @@ pg_stop_backup(pgBackup *backup, PGconn *pg_startbackup_conn, * postgres archive_command problem and in this case we will * wait for pg_stop_backup() forever. */ + sent = pgut_send(conn, stop_backup_query, 0, NULL, WARNING); + if (!sent) +#if PG_VERSION_NUM >= 150000 + elog(ERROR, "Failed to send pg_backup_stop query"); +#else + elog(ERROR, "Failed to send pg_stop_backup query"); +#endif + + /* After we have sent pg_stop_backup, we don't need this callback anymore */ + pgut_atexit_pop(backup_stopbackup_callback, &stop_callback_params); + + if (query_text) + *query_text = pgut_strdup(stop_backup_query); +} - if (!pg_stop_backup_is_sent) +/* + * pg_stop_backup_consume -- get 'pg_stop_backup' query results + * side effects: + * - allocates memory for tablespace_map and backup_label contents, so it must freed by caller (if its not null) + * parameters: + * - + */ +void +pg_stop_backup_consume(PGconn *conn, int server_version, + bool is_exclusive, uint32 timeout, const char *query_text, + PGStopBackupResult *result) +{ + PGresult *query_result; + uint32 pg_stop_backup_timeout = 0; + enum stop_backup_query_result_column_numbers { + recovery_xid_colno = 0, + recovery_time_colno, + lsn_colno, + backup_label_colno, + tablespace_map_colno + }; + + /* and now wait */ + while (1) { - bool sent = false; + if (!PQconsumeInput(conn)) + elog(ERROR, "pg_stop backup() failed: %s", + PQerrorMessage(conn)); - if (!exclusive_backup) + if (PQisBusy(conn)) { - /* - * Stop the non-exclusive backup. Besides stop_lsn it returns from - * pg_stop_backup(false) copy of the backup label and tablespace map - * so they can be written to disk by the caller. - * In case of backup from replica >= 9.6 we do not trust minRecPoint - * and stop_backup LSN, so we use latest replayed LSN as STOP LSN. - */ - if (current.from_replica) - stop_backup_query = "SELECT" - " pg_catalog.txid_snapshot_xmax(pg_catalog.txid_current_snapshot())," - " current_timestamp(0)::timestamptz," -#if PG_VERSION_NUM >= 100000 - " pg_catalog.pg_last_wal_replay_lsn()," -#else - " pg_catalog.pg_last_xlog_replay_location()," -#endif - " labelfile," - " spcmapfile" -#if PG_VERSION_NUM >= 100000 - " FROM pg_catalog.pg_stop_backup(false, false)"; + pg_stop_backup_timeout++; + sleep(1); + + if (interrupted) + { + pgut_cancel(conn); +#if PG_VERSION_NUM >= 150000 + elog(ERROR, "Interrupted during waiting for pg_backup_stop"); #else - " FROM pg_catalog.pg_stop_backup(false)"; + elog(ERROR, "Interrupted during waiting for pg_stop_backup"); #endif - else - stop_backup_query = "SELECT" - " pg_catalog.txid_snapshot_xmax(pg_catalog.txid_current_snapshot())," - " current_timestamp(0)::timestamptz," - " lsn," - " labelfile," - " spcmapfile" -#if PG_VERSION_NUM >= 100000 - " FROM pg_catalog.pg_stop_backup(false, false)"; + } + + if (pg_stop_backup_timeout == 1) + elog(INFO, "wait for pg_stop_backup()"); + + /* + * If postgres haven't answered in archive_timeout seconds, + * send an interrupt. + */ + if (pg_stop_backup_timeout > timeout) + { + pgut_cancel(conn); +#if PG_VERSION_NUM >= 150000 + elog(ERROR, "pg_backup_stop doesn't answer in %d seconds, cancel it", timeout); #else - " FROM pg_catalog.pg_stop_backup(false)"; + elog(ERROR, "pg_stop_backup doesn't answer in %d seconds, cancel it", timeout); #endif - + } } else { - stop_backup_query = "SELECT" - " pg_catalog.txid_snapshot_xmax(pg_catalog.txid_current_snapshot())," - " current_timestamp(0)::timestamptz," - " pg_catalog.pg_stop_backup() as lsn"; + query_result = PQgetResult(conn); + break; } - - sent = pgut_send(conn, stop_backup_query, 0, NULL, WARNING); - pg_stop_backup_is_sent = true; - if (!sent) - elog(ERROR, "Failed to send pg_stop_backup query"); } - /* After we have sent pg_stop_backup, we don't need this callback anymore */ - pgut_atexit_pop(backup_stopbackup_callback, pg_startbackup_conn); - - /* - * Wait for the result of pg_stop_backup(), but no longer than - * archive_timeout seconds - */ - if (pg_stop_backup_is_sent && !in_cleanup) + /* Check successfull execution of pg_stop_backup() */ + if (!query_result) +#if PG_VERSION_NUM >= 150000 + elog(ERROR, "pg_backup_stop() failed"); +#else + elog(ERROR, "pg_stop_backup() failed"); +#endif + else { - res = NULL; - - while (1) + switch (PQresultStatus(query_result)) { - if (!PQconsumeInput(conn)) - elog(ERROR, "pg_stop backup() failed: %s", - PQerrorMessage(conn)); + /* + * We should expect only PGRES_TUPLES_OK since pg_stop_backup + * returns tuples. + */ + case PGRES_TUPLES_OK: + break; + default: + elog(ERROR, "Query failed: %s query was: %s", + PQerrorMessage(conn), query_text); + } + backup_in_progress = false; + elog(INFO, "pg_stop backup() successfully executed"); + } - if (PQisBusy(conn)) - { - pg_stop_backup_timeout++; - sleep(1); + /* get results and fill result structure */ + /* get&check recovery_xid */ + if (sscanf(PQgetvalue(query_result, 0, recovery_xid_colno), XID_FMT, &result->snapshot_xid) != 1) + elog(ERROR, + "Result of txid_snapshot_xmax() is invalid: %s", + PQgetvalue(query_result, 0, recovery_xid_colno)); - if (interrupted) - { - pgut_cancel(conn); - elog(ERROR, "interrupted during waiting for pg_stop_backup"); - } + /* get&check recovery_time */ + if (!parse_time(PQgetvalue(query_result, 0, recovery_time_colno), &result->invocation_time, true)) + elog(ERROR, + "Result of current_timestamp is invalid: %s", + PQgetvalue(query_result, 0, recovery_time_colno)); - if (pg_stop_backup_timeout == 1) - elog(INFO, "wait for pg_stop_backup()"); - - /* - * If postgres haven't answered in archive_timeout seconds, - * send an interrupt. - */ - if (pg_stop_backup_timeout > instance_config.archive_timeout) - { - pgut_cancel(conn); - elog(ERROR, "pg_stop_backup doesn't answer in %d seconds, cancel it", - instance_config.archive_timeout); - } - } - else - { - res = PQgetResult(conn); - break; - } - } - - /* Check successfull execution of pg_stop_backup() */ - if (!res) - elog(ERROR, "pg_stop backup() failed"); - else - { - switch (PQresultStatus(res)) - { - /* - * We should expect only PGRES_TUPLES_OK since pg_stop_backup - * returns tuples. - */ - case PGRES_TUPLES_OK: - break; - default: - elog(ERROR, "query failed: %s query was: %s", - PQerrorMessage(conn), stop_backup_query); - } - elog(INFO, "pg_stop backup() successfully executed"); - } + /* get stop_backup_lsn */ + { + uint32 lsn_hi; + uint32 lsn_lo; - backup_in_progress = false; +// char *target_lsn = "2/F578A000"; +// XLogDataFromLSN(target_lsn, &lsn_hi, &lsn_lo); /* Extract timeline and LSN from results of pg_stop_backup() */ - XLogDataFromLSN(PQgetvalue(res, 0, 2), &lsn_hi, &lsn_lo); + XLogDataFromLSN(PQgetvalue(query_result, 0, lsn_colno), &lsn_hi, &lsn_lo); /* Calculate LSN */ - stop_backup_lsn = ((uint64) lsn_hi) << 32 | lsn_lo; + result->lsn = ((uint64) lsn_hi) << 32 | lsn_lo; + } - if (!XRecOffIsValid(stop_backup_lsn)) - { - if (XRecOffIsNull(stop_backup_lsn)) - { - char *xlog_path, - stream_xlog_path[MAXPGPATH]; + /* get backup_label_content */ + result->backup_label_content = NULL; + // if (!PQgetisnull(query_result, 0, backup_label_colno)) + if (!is_exclusive) + { + result->backup_label_content_len = PQgetlength(query_result, 0, backup_label_colno); + if (result->backup_label_content_len > 0) + result->backup_label_content = pgut_strndup(PQgetvalue(query_result, 0, backup_label_colno), + result->backup_label_content_len); + } else { + result->backup_label_content_len = 0; + } - elog(WARNING, "Invalid stop_backup_lsn value %X/%X", - (uint32) (stop_backup_lsn >> 32), (uint32) (stop_backup_lsn)); + /* get tablespace_map_content */ + result->tablespace_map_content = NULL; + // if (!PQgetisnull(query_result, 0, tablespace_map_colno)) + if (!is_exclusive) + { + result->tablespace_map_content_len = PQgetlength(query_result, 0, tablespace_map_colno); + if (result->tablespace_map_content_len > 0) + result->tablespace_map_content = pgut_strndup(PQgetvalue(query_result, 0, tablespace_map_colno), + result->tablespace_map_content_len); + } else { + result->tablespace_map_content_len = 0; + } +} - if (stream_wal) - { - pgBackupGetPath2(backup, stream_xlog_path, - lengthof(stream_xlog_path), - DATABASE_DIR, PG_XLOG_DIR); - xlog_path = stream_xlog_path; - } - else - xlog_path = arclog_path; +/* + * helper routine used to write backup_label and tablespace_map in pg_stop_backup() + */ +void +pg_stop_backup_write_file_helper(const char *path, const char *filename, const char *error_msg_filename, + const void *data, size_t len, parray *file_list) +{ + FILE *fp; + pgFile *file; + char full_filename[MAXPGPATH]; + + join_path_components(full_filename, path, filename); + fp = fio_fopen(full_filename, PG_BINARY_W, FIO_BACKUP_HOST); + if (fp == NULL) + elog(ERROR, "Can't open %s file \"%s\": %s", + error_msg_filename, full_filename, strerror(errno)); + + if (fio_fwrite(fp, data, len) != len || + fio_fflush(fp) != 0 || + fio_fclose(fp)) + elog(ERROR, "Can't write %s file \"%s\": %s", + error_msg_filename, full_filename, strerror(errno)); - wait_wal_lsn(stop_backup_lsn, false, true); - stop_backup_lsn = get_last_wal_lsn(xlog_path, backup->start_lsn, - stop_backup_lsn, backup->tli, - true, instance_config.xlog_seg_size); - /* - * Do not check existance of LSN again below using - * wait_wal_lsn(). - */ - stop_lsn_exists = true; - } - else - elog(ERROR, "Invalid stop_backup_lsn value %X/%X", - (uint32) (stop_backup_lsn >> 32), (uint32) (stop_backup_lsn)); - } + /* + * It's vital to check if files_list is initialized, + * because we could get here because the backup was interrupted + */ + if (file_list) + { + file = pgFileNew(full_filename, filename, true, 0, + FIO_BACKUP_HOST); - /* Write backup_label and tablespace_map */ - if (!exclusive_backup) + if (S_ISREG(file->mode)) { - Assert(PQnfields(res) >= 4); - pgBackupGetPath(¤t, path, lengthof(path), DATABASE_DIR); - - /* Write backup_label */ - join_path_components(backup_label, path, PG_BACKUP_LABEL_FILE); - fp = fio_fopen(backup_label, PG_BINARY_W, FIO_BACKUP_HOST); - if (fp == NULL) - elog(ERROR, "can't open backup label file \"%s\": %s", - backup_label, strerror(errno)); - - len = strlen(PQgetvalue(res, 0, 3)); - if (fio_fwrite(fp, PQgetvalue(res, 0, 3), len) != len || - fio_fflush(fp) != 0 || - fio_fclose(fp)) - elog(ERROR, "can't write backup label file \"%s\": %s", - backup_label, strerror(errno)); + file->crc = pgFileGetCRC(full_filename, true, false); - /* - * It's vital to check if backup_files_list is initialized, - * because we could get here because the backup was interrupted - */ - if (backup_files_list) - { - file = pgFileNew(backup_label, PG_BACKUP_LABEL_FILE, true, 0, - FIO_BACKUP_HOST); - file->crc = pgFileGetCRC(file->path, true, false, - &file->read_size, FIO_BACKUP_HOST); - file->write_size = file->read_size; - free(file->path); - file->path = strdup(PG_BACKUP_LABEL_FILE); - parray_append(backup_files_list, file); - } + file->write_size = file->size; + file->uncompressed_size = file->size; } + parray_append(file_list, file); + } +} - if (sscanf(PQgetvalue(res, 0, 0), XID_FMT, &recovery_xid) != 1) - elog(ERROR, - "result of txid_snapshot_xmax() is invalid: %s", - PQgetvalue(res, 0, 0)); - if (!parse_time(PQgetvalue(res, 0, 1), &recovery_time, true)) - elog(ERROR, - "result of current_timestamp is invalid: %s", - PQgetvalue(res, 0, 1)); - - /* Get content for tablespace_map from stop_backup results - * in case of non-exclusive backup - */ - if (!exclusive_backup) - val = PQgetvalue(res, 0, 4); +/* + * Notify end of backup to PostgreSQL server. + */ +static void +pg_stop_backup(InstanceState *instanceState, pgBackup *backup, PGconn *pg_startbackup_conn, + PGNodeInfo *nodeInfo) +{ + PGStopBackupResult stop_backup_result; + char *xlog_path, stream_xlog_path[MAXPGPATH]; + /* kludge against some old bug in archive_timeout. TODO: remove in 3.0.0 */ + int timeout = (instance_config.archive_timeout > 0) ? + instance_config.archive_timeout : ARCHIVE_TIMEOUT_DEFAULT; + char *query_text = NULL; + + /* Remove it ? */ + if (!backup_in_progress) + elog(ERROR, "Backup is not in progress"); - /* Write tablespace_map */ - if (!exclusive_backup && val && strlen(val) > 0) - { - char tablespace_map[MAXPGPATH]; - - join_path_components(tablespace_map, path, PG_TABLESPACE_MAP_FILE); - fp = fio_fopen(tablespace_map, PG_BINARY_W, FIO_BACKUP_HOST); - if (fp == NULL) - elog(ERROR, "can't open tablespace map file \"%s\": %s", - tablespace_map, strerror(errno)); - - len = strlen(val); - if (fio_fwrite(fp, val, len) != len || - fio_fflush(fp) != 0 || - fio_fclose(fp)) - elog(ERROR, "can't write tablespace map file \"%s\": %s", - tablespace_map, strerror(errno)); - - if (backup_files_list) - { - file = pgFileNew(tablespace_map, PG_TABLESPACE_MAP_FILE, true, 0, - FIO_BACKUP_HOST); - if (S_ISREG(file->mode)) - { - file->crc = pgFileGetCRC(file->path, true, false, - &file->read_size, FIO_BACKUP_HOST); - file->write_size = file->read_size; - } - free(file->path); - file->path = strdup(PG_TABLESPACE_MAP_FILE); - parray_append(backup_files_list, file); - } - } + pg_silent_client_messages(pg_startbackup_conn); - if (tablespace_map_content) - PQclear(tablespace_map_content); - PQclear(res); + /* Create restore point + * Only if backup is from master. + * For PG 9.5 create restore point only if pguser is superuser. + */ + if (!backup->from_replica && + !(nodeInfo->server_version < 90600 && + !nodeInfo->is_superuser)) //TODO: check correctness + pg_create_restore_point(pg_startbackup_conn, backup->start_time); - if (stream_wal) - { - /* Wait for the completion of stream */ - pthread_join(stream_thread, NULL); - if (stream_thread_arg.ret == 1) - elog(ERROR, "WAL streaming failed"); - } - } + /* Execute pg_stop_backup using PostgreSQL connection */ + pg_stop_backup_send(pg_startbackup_conn, nodeInfo->server_version, backup->from_replica, exclusive_backup, &query_text); - /* Fill in fields if that is the correct end of backup. */ - if (backup != NULL) - { - char *xlog_path, - stream_xlog_path[MAXPGPATH]; + /* + * Wait for the result of pg_stop_backup(), but no longer than + * archive_timeout seconds + */ + pg_stop_backup_consume(pg_startbackup_conn, nodeInfo->server_version, exclusive_backup, timeout, query_text, &stop_backup_result); - /* - * Wait for stop_lsn to be archived or streamed. - * We wait for stop_lsn in stream mode just in case. - */ - if (!stop_lsn_exists) - stop_backup_lsn = wait_wal_lsn(stop_backup_lsn, false, false); + if (backup->stream) + { + join_path_components(stream_xlog_path, backup->database_dir, PG_XLOG_DIR); + xlog_path = stream_xlog_path; + } + else + xlog_path = instanceState->instance_wal_subdir_path; - if (stream_wal) - { - pgBackupGetPath2(backup, stream_xlog_path, - lengthof(stream_xlog_path), - DATABASE_DIR, PG_XLOG_DIR); - xlog_path = stream_xlog_path; - } - else - xlog_path = arclog_path; + wait_wal_and_calculate_stop_lsn(xlog_path, stop_backup_result.lsn, backup); - backup->tli = get_current_timeline(false); - backup->stop_lsn = stop_backup_lsn; + /* Write backup_label and tablespace_map */ + if (!exclusive_backup) + { + Assert(stop_backup_result.backup_label_content != NULL); - elog(LOG, "Getting the Recovery Time from WAL"); + /* Write backup_label */ + pg_stop_backup_write_file_helper(backup->database_dir, PG_BACKUP_LABEL_FILE, "backup label", + stop_backup_result.backup_label_content, stop_backup_result.backup_label_content_len, + backup_files_list); + free(stop_backup_result.backup_label_content); + stop_backup_result.backup_label_content = NULL; + stop_backup_result.backup_label_content_len = 0; - /* iterate over WAL from stop_backup lsn to start_backup lsn */ - if (!read_recovery_info(xlog_path, backup->tli, - instance_config.xlog_seg_size, - backup->start_lsn, backup->stop_lsn, - &backup->recovery_time, &backup->recovery_xid)) + /* Write tablespace_map */ + if (stop_backup_result.tablespace_map_content != NULL) { - elog(LOG, "Failed to find Recovery Time in WAL. Forced to trust current_timestamp"); - backup->recovery_time = recovery_time; - backup->recovery_xid = recovery_xid; + pg_stop_backup_write_file_helper(backup->database_dir, PG_TABLESPACE_MAP_FILE, "tablespace map", + stop_backup_result.tablespace_map_content, stop_backup_result.tablespace_map_content_len, + backup_files_list); + free(stop_backup_result.tablespace_map_content); + stop_backup_result.tablespace_map_content = NULL; + stop_backup_result.tablespace_map_content_len = 0; } } -} -/* - * Retrieve checkpoint_timeout GUC value in seconds. - */ -static int -checkpoint_timeout(PGconn *backup_conn) -{ - PGresult *res; - const char *val; - const char *hintmsg; - int val_int; + if (backup->stream) + { + /* This function will also add list of xlog files + * to the passed filelist */ + if(wait_WAL_streaming_end(backup_files_list)) + elog(ERROR, "WAL streaming failed"); + } + + backup->recovery_xid = stop_backup_result.snapshot_xid; - res = pgut_execute(backup_conn, "show checkpoint_timeout", 0, NULL); - val = PQgetvalue(res, 0, 0); + elog(INFO, "Getting the Recovery Time from WAL"); - if (!parse_int(val, &val_int, OPTION_UNIT_S, &hintmsg)) + /* iterate over WAL from stop_backup lsn to start_backup lsn */ + if (!read_recovery_info(xlog_path, backup->tli, + instance_config.xlog_seg_size, + backup->start_lsn, backup->stop_lsn, + &backup->recovery_time)) { - PQclear(res); - if (hintmsg) - elog(ERROR, "Invalid value of checkout_timeout %s: %s", val, - hintmsg); - else - elog(ERROR, "Invalid value of checkout_timeout %s", val); + elog(INFO, "Failed to find Recovery Time in WAL, forced to trust current_timestamp"); + backup->recovery_time = stop_backup_result.invocation_time; } - PQclear(res); - - return val_int; + /* Cleanup */ + pg_free(query_text); } /* @@ -1905,10 +2073,10 @@ backup_cleanup(bool fatal, void *userdata) if (current.status == BACKUP_STATUS_RUNNING && current.end_time == 0) { elog(WARNING, "Backup %s is running, setting its status to ERROR", - base36enc(current.start_time)); + backup_id_of(¤t)); current.end_time = time(NULL); current.status = BACKUP_STATUS_ERROR; - write_backup(¤t); + write_backup(¤t, true); } } @@ -1924,183 +2092,216 @@ static void * backup_files(void *arg) { int i; - backup_files_arg *arguments = (backup_files_arg *) arg; - int n_backup_files_list = parray_num(arguments->files_list); static time_t prev_time; + backup_files_arg *arguments = (backup_files_arg *) arg; + int n_backup_files_list = parray_num(arguments->files_list); + prev_time = current.start_time; /* backup a file */ for (i = 0; i < n_backup_files_list; i++) { - int ret; - struct stat buf; - pgFile *file = (pgFile *) parray_get(arguments->files_list, i); + pgFile *file = (pgFile *) parray_get(arguments->files_list, i); + + /* We have already copied all directories */ + if (S_ISDIR(file->mode)) + continue; + /* + * Don't copy the pg_control file now, we'll copy it last + */ + if(file->external_dir_num == 0 && pg_strcasecmp(file->rel_path, XLOG_CONTROL_FILE) == 0) + { + continue; + } if (arguments->thread_num == 1) { - /* update backup_content.control every 10 seconds */ - if ((difftime(time(NULL), prev_time)) > 10) + /* update backup_content.control every 60 seconds */ + if ((difftime(time(NULL), prev_time)) > 60) { - prev_time = time(NULL); - write_backup_filelist(¤t, arguments->files_list, arguments->from_root, - arguments->external_dirs); + arguments->external_dirs, false); /* update backup control file to update size info */ - write_backup(¤t); + write_backup(¤t, true); + + prev_time = time(NULL); } } + if (file->skip_cfs_nested) + continue; + if (!pg_atomic_test_set_flag(&file->lock)) continue; - elog(VERBOSE, "Copying file: \"%s\" ", file->path); /* check for interrupt */ if (interrupted || thread_interrupted) - elog(ERROR, "interrupted during backup"); + elog(ERROR, "Interrupted during backup"); - if (progress) - elog(INFO, "Progress: (%d/%d). Process file \"%s\"", - i + 1, n_backup_files_list, file->path); + elog(progress ? INFO : LOG, "Progress: (%d/%d). Process file \"%s\"", + i + 1, n_backup_files_list, file->rel_path); - /* stat file to check its current state */ - ret = fio_stat(file->path, &buf, true, FIO_DB_HOST); - if (ret == -1) + if (file->is_cfs) { - if (errno == ENOENT) - { - /* - * If file is not found, this is not en error. - * It could have been deleted by concurrent postgres transaction. - */ - file->write_size = FILE_NOT_FOUND; - elog(LOG, "File \"%s\" is not found", file->path); - continue; - } - else - { - elog(ERROR, - "can't stat file to backup \"%s\": %s", - file->path, strerror(errno)); - } + backup_cfs_segment(i, file, arguments); + } + else + { + process_file(i, file, arguments); } + } - /* We have already copied all directories */ - if (S_ISDIR(buf.st_mode)) - continue; + /* ssh connection to longer needed */ + fio_disconnect(); - if (S_ISREG(buf.st_mode)) - { - pgFile **prev_file = NULL; - char *external_path = NULL; + /* Data files transferring is successful */ + arguments->ret = 0; - if (file->external_dir_num) - external_path = parray_get(arguments->external_dirs, + return NULL; +} + +static void +process_file(int i, pgFile *file, backup_files_arg *arguments) +{ + char from_fullpath[MAXPGPATH]; + char to_fullpath[MAXPGPATH]; + pgFile *prev_file = NULL; + + elog(progress ? INFO : LOG, "Progress: (%d/%zu). Process file \"%s\"", + i + 1, parray_num(arguments->files_list), file->rel_path); + + /* Handle zero sized files */ + if (file->size == 0) + { + file->write_size = 0; + return; + } + + /* construct from_fullpath & to_fullpath */ + if (file->external_dir_num == 0) + { + join_path_components(from_fullpath, arguments->from_root, file->rel_path); + join_path_components(to_fullpath, arguments->to_root, file->rel_path); + } + else + { + char external_dst[MAXPGPATH]; + char *external_path = parray_get(arguments->external_dirs, file->external_dir_num - 1); - /* Check that file exist in previous backup */ - if (current.backup_mode != BACKUP_MODE_FULL) - { - char *relative; - pgFile key; - - relative = GetRelativePath(file->path, file->external_dir_num ? - external_path : arguments->from_root); - key.path = relative; - key.external_dir_num = file->external_dir_num; - - prev_file = (pgFile **) parray_bsearch(arguments->prev_filelist, - &key, pgFileComparePathWithExternal); - if (prev_file) - /* File exists in previous backup */ - file->exists_in_prev = true; - } + makeExternalDirPathByNum(external_dst, + arguments->external_prefix, + file->external_dir_num); - /* copy the file into backup */ - if (file->is_datafile && !file->is_cfs) - { - char to_path[MAXPGPATH]; - - join_path_components(to_path, arguments->to_root, - file->path + strlen(arguments->from_root) + 1); - - /* backup block by block if datafile AND not compressed by cfs*/ - if (!backup_data_file(arguments, to_path, file, - arguments->prev_start_lsn, - current.backup_mode, - instance_config.compress_alg, - instance_config.compress_level, - true)) - { - /* disappeared file not to be confused with 'not changed' */ - if (file->write_size != FILE_NOT_FOUND) - file->write_size = BYTES_INVALID; - elog(VERBOSE, "File \"%s\" was not copied to backup", file->path); - continue; - } - } - else if (!file->external_dir_num && - strcmp(file->name, "pg_control") == 0) - copy_pgcontrol_file(arguments->from_root, FIO_DB_HOST, - arguments->to_root, FIO_BACKUP_HOST, - file); - else - { - const char *dst; - bool skip = false; - char external_dst[MAXPGPATH]; + join_path_components(to_fullpath, external_dst, file->rel_path); + join_path_components(from_fullpath, external_path, file->rel_path); + } - /* If non-data file has not changed since last backup... */ - if (prev_file && file->exists_in_prev && - buf.st_mtime < current.parent_backup) - { - file->crc = pgFileGetCRC(file->path, true, false, - &file->read_size, FIO_DB_HOST); - file->write_size = file->read_size; - /* ...and checksum is the same... */ - if (EQ_TRADITIONAL_CRC32(file->crc, (*prev_file)->crc)) - skip = true; /* ...skip copying file. */ - } - /* Set file paths */ - if (file->external_dir_num) - { - makeExternalDirPathByNum(external_dst, - arguments->external_prefix, - file->external_dir_num); - dst = external_dst; - } - else - dst = arguments->to_root; - if (skip || - !copy_file(FIO_DB_HOST, dst, FIO_BACKUP_HOST, file, true)) - { - /* disappeared file not to be confused with 'not changed' */ - if (file->write_size != FILE_NOT_FOUND) - file->write_size = BYTES_INVALID; - elog(VERBOSE, "File \"%s\" was not copied to backup", - file->path); - continue; - } - } + /* Encountered some strange beast */ + if (!S_ISREG(file->mode)) + { + elog(WARNING, "Unexpected type %d of file \"%s\", skipping", + file->mode, from_fullpath); + return; + } - elog(VERBOSE, "File \"%s\". Copied "INT64_FORMAT " bytes", - file->path, file->write_size); + /* Check that file exist in previous backup */ + if (current.backup_mode != BACKUP_MODE_FULL) + { + pgFile **prevFileTmp = NULL; + prevFileTmp = (pgFile **) parray_bsearch(arguments->prev_filelist, + file, pgFileCompareRelPathWithExternal); + if (prevFileTmp) + { + /* File exists in previous backup */ + file->exists_in_prev = true; + prev_file = *prevFileTmp; } - else - elog(WARNING, "unexpected file type %d", buf.st_mode); } - /* ssh connection to longer needed */ - fio_disconnect(); + /* backup file */ + if (file->is_datafile && !file->is_cfs) + { + backup_data_file(file, from_fullpath, to_fullpath, + arguments->prev_start_lsn, + current.backup_mode, + instance_config.compress_alg, + instance_config.compress_level, + arguments->nodeInfo->checksum_version, + arguments->hdr_map, false); + } + else + { + backup_non_data_file(file, prev_file, from_fullpath, to_fullpath, + current.backup_mode, current.parent_backup, true); + } - /* Close connection */ - if (arguments->conn_arg.conn) - pgut_disconnect(arguments->conn_arg.conn); + if (file->write_size == FILE_NOT_FOUND) + return; - /* Data files transferring is successful */ - arguments->ret = 0; + if (file->write_size == BYTES_INVALID) + { + elog(LOG, "Skipping the unchanged file: \"%s\"", from_fullpath); + return; + } - return NULL; + elog(LOG, "File \"%s\". Copied "INT64_FORMAT " bytes", + from_fullpath, file->write_size); + +} + +static void +backup_cfs_segment(int i, pgFile *file, backup_files_arg *arguments) { + pgFile *data_file = file; + pgFile *cfm_file = NULL; + pgFile *data_bck_file = NULL; + pgFile *cfm_bck_file = NULL; + + while (data_file->cfs_chain) + { + data_file = data_file->cfs_chain; + if (data_file->forkName == cfm) + cfm_file = data_file; + if (data_file->forkName == cfs_bck) + data_bck_file = data_file; + if (data_file->forkName == cfm_bck) + cfm_bck_file = data_file; + } + data_file = file; + if (data_file->relOid >= FirstNormalObjectId && cfm_file == NULL) + { + elog(ERROR, "'CFS' file '%s' have to have '%s.cfm' companion file", + data_file->rel_path, data_file->name); + } + + elog(LOG, "backup CFS segment %s, data_file=%s, cfm_file=%s, data_bck_file=%s, cfm_bck_file=%s", + data_file->name, data_file->name, cfm_file->name, data_bck_file == NULL? "NULL": data_bck_file->name, cfm_bck_file == NULL? "NULL": cfm_bck_file->name); + + /* storing cfs segment. processing corner case [PBCKP-287] stage 1. + * - when we do have data_bck_file we should skip both data_bck_file and cfm_bck_file if exists. + * they are removed by cfs_recover() during postgres start. + */ + if (data_bck_file) + { + if (cfm_bck_file) + cfm_bck_file->write_size = FILE_NOT_FOUND; + data_bck_file->write_size = FILE_NOT_FOUND; + } + /* else we store cfm_bck_file. processing corner case [PBCKP-287] stage 2. + * - when we do have cfm_bck_file only we should store it. + * it will replace cfm_file after postgres start. + */ + else if (cfm_bck_file) + process_file(i, cfm_bck_file, arguments); + + /* storing cfs segment in order cfm_file -> datafile to guarantee their consistency */ + /* cfm_file could be NULL for system tables. But we don't clear is_cfs flag + * for compatibility with older pg_probackup. */ + if (cfm_file) + process_file(i, cfm_file, arguments); + process_file(i, data_file, arguments); + elog(LOG, "Backup CFS segment %s done", data_file->name); } /* @@ -2119,13 +2320,10 @@ parse_filelist_filenames(parray *files, const char *root) while (i < parray_num(files)) { pgFile *file = (pgFile *) parray_get(files, i); - char *relative; int sscanf_result; - relative = GetRelativePath(file->path, root); - if (S_ISREG(file->mode) && - path_is_prefix_of_path(PG_TBLSPC_DIR, relative)) + path_is_prefix_of_path(PG_TBLSPC_DIR, file->rel_path)) { /* * Found file in pg_tblspc/tblsOid/TABLESPACE_VERSION_DIRECTORY @@ -2133,28 +2331,31 @@ parse_filelist_filenames(parray *files, const char *root) */ if (strcmp(file->name, "pg_compression") == 0) { + /* processing potential cfs tablespace */ Oid tblspcOid; Oid dbOid; char tmp_rel_path[MAXPGPATH]; /* - * Check that the file is located under + * Check that pg_compression is located under * TABLESPACE_VERSION_DIRECTORY */ - sscanf_result = sscanf(relative, PG_TBLSPC_DIR "/%u/%s/%u", + sscanf_result = sscanf(file->rel_path, PG_TBLSPC_DIR "/%u/%s/%u", &tblspcOid, tmp_rel_path, &dbOid); /* Yes, it is */ if (sscanf_result == 2 && strncmp(tmp_rel_path, TABLESPACE_VERSION_DIRECTORY, - strlen(TABLESPACE_VERSION_DIRECTORY)) == 0) - set_cfs_datafiles(files, root, relative, i); + strlen(TABLESPACE_VERSION_DIRECTORY)) == 0) { + /* rewind index to the beginning of cfs tablespace */ + rewind_and_mark_cfs_datafiles(files, root, file->rel_path, i); + } } } if (S_ISREG(file->mode) && file->tblspcOid != 0 && file->name && file->name[0]) { - if (strcmp(file->forkName, "init") == 0) + if (file->forkName == init) { /* * Do not backup files of unlogged relations. @@ -2162,7 +2363,7 @@ parse_filelist_filenames(parray *files, const char *root) */ int unlogged_file_num = i - 1; pgFile *unlogged_file = (pgFile *) parray_get(files, - unlogged_file_num); + unlogged_file_num); unlogged_file_reloid = file->relOid; @@ -2170,11 +2371,10 @@ parse_filelist_filenames(parray *files, const char *root) (unlogged_file_reloid != 0) && (unlogged_file->relOid == unlogged_file_reloid)) { - pgFileFree(unlogged_file); - parray_remove(files, unlogged_file_num); + /* flagged to remove from list on stage 2 */ + unlogged_file->remove_from_list = true; unlogged_file_num--; - i--; unlogged_file = (pgFile *) parray_get(files, unlogged_file_num); @@ -2184,6 +2384,22 @@ parse_filelist_filenames(parray *files, const char *root) i++; } + + /* stage 2. clean up from temporary tables */ + parray_remove_if(files, remove_excluded_files_criterion, NULL, pgFileFree); +} + +static bool +remove_excluded_files_criterion(void *value, void *exclude_args) { + pgFile *file = (pgFile*)value; + return file->remove_from_list; +} + +static uint32 +hash_rel_seg(pgFile* file) +{ + uint32 hash = hash_mix32_2(file->relOid, file->segno); + return hash_mix32_2(hash, 0xcf5); } /* If file is equal to pg_compression, then we consider this tablespace as @@ -2197,45 +2413,95 @@ parse_filelist_filenames(parray *files, const char *root) * tblspcOid/TABLESPACE_VERSION_DIRECTORY/dboid/1 * tblspcOid/TABLESPACE_VERSION_DIRECTORY/dboid/1.cfm * tblspcOid/TABLESPACE_VERSION_DIRECTORY/pg_compression + * + * @returns index of first tablespace entry, i.e tblspcOid/TABLESPACE_VERSION_DIRECTORY */ static void -set_cfs_datafiles(parray *files, const char *root, char *relative, size_t i) +rewind_and_mark_cfs_datafiles(parray *files, const char *root, char *relative, size_t i) { int len; int p; + int j; pgFile *prev_file; + pgFile *tmp_file; char *cfs_tblspc_path; - char *relative_prev_file; + uint32 h; + + /* hash table for cfm files */ +#define HASHN 128 + parray *hashtab[HASHN] = {NULL}; + parray *bucket; + for (p = 0; p < HASHN; p++) + hashtab[p] = parray_new(); + cfs_tblspc_path = strdup(relative); if(!cfs_tblspc_path) elog(ERROR, "Out of memory"); len = strlen("/pg_compression"); cfs_tblspc_path[strlen(cfs_tblspc_path) - len] = 0; - elog(VERBOSE, "CFS DIRECTORY %s, pg_compression path: %s", cfs_tblspc_path, relative); + elog(LOG, "CFS DIRECTORY %s, pg_compression path: %s", cfs_tblspc_path, relative); for (p = (int) i; p >= 0; p--) { prev_file = (pgFile *) parray_get(files, (size_t) p); - relative_prev_file = GetRelativePath(prev_file->path, root); - elog(VERBOSE, "Checking file in cfs tablespace %s", relative_prev_file); + elog(LOG, "Checking file in cfs tablespace %s", prev_file->rel_path); - if (strstr(relative_prev_file, cfs_tblspc_path) != NULL) + if (strstr(prev_file->rel_path, cfs_tblspc_path) == NULL) { - if (S_ISREG(prev_file->mode) && prev_file->is_datafile) + elog(LOG, "Breaking on %s", prev_file->rel_path); + break; + } + + if (!S_ISREG(prev_file->mode)) + continue; + + h = hash_rel_seg(prev_file); + bucket = hashtab[h % HASHN]; + + if (prev_file->forkName == cfm || prev_file->forkName == cfm_bck || + prev_file->forkName == cfs_bck) + { + prev_file->skip_cfs_nested = true; + parray_append(bucket, prev_file); + } + else if (prev_file->is_datafile && prev_file->forkName == none) + { + elog(LOG, "Processing 'cfs' file %s", prev_file->rel_path); + /* have to mark as is_cfs even for system-tables for compatibility + * with older pg_probackup */ + prev_file->is_cfs = true; + prev_file->cfs_chain = NULL; + for (j = 0; j < parray_num(bucket); j++) { - elog(VERBOSE, "Setting 'is_cfs' on file %s, name %s", - relative_prev_file, prev_file->name); - prev_file->is_cfs = true; + tmp_file = parray_get(bucket, j); + elog(LOG, "Linking 'cfs' file '%s' to '%s'", + tmp_file->rel_path, prev_file->rel_path); + if (tmp_file->relOid == prev_file->relOid && + tmp_file->segno == prev_file->segno) + { + tmp_file->cfs_chain = prev_file->cfs_chain; + prev_file->cfs_chain = tmp_file; + parray_remove(bucket, j); + j--; + } } } - else + } + + for (p = 0; p < HASHN; p++) + { + bucket = hashtab[p]; + for (j = 0; j < parray_num(bucket); j++) { - elog(VERBOSE, "Breaking on %s", relative_prev_file); - break; + tmp_file = parray_get(bucket, j); + elog(WARNING, "Orphaned cfs related file '%s'", tmp_file->rel_path); } + parray_free(bucket); + hashtab[p] = NULL; } +#undef HASHN free(cfs_tblspc_path); } @@ -2246,7 +2512,7 @@ set_cfs_datafiles(parray *files, const char *root, char *relative, size_t i) void process_block_change(ForkNumber forknum, RelFileNode rnode, BlockNumber blkno) { - char *path; +// char *path; char *rel_path; BlockNumber blkno_inseg; int segno; @@ -2258,16 +2524,15 @@ process_block_change(ForkNumber forknum, RelFileNode rnode, BlockNumber blkno) rel_path = relpathperm(rnode, forknum); if (segno > 0) - path = psprintf("%s/%s.%u", instance_config.pgdata, rel_path, segno); + f.rel_path = psprintf("%s.%u", rel_path, segno); else - path = psprintf("%s/%s", instance_config.pgdata, rel_path); + f.rel_path = rel_path; - pg_free(rel_path); + f.external_dir_num = 0; - f.path = path; /* backup_files_list should be sorted before */ file_item = (pgFile **) parray_bsearch(backup_files_list, &f, - pgFileComparePath); + pgFileCompareRelPathWithExternal); /* * If we don't have any record of this file in the file map, it means @@ -2287,381 +2552,13 @@ process_block_change(ForkNumber forknum, RelFileNode rnode, BlockNumber blkno) pthread_mutex_unlock(&backup_pagemap_mutex); } - pg_free(path); -} - -/* - * Given a list of files in the instance to backup, build a pagemap for each - * data file that has ptrack. Result is saved in the pagemap field of pgFile. - * NOTE we rely on the fact that provided parray is sorted by file->path. - */ -static void -make_pagemap_from_ptrack(parray *files, PGconn *backup_conn) -{ - size_t i; - Oid dbOid_with_ptrack_init = 0; - Oid tblspcOid_with_ptrack_init = 0; - char *ptrack_nonparsed = NULL; - size_t ptrack_nonparsed_size = 0; - - elog(LOG, "Compiling pagemap"); - for (i = 0; i < parray_num(files); i++) - { - pgFile *file = (pgFile *) parray_get(files, i); - size_t start_addr; - - /* - * If there is a ptrack_init file in the database, - * we must backup all its files, ignoring ptrack files for relations. - */ - if (file->is_database) - { - char *filename = strrchr(file->path, '/'); - - Assert(filename != NULL); - filename++; - - /* - * The function pg_ptrack_get_and_clear_db returns true - * if there was a ptrack_init file. - * Also ignore ptrack files for global tablespace, - * to avoid any possible specific errors. - */ - if ((file->tblspcOid == GLOBALTABLESPACE_OID) || - pg_ptrack_get_and_clear_db(file->dbOid, file->tblspcOid, backup_conn)) - { - dbOid_with_ptrack_init = file->dbOid; - tblspcOid_with_ptrack_init = file->tblspcOid; - } - } - - if (file->is_datafile) - { - if (file->tblspcOid == tblspcOid_with_ptrack_init && - file->dbOid == dbOid_with_ptrack_init) - { - /* ignore ptrack if ptrack_init exists */ - elog(VERBOSE, "Ignoring ptrack because of ptrack_init for file: %s", file->path); - file->pagemap_isabsent = true; - continue; - } - - /* get ptrack bitmap once for all segments of the file */ - if (file->segno == 0) - { - /* release previous value */ - pg_free(ptrack_nonparsed); - ptrack_nonparsed_size = 0; - - ptrack_nonparsed = pg_ptrack_get_and_clear(file->tblspcOid, file->dbOid, - file->relOid, &ptrack_nonparsed_size, backup_conn); - } - - if (ptrack_nonparsed != NULL) - { - /* - * pg_ptrack_get_and_clear() returns ptrack with VARHDR cut out. - * Compute the beginning of the ptrack map related to this segment - * - * HEAPBLOCKS_PER_BYTE. Number of heap pages one ptrack byte can track: 8 - * RELSEG_SIZE. Number of Pages per segment: 131072 - * RELSEG_SIZE/HEAPBLOCKS_PER_BYTE. number of bytes in ptrack file needed - * to keep track on one relsegment: 16384 - */ - start_addr = (RELSEG_SIZE/HEAPBLOCKS_PER_BYTE)*file->segno; - - /* - * If file segment was created after we have read ptrack, - * we won't have a bitmap for this segment. - */ - if (start_addr > ptrack_nonparsed_size) - { - elog(VERBOSE, "Ptrack is missing for file: %s", file->path); - file->pagemap_isabsent = true; - } - else - { - - if (start_addr + RELSEG_SIZE/HEAPBLOCKS_PER_BYTE > ptrack_nonparsed_size) - { - file->pagemap.bitmapsize = ptrack_nonparsed_size - start_addr; - elog(VERBOSE, "pagemap size: %i", file->pagemap.bitmapsize); - } - else - { - file->pagemap.bitmapsize = RELSEG_SIZE/HEAPBLOCKS_PER_BYTE; - elog(VERBOSE, "pagemap size: %i", file->pagemap.bitmapsize); - } - - file->pagemap.bitmap = pg_malloc(file->pagemap.bitmapsize); - memcpy(file->pagemap.bitmap, ptrack_nonparsed+start_addr, file->pagemap.bitmapsize); - } - } - else - { - /* - * If ptrack file is missing, try to copy the entire file. - * It can happen in two cases: - * - files were created by commands that bypass buffer manager - * and, correspondingly, ptrack mechanism. - * i.e. CREATE DATABASE - * - target relation was deleted. - */ - elog(VERBOSE, "Ptrack is missing for file: %s", file->path); - file->pagemap_isabsent = true; - } - } - } - elog(LOG, "Pagemap compiled"); -} - - -/* - * Stop WAL streaming if current 'xlogpos' exceeds 'stop_backup_lsn', which is - * set by pg_stop_backup(). - */ -static bool -stop_streaming(XLogRecPtr xlogpos, uint32 timeline, bool segment_finished) -{ - static uint32 prevtimeline = 0; - static XLogRecPtr prevpos = InvalidXLogRecPtr; - - /* check for interrupt */ - if (interrupted || thread_interrupted) - elog(ERROR, "Interrupted during backup stop_streaming"); - - /* we assume that we get called once at the end of each segment */ - if (segment_finished) - elog(VERBOSE, _("finished segment at %X/%X (timeline %u)"), - (uint32) (xlogpos >> 32), (uint32) xlogpos, timeline); - - /* - * Note that we report the previous, not current, position here. After a - * timeline switch, xlogpos points to the beginning of the segment because - * that's where we always begin streaming. Reporting the end of previous - * timeline isn't totally accurate, because the next timeline can begin - * slightly before the end of the WAL that we received on the previous - * timeline, but it's close enough for reporting purposes. - */ - if (prevtimeline != 0 && prevtimeline != timeline) - elog(LOG, _("switched to timeline %u at %X/%X\n"), - timeline, (uint32) (prevpos >> 32), (uint32) prevpos); - - if (!XLogRecPtrIsInvalid(stop_backup_lsn)) - { - if (xlogpos >= stop_backup_lsn) - { - stop_stream_lsn = xlogpos; - return true; - } - - /* pg_stop_backup() was executed, wait for the completion of stream */ - if (stream_stop_begin == 0) - { - elog(INFO, "Wait for LSN %X/%X to be streamed", - (uint32) (stop_backup_lsn >> 32), (uint32) stop_backup_lsn); - - stream_stop_begin = time(NULL); - } - - if (time(NULL) - stream_stop_begin > stream_stop_timeout) - elog(ERROR, "Target LSN %X/%X could not be streamed in %d seconds", - (uint32) (stop_backup_lsn >> 32), (uint32) stop_backup_lsn, - stream_stop_timeout); - } - - prevtimeline = timeline; - prevpos = xlogpos; - - return false; -} - -/* - * Start the log streaming - */ -static void * -StreamLog(void *arg) -{ - StreamThreadArg *stream_arg = (StreamThreadArg *) arg; - - /* - * Always start streaming at the beginning of a segment - */ - stream_arg->startpos -= stream_arg->startpos % instance_config.xlog_seg_size; - - /* Initialize timeout */ - stream_stop_timeout = 0; - stream_stop_begin = 0; - -#if PG_VERSION_NUM >= 100000 - /* if slot name was not provided for temp slot, use default slot name */ - if (!replication_slot && temp_slot) - replication_slot = "pg_probackup_slot"; -#endif - - -#if PG_VERSION_NUM >= 110000 - /* Create temp repslot */ - if (temp_slot) - CreateReplicationSlot(stream_arg->conn, replication_slot, - NULL, temp_slot, true, true, false); -#endif - - /* - * Start the replication - */ - elog(LOG, "started streaming WAL at %X/%X (timeline %u)", - (uint32) (stream_arg->startpos >> 32), (uint32) stream_arg->startpos, - stream_arg->starttli); - -#if PG_VERSION_NUM >= 90600 - { - StreamCtl ctl; - - MemSet(&ctl, 0, sizeof(ctl)); - - ctl.startpos = stream_arg->startpos; - ctl.timeline = stream_arg->starttli; - ctl.sysidentifier = NULL; - -#if PG_VERSION_NUM >= 100000 - ctl.walmethod = CreateWalDirectoryMethod(stream_arg->basedir, 0, true); - ctl.replication_slot = replication_slot; - ctl.stop_socket = PGINVALID_SOCKET; -#if PG_VERSION_NUM >= 100000 && PG_VERSION_NUM < 110000 - ctl.temp_slot = temp_slot; -#endif -#else - ctl.basedir = (char *) stream_arg->basedir; -#endif - - ctl.stream_stop = stop_streaming; - ctl.standby_message_timeout = standby_message_timeout; - ctl.partial_suffix = NULL; - ctl.synchronous = false; - ctl.mark_done = false; - - if(ReceiveXlogStream(stream_arg->conn, &ctl) == false) - elog(ERROR, "Problem in receivexlog"); - -#if PG_VERSION_NUM >= 100000 - if (!ctl.walmethod->finish()) - elog(ERROR, "Could not finish writing WAL files: %s", - strerror(errno)); -#endif - } -#else - if(ReceiveXlogStream(stream_arg->conn, stream_arg->startpos, stream_arg->starttli, - NULL, (char *) stream_arg->basedir, stop_streaming, - standby_message_timeout, NULL, false, false) == false) - elog(ERROR, "Problem in receivexlog"); -#endif - - elog(LOG, "finished streaming WAL at %X/%X (timeline %u)", - (uint32) (stop_stream_lsn >> 32), (uint32) stop_stream_lsn, stream_arg->starttli); - stream_arg->ret = 0; - - PQfinish(stream_arg->conn); - stream_arg->conn = NULL; - - return NULL; -} - -/* - * Get lsn of the moment when ptrack was enabled the last time. - */ -static XLogRecPtr -get_last_ptrack_lsn(PGconn *backup_conn) - -{ - PGresult *res; - uint32 lsn_hi; - uint32 lsn_lo; - XLogRecPtr lsn; - - res = pgut_execute(backup_conn, "select pg_catalog.pg_ptrack_control_lsn()", - 0, NULL); - - /* Extract timeline and LSN from results of pg_start_backup() */ - XLogDataFromLSN(PQgetvalue(res, 0, 0), &lsn_hi, &lsn_lo); - /* Calculate LSN */ - lsn = ((uint64) lsn_hi) << 32 | lsn_lo; - - PQclear(res); - return lsn; -} - -char * -pg_ptrack_get_block(ConnectionArgs *arguments, - Oid dbOid, - Oid tblsOid, - Oid relOid, - BlockNumber blknum, - size_t *result_size) -{ - PGresult *res; - char *params[4]; - char *result; - - params[0] = palloc(64); - params[1] = palloc(64); - params[2] = palloc(64); - params[3] = palloc(64); - - /* - * Use tmp_conn, since we may work in parallel threads. - * We can connect to any database. - */ - sprintf(params[0], "%i", tblsOid); - sprintf(params[1], "%i", dbOid); - sprintf(params[2], "%i", relOid); - sprintf(params[3], "%u", blknum); - - if (arguments->conn == NULL) - { - arguments->conn = pgut_connect(instance_config.conn_opt.pghost, - instance_config.conn_opt.pgport, - instance_config.conn_opt.pgdatabase, - instance_config.conn_opt.pguser); - } - - if (arguments->cancel_conn == NULL) - arguments->cancel_conn = PQgetCancel(arguments->conn); - - //elog(LOG, "db %i pg_ptrack_get_block(%i, %i, %u)",dbOid, tblsOid, relOid, blknum); - res = pgut_execute_parallel(arguments->conn, - arguments->cancel_conn, - "SELECT pg_catalog.pg_ptrack_get_block_2($1, $2, $3, $4)", - 4, (const char **)params, true, false, false); - - if (PQnfields(res) != 1) - { - elog(VERBOSE, "cannot get file block for relation oid %u", - relOid); - return NULL; - } - - if (PQgetisnull(res, 0, 0)) - { - elog(VERBOSE, "cannot get file block for relation oid %u", - relOid); - return NULL; - } - - result = (char *) PQunescapeBytea((unsigned char *) PQgetvalue(res, 0, 0), - result_size); - - PQclear(res); - - pfree(params[0]); - pfree(params[1]); - pfree(params[2]); - pfree(params[3]); + if (segno > 0) + pg_free(f.rel_path); + pg_free(rel_path); - return result; } -static void +void check_external_for_tablespaces(parray *external_list, PGconn *backup_conn) { PGresult *res; @@ -2727,60 +2624,34 @@ check_external_for_tablespaces(parray *external_list, PGconn *backup_conn) } /* - * Run IDENTIFY_SYSTEM through a given connection and - * check system identifier and timeline are matching + * Calculate pgdata_bytes + * accepts (parray *) of (pgFile *) */ -void -IdentifySystem(StreamThreadArg *stream_thread_arg) +int64 +calculate_datasize_of_filelist(parray *filelist) { - PGresult *res; - - uint64 stream_conn_sysidentifier = 0; - char *stream_conn_sysidentifier_str; - TimeLineID stream_conn_tli = 0; + int64 bytes = 0; + int i; - if (!CheckServerVersionForStreaming(stream_thread_arg->conn)) - { - PQfinish(stream_thread_arg->conn); - /* - * Error message already written in CheckServerVersionForStreaming(). - * There's no hope of recovering from a version mismatch, so don't - * retry. - */ - elog(ERROR, "Cannot continue backup because stream connect has failed."); - } - - /* - * Identify server, obtain server system identifier and timeline - */ - res = pgut_execute(stream_thread_arg->conn, "IDENTIFY_SYSTEM", 0, NULL); + /* parray_num don't check for NULL */ + if (filelist == NULL) + return 0; - if (PQresultStatus(res) != PGRES_TUPLES_OK) + for (i = 0; i < parray_num(filelist); i++) { - elog(WARNING,"Could not send replication command \"%s\": %s", - "IDENTIFY_SYSTEM", PQerrorMessage(stream_thread_arg->conn)); - PQfinish(stream_thread_arg->conn); - elog(ERROR, "Cannot continue backup because stream connect has failed."); - } - - stream_conn_sysidentifier_str = PQgetvalue(res, 0, 0); - stream_conn_tli = atoi(PQgetvalue(res, 0, 1)); + pgFile *file = (pgFile *) parray_get(filelist, i); - /* Additional sanity, primary for PG 9.5, - * where system id can be obtained only via "IDENTIFY SYSTEM" - */ - if (!parse_uint64(stream_conn_sysidentifier_str, &stream_conn_sysidentifier, 0)) - elog(ERROR, "%s is not system_identifier", stream_conn_sysidentifier_str); - - if (stream_conn_sysidentifier != instance_config.system_identifier) - elog(ERROR, "System identifier mismatch. Connected PostgreSQL instance has system id: " - "" UINT64_FORMAT ". Expected: " UINT64_FORMAT ".", - stream_conn_sysidentifier, instance_config.system_identifier); + if (file->external_dir_num != 0 || file->excluded) + continue; - if (stream_conn_tli != current.tli) - elog(ERROR, "Timeline identifier mismatch. " - "Connected PostgreSQL instance has timeline id: %X. Expected: %X.", - stream_conn_tli, current.tli); + if (S_ISDIR(file->mode)) + { + // TODO is a dir always 4K? + bytes += 4096; + continue; + } - PQclear(res); + bytes += file->size; + } + return bytes; } diff --git a/src/catalog.c b/src/catalog.c index ba7fa5aee..b29090789 100644 --- a/src/catalog.c +++ b/src/catalog.c @@ -9,6 +9,7 @@ */ #include "pg_probackup.h" +#include "access/timeline.h" #include #include @@ -18,33 +19,89 @@ #include "utils/file.h" #include "utils/configuration.h" +static pgBackup* get_closest_backup(timelineInfo *tlinfo); +static pgBackup* get_oldest_backup(timelineInfo *tlinfo); static const char *backupModes[] = {"", "PAGE", "PTRACK", "DELTA", "FULL"}; static pgBackup *readBackupControlFile(const char *path); +static int create_backup_dir(pgBackup *backup, const char *backup_instance_path); -static bool exit_hook_registered = false; -static parray *lock_files = NULL; +static bool backup_lock_exit_hook_registered = false; +static parray *locks = NULL; -static void -unlink_lock_atexit(void) +static int grab_excl_lock_file(const char *backup_dir, const char *backup_id, bool strict); +static int grab_shared_lock_file(pgBackup *backup); +static int wait_shared_owners(pgBackup *backup); + + +static void unlink_lock_atexit(bool fatal, void *userdata); +static void unlock_backup(const char *backup_dir, const char *backup_id, bool exclusive); +static void release_excl_lock_file(const char *backup_dir); +static void release_shared_lock_file(const char *backup_dir); + +#define LOCK_OK 0 +#define LOCK_FAIL_TIMEOUT 1 +#define LOCK_FAIL_ENOSPC 2 +#define LOCK_FAIL_EROFS 3 + +typedef struct LockInfo { - int i; + char backup_id[10]; + char backup_dir[MAXPGPATH]; + bool exclusive; +} LockInfo; - if (lock_files == NULL) - return; +timelineInfo * +timelineInfoNew(TimeLineID tli) +{ + timelineInfo *tlinfo = (timelineInfo *) pgut_malloc(sizeof(timelineInfo)); + MemSet(tlinfo, 0, sizeof(timelineInfo)); + tlinfo->tli = tli; + tlinfo->switchpoint = InvalidXLogRecPtr; + tlinfo->parent_link = NULL; + tlinfo->xlog_filelist = parray_new(); + tlinfo->anchor_lsn = InvalidXLogRecPtr; + tlinfo->anchor_tli = 0; + tlinfo->n_xlog_files = 0; + return tlinfo; +} - for (i = 0; i < parray_num(lock_files); i++) +/* free timelineInfo object */ +void +timelineInfoFree(void *tliInfo) +{ + timelineInfo *tli = (timelineInfo *) tliInfo; + + parray_walk(tli->xlog_filelist, pgFileFree); + parray_free(tli->xlog_filelist); + + if (tli->backups) { - char *lock_file = (char *) parray_get(lock_files, i); - int res; + /* backups themselves should freed separately */ +// parray_walk(tli->backups, pgBackupFree); + parray_free(tli->backups); + } + + pfree(tliInfo); +} + +/* Iterate over locked backups and unlock them */ +void +unlink_lock_atexit(bool unused_fatal, void *unused_userdata) +{ + int i; + + if (locks == NULL) + return; - res = fio_unlink(lock_file, FIO_BACKUP_HOST); - if (res != 0 && errno != ENOENT) - elog(WARNING, "%s: %s", lock_file, strerror(errno)); + for (i = 0; i < parray_num(locks); i++) + { + LockInfo *lock = (LockInfo *) parray_get(locks, i); + unlock_backup(lock->backup_dir, lock->backup_id, lock->exclusive); } - parray_walk(lock_files, pfree); - parray_free(lock_files); - lock_files = NULL; + parray_walk(locks, pg_free); + parray_free(locks); + locks = NULL; } /* @@ -52,13 +109,11 @@ unlink_lock_atexit(void) * If no backup matches, return NULL. */ pgBackup * -read_backup(time_t timestamp) +read_backup(const char *root_dir) { - pgBackup tmp; char conf_path[MAXPGPATH]; - tmp.start_time = timestamp; - pgBackupGetPath(&tmp, conf_path, lengthof(conf_path), BACKUP_CONTROL_FILE); + join_path_components(conf_path, root_dir, BACKUP_CONTROL_FILE); return readBackupControlFile(conf_path); } @@ -70,11 +125,12 @@ read_backup(time_t timestamp) * status. */ void -write_backup_status(pgBackup *backup, BackupStatus status) +write_backup_status(pgBackup *backup, BackupStatus status, + bool strict) { pgBackup *tmp; - tmp = read_backup(backup->start_time); + tmp = read_backup(backup->root_dir); if (!tmp) { /* @@ -84,67 +140,186 @@ write_backup_status(pgBackup *backup, BackupStatus status) return; } + /* overwrite control file only if status has changed */ + if (tmp->status == status) + { + pgBackupFree(tmp); + return; + } + backup->status = status; tmp->status = backup->status; - write_backup(tmp); + tmp->root_dir = pgut_strdup(backup->root_dir); + + /* lock backup in exclusive mode */ + if (!lock_backup(tmp, strict, true)) + elog(ERROR, "Cannot lock backup %s directory", backup_id_of(backup)); + + write_backup(tmp, strict); pgBackupFree(tmp); } /* - * Create exclusive lockfile in the backup's directory. + * Lock backup in either exclusive or shared mode. + * "strict" flag allows to ignore "out of space" errors and should be + * used only by DELETE command to free disk space on filled up + * filesystem. + * + * Only read only tasks (validate, restore) are allowed to take shared locks. + * Changing backup metadata must be done with exclusive lock. + * + * Only one process can hold exclusive lock at any time. + * Exlusive lock - PID of process, holding the lock - is placed in + * lock file: BACKUP_LOCK_FILE. + * + * Multiple proccess are allowed to take shared locks simultaneously. + * Shared locks - PIDs of proccesses, holding the lock - are placed in + * separate lock file: BACKUP_RO_LOCK_FILE. + * When taking shared lock, a brief exclusive lock is taken. + * + * -> exclusive -> grab exclusive lock file and wait until all shared lockers are gone, return + * -> shared -> grab exclusive lock file, grab shared lock file, release exclusive lock file, return + * + * TODO: lock-timeout as parameter */ bool -lock_backup(pgBackup *backup) +lock_backup(pgBackup *backup, bool strict, bool exclusive) { - char lock_file[MAXPGPATH]; - int fd; - char buffer[MAXPGPATH * 2 + 256]; - int ntries; - int len; - int encoded_pid; - pid_t my_pid, - my_p_pid; + int rc; + char lock_file[MAXPGPATH]; + bool enospc_detected = false; + LockInfo *lock = NULL; + + join_path_components(lock_file, backup->root_dir, BACKUP_LOCK_FILE); + + rc = grab_excl_lock_file(backup->root_dir, backup_id_of(backup), strict); + + if (rc == LOCK_FAIL_TIMEOUT) + return false; + else if (rc == LOCK_FAIL_ENOSPC) + { + /* + * If we failed to take exclusive lock due to ENOSPC, + * then in lax mode treat such condition as if lock was taken. + */ - pgBackupGetPath(backup, lock_file, lengthof(lock_file), BACKUP_CATALOG_PID); + enospc_detected = true; + if (strict) + return false; + } + else if (rc == LOCK_FAIL_EROFS) + { + /* + * If we failed to take exclusive lock due to EROFS, + * then in shared mode treat such condition as if lock was taken. + */ + return !exclusive; + } /* - * If the PID in the lockfile is our own PID or our parent's or - * grandparent's PID, then the file must be stale (probably left over from - * a previous system boot cycle). We need to check this because of the - * likelihood that a reboot will assign exactly the same PID as we had in - * the previous reboot, or one that's only one or two counts larger and - * hence the lockfile's PID now refers to an ancestor shell process. We - * allow pg_ctl to pass down its parent shell PID (our grandparent PID) - * via the environment variable PG_GRANDPARENT_PID; this is so that - * launching the postmaster via pg_ctl can be just as reliable as - * launching it directly. There is no provision for detecting - * further-removed ancestor processes, but if the init script is written - * carefully then all but the immediate parent shell will be root-owned - * processes and so the kill test will fail with EPERM. Note that we - * cannot get a false negative this way, because an existing postmaster - * would surely never launch a competing postmaster or pg_ctl process - * directly. + * We have exclusive lock, now there are following scenarios: + * + * 1. If we are for exlusive lock, then we must open the shared lock file + * and check if any of the processes listed there are still alive. + * If some processes are alive and are not going away in lock_timeout, + * then return false. + * + * 2. If we are here for non-exlusive lock, then write the pid + * into shared lock file and release the exclusive lock. */ - my_pid = getpid(); -#ifndef WIN32 - my_p_pid = getppid(); -#else + + if (exclusive) + rc = wait_shared_owners(backup); + else + rc = grab_shared_lock_file(backup); + + if (rc != 0) + { + /* + * Failed to grab shared lock or (in case of exclusive mode) shared lock owners + * are not going away in time, release the exclusive lock file and return in shame. + */ + release_excl_lock_file(backup->root_dir); + return false; + } + + if (!exclusive) + { + /* Shared lock file is grabbed, now we can release exclusive lock file */ + release_excl_lock_file(backup->root_dir); + } + + if (exclusive && !strict && enospc_detected) + { + /* We are in lax exclusive mode and EONSPC was encountered: + * once again try to grab exclusive lock file, + * because there is a chance that release of shared lock file in wait_shared_owners may have + * freed some space on filesystem, thanks to unlinking of BACKUP_RO_LOCK_FILE. + * If somebody concurrently acquired exclusive lock file first, then we should give up. + */ + if (grab_excl_lock_file(backup->root_dir, backup_id_of(backup), strict) == LOCK_FAIL_TIMEOUT) + return false; + + return true; + } /* - * Windows hasn't got getppid(), but doesn't need it since it's not using - * real kill() either... + * Arrange the unlocking at proc_exit. */ - my_p_pid = 0; -#endif + if (!backup_lock_exit_hook_registered) + { + pgut_atexit_push(unlink_lock_atexit, NULL); + backup_lock_exit_hook_registered = true; + } + + /* save lock metadata for later unlocking */ + lock = pgut_malloc(sizeof(LockInfo)); + snprintf(lock->backup_id, 10, "%s", backup_id_of(backup)); + snprintf(lock->backup_dir, MAXPGPATH, "%s", backup->root_dir); + lock->exclusive = exclusive; + + /* Use parray for lock release */ + if (locks == NULL) + locks = parray_new(); + parray_append(locks, lock); + + return true; +} + +/* + * Lock backup in exclusive mode + * Result codes: + * LOCK_OK Success + * LOCK_FAIL_TIMEOUT Failed to acquire lock in lock_timeout time + * LOCK_FAIL_ENOSPC Failed to acquire lock due to ENOSPC + * LOCK_FAIL_EROFS Failed to acquire lock due to EROFS + */ +int +grab_excl_lock_file(const char *root_dir, const char *backup_id, bool strict) +{ + char lock_file[MAXPGPATH]; + int fd = 0; + char buffer[256]; + int ntries = LOCK_TIMEOUT; + int empty_tries = LOCK_STALE_TIMEOUT; + int len; + int encoded_pid; + + join_path_components(lock_file, root_dir, BACKUP_LOCK_FILE); /* * We need a loop here because of race conditions. But don't loop forever * (for example, a non-writable $backup_instance_path directory might cause a failure - * that won't go away). 100 tries seems like plenty. + * that won't go away). */ - for (ntries = 0;; ntries++) + do { + FILE *fp_out = NULL; + + if (interrupted) + elog(ERROR, "Interrupted while locking backup %s", backup_id); + /* * Try to create the lock file --- O_EXCL makes this atomic. * @@ -155,10 +330,21 @@ lock_backup(pgBackup *backup) if (fd >= 0) break; /* Success; exit the retry loop */ + /* read-only fs is a special case */ + if (errno == EROFS) + { + elog(WARNING, "Could not create lock file \"%s\": %s", + lock_file, strerror(errno)); + return LOCK_FAIL_EROFS; + } + /* * Couldn't create the pid file. Probably it already exists. + * If file already exists or we have some permission problem (???), + * then retry; */ - if ((errno != EEXIST && errno != EACCES) || ntries > 100) +// if ((errno != EEXIST && errno != EACCES)) + if (errno != EEXIST) elog(ERROR, "Could not create lock file \"%s\": %s", lock_file, strerror(errno)); @@ -166,66 +352,118 @@ lock_backup(pgBackup *backup) * Read the file to get the old owner's PID. Note race condition * here: file might have been deleted since we tried to create it. */ - fd = fio_open(lock_file, O_RDONLY, FIO_BACKUP_HOST); - if (fd < 0) + + fp_out = fopen(lock_file, "r"); + if (fp_out == NULL) { if (errno == ENOENT) - continue; /* race condition; try again */ - elog(ERROR, "Could not open lock file \"%s\": %s", - lock_file, strerror(errno)); + continue; /* race condition; try again */ + elog(ERROR, "Cannot open lock file \"%s\": %s", lock_file, strerror(errno)); } - if ((len = fio_read(fd, buffer, sizeof(buffer) - 1)) < 0) - elog(ERROR, "Could not read lock file \"%s\": %s", - lock_file, strerror(errno)); - fio_close(fd); + len = fread(buffer, 1, sizeof(buffer) - 1, fp_out); + if (ferror(fp_out)) + elog(ERROR, "Cannot read from lock file: \"%s\"", lock_file); + fclose(fp_out); + + /* + * There are several possible reasons for lock file + * to be empty: + * - system crash + * - process crash + * - race between writer and reader + * + * Consider empty file to be stale after LOCK_STALE_TIMEOUT attempts. + * + * TODO: alternatively we can write into temp file (lock_file_%pid), + * rename it and then re-read lock file to make sure, + * that we are successfully acquired the lock. + */ if (len == 0) - elog(ERROR, "Lock file \"%s\" is empty", lock_file); + { + if (empty_tries == 0) + { + elog(WARNING, "Lock file \"%s\" is empty", lock_file); + goto grab_lock; + } + + if ((empty_tries % LOG_FREQ) == 0) + elog(WARNING, "Waiting %u seconds on empty exclusive lock for backup %s", + empty_tries, backup_id); + + sleep(1); + /* + * waiting on empty lock file should not affect + * the timer for concurrent lockers (ntries). + */ + empty_tries--; + ntries++; + continue; + } - buffer[len] = '\0'; encoded_pid = atoi(buffer); if (encoded_pid <= 0) - elog(ERROR, "Bogus data in lock file \"%s\": \"%s\"", + { + elog(WARNING, "Bogus data in lock file \"%s\": \"%s\"", lock_file, buffer); + goto grab_lock; + } /* * Check to see if the other process still exists - * - * Per discussion above, my_pid, my_p_pid can be - * ignored as false matches. - * * Normally kill() will fail with ESRCH if the given PID doesn't * exist. */ - if (encoded_pid != my_pid && encoded_pid != my_p_pid) + if (encoded_pid == my_pid) + return LOCK_OK; + + if (kill(encoded_pid, 0) == 0) { - if (kill(encoded_pid, 0) == 0) + /* complain every fifth interval */ + if ((ntries % LOG_FREQ) == 0) { - elog(WARNING, "Process %d is using backup %s and still is running", - encoded_pid, base36enc(backup->start_time)); - return false; + elog(WARNING, "Process %d is using backup %s, and is still running", + encoded_pid, backup_id); + + elog(WARNING, "Waiting %u seconds on exclusive lock for backup %s", + ntries, backup_id); } + + sleep(1); + + /* try again */ + continue; + } + else + { + if (errno == ESRCH) + elog(WARNING, "Process %d which used backup %s no longer exists", + encoded_pid, backup_id); else - { - if (errno == ESRCH) - elog(WARNING, "Process %d which used backup %s no longer exists", - encoded_pid, base36enc(backup->start_time)); - else - elog(ERROR, "Failed to send signal 0 to a process %d: %s", - encoded_pid, strerror(errno)); - } + elog(ERROR, "Failed to send signal 0 to a process %d: %s", + encoded_pid, strerror(errno)); } +grab_lock: /* * Looks like nobody's home. Unlink the file and try again to create * it. Need a loop because of possible race condition against other * would-be creators. */ if (fio_unlink(lock_file, FIO_BACKUP_HOST) < 0) + { + if (errno == ENOENT) + continue; /* race condition, again */ elog(ERROR, "Could not remove old lock file \"%s\": %s", lock_file, strerror(errno)); - } + } + + } while (ntries--); + + /* Failed to acquire exclusive lock in time */ + if (fd <= 0) + return LOCK_FAIL_TIMEOUT; /* * Successfully created the file, now fill it. @@ -235,59 +473,369 @@ lock_backup(pgBackup *backup) errno = 0; if (fio_write(fd, buffer, strlen(buffer)) != strlen(buffer)) { - int save_errno = errno; + int save_errno = errno; fio_close(fd); fio_unlink(lock_file, FIO_BACKUP_HOST); - /* if write didn't set errno, assume problem is no disk space */ - errno = save_errno ? save_errno : ENOSPC; - elog(ERROR, "Could not write lock file \"%s\": %s", - lock_file, strerror(errno)); + + /* In lax mode if we failed to grab lock because of 'out of space error', + * then treat backup as locked. + * Only delete command should be run in lax mode. + */ + if (!strict && save_errno == ENOSPC) + return LOCK_FAIL_ENOSPC; + else + elog(ERROR, "Could not write lock file \"%s\": %s", + lock_file, strerror(save_errno)); } + if (fio_flush(fd) != 0) { - int save_errno = errno; + int save_errno = errno; fio_close(fd); fio_unlink(lock_file, FIO_BACKUP_HOST); - errno = save_errno; - elog(ERROR, "Could not write lock file \"%s\": %s", - lock_file, strerror(errno)); + + /* In lax mode if we failed to grab lock because of 'out of space error', + * then treat backup as locked. + * Only delete command should be run in lax mode. + */ + if (!strict && save_errno == ENOSPC) + return LOCK_FAIL_ENOSPC; + else + elog(ERROR, "Could not flush lock file \"%s\": %s", + lock_file, strerror(save_errno)); } + if (fio_close(fd) != 0) { - int save_errno = errno; + int save_errno = errno; fio_unlink(lock_file, FIO_BACKUP_HOST); - errno = save_errno; - elog(ERROR, "Could not write lock file \"%s\": %s", - lock_file, strerror(errno)); + + if (!strict && errno == ENOSPC) + return LOCK_FAIL_ENOSPC; + else + elog(ERROR, "Could not close lock file \"%s\": %s", + lock_file, strerror(save_errno)); } - /* - * Arrange to unlink the lock file(s) at proc_exit. - */ - if (!exit_hook_registered) +// elog(LOG, "Acquired exclusive lock for backup %s after %ds", +// backup_id_of(backup), +// LOCK_TIMEOUT - ntries + LOCK_STALE_TIMEOUT - empty_tries); + + return LOCK_OK; +} + +/* Wait until all shared lock owners are gone + * 0 - successs + * 1 - fail + */ +int +wait_shared_owners(pgBackup *backup) +{ + FILE *fp = NULL; + char buffer[256]; + pid_t encoded_pid = 0; + int ntries = LOCK_TIMEOUT; + char lock_file[MAXPGPATH]; + + join_path_components(lock_file, backup->root_dir, BACKUP_RO_LOCK_FILE); + + fp = fopen(lock_file, "r"); + if (fp == NULL && errno != ENOENT) + elog(ERROR, "Cannot open lock file \"%s\": %s", lock_file, strerror(errno)); + + /* iterate over pids in lock file */ + while (fp && fgets(buffer, sizeof(buffer), fp)) + { + encoded_pid = atoi(buffer); + if (encoded_pid <= 0) + { + elog(WARNING, "Bogus data in lock file \"%s\": \"%s\"", lock_file, buffer); + continue; + } + + /* wait until shared lock owners go away */ + do + { + if (interrupted) + elog(ERROR, "Interrupted while locking backup %s", + backup_id_of(backup)); + + if (encoded_pid == my_pid) + break; + + /* check if lock owner is still alive */ + if (kill(encoded_pid, 0) == 0) + { + /* complain from time to time */ + if ((ntries % LOG_FREQ) == 0) + { + elog(WARNING, "Process %d is using backup %s in shared mode, and is still running", + encoded_pid, backup_id_of(backup)); + + elog(WARNING, "Waiting %u seconds on lock for backup %s", ntries, + backup_id_of(backup)); + } + + sleep(1); + + /* try again */ + continue; + } + else if (errno != ESRCH) + elog(ERROR, "Failed to send signal 0 to a process %d: %s", + encoded_pid, strerror(errno)); + + /* locker is dead */ + break; + + } while (ntries--); + } + + if (fp && ferror(fp)) + elog(ERROR, "Cannot read from lock file: \"%s\"", lock_file); + + if (fp) + fclose(fp); + + /* some shared owners are still alive */ + if (ntries <= 0) + { + elog(WARNING, "Cannot to lock backup %s in exclusive mode, because process %u owns shared lock", + backup_id_of(backup), encoded_pid); + return 1; + } + + /* unlink shared lock file */ + fio_unlink(lock_file, FIO_BACKUP_HOST); + return 0; +} + +/* + * Lock backup in shared mode + * 0 - successs + * 1 - fail + */ +int +grab_shared_lock_file(pgBackup *backup) +{ + FILE *fp_in = NULL; + FILE *fp_out = NULL; + char buf_in[256]; + pid_t encoded_pid; + char lock_file[MAXPGPATH]; + + char buffer[8192]; /*TODO: should be enough, but maybe malloc+realloc is better ? */ + char lock_file_tmp[MAXPGPATH]; + int buffer_len = 0; + + join_path_components(lock_file, backup->root_dir, BACKUP_RO_LOCK_FILE); + snprintf(lock_file_tmp, MAXPGPATH, "%s%s", lock_file, "tmp"); + + /* open already existing lock files */ + fp_in = fopen(lock_file, "r"); + if (fp_in == NULL && errno != ENOENT) + elog(ERROR, "Cannot open lock file \"%s\": %s", lock_file, strerror(errno)); + + /* read PIDs of owners */ + while (fp_in && fgets(buf_in, sizeof(buf_in), fp_in)) { - atexit(unlink_lock_atexit); - exit_hook_registered = true; + encoded_pid = atoi(buf_in); + if (encoded_pid <= 0) + { + elog(WARNING, "Bogus data in lock file \"%s\": \"%s\"", lock_file, buf_in); + continue; + } + + if (encoded_pid == my_pid) + continue; + + if (kill(encoded_pid, 0) == 0) + { + /* + * Somebody is still using this backup in shared mode, + * copy this pid into a new file. + */ + buffer_len += snprintf(buffer+buffer_len, 4096, "%u\n", encoded_pid); + } + else if (errno != ESRCH) + elog(ERROR, "Failed to send signal 0 to a process %d: %s", + encoded_pid, strerror(errno)); + } + + if (fp_in) + { + if (ferror(fp_in)) + elog(ERROR, "Cannot read from lock file: \"%s\"", lock_file); + fclose(fp_in); + } + + fp_out = fopen(lock_file_tmp, "w"); + if (fp_out == NULL) + { + if (errno == EROFS) + return 0; + + elog(ERROR, "Cannot open temp lock file \"%s\": %s", lock_file_tmp, strerror(errno)); } - /* Use parray so that the lock files are unlinked in a loop */ - if (lock_files == NULL) - lock_files = parray_new(); - parray_append(lock_files, pgut_strdup(lock_file)); + /* add my own pid */ + buffer_len += snprintf(buffer+buffer_len, sizeof(buffer), "%u\n", my_pid); - return true; + /* write out the collected PIDs to temp lock file */ + fwrite(buffer, 1, buffer_len, fp_out); + + if (ferror(fp_out)) + elog(ERROR, "Cannot write to lock file: \"%s\"", lock_file_tmp); + + if (fclose(fp_out) != 0) + elog(ERROR, "Cannot close temp lock file \"%s\": %s", lock_file_tmp, strerror(errno)); + + if (rename(lock_file_tmp, lock_file) < 0) + elog(ERROR, "Cannot rename file \"%s\" to \"%s\": %s", + lock_file_tmp, lock_file, strerror(errno)); + + return 0; +} + +void +unlock_backup(const char *backup_dir, const char *backup_id, bool exclusive) +{ + if (exclusive) + { + release_excl_lock_file(backup_dir); + return; + } + + /* To remove shared lock, we must briefly obtain exclusive lock, ... */ + if (grab_excl_lock_file(backup_dir, backup_id, false) != LOCK_OK) + /* ... if it's not possible then leave shared lock */ + return; + + release_shared_lock_file(backup_dir); + release_excl_lock_file(backup_dir); +} + +void +release_excl_lock_file(const char *backup_dir) +{ + char lock_file[MAXPGPATH]; + + join_path_components(lock_file, backup_dir, BACKUP_LOCK_FILE); + + /* TODO Sanity check: maybe we should check, that pid in lock file is my_pid */ + + /* unlink pid file */ + fio_unlink(lock_file, FIO_BACKUP_HOST); +} + +void +release_shared_lock_file(const char *backup_dir) +{ + FILE *fp_in = NULL; + FILE *fp_out = NULL; + char buf_in[256]; + pid_t encoded_pid; + char lock_file[MAXPGPATH]; + + char buffer[8192]; /*TODO: should be enough, but maybe malloc+realloc is better ? */ + char lock_file_tmp[MAXPGPATH]; + int buffer_len = 0; + + join_path_components(lock_file, backup_dir, BACKUP_RO_LOCK_FILE); + snprintf(lock_file_tmp, MAXPGPATH, "%s%s", lock_file, "tmp"); + + /* open lock file */ + fp_in = fopen(lock_file, "r"); + if (fp_in == NULL) + { + if (errno == ENOENT) + return; + else + elog(ERROR, "Cannot open lock file \"%s\": %s", lock_file, strerror(errno)); + } + + /* read PIDs of owners */ + while (fgets(buf_in, sizeof(buf_in), fp_in)) + { + encoded_pid = atoi(buf_in); + + if (encoded_pid <= 0) + { + elog(WARNING, "Bogus data in lock file \"%s\": \"%s\"", lock_file, buf_in); + continue; + } + + /* remove my pid */ + if (encoded_pid == my_pid) + continue; + + if (kill(encoded_pid, 0) == 0) + { + /* + * Somebody is still using this backup in shared mode, + * copy this pid into a new file. + */ + buffer_len += snprintf(buffer+buffer_len, 4096, "%u\n", encoded_pid); + } + else if (errno != ESRCH) + elog(ERROR, "Failed to send signal 0 to a process %d: %s", + encoded_pid, strerror(errno)); + } + + if (ferror(fp_in)) + elog(ERROR, "Cannot read from lock file: \"%s\"", lock_file); + fclose(fp_in); + + /* if there is no active pid left, then there is nothing to do */ + if (buffer_len == 0) + { + fio_unlink(lock_file, FIO_BACKUP_HOST); + return; + } + + fp_out = fopen(lock_file_tmp, "w"); + if (fp_out == NULL) + elog(ERROR, "Cannot open temp lock file \"%s\": %s", lock_file_tmp, strerror(errno)); + + /* write out the collected PIDs to temp lock file */ + fwrite(buffer, 1, buffer_len, fp_out); + + if (ferror(fp_out)) + elog(ERROR, "Cannot write to lock file: \"%s\"", lock_file_tmp); + + if (fclose(fp_out) != 0) + elog(ERROR, "Cannot close temp lock file \"%s\": %s", lock_file_tmp, strerror(errno)); + + if (rename(lock_file_tmp, lock_file) < 0) + elog(ERROR, "Cannot rename file \"%s\" to \"%s\": %s", + lock_file_tmp, lock_file, strerror(errno)); + + return; } /* * Get backup_mode in string representation. */ const char * -pgBackupGetBackupMode(pgBackup *backup) +pgBackupGetBackupMode(pgBackup *backup, bool show_color) { - return backupModes[backup->backup_mode]; + if (show_color) + { + /* color the Backup mode */ + char *mode = pgut_malloc(24); /* leaking memory here */ + + if (backup->backup_mode == BACKUP_MODE_FULL) + snprintf(mode, 24, "%s%s%s", TC_GREEN_BOLD, backupModes[backup->backup_mode], TC_RESET); + else + snprintf(mode, 24, "%s%s%s", TC_BLUE_BOLD, backupModes[backup->backup_mode], TC_RESET); + + return mode; + } + else + return backupModes[backup->backup_mode]; } static bool @@ -296,11 +844,80 @@ IsDir(const char *dirpath, const char *entry, fio_location location) char path[MAXPGPATH]; struct stat st; - snprintf(path, MAXPGPATH, "%s/%s", dirpath, entry); + join_path_components(path, dirpath, entry); return fio_stat(path, &st, false, location) == 0 && S_ISDIR(st.st_mode); } +/* + * Create list of instances in given backup catalog. + * + * Returns parray of InstanceState structures. + */ +parray * +catalog_get_instance_list(CatalogState *catalogState) +{ + DIR *dir; + struct dirent *dent; + parray *instances; + + instances = parray_new(); + + /* open directory and list contents */ + dir = opendir(catalogState->backup_subdir_path); + if (dir == NULL) + elog(ERROR, "Cannot open directory \"%s\": %s", + catalogState->backup_subdir_path, strerror(errno)); + + while (errno = 0, (dent = readdir(dir)) != NULL) + { + char child[MAXPGPATH]; + struct stat st; + InstanceState *instanceState = NULL; + + /* skip entries point current dir or parent dir */ + if (strcmp(dent->d_name, ".") == 0 || + strcmp(dent->d_name, "..") == 0) + continue; + + join_path_components(child, catalogState->backup_subdir_path, dent->d_name); + + if (lstat(child, &st) == -1) + elog(ERROR, "Cannot stat file \"%s\": %s", + child, strerror(errno)); + + if (!S_ISDIR(st.st_mode)) + continue; + + instanceState = pgut_new(InstanceState); + + strlcpy(instanceState->instance_name, dent->d_name, MAXPGPATH); + join_path_components(instanceState->instance_backup_subdir_path, + catalogState->backup_subdir_path, instanceState->instance_name); + join_path_components(instanceState->instance_wal_subdir_path, + catalogState->wal_subdir_path, instanceState->instance_name); + join_path_components(instanceState->instance_config_path, + instanceState->instance_backup_subdir_path, BACKUP_CATALOG_CONF_FILE); + + instanceState->config = readInstanceConfigFile(instanceState); + parray_append(instances, instanceState); + } + + /* TODO 3.0: switch to ERROR */ + if (parray_num(instances) == 0) + elog(WARNING, "This backup catalog contains no backup instances. Backup instance can be added via 'add-instance' command."); + + if (errno) + elog(ERROR, "Cannot read directory \"%s\": %s", + catalogState->backup_subdir_path, strerror(errno)); + + if (closedir(dir)) + elog(ERROR, "Cannot close directory \"%s\": %s", + catalogState->backup_subdir_path, strerror(errno)); + + return instances; +} + /* * Create list of backups. * If 'requested_backup_id' is INVALID_BACKUP_ID, return list of all backups. @@ -308,7 +925,7 @@ IsDir(const char *dirpath, const char *entry, fio_location location) * If valid backup id is passed only matching backup will be added to the list. */ parray * -catalog_get_backup_list(time_t requested_backup_id) +catalog_get_backup_list(InstanceState *instanceState, time_t requested_backup_id) { DIR *data_dir = NULL; struct dirent *data_ent = NULL; @@ -316,10 +933,10 @@ catalog_get_backup_list(time_t requested_backup_id) int i; /* open backup instance backups directory */ - data_dir = fio_opendir(backup_instance_path, FIO_BACKUP_HOST); + data_dir = fio_opendir(instanceState->instance_backup_subdir_path, FIO_BACKUP_HOST); if (data_dir == NULL) { - elog(WARNING, "cannot open directory \"%s\": %s", backup_instance_path, + elog(WARNING, "cannot open directory \"%s\": %s", instanceState->instance_backup_subdir_path, strerror(errno)); goto err_proc; } @@ -333,30 +950,42 @@ catalog_get_backup_list(time_t requested_backup_id) pgBackup *backup = NULL; /* skip not-directory entries and hidden entries */ - if (!IsDir(backup_instance_path, data_ent->d_name, FIO_BACKUP_HOST) + if (!IsDir(instanceState->instance_backup_subdir_path, data_ent->d_name, FIO_BACKUP_HOST) || data_ent->d_name[0] == '.') continue; /* open subdirectory of specific backup */ - join_path_components(data_path, backup_instance_path, data_ent->d_name); + join_path_components(data_path, instanceState->instance_backup_subdir_path, data_ent->d_name); /* read backup information from BACKUP_CONTROL_FILE */ - snprintf(backup_conf_path, MAXPGPATH, "%s/%s", data_path, BACKUP_CONTROL_FILE); + join_path_components(backup_conf_path, data_path, BACKUP_CONTROL_FILE); backup = readBackupControlFile(backup_conf_path); if (!backup) { - backup = pgut_new(pgBackup); + backup = pgut_new0(pgBackup); pgBackupInit(backup); backup->start_time = base36dec(data_ent->d_name); + /* XXX BACKUP_ID change it when backup_id wouldn't match start_time */ + Assert(backup->backup_id == 0 || backup->backup_id == backup->start_time); + backup->backup_id = backup->start_time; } - else if (strcmp(base36enc(backup->start_time), data_ent->d_name) != 0) + else if (strcmp(backup_id_of(backup), data_ent->d_name) != 0) { + /* TODO there is no such guarantees */ elog(WARNING, "backup ID in control file \"%s\" doesn't match name of the backup folder \"%s\"", - base36enc(backup->start_time), backup_conf_path); + backup_id_of(backup), backup_conf_path); } - backup->backup_id = backup->start_time; + backup->root_dir = pgut_strdup(data_path); + + backup->database_dir = pgut_malloc(MAXPGPATH); + join_path_components(backup->database_dir, backup->root_dir, DATABASE_DIR); + + /* Initialize page header map */ + init_header_map(backup); + + /* TODO: save encoded backup id */ if (requested_backup_id != INVALID_BACKUP_ID && requested_backup_id != backup->start_time) { @@ -364,18 +993,12 @@ catalog_get_backup_list(time_t requested_backup_id) continue; } parray_append(backups, backup); - - if (errno && errno != ENOENT) - { - elog(WARNING, "cannot read data directory \"%s\": %s", - data_ent->d_name, strerror(errno)); - goto err_proc; - } } + if (errno) { - elog(WARNING, "cannot read backup root directory \"%s\": %s", - backup_instance_path, strerror(errno)); + elog(WARNING, "Cannot read backup root directory \"%s\": %s", + instanceState->instance_backup_subdir_path, strerror(errno)); goto err_proc; } @@ -384,205 +1007,1354 @@ catalog_get_backup_list(time_t requested_backup_id) parray_qsort(backups, pgBackupCompareIdDesc); - /* Link incremental backups with their ancestors.*/ - for (i = 0; i < parray_num(backups); i++) + /* Link incremental backups with their ancestors.*/ + for (i = 0; i < parray_num(backups); i++) + { + pgBackup *curr = parray_get(backups, i); + pgBackup **ancestor; + pgBackup key = {0}; + + if (curr->backup_mode == BACKUP_MODE_FULL) + continue; + + key.start_time = curr->parent_backup; + ancestor = (pgBackup **) parray_bsearch(backups, &key, + pgBackupCompareIdDesc); + if (ancestor) + curr->parent_backup_link = *ancestor; + } + + return backups; + +err_proc: + if (data_dir) + fio_closedir(data_dir); + if (backups) + parray_walk(backups, pgBackupFree); + parray_free(backups); + + elog(ERROR, "Failed to get backup list"); + + return NULL; +} + +/* + * Get list of files in the backup from the DATABASE_FILE_LIST. + */ +parray * +get_backup_filelist(pgBackup *backup, bool strict) +{ + parray *files = NULL; + char backup_filelist_path[MAXPGPATH]; + FILE *fp; + char buf[BLCKSZ]; + char stdio_buf[STDIO_BUFSIZE]; + pg_crc32 content_crc = 0; + + join_path_components(backup_filelist_path, backup->root_dir, DATABASE_FILE_LIST); + + fp = fio_open_stream(backup_filelist_path, FIO_BACKUP_HOST); + if (fp == NULL) + elog(ERROR, "Cannot open \"%s\": %s", backup_filelist_path, strerror(errno)); + + /* enable stdio buffering for local file */ + if (!fio_is_remote(FIO_BACKUP_HOST)) + setvbuf(fp, stdio_buf, _IOFBF, STDIO_BUFSIZE); + + files = parray_new(); + + INIT_FILE_CRC32(true, content_crc); + + while (fgets(buf, lengthof(buf), fp)) + { + char path[MAXPGPATH]; + char linked[MAXPGPATH]; + char compress_alg_string[MAXPGPATH]; + int64 write_size, + uncompressed_size, + mode, /* bit length of mode_t depends on platforms */ + is_datafile, + is_cfs, + external_dir_num, + crc, + segno, + n_blocks, + n_headers, + dbOid, /* used for partial restore */ + hdr_crc, + hdr_off, + hdr_size; + pgFile *file; + + COMP_FILE_CRC32(true, content_crc, buf, strlen(buf)); + + get_control_value_str(buf, "path", path, sizeof(path),true); + get_control_value_int64(buf, "size", &write_size, true); + get_control_value_int64(buf, "mode", &mode, true); + get_control_value_int64(buf, "is_datafile", &is_datafile, true); + get_control_value_int64(buf, "is_cfs", &is_cfs, false); + get_control_value_int64(buf, "crc", &crc, true); + get_control_value_str(buf, "compress_alg", compress_alg_string, sizeof(compress_alg_string), false); + get_control_value_int64(buf, "external_dir_num", &external_dir_num, false); + get_control_value_int64(buf, "dbOid", &dbOid, false); + + file = pgFileInit(path); + file->write_size = (int64) write_size; + file->mode = (mode_t) mode; + file->is_datafile = is_datafile ? true : false; + file->is_cfs = is_cfs ? true : false; + file->crc = (pg_crc32) crc; + file->compress_alg = parse_compress_alg(compress_alg_string); + file->external_dir_num = external_dir_num; + file->dbOid = dbOid ? dbOid : 0; + + /* + * Optional fields + */ + if (get_control_value_str(buf, "linked", linked, sizeof(linked), false) && linked[0]) + { + file->linked = pgut_strdup(linked); + canonicalize_path(file->linked); + } + + if (get_control_value_int64(buf, "segno", &segno, false)) + file->segno = (int) segno; + + if (get_control_value_int64(buf, "n_blocks", &n_blocks, false)) + file->n_blocks = (int) n_blocks; + + if (get_control_value_int64(buf, "n_headers", &n_headers, false)) + file->n_headers = (int) n_headers; + + if (get_control_value_int64(buf, "hdr_crc", &hdr_crc, false)) + file->hdr_crc = (pg_crc32) hdr_crc; + + if (get_control_value_int64(buf, "hdr_off", &hdr_off, false)) + file->hdr_off = hdr_off; + + if (get_control_value_int64(buf, "hdr_size", &hdr_size, false)) + file->hdr_size = (int) hdr_size; + + if (get_control_value_int64(buf, "full_size", &uncompressed_size, false)) + file->uncompressed_size = uncompressed_size; + else + file->uncompressed_size = write_size; + if (!file->is_datafile || file->is_cfs) + file->size = file->uncompressed_size; + + if (file->external_dir_num == 0 && + (file->dbOid != 0 || + path_is_prefix_of_path("global", file->rel_path)) && + S_ISREG(file->mode)) + { + bool is_datafile = file->is_datafile; + set_forkname(file); + if (is_datafile != file->is_datafile) + { + if (is_datafile) + elog(WARNING, "File '%s' was stored as datafile, but looks like it is not", + file->rel_path); + else + elog(WARNING, "File '%s' was stored as non-datafile, but looks like it is", + file->rel_path); + /* Lets fail in tests */ + Assert(file->is_datafile == file->is_datafile); + file->is_datafile = is_datafile; + } + } + + parray_append(files, file); + } + + FIN_FILE_CRC32(true, content_crc); + + if (ferror(fp)) + elog(ERROR, "Failed to read from file: \"%s\"", backup_filelist_path); + + fio_close_stream(fp); + + if (backup->content_crc != 0 && + backup->content_crc != content_crc) + { + elog(WARNING, "Invalid CRC of backup control file '%s': %u. Expected: %u", + backup_filelist_path, content_crc, backup->content_crc); + parray_free(files); + files = NULL; + + } + + /* redundant sanity? */ + if (!files) + elog(strict ? ERROR : WARNING, "Failed to get file list for backup %s", backup_id_of(backup)); + + return files; +} + +/* + * Lock list of backups. Function goes in backward direction. + */ +void +catalog_lock_backup_list(parray *backup_list, int from_idx, int to_idx, bool strict, bool exclusive) +{ + int start_idx, + end_idx; + int i; + + if (parray_num(backup_list) == 0) + return; + + start_idx = Max(from_idx, to_idx); + end_idx = Min(from_idx, to_idx); + + for (i = start_idx; i >= end_idx; i--) + { + pgBackup *backup = (pgBackup *) parray_get(backup_list, i); + if (!lock_backup(backup, strict, exclusive)) + elog(ERROR, "Cannot lock backup %s directory", + backup_id_of(backup)); + } +} + +/* + * Find the latest valid child of latest valid FULL backup on given timeline + */ +pgBackup * +catalog_get_last_data_backup(parray *backup_list, TimeLineID tli, time_t current_start_time) +{ + int i; + pgBackup *full_backup = NULL; + pgBackup *tmp_backup = NULL; + + /* backup_list is sorted in order of descending ID */ + for (i = 0; i < parray_num(backup_list); i++) + { + pgBackup *backup = (pgBackup *) parray_get(backup_list, i); + + if ((backup->backup_mode == BACKUP_MODE_FULL && + (backup->status == BACKUP_STATUS_OK || + backup->status == BACKUP_STATUS_DONE)) && backup->tli == tli) + { + full_backup = backup; + break; + } + } + + /* Failed to find valid FULL backup to fulfill ancestor role */ + if (!full_backup) + return NULL; + + elog(LOG, "Latest valid FULL backup: %s", + backup_id_of(full_backup)); + + /* FULL backup is found, lets find his latest child */ + for (i = 0; i < parray_num(backup_list); i++) + { + pgBackup *backup = (pgBackup *) parray_get(backup_list, i); + + /* only valid descendants are acceptable for evaluation */ + if ((backup->status == BACKUP_STATUS_OK || + backup->status == BACKUP_STATUS_DONE)) + { + switch (scan_parent_chain(backup, &tmp_backup)) + { + /* broken chain */ + case ChainIsBroken: + elog(WARNING, "Backup %s has missing parent: %s. Cannot be a parent", + backup_id_of(backup), base36enc(tmp_backup->parent_backup)); + continue; + + /* chain is intact, but at least one parent is invalid */ + case ChainIsInvalid: + elog(WARNING, "Backup %s has invalid parent: %s. Cannot be a parent", + backup_id_of(backup), backup_id_of(tmp_backup)); + continue; + + /* chain is ok */ + case ChainIsOk: + /* Yes, we could call is_parent() earlier - after choosing the ancestor, + * but this way we have an opportunity to detect and report all possible + * anomalies. + */ + if (is_parent(full_backup->start_time, backup, true)) + return backup; + } + } + /* skip yourself */ + else if (backup->start_time == current_start_time) + continue; + else + { + elog(WARNING, "Backup %s has status: %s. Cannot be a parent.", + backup_id_of(backup), status2str(backup->status)); + } + } + + return NULL; +} + +/* + * For multi-timeline chain, look up suitable parent for incremental backup. + * Multi-timeline chain has full backup and one or more descendants located + * on different timelines. + */ +pgBackup * +get_multi_timeline_parent(parray *backup_list, parray *tli_list, + TimeLineID current_tli, time_t current_start_time, + InstanceConfig *instance) +{ + int i; + timelineInfo *my_tlinfo = NULL; + timelineInfo *tmp_tlinfo = NULL; + pgBackup *ancestor_backup = NULL; + + /* there are no timelines in the archive */ + if (parray_num(tli_list) == 0) + return NULL; + + /* look for current timelineInfo */ + for (i = 0; i < parray_num(tli_list); i++) + { + timelineInfo *tlinfo = (timelineInfo *) parray_get(tli_list, i); + + if (tlinfo->tli == current_tli) + { + my_tlinfo = tlinfo; + break; + } + } + + if (my_tlinfo == NULL) + return NULL; + + /* Locate tlinfo of suitable full backup. + * Consider this example: + * t3 s2-------X <-! We are here + * / + * t2 s1----D---*----E---> + * / + * t1--A--B--*---C-------> + * + * A, E - full backups + * B, C, D - incremental backups + * + * We must find A. + */ + tmp_tlinfo = my_tlinfo; + while (tmp_tlinfo->parent_link) + { + /* if timeline has backups, iterate over them */ + if (tmp_tlinfo->parent_link->backups) + { + for (i = 0; i < parray_num(tmp_tlinfo->parent_link->backups); i++) + { + pgBackup *backup = (pgBackup *) parray_get(tmp_tlinfo->parent_link->backups, i); + + if (backup->backup_mode == BACKUP_MODE_FULL && + (backup->status == BACKUP_STATUS_OK || + backup->status == BACKUP_STATUS_DONE) && + backup->stop_lsn <= tmp_tlinfo->switchpoint) + { + ancestor_backup = backup; + break; + } + } + } + + if (ancestor_backup) + break; + + tmp_tlinfo = tmp_tlinfo->parent_link; + } + + /* failed to find valid FULL backup on parent timelines */ + if (!ancestor_backup) + return NULL; + else + elog(LOG, "Latest valid full backup: %s, tli: %i", + backup_id_of(ancestor_backup), ancestor_backup->tli); + + /* At this point we found suitable full backup, + * now we must find his latest child, suitable to be + * parent of current incremental backup. + * Consider this example: + * t3 s2-------X <-! We are here + * / + * t2 s1----D---*----E---> + * / + * t1--A--B--*---C-------> + * + * A, E - full backups + * B, C, D - incremental backups + * + * We found A, now we must find D. + */ + + /* Optimistically, look on current timeline for valid incremental backup, child of ancestor */ + if (my_tlinfo->backups) + { + /* backups are sorted in descending order and we need latest valid */ + for (i = 0; i < parray_num(my_tlinfo->backups); i++) + { + pgBackup *tmp_backup = NULL; + pgBackup *backup = (pgBackup *) parray_get(my_tlinfo->backups, i); + + /* found suitable parent */ + if (scan_parent_chain(backup, &tmp_backup) == ChainIsOk && + is_parent(ancestor_backup->start_time, backup, false)) + return backup; + } + } + + /* Iterate over parent timelines and look for a valid backup, child of ancestor */ + tmp_tlinfo = my_tlinfo; + while (tmp_tlinfo->parent_link) + { + + /* if timeline has backups, iterate over them */ + if (tmp_tlinfo->parent_link->backups) + { + for (i = 0; i < parray_num(tmp_tlinfo->parent_link->backups); i++) + { + pgBackup *tmp_backup = NULL; + pgBackup *backup = (pgBackup *) parray_get(tmp_tlinfo->parent_link->backups, i); + + /* We are not interested in backups + * located outside of our timeline history + */ + if (backup->stop_lsn > tmp_tlinfo->switchpoint) + continue; + + if (scan_parent_chain(backup, &tmp_backup) == ChainIsOk && + is_parent(ancestor_backup->start_time, backup, true)) + return backup; + } + } + + tmp_tlinfo = tmp_tlinfo->parent_link; + } + + return NULL; +} + +/* + * Create backup directory in $BACKUP_PATH + * (with proposed backup->backup_id) + * and initialize this directory. + * If creation of directory fails, then + * backup_id will be cleared (set to INVALID_BACKUP_ID). + * It is possible to get diffrent values in + * pgBackup.start_time and pgBackup.backup_id. + * It may be ok or maybe not, so it's up to the caller + * to fix it or let it be. + */ + +void +pgBackupInitDir(pgBackup *backup, const char *backup_instance_path) +{ + int i; + char temp[MAXPGPATH]; + parray *subdirs; + + /* Try to create backup directory at first */ + if (create_backup_dir(backup, backup_instance_path) != 0) + { + /* Clear backup_id as indication of error */ + reset_backup_id(backup); + return; + } + + subdirs = parray_new(); + parray_append(subdirs, pg_strdup(DATABASE_DIR)); + + /* Add external dirs containers */ + if (backup->external_dir_str) + { + parray *external_list; + + external_list = make_external_directory_list(backup->external_dir_str, + false); + for (i = 0; i < parray_num(external_list); i++) + { + /* Numeration of externaldirs starts with 1 */ + makeExternalDirPathByNum(temp, EXTERNAL_DIR, i+1); + parray_append(subdirs, pg_strdup(temp)); + } + free_dir_list(external_list); + } + + backup->database_dir = pgut_malloc(MAXPGPATH); + join_path_components(backup->database_dir, backup->root_dir, DATABASE_DIR); + + /* block header map */ + init_header_map(backup); + + /* create directories for actual backup files */ + for (i = 0; i < parray_num(subdirs); i++) + { + join_path_components(temp, backup->root_dir, parray_get(subdirs, i)); + fio_mkdir(temp, DIR_PERMISSION, FIO_BACKUP_HOST); + } + + free_dir_list(subdirs); +} + +/* + * Create root directory for backup, + * update pgBackup.root_dir if directory creation was a success + * Return values (same as dir_create_dir()): + * 0 - ok + * -1 - error (warning message already emitted) + */ +int +create_backup_dir(pgBackup *backup, const char *backup_instance_path) +{ + int rc; + char path[MAXPGPATH]; + + join_path_components(path, backup_instance_path, backup_id_of(backup)); + + /* TODO: add wrapper for remote mode */ + rc = dir_create_dir(path, DIR_PERMISSION, true); + + if (rc == 0) + backup->root_dir = pgut_strdup(path); + else + elog(WARNING, "Cannot create directory \"%s\": %s", path, strerror(errno)); + return rc; +} + +/* + * Create list of timelines. + * TODO: '.partial' and '.part' segno information should be added to tlinfo. + */ +parray * +catalog_get_timelines(InstanceState *instanceState, InstanceConfig *instance) +{ + int i,j,k; + parray *xlog_files_list = parray_new(); + parray *timelineinfos; + parray *backups; + timelineInfo *tlinfo; + + /* for fancy reporting */ + char begin_segno_str[MAXFNAMELEN]; + char end_segno_str[MAXFNAMELEN]; + + /* read all xlog files that belong to this archive */ + dir_list_file(xlog_files_list, instanceState->instance_wal_subdir_path, + false, true, false, false, true, 0, FIO_BACKUP_HOST); + parray_qsort(xlog_files_list, pgFileCompareName); + + timelineinfos = parray_new(); + tlinfo = NULL; + + /* walk through files and collect info about timelines */ + for (i = 0; i < parray_num(xlog_files_list); i++) + { + pgFile *file = (pgFile *) parray_get(xlog_files_list, i); + TimeLineID tli; + parray *timelines; + xlogFile *wal_file = NULL; + + /* + * Regular WAL file. + * IsXLogFileName() cannot be used here + */ + if (strspn(file->name, "0123456789ABCDEF") == XLOG_FNAME_LEN) + { + int result = 0; + uint32 log, seg; + XLogSegNo segno = 0; + char suffix[MAXFNAMELEN]; + + result = sscanf(file->name, "%08X%08X%08X.%s", + &tli, &log, &seg, (char *) &suffix); + + /* sanity */ + if (result < 3) + { + elog(WARNING, "unexpected WAL file name \"%s\"", file->name); + continue; + } + + /* get segno from log */ + GetXLogSegNoFromScrath(segno, log, seg, instance->xlog_seg_size); + + /* regular WAL file with suffix */ + if (result == 4) + { + /* backup history file. Currently we don't use them */ + if (IsBackupHistoryFileName(file->name)) + { + elog(VERBOSE, "backup history file \"%s\"", file->name); + + if (!tlinfo || tlinfo->tli != tli) + { + tlinfo = timelineInfoNew(tli); + parray_append(timelineinfos, tlinfo); + } + + /* append file to xlog file list */ + wal_file = palloc(sizeof(xlogFile)); + wal_file->file = *file; + wal_file->segno = segno; + wal_file->type = BACKUP_HISTORY_FILE; + wal_file->keep = false; + parray_append(tlinfo->xlog_filelist, wal_file); + continue; + } + /* partial WAL segment */ + else if (IsPartialXLogFileName(file->name) || + IsPartialCompressXLogFileName(file->name)) + { + elog(VERBOSE, "partial WAL file \"%s\"", file->name); + + if (!tlinfo || tlinfo->tli != tli) + { + tlinfo = timelineInfoNew(tli); + parray_append(timelineinfos, tlinfo); + } + + /* append file to xlog file list */ + wal_file = palloc(sizeof(xlogFile)); + wal_file->file = *file; + wal_file->segno = segno; + wal_file->type = PARTIAL_SEGMENT; + wal_file->keep = false; + parray_append(tlinfo->xlog_filelist, wal_file); + continue; + } + /* temp WAL segment */ + else if (IsTempXLogFileName(file->name) || + IsTempCompressXLogFileName(file->name) || + IsTempPartialXLogFileName(file->name)) + { + elog(VERBOSE, "temp WAL file \"%s\"", file->name); + + if (!tlinfo || tlinfo->tli != tli) + { + tlinfo = timelineInfoNew(tli); + parray_append(timelineinfos, tlinfo); + } + + /* append file to xlog file list */ + wal_file = palloc(sizeof(xlogFile)); + wal_file->file = *file; + wal_file->segno = segno; + wal_file->type = TEMP_SEGMENT; + wal_file->keep = false; + parray_append(tlinfo->xlog_filelist, wal_file); + continue; + } + /* we only expect compressed wal files with .gz suffix */ + else if (strcmp(suffix, "gz") != 0) + { + elog(WARNING, "unexpected WAL file name \"%s\"", file->name); + continue; + } + } + + /* new file belongs to new timeline */ + if (!tlinfo || tlinfo->tli != tli) + { + tlinfo = timelineInfoNew(tli); + parray_append(timelineinfos, tlinfo); + } + /* + * As it is impossible to detect if segments before segno are lost, + * or just do not exist, do not report them as lost. + */ + else if (tlinfo->n_xlog_files != 0) + { + /* check, if segments are consequent */ + XLogSegNo expected_segno = tlinfo->end_segno + 1; + + /* + * Some segments are missing. remember them in lost_segments to report. + * Normally we expect that segment numbers form an increasing sequence, + * though it's legal to find two files with equal segno in case there + * are both compressed and non-compessed versions. For example + * 000000010000000000000002 and 000000010000000000000002.gz + * + */ + if (segno != expected_segno && segno != tlinfo->end_segno) + { + xlogInterval *interval = palloc(sizeof(xlogInterval));; + interval->begin_segno = expected_segno; + interval->end_segno = segno - 1; + + if (tlinfo->lost_segments == NULL) + tlinfo->lost_segments = parray_new(); + + parray_append(tlinfo->lost_segments, interval); + } + } + + if (tlinfo->begin_segno == 0) + tlinfo->begin_segno = segno; + + /* this file is the last for this timeline so far */ + tlinfo->end_segno = segno; + /* update counters */ + tlinfo->n_xlog_files++; + tlinfo->size += file->size; + + /* append file to xlog file list */ + wal_file = palloc(sizeof(xlogFile)); + wal_file->file = *file; + wal_file->segno = segno; + wal_file->type = SEGMENT; + wal_file->keep = false; + parray_append(tlinfo->xlog_filelist, wal_file); + } + /* timeline history file */ + else if (IsTLHistoryFileName(file->name)) + { + TimeLineHistoryEntry *tln; + + sscanf(file->name, "%08X.history", &tli); + timelines = read_timeline_history(instanceState->instance_wal_subdir_path, tli, true); + + /* History file is empty or corrupted, disregard it */ + if (!timelines) + continue; + + if (!tlinfo || tlinfo->tli != tli) + { + tlinfo = timelineInfoNew(tli); + parray_append(timelineinfos, tlinfo); + /* + * 1 is the latest timeline in the timelines list. + * 0 - is our timeline, which is of no interest here + */ + tln = (TimeLineHistoryEntry *) parray_get(timelines, 1); + tlinfo->switchpoint = tln->end; + tlinfo->parent_tli = tln->tli; + + /* find parent timeline to link it with this one */ + for (j = 0; j < parray_num(timelineinfos); j++) + { + timelineInfo *cur = (timelineInfo *) parray_get(timelineinfos, j); + if (cur->tli == tlinfo->parent_tli) + { + tlinfo->parent_link = cur; + break; + } + } + } + + parray_walk(timelines, pfree); + parray_free(timelines); + } + else + elog(WARNING, "unexpected WAL file name \"%s\"", file->name); + } + + /* save information about backups belonging to each timeline */ + backups = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); + + for (i = 0; i < parray_num(timelineinfos); i++) + { + timelineInfo *tlinfo = parray_get(timelineinfos, i); + for (j = 0; j < parray_num(backups); j++) + { + pgBackup *backup = parray_get(backups, j); + if (tlinfo->tli == backup->tli) + { + if (tlinfo->backups == NULL) + tlinfo->backups = parray_new(); + + parray_append(tlinfo->backups, backup); + } + } + } + + /* determine oldest backup and closest backup for every timeline */ + for (i = 0; i < parray_num(timelineinfos); i++) + { + timelineInfo *tlinfo = parray_get(timelineinfos, i); + + tlinfo->oldest_backup = get_oldest_backup(tlinfo); + tlinfo->closest_backup = get_closest_backup(tlinfo); + } + + /* determine which WAL segments must be kept because of wal retention */ + if (instance->wal_depth <= 0) + return timelineinfos; + + /* + * WAL retention for now is fairly simple. + * User can set only one parameter - 'wal-depth'. + * It determines how many latest valid(!) backups on timeline + * must have an ability to perform PITR: + * Consider the example: + * + * ---B1-------B2-------B3-------B4--------> WAL timeline1 + * + * If 'wal-depth' is set to 2, then WAL purge should produce the following result: + * + * B1 B2 B3-------B4--------> WAL timeline1 + * + * Only valid backup can satisfy 'wal-depth' condition, so if B3 is not OK or DONE, + * then WAL purge should produce the following result: + * B1 B2-------B3-------B4--------> WAL timeline1 + * + * Complicated cases, such as branched timelines are taken into account. + * wal-depth is applied to each timeline independently: + * + * |---------> WAL timeline2 + * ---B1---|---B2-------B3-------B4--------> WAL timeline1 + * + * after WAL purge with wal-depth=2: + * + * |---------> WAL timeline2 + * B1---| B2 B3-------B4--------> WAL timeline1 + * + * In this example WAL retention prevents purge of WAL required by tli2 + * to stay reachable from backup B on tli1. + * + * To protect WAL from purge we try to set 'anchor_lsn' and 'anchor_tli' in every timeline. + * They are usually comes from 'start-lsn' and 'tli' attributes of backup + * calculated by 'wal-depth' parameter. + * With 'wal-depth=2' anchor_backup in tli1 is B3. + + * If timeline has not enough valid backups to satisfy 'wal-depth' condition, + * then 'anchor_lsn' and 'anchor_tli' taken from from 'start-lsn' and 'tli + * attribute of closest_backup. + * The interval of WAL starting from closest_backup to switchpoint is + * saved into 'keep_segments' attribute. + * If there is several intermediate timelines between timeline and its closest_backup + * then on every intermediate timeline WAL interval between switchpoint + * and starting segment is placed in 'keep_segments' attributes: + * + * |---------> WAL timeline3 + * |------| B5-----B6--> WAL timeline2 + * B1---| B2 B3-------B4------------> WAL timeline1 + * + * On timeline where closest_backup is located the WAL interval between + * closest_backup and switchpoint is placed into 'keep_segments'. + * If timeline has no 'closest_backup', then 'wal-depth' rules cannot be applied + * to this timeline and its WAL must be purged by following the basic rules of WAL purging. + * + * Third part is handling of ARCHIVE backups. + * If B1 and B2 have ARCHIVE wal-mode, then we must preserve WAL intervals + * between start_lsn and stop_lsn for each of them in 'keep_segments'. + */ + + /* determine anchor_lsn and keep_segments for every timeline */ + for (i = 0; i < parray_num(timelineinfos); i++) + { + int count = 0; + timelineInfo *tlinfo = parray_get(timelineinfos, i); + + /* + * Iterate backward on backups belonging to this timeline to find + * anchor_backup. NOTE Here we rely on the fact that backups list + * is ordered by start_lsn DESC. + */ + if (tlinfo->backups) + { + for (j = 0; j < parray_num(tlinfo->backups); j++) + { + pgBackup *backup = parray_get(tlinfo->backups, j); + + /* sanity */ + if (XLogRecPtrIsInvalid(backup->start_lsn) || + backup->tli <= 0) + continue; + + /* skip invalid backups */ + if (backup->status != BACKUP_STATUS_OK && + backup->status != BACKUP_STATUS_DONE) + continue; + + /* + * Pinned backups should be ignored for the + * purpose of retention fulfillment, so skip them. + */ + if (backup->expire_time > 0 && + backup->expire_time > current_time) + { + elog(LOG, "Pinned backup %s is ignored for the " + "purpose of WAL retention", + backup_id_of(backup)); + continue; + } + + count++; + + if (count == instance->wal_depth) + { + elog(LOG, "On timeline %i WAL is protected from purge at %X/%X", + tlinfo->tli, + (uint32) (backup->start_lsn >> 32), + (uint32) (backup->start_lsn)); + + tlinfo->anchor_lsn = backup->start_lsn; + tlinfo->anchor_tli = backup->tli; + break; + } + } + } + + /* + * Failed to find anchor backup for this timeline. + * We cannot just thrown it to the wolves, because by + * doing that we will violate our own guarantees. + * So check the existence of closest_backup for + * this timeline. If there is one, then + * set the 'anchor_lsn' and 'anchor_tli' to closest_backup + * 'start-lsn' and 'tli' respectively. + * |-------------B5----------> WAL timeline3 + * |-----|-------------------------> WAL timeline2 + * B1 B2---| B3 B4-------B6-----> WAL timeline1 + * + * wal-depth=2 + * + * If number of valid backups on timelines is less than 'wal-depth' + * then timeline must(!) stay reachable via parent timelines if any. + * If closest_backup is not available, then general WAL purge rules + * are applied. + */ + if (XLogRecPtrIsInvalid(tlinfo->anchor_lsn)) + { + /* + * Failed to find anchor_lsn in our own timeline. + * Consider the case: + * -------------------------------------> tli5 + * ----------------------------B4-------> tli4 + * S3`--------------> tli3 + * S1`------------S3---B3-------B6-> tli2 + * B1---S1-------------B2--------B5-----> tli1 + * + * B* - backups + * S* - switchpoints + * wal-depth=2 + * + * Expected result: + * TLI5 will be purged entirely + * B4-------> tli4 + * S2`--------------> tli3 + * S1`------------S2 B3-------B6-> tli2 + * B1---S1 B2--------B5-----> tli1 + */ + pgBackup *closest_backup = NULL; + xlogInterval *interval = NULL; + TimeLineID tli = 0; + /* check if tli has closest_backup */ + if (!tlinfo->closest_backup) + /* timeline has no closest_backup, wal retention cannot be + * applied to this timeline. + * Timeline will be purged up to oldest_backup if any or + * purge entirely if there is none. + * In example above: tli5 and tli4. + */ + continue; + + /* sanity for closest_backup */ + if (XLogRecPtrIsInvalid(tlinfo->closest_backup->start_lsn) || + tlinfo->closest_backup->tli <= 0) + continue; + + /* + * Set anchor_lsn and anchor_tli to protect whole timeline from purge + * In the example above: tli3. + */ + tlinfo->anchor_lsn = tlinfo->closest_backup->start_lsn; + tlinfo->anchor_tli = tlinfo->closest_backup->tli; + + /* closest backup may be located not in parent timeline */ + closest_backup = tlinfo->closest_backup; + + tli = tlinfo->tli; + + /* + * Iterate over parent timeline chain and + * look for timeline where closest_backup belong + */ + while (tlinfo->parent_link) + { + /* In case of intermediate timeline save to keep_segments + * begin_segno and switchpoint segment. + * In case of final timelines save to keep_segments + * closest_backup start_lsn segment and switchpoint segment. + */ + XLogRecPtr switchpoint = tlinfo->switchpoint; + + tlinfo = tlinfo->parent_link; + + if (tlinfo->keep_segments == NULL) + tlinfo->keep_segments = parray_new(); + + /* in any case, switchpoint segment must be added to interval */ + interval = palloc(sizeof(xlogInterval)); + GetXLogSegNo(switchpoint, interval->end_segno, instance->xlog_seg_size); + + /* Save [S1`, S2] to keep_segments */ + if (tlinfo->tli != closest_backup->tli) + interval->begin_segno = tlinfo->begin_segno; + /* Save [B1, S1] to keep_segments */ + else + GetXLogSegNo(closest_backup->start_lsn, interval->begin_segno, instance->xlog_seg_size); + + /* + * TODO: check, maybe this interval is already here or + * covered by other larger interval. + */ + + GetXLogFileName(begin_segno_str, tlinfo->tli, interval->begin_segno, instance->xlog_seg_size); + GetXLogFileName(end_segno_str, tlinfo->tli, interval->end_segno, instance->xlog_seg_size); + + elog(LOG, "Timeline %i to stay reachable from timeline %i " + "protect from purge WAL interval between " + "%s and %s on timeline %i", + tli, closest_backup->tli, begin_segno_str, + end_segno_str, tlinfo->tli); + + parray_append(tlinfo->keep_segments, interval); + continue; + } + continue; + } + + /* Iterate over backups left */ + for (j = count; j < parray_num(tlinfo->backups); j++) + { + XLogSegNo segno = 0; + xlogInterval *interval = NULL; + pgBackup *backup = parray_get(tlinfo->backups, j); + + /* + * We must calculate keep_segments intervals for ARCHIVE backups + * with start_lsn less than anchor_lsn. + */ + + /* STREAM backups cannot contribute to keep_segments */ + if (backup->stream) + continue; + + /* sanity */ + if (XLogRecPtrIsInvalid(backup->start_lsn) || + backup->tli <= 0) + continue; + + /* no point in clogging keep_segments by backups protected by anchor_lsn */ + if (backup->start_lsn >= tlinfo->anchor_lsn) + continue; + + /* append interval to keep_segments */ + interval = palloc(sizeof(xlogInterval)); + GetXLogSegNo(backup->start_lsn, segno, instance->xlog_seg_size); + interval->begin_segno = segno; + GetXLogSegNo(backup->stop_lsn, segno, instance->xlog_seg_size); + + /* + * On replica it is possible to get STOP_LSN pointing to contrecord, + * so set end_segno to the next segment after STOP_LSN just to be safe. + */ + if (backup->from_replica) + interval->end_segno = segno + 1; + else + interval->end_segno = segno; + + GetXLogFileName(begin_segno_str, tlinfo->tli, interval->begin_segno, instance->xlog_seg_size); + GetXLogFileName(end_segno_str, tlinfo->tli, interval->end_segno, instance->xlog_seg_size); + + elog(LOG, "Archive backup %s to stay consistent " + "protect from purge WAL interval " + "between %s and %s on timeline %i", + backup_id_of(backup), + begin_segno_str, end_segno_str, backup->tli); + + if (tlinfo->keep_segments == NULL) + tlinfo->keep_segments = parray_new(); + + parray_append(tlinfo->keep_segments, interval); + } + } + + /* + * Protect WAL segments from deletion by setting 'keep' flag. + * We must keep all WAL segments after anchor_lsn (including), and also segments + * required by ARCHIVE backups for consistency - WAL between [start_lsn, stop_lsn]. + */ + for (i = 0; i < parray_num(timelineinfos); i++) { - pgBackup *curr = parray_get(backups, i); - pgBackup **ancestor; - pgBackup key; + XLogSegNo anchor_segno = 0; + timelineInfo *tlinfo = parray_get(timelineinfos, i); - if (curr->backup_mode == BACKUP_MODE_FULL) + /* + * At this point invalid anchor_lsn can be only in one case: + * timeline is going to be purged by regular WAL purge rules. + */ + if (XLogRecPtrIsInvalid(tlinfo->anchor_lsn)) continue; - key.start_time = curr->parent_backup; - ancestor = (pgBackup **) parray_bsearch(backups, &key, - pgBackupCompareIdDesc); - if (ancestor) - curr->parent_backup_link = *ancestor; - } + /* + * anchor_lsn is located in another timeline, it means that the timeline + * will be protected from purge entirely. + */ + if (tlinfo->anchor_tli > 0 && tlinfo->anchor_tli != tlinfo->tli) + continue; - return backups; + GetXLogSegNo(tlinfo->anchor_lsn, anchor_segno, instance->xlog_seg_size); -err_proc: - if (data_dir) - fio_closedir(data_dir); - if (backups) - parray_walk(backups, pgBackupFree); - parray_free(backups); + for (j = 0; j < parray_num(tlinfo->xlog_filelist); j++) + { + xlogFile *wal_file = (xlogFile *) parray_get(tlinfo->xlog_filelist, j); - elog(ERROR, "Failed to get backup list"); + if (wal_file->segno >= anchor_segno) + { + wal_file->keep = true; + continue; + } - return NULL; + /* no keep segments */ + if (!tlinfo->keep_segments) + continue; + + /* Protect segments belonging to one of the keep invervals */ + for (k = 0; k < parray_num(tlinfo->keep_segments); k++) + { + xlogInterval *keep_segments = (xlogInterval *) parray_get(tlinfo->keep_segments, k); + + if ((wal_file->segno >= keep_segments->begin_segno) && + wal_file->segno <= keep_segments->end_segno) + { + wal_file->keep = true; + break; + } + } + } + } + + return timelineinfos; } /* - * Lock list of backups. Function goes in backward direction. + * Iterate over parent timelines and look for valid backup + * closest to given timeline switchpoint. + * + * If such backup doesn't exist, it means that + * timeline is unreachable. Return NULL. */ -void -catalog_lock_backup_list(parray *backup_list, int from_idx, int to_idx) +pgBackup* +get_closest_backup(timelineInfo *tlinfo) { - int start_idx, - end_idx; - int i; - - if (parray_num(backup_list) == 0) - return; - - start_idx = Max(from_idx, to_idx); - end_idx = Min(from_idx, to_idx); + pgBackup *closest_backup = NULL; + int i; - for (i = start_idx; i >= end_idx; i--) + /* + * Iterate over backups belonging to parent timelines + * and look for candidates. + */ + while (tlinfo->parent_link && !closest_backup) { - pgBackup *backup = (pgBackup *) parray_get(backup_list, i); - if (!lock_backup(backup)) - elog(ERROR, "Cannot lock backup %s directory", - base36enc(backup->start_time)); + parray *backup_list = tlinfo->parent_link->backups; + if (backup_list != NULL) + { + for (i = 0; i < parray_num(backup_list); i++) + { + pgBackup *backup = parray_get(backup_list, i); + + /* + * Only valid backups made before switchpoint + * should be considered. + */ + if (!XLogRecPtrIsInvalid(backup->stop_lsn) && + XRecOffIsValid(backup->stop_lsn) && + backup->stop_lsn <= tlinfo->switchpoint && + (backup->status == BACKUP_STATUS_OK || + backup->status == BACKUP_STATUS_DONE)) + { + /* Check if backup is closer to switchpoint than current candidate */ + if (!closest_backup || backup->stop_lsn > closest_backup->stop_lsn) + closest_backup = backup; + } + } + } + + /* Continue with parent */ + tlinfo = tlinfo->parent_link; } + + return closest_backup; } /* - * Find the latest valid child of latest valid FULL backup on given timeline + * Find oldest backup in given timeline + * to determine what WAL segments of this timeline + * are reachable from backups belonging to it. + * + * If such backup doesn't exist, it means that + * there is no backups on this timeline. Return NULL. */ -pgBackup * -catalog_get_last_data_backup(parray *backup_list, TimeLineID tli, time_t current_start_time) +pgBackup* +get_oldest_backup(timelineInfo *tlinfo) { - int i; - pgBackup *full_backup = NULL; - pgBackup *tmp_backup = NULL; - char *invalid_backup_id; + pgBackup *oldest_backup = NULL; + int i; + parray *backup_list = tlinfo->backups; - /* backup_list is sorted in order of descending ID */ - for (i = 0; i < parray_num(backup_list); i++) + if (backup_list != NULL) { - pgBackup *backup = (pgBackup *) parray_get(backup_list, i); - - if ((backup->backup_mode == BACKUP_MODE_FULL && - (backup->status == BACKUP_STATUS_OK || - backup->status == BACKUP_STATUS_DONE)) && backup->tli == tli) + for (i = 0; i < parray_num(backup_list); i++) { - full_backup = backup; - break; + pgBackup *backup = parray_get(backup_list, i); + + /* Backups with invalid START LSN can be safely skipped */ + if (XLogRecPtrIsInvalid(backup->start_lsn) || + !XRecOffIsValid(backup->start_lsn)) + continue; + + /* + * Check if backup is older than current candidate. + * Here we use start_lsn for comparison, because backup that + * started earlier needs more WAL. + */ + if (!oldest_backup || backup->start_lsn < oldest_backup->start_lsn) + oldest_backup = backup; } } - /* Failed to find valid FULL backup to fulfill ancestor role */ - if (!full_backup) - return NULL; - - elog(LOG, "Latest valid FULL backup: %s", - base36enc(full_backup->start_time)); + return oldest_backup; +} - /* FULL backup is found, lets find his latest child */ - for (i = 0; i < parray_num(backup_list); i++) - { - pgBackup *backup = (pgBackup *) parray_get(backup_list, i); +/* + * Overwrite backup metadata. + */ +void +do_set_backup(InstanceState *instanceState, time_t backup_id, + pgSetBackupParams *set_backup_params) +{ + pgBackup *target_backup = NULL; + parray *backup_list = NULL; - /* only valid descendants are acceptable for evaluation */ - if ((backup->status == BACKUP_STATUS_OK || - backup->status == BACKUP_STATUS_DONE)) - { - switch (scan_parent_chain(backup, &tmp_backup)) - { - /* broken chain */ - case 0: - invalid_backup_id = base36enc_dup(tmp_backup->parent_backup); + if (!set_backup_params) + elog(ERROR, "Nothing to set by 'set-backup' command"); - elog(WARNING, "Backup %s has missing parent: %s. Cannot be a parent", - base36enc(backup->start_time), invalid_backup_id); - pg_free(invalid_backup_id); - continue; + backup_list = catalog_get_backup_list(instanceState, backup_id); + if (parray_num(backup_list) != 1) + elog(ERROR, "Failed to find backup %s", base36enc(backup_id)); - /* chain is intact, but at least one parent is invalid */ - case 1: - invalid_backup_id = base36enc_dup(tmp_backup->start_time); + target_backup = (pgBackup *) parray_get(backup_list, 0); - elog(WARNING, "Backup %s has invalid parent: %s. Cannot be a parent", - base36enc(backup->start_time), invalid_backup_id); - pg_free(invalid_backup_id); - continue; + /* Pin or unpin backup if requested */ + if (set_backup_params->ttl >= 0 || set_backup_params->expire_time > 0) + pin_backup(target_backup, set_backup_params); - /* chain is ok */ - case 2: - /* Yes, we could call is_parent() earlier - after choosing the ancestor, - * but this way we have an opportunity to detect and report all possible - * anomalies. - */ - if (is_parent(full_backup->start_time, backup, true)) - { - elog(INFO, "Parent backup: %s", - base36enc(backup->start_time)); - return backup; - } - } - } - /* skip yourself */ - else if (backup->start_time == current_start_time) - continue; - else - { - elog(WARNING, "Backup %s has status: %s. Cannot be a parent.", - base36enc(backup->start_time), status2str(backup->status)); - } + if (set_backup_params->note) + add_note(target_backup, set_backup_params->note); + /* Cleanup */ + if (backup_list) + { + parray_walk(backup_list, pgBackupFree); + parray_free(backup_list); } - - return NULL; } -/* create backup directory in $BACKUP_PATH */ -int -pgBackupCreateDir(pgBackup *backup) +/* + * Set 'expire-time' attribute based on set_backup_params, or unpin backup + * if ttl is equal to zero. + */ +void +pin_backup(pgBackup *target_backup, pgSetBackupParams *set_backup_params) { - int i; - char path[MAXPGPATH]; - parray *subdirs = parray_new(); - parray_append(subdirs, pg_strdup(DATABASE_DIR)); + /* sanity, backup must have positive recovery-time */ + if (target_backup->recovery_time <= 0) + elog(ERROR, "Failed to set 'expire-time' for backup %s: invalid 'recovery-time'", + backup_id_of(target_backup)); - /* Add external dirs containers */ - if (backup->external_dir_str) + /* Pin comes from ttl */ + if (set_backup_params->ttl > 0) + target_backup->expire_time = target_backup->recovery_time + set_backup_params->ttl; + /* Unpin backup */ + else if (set_backup_params->ttl == 0) { - parray *external_list; - - external_list = make_external_directory_list(backup->external_dir_str, - false); - for (i = 0; i < parray_num(external_list); i++) + /* If backup was not pinned in the first place, + * then there is nothing to unpin. + */ + if (target_backup->expire_time == 0) { - char temp[MAXPGPATH]; - /* Numeration of externaldirs starts with 1 */ - makeExternalDirPathByNum(temp, EXTERNAL_DIR, i+1); - parray_append(subdirs, pg_strdup(temp)); + elog(WARNING, "Backup %s is not pinned, nothing to unpin", + backup_id_of(target_backup)); + return; } - free_dir_list(external_list); + target_backup->expire_time = 0; } + /* Pin comes from expire-time */ + else if (set_backup_params->expire_time > 0) + target_backup->expire_time = set_backup_params->expire_time; + else + /* nothing to do */ + return; - pgBackupGetPath(backup, path, lengthof(path), NULL); + /* Update backup.control */ + write_backup(target_backup, true); - if (!dir_is_empty(path, FIO_BACKUP_HOST)) - elog(ERROR, "backup destination is not empty \"%s\"", path); + if (set_backup_params->ttl > 0 || set_backup_params->expire_time > 0) + { + char expire_timestamp[100]; - fio_mkdir(path, DIR_PERMISSION, FIO_BACKUP_HOST); + time2iso(expire_timestamp, lengthof(expire_timestamp), target_backup->expire_time, false); + elog(INFO, "Backup %s is pinned until '%s'", backup_id_of(target_backup), + expire_timestamp); + } + else + elog(INFO, "Backup %s is unpinned", backup_id_of(target_backup)); - /* create directories for actual backup files */ - for (i = 0; i < parray_num(subdirs); i++) + return; +} + +/* + * Add note to backup metadata or unset already existing note. + * It is a job of the caller to make sure that note is not NULL. + */ +void +add_note(pgBackup *target_backup, char *note) +{ + + char *note_string; + char *p; + + /* unset note */ + if (pg_strcasecmp(note, "none") == 0) { - pgBackupGetPath(backup, path, lengthof(path), parray_get(subdirs, i)); - fio_mkdir(path, DIR_PERMISSION, FIO_BACKUP_HOST); + target_backup->note = NULL; + elog(INFO, "Removing note from backup %s", + backup_id_of(target_backup)); } + else + { + /* Currently we do not allow string with newlines as note, + * because it will break parsing of backup.control. + * So if user provides string like this "aaa\nbbbbb", + * we save only "aaa" + * Example: tests.set_backup.SetBackupTest.test_add_note_newlines + */ + p = strchr(note, '\n'); + note_string = pgut_strndup(note, p ? (p-note) : MAX_NOTE_SIZE); - free_dir_list(subdirs); - return 0; + target_backup->note = note_string; + elog(INFO, "Adding note to backup %s: '%s'", + backup_id_of(target_backup), target_backup->note); + } + + /* Update backup.control */ + write_backup(target_backup, true); } /* * Write information about backup.in to stream "out". */ void -pgBackupWriteControl(FILE *out, pgBackup *backup) +pgBackupWriteControl(FILE *out, pgBackup *backup, bool utc) { char timestamp[100]; fio_fprintf(out, "#Configuration\n"); - fio_fprintf(out, "backup-mode = %s\n", pgBackupGetBackupMode(backup)); + fio_fprintf(out, "backup-mode = %s\n", pgBackupGetBackupMode(backup, false)); fio_fprintf(out, "stream = %s\n", backup->stream ? "true" : "false"); fio_fprintf(out, "compress-alg = %s\n", deparse_compress_alg(backup->compress_alg)); @@ -609,24 +2381,32 @@ pgBackupWriteControl(FILE *out, pgBackup *backup) (uint32) (backup->stop_lsn >> 32), (uint32) backup->stop_lsn); - time2iso(timestamp, lengthof(timestamp), backup->start_time); + time2iso(timestamp, lengthof(timestamp), backup->start_time, utc); fio_fprintf(out, "start-time = '%s'\n", timestamp); if (backup->merge_time > 0) { - time2iso(timestamp, lengthof(timestamp), backup->merge_time); + time2iso(timestamp, lengthof(timestamp), backup->merge_time, utc); fio_fprintf(out, "merge-time = '%s'\n", timestamp); } if (backup->end_time > 0) { - time2iso(timestamp, lengthof(timestamp), backup->end_time); + time2iso(timestamp, lengthof(timestamp), backup->end_time, utc); fio_fprintf(out, "end-time = '%s'\n", timestamp); } fio_fprintf(out, "recovery-xid = " XID_FMT "\n", backup->recovery_xid); if (backup->recovery_time > 0) { - time2iso(timestamp, lengthof(timestamp), backup->recovery_time); + time2iso(timestamp, lengthof(timestamp), backup->recovery_time, utc); fio_fprintf(out, "recovery-time = '%s'\n", timestamp); } + if (backup->expire_time > 0) + { + time2iso(timestamp, lengthof(timestamp), backup->expire_time, utc); + fio_fprintf(out, "expire-time = '%s'\n", timestamp); + } + + if (backup->merge_dest_backup != 0) + fio_fprintf(out, "merge-dest-id = '%s'\n", base36enc(backup->merge_dest_backup)); /* * Size of PGDATA directory. The size does not include size of related @@ -638,6 +2418,12 @@ pgBackupWriteControl(FILE *out, pgBackup *backup) if (backup->wal_bytes != BYTES_INVALID) fio_fprintf(out, "wal-bytes = " INT64_FORMAT "\n", backup->wal_bytes); + if (backup->uncompressed_bytes >= 0) + fio_fprintf(out, "uncompressed-bytes = " INT64_FORMAT "\n", backup->uncompressed_bytes); + + if (backup->pgdata_bytes >= 0) + fio_fprintf(out, "pgdata-bytes = " INT64_FORMAT "\n", backup->pgdata_bytes); + fio_fprintf(out, "status = %s\n", status2str(backup->status)); /* 'parent_backup' is set if it is incremental backup */ @@ -651,44 +2437,76 @@ pgBackupWriteControl(FILE *out, pgBackup *backup) /* print external directories list */ if (backup->external_dir_str) fio_fprintf(out, "external-dirs = '%s'\n", backup->external_dir_str); + + if (backup->note) + fio_fprintf(out, "note = '%s'\n", backup->note); + + if (backup->content_crc != 0) + fio_fprintf(out, "content-crc = %u\n", backup->content_crc); + } /* * Save the backup content into BACKUP_CONTROL_FILE. + * Flag strict allows to ignore "out of space" error + * when attempting to lock backup. Only delete is allowed + * to use this functionality. */ void -write_backup(pgBackup *backup) +write_backup(pgBackup *backup, bool strict) { - FILE *fp = NULL; - char path[MAXPGPATH]; - char path_temp[MAXPGPATH]; - int errno_temp; + FILE *fp = NULL; + char path[MAXPGPATH]; + char path_temp[MAXPGPATH]; + char buf[8192]; - pgBackupGetPath(backup, path, lengthof(path), BACKUP_CONTROL_FILE); + join_path_components(path, backup->root_dir, BACKUP_CONTROL_FILE); snprintf(path_temp, sizeof(path_temp), "%s.tmp", path); - fp = fio_fopen(path_temp, PG_BINARY_W, FIO_BACKUP_HOST); + fp = fopen(path_temp, PG_BINARY_W); if (fp == NULL) - elog(ERROR, "Cannot open configuration file \"%s\": %s", + elog(ERROR, "Cannot open control file \"%s\": %s", path_temp, strerror(errno)); - pgBackupWriteControl(fp, backup); + if (chmod(path_temp, FILE_PERMISSION) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", path_temp, + strerror(errno)); - if (fio_fflush(fp) || fio_fclose(fp)) - { - errno_temp = errno; - fio_unlink(path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot write configuration file \"%s\": %s", - path_temp, strerror(errno_temp)); - } + setvbuf(fp, buf, _IOFBF, sizeof(buf)); + + pgBackupWriteControl(fp, backup, true); - if (fio_rename(path_temp, path, FIO_BACKUP_HOST) < 0) + /* Ignore 'out of space' error in lax mode */ + if (fflush(fp) != 0) { - errno_temp = errno; - fio_unlink(path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot rename configuration file \"%s\" to \"%s\": %s", - path_temp, path, strerror(errno_temp)); + int elevel = ERROR; + int save_errno = errno; + + if (!strict && (errno == ENOSPC)) + elevel = WARNING; + + elog(elevel, "Cannot flush control file \"%s\": %s", + path_temp, strerror(save_errno)); + + if (!strict && (save_errno == ENOSPC)) + { + fclose(fp); + fio_unlink(path_temp, FIO_BACKUP_HOST); + return; + } } + + if (fclose(fp) != 0) + elog(ERROR, "Cannot close control file \"%s\": %s", + path_temp, strerror(errno)); + + if (fio_sync(path_temp, FIO_BACKUP_HOST) < 0) + elog(ERROR, "Cannot sync control file \"%s\": %s", + path_temp, strerror(errno)); + + if (rename(path_temp, path) < 0) + elog(ERROR, "Cannot rename file \"%s\" to \"%s\": %s", + path_temp, path, strerror(errno)); } /* @@ -696,61 +2514,86 @@ write_backup(pgBackup *backup) */ void write_backup_filelist(pgBackup *backup, parray *files, const char *root, - parray *external_list) + parray *external_list, bool sync) { FILE *out; - char path[MAXPGPATH]; - char path_temp[MAXPGPATH]; - int errno_temp; + char control_path[MAXPGPATH]; + char control_path_temp[MAXPGPATH]; size_t i = 0; - #define BUFFERSZ BLCKSZ*500 - char buf[BUFFERSZ]; - size_t write_len = 0; + #define BUFFERSZ (1024*1024) + char *buf; int64 backup_size_on_disk = 0; + int64 uncompressed_size_on_disk = 0; + int64 wal_size_on_disk = 0; - pgBackupGetPath(backup, path, lengthof(path), DATABASE_FILE_LIST); - snprintf(path_temp, sizeof(path_temp), "%s.tmp", path); + join_path_components(control_path, backup->root_dir, DATABASE_FILE_LIST); + snprintf(control_path_temp, sizeof(control_path_temp), "%s.tmp", control_path); - out = fio_fopen(path_temp, PG_BINARY_W, FIO_BACKUP_HOST); + out = fopen(control_path_temp, PG_BINARY_W); if (out == NULL) - elog(ERROR, "Cannot open file list \"%s\": %s", path_temp, + elog(ERROR, "Cannot open file list \"%s\": %s", control_path_temp, + strerror(errno)); + + if (chmod(control_path_temp, FILE_PERMISSION) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", control_path_temp, strerror(errno)); + buf = pgut_malloc(BUFFERSZ); + setvbuf(out, buf, _IOFBF, BUFFERSZ); + + if (sync) + INIT_FILE_CRC32(true, backup->content_crc); + /* print each file in the list */ - while(i < parray_num(files)) + for (i = 0; i < parray_num(files); i++) { - pgFile *file = (pgFile *) parray_get(files, i); - char *path = file->path; /* for streamed WAL files */ - char line[BLCKSZ]; - int len = 0; + int len = 0; + char line[BLCKSZ]; + pgFile *file = (pgFile *) parray_get(files, i); - i++; + /* Ignore disappeared file */ + if (file->write_size == FILE_NOT_FOUND) + continue; if (S_ISDIR(file->mode)) + { backup_size_on_disk += 4096; + uncompressed_size_on_disk += 4096; + } /* Count the amount of the data actually copied */ if (S_ISREG(file->mode) && file->write_size > 0) - backup_size_on_disk += file->write_size; - - /* for files from PGDATA and external files use rel_path - * streamed WAL files has rel_path relative not to "database/" - * but to "database/pg_wal", so for them use path. - */ - if ((root && strstr(path, root) == path) || - (file->external_dir_num && external_list)) - path = file->rel_path; + { + /* + * Size of WAL files in 'pg_wal' is counted separately + * TODO: in 3.0 add attribute is_walfile + */ + if (IsXLogFileName(file->name) && file->external_dir_num == 0) + wal_size_on_disk += file->write_size; + else + { + backup_size_on_disk += file->write_size; + uncompressed_size_on_disk += file->uncompressed_size; + } + } len = sprintf(line, "{\"path\":\"%s\", \"size\":\"" INT64_FORMAT "\", " "\"mode\":\"%u\", \"is_datafile\":\"%u\", " "\"is_cfs\":\"%u\", \"crc\":\"%u\", " - "\"compress_alg\":\"%s\", \"external_dir_num\":\"%d\"", - path, file->write_size, file->mode, + "\"compress_alg\":\"%s\", \"external_dir_num\":\"%d\", " + "\"dbOid\":\"%u\"", + file->rel_path, file->write_size, file->mode, file->is_datafile ? 1 : 0, file->is_cfs ? 1 : 0, file->crc, deparse_compress_alg(file->compress_alg), - file->external_dir_num); + file->external_dir_num, + file->dbOid); + + if (file->uncompressed_size != 0 && + file->uncompressed_size != file->write_size) + len += sprintf(line+len, ",\"full_size\":\"" INT64_FORMAT "\"", + file->uncompressed_size); if (file->is_datafile) len += sprintf(line+len, ",\"segno\":\"%d\"", file->segno); @@ -758,59 +2601,52 @@ write_backup_filelist(pgBackup *backup, parray *files, const char *root, if (file->linked) len += sprintf(line+len, ",\"linked\":\"%s\"", file->linked); - if (file->n_blocks != BLOCKNUM_INVALID) + if (file->n_blocks > 0) len += sprintf(line+len, ",\"n_blocks\":\"%i\"", file->n_blocks); - len += sprintf(line+len, "}\n"); - - if (write_len + len <= BUFFERSZ) - { - memcpy(buf+write_len, line, len); - write_len += len; - } - else + if (file->n_headers > 0) { - /* write buffer to file */ - if (fio_fwrite(out, buf, write_len) != write_len) - { - errno_temp = errno; - fio_unlink(path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot write file list \"%s\": %s", - path_temp, strerror(errno)); - } - /* reset write_len */ - write_len = 0; + len += sprintf(line+len, ",\"n_headers\":\"%i\"", file->n_headers); + len += sprintf(line+len, ",\"hdr_crc\":\"%u\"", file->hdr_crc); + len += sprintf(line+len, ",\"hdr_off\":\"%llu\"", file->hdr_off); + len += sprintf(line+len, ",\"hdr_size\":\"%i\"", file->hdr_size); } - } - /* write what is left in the buffer to file */ - if (write_len > 0) - if (fio_fwrite(out, buf, write_len) != write_len) - { - errno_temp = errno; - fio_unlink(path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot write file list \"%s\": %s", - path_temp, strerror(errno)); - } + sprintf(line+len, "}\n"); - if (fio_fflush(out) || fio_fclose(out)) - { - errno_temp = errno; - fio_unlink(path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot write file list \"%s\": %s", - path_temp, strerror(errno)); - } + if (sync) + COMP_FILE_CRC32(true, backup->content_crc, line, strlen(line)); - if (fio_rename(path_temp, path, FIO_BACKUP_HOST) < 0) - { - errno_temp = errno; - fio_unlink(path_temp, FIO_BACKUP_HOST); - elog(ERROR, "Cannot rename configuration file \"%s\" to \"%s\": %s", - path_temp, path, strerror(errno_temp)); + fprintf(out, "%s", line); } + if (sync) + FIN_FILE_CRC32(true, backup->content_crc); + + if (fflush(out) != 0) + elog(ERROR, "Cannot flush file list \"%s\": %s", + control_path_temp, strerror(errno)); + + if (sync && fsync(fileno(out)) < 0) + elog(ERROR, "Cannot sync file list \"%s\": %s", + control_path_temp, strerror(errno)); + + if (fclose(out) != 0) + elog(ERROR, "Cannot close file list \"%s\": %s", + control_path_temp, strerror(errno)); + + if (rename(control_path_temp, control_path) < 0) + elog(ERROR, "Cannot rename file \"%s\" to \"%s\": %s", + control_path_temp, control_path, strerror(errno)); + /* use extra variable to avoid reset of previous data_bytes value in case of error */ backup->data_bytes = backup_size_on_disk; + backup->uncompressed_bytes = uncompressed_size_on_disk; + + if (backup->stream) + backup->wal_bytes = wal_size_on_disk; + + free(buf); } /* @@ -821,12 +2657,13 @@ write_backup_filelist(pgBackup *backup, parray *files, const char *root, static pgBackup * readBackupControlFile(const char *path) { - pgBackup *backup = pgut_new(pgBackup); + pgBackup *backup = pgut_new0(pgBackup); char *backup_mode = NULL; char *start_lsn = NULL; char *stop_lsn = NULL; char *status = NULL; char *parent_backup = NULL; + char *merge_dest_backup = NULL; char *program_version = NULL; char *server_version = NULL; char *compress_alg = NULL; @@ -843,8 +2680,11 @@ readBackupControlFile(const char *path) {'t', 0, "end-time", &backup->end_time, SOURCE_FILE_STRICT}, {'U', 0, "recovery-xid", &backup->recovery_xid, SOURCE_FILE_STRICT}, {'t', 0, "recovery-time", &backup->recovery_time, SOURCE_FILE_STRICT}, + {'t', 0, "expire-time", &backup->expire_time, SOURCE_FILE_STRICT}, {'I', 0, "data-bytes", &backup->data_bytes, SOURCE_FILE_STRICT}, {'I', 0, "wal-bytes", &backup->wal_bytes, SOURCE_FILE_STRICT}, + {'I', 0, "uncompressed-bytes", &backup->uncompressed_bytes, SOURCE_FILE_STRICT}, + {'I', 0, "pgdata-bytes", &backup->pgdata_bytes, SOURCE_FILE_STRICT}, {'u', 0, "block-size", &backup->block_size, SOURCE_FILE_STRICT}, {'u', 0, "xlog-block-size", &backup->wal_block_size, SOURCE_FILE_STRICT}, {'u', 0, "checksum-version", &backup->checksum_version, SOURCE_FILE_STRICT}, @@ -853,11 +2693,14 @@ readBackupControlFile(const char *path) {'b', 0, "stream", &backup->stream, SOURCE_FILE_STRICT}, {'s', 0, "status", &status, SOURCE_FILE_STRICT}, {'s', 0, "parent-backup-id", &parent_backup, SOURCE_FILE_STRICT}, + {'s', 0, "merge-dest-id", &merge_dest_backup, SOURCE_FILE_STRICT}, {'s', 0, "compress-alg", &compress_alg, SOURCE_FILE_STRICT}, {'u', 0, "compress-level", &backup->compress_level, SOURCE_FILE_STRICT}, {'b', 0, "from-replica", &backup->from_replica, SOURCE_FILE_STRICT}, {'s', 0, "primary-conninfo", &backup->primary_conninfo, SOURCE_FILE_STRICT}, {'s', 0, "external-dirs", &backup->external_dir_str, SOURCE_FILE_STRICT}, + {'s', 0, "note", &backup->note, SOURCE_FILE_STRICT}, + {'u', 0, "content-crc", &backup->content_crc, SOURCE_FILE_STRICT}, {0} }; @@ -884,6 +2727,9 @@ readBackupControlFile(const char *path) pgBackupFree(backup); return NULL; } + /* XXX BACKUP_ID change it when backup_id wouldn't match start_time */ + Assert(backup->backup_id == 0 || backup->backup_id == backup->start_time); + backup->backup_id = backup->start_time; if (backup_mode) { @@ -925,6 +2771,8 @@ readBackupControlFile(const char *path) backup->status = BACKUP_STATUS_RUNNING; else if (strcmp(status, "MERGING") == 0) backup->status = BACKUP_STATUS_MERGING; + else if (strcmp(status, "MERGED") == 0) + backup->status = BACKUP_STATUS_MERGED; else if (strcmp(status, "DELETING") == 0) backup->status = BACKUP_STATUS_DELETING; else if (strcmp(status, "DELETED") == 0) @@ -946,16 +2794,22 @@ readBackupControlFile(const char *path) free(parent_backup); } + if (merge_dest_backup) + { + backup->merge_dest_backup = base36dec(merge_dest_backup); + free(merge_dest_backup); + } + if (program_version) { - StrNCpy(backup->program_version, program_version, + strlcpy(backup->program_version, program_version, sizeof(backup->program_version)); pfree(program_version); } if (server_version) { - StrNCpy(backup->server_version, server_version, + strlcpy(backup->server_version, server_version, sizeof(backup->server_version)); pfree(server_version); } @@ -987,7 +2841,7 @@ parse_backup_mode(const char *value) return BACKUP_MODE_DIFF_DELTA; /* Backup mode is invalid, so leave with an error */ - elog(ERROR, "invalid backup-mode \"%s\"", value); + elog(ERROR, "Invalid backup-mode \"%s\"", value); return BACKUP_MODE_INVALID; } @@ -1022,7 +2876,7 @@ parse_compress_alg(const char *arg) len = strlen(arg); if (len == 0) - elog(ERROR, "compress algorithm is empty"); + elog(ERROR, "Compress algorithm is empty"); if (pg_strncasecmp("zlib", arg, len) == 0) return ZLIB_COMPRESS; @@ -1031,7 +2885,7 @@ parse_compress_alg(const char *arg) else if (pg_strncasecmp("none", arg, len) == 0) return NONE_COMPRESS; else - elog(ERROR, "invalid compress algorithm value \"%s\"", arg); + elog(ERROR, "Invalid compress algorithm value \"%s\"", arg); return NOT_DEFINED_COMPRESS; } @@ -1064,9 +2918,14 @@ pgNodeInit(PGNodeInfo *node) node->checksum_version = 0; node->is_superuser = false; + node->pgpro_support = false; node->server_version = 0; node->server_version_str[0] = '\0'; + + node->ptrack_version_num = 0; + node->is_ptrack_enabled = false; + node->ptrack_schema = NULL; } /* @@ -1086,9 +2945,12 @@ pgBackupInit(pgBackup *backup) backup->end_time = (time_t) 0; backup->recovery_xid = 0; backup->recovery_time = (time_t) 0; + backup->expire_time = (time_t) 0; backup->data_bytes = BYTES_INVALID; backup->wal_bytes = BYTES_INVALID; + backup->uncompressed_bytes = 0; + backup->pgdata_bytes = 0; backup->compress_alg = COMPRESS_ALG_DEFAULT; backup->compress_level = COMPRESS_LEVEL_DEFAULT; @@ -1100,11 +2962,17 @@ pgBackupInit(pgBackup *backup) backup->stream = false; backup->from_replica = false; backup->parent_backup = INVALID_BACKUP_ID; + backup->merge_dest_backup = INVALID_BACKUP_ID; backup->parent_backup_link = NULL; backup->primary_conninfo = NULL; backup->program_version[0] = '\0'; backup->server_version[0] = '\0'; backup->external_dir_str = NULL; + backup->root_dir = NULL; + backup->database_dir = NULL; + backup->files = NULL; + backup->note = NULL; + backup->content_crc = 0; } /* free pgBackup object */ @@ -1113,9 +2981,12 @@ pgBackupFree(void *backup) { pgBackup *b = (pgBackup *) backup; - pfree(b->primary_conninfo); - pfree(b->external_dir_str); - pfree(backup); + pg_free(b->primary_conninfo); + pg_free(b->external_dir_str); + pg_free(b->root_dir); + pg_free(b->database_dir); + pg_free(b->note); + pg_free(backup); } /* Compare two pgBackup with their IDs (start time) in ascending order */ @@ -1140,37 +3011,6 @@ pgBackupCompareIdDesc(const void *l, const void *r) return -pgBackupCompareId(l, r); } -/* - * Construct absolute path of the backup directory. - * If subdir is not NULL, it will be appended after the path. - */ -void -pgBackupGetPath(const pgBackup *backup, char *path, size_t len, const char *subdir) -{ - pgBackupGetPath2(backup, path, len, subdir, NULL); -} - -/* - * Construct absolute path of the backup directory. - * Append "subdir1" and "subdir2" to the backup directory. - */ -void -pgBackupGetPath2(const pgBackup *backup, char *path, size_t len, - const char *subdir1, const char *subdir2) -{ - /* If "subdir1" is NULL do not check "subdir2" */ - if (!subdir1) - snprintf(path, len, "%s/%s", backup_instance_path, - base36enc(backup->start_time)); - else if (!subdir2) - snprintf(path, len, "%s/%s/%s", backup_instance_path, - base36enc(backup->start_time), subdir1); - /* "subdir1" and "subdir2" is not NULL */ - else - snprintf(path, len, "%s/%s/%s/%s", backup_instance_path, - base36enc(backup->start_time), subdir1, subdir2); -} - /* * Check if multiple backups consider target backup to be their direct parent */ @@ -1223,7 +3063,7 @@ find_parent_full_backup(pgBackup *current_backup) base36enc(base_full_backup->parent_backup)); else elog(WARNING, "Failed to find parent FULL backup for %s", - base36enc(current_backup->start_time)); + backup_id_of(current_backup)); return NULL; } @@ -1275,18 +3115,18 @@ scan_parent_chain(pgBackup *current_backup, pgBackup **result_backup) { /* Set oldest child backup in chain */ *result_backup = target_backup; - return 0; + return ChainIsBroken; } /* chain is ok, but some backups are invalid */ if (invalid_backup) { *result_backup = invalid_backup; - return 1; + return ChainIsInvalid; } *result_backup = target_backup; - return 2; + return ChainIsOk; } /* @@ -1320,22 +3160,22 @@ is_parent(time_t parent_backup_time, pgBackup *child_backup, bool inclusive) return false; } -/* - * Return backup index number. - * Note: this index number holds true until new sorting of backup list - */ -int -get_backup_index_number(parray *backup_list, pgBackup *backup) +/* On backup_list lookup children of target_backup and append them to append_list */ +void +append_children(parray *backup_list, pgBackup *target_backup, parray *append_list) { int i; for (i = 0; i < parray_num(backup_list); i++) { - pgBackup *tmp_backup = (pgBackup *) parray_get(backup_list, i); + pgBackup *backup = (pgBackup *) parray_get(backup_list, i); - if (tmp_backup->start_time == backup->start_time) - return i; + /* check if backup is descendant of target backup */ + if (is_parent(target_backup->start_time, backup, false)) + { + /* if backup is already in the list, then skip it */ + if (!parray_contains(append_list, backup)) + parray_append(append_list, backup); + } } - elog(WARNING, "Failed to find backup %s", base36enc(backup->start_time)); - return -1; } diff --git a/src/catchup.c b/src/catchup.c new file mode 100644 index 000000000..00752b194 --- /dev/null +++ b/src/catchup.c @@ -0,0 +1,1170 @@ +/*------------------------------------------------------------------------- + * + * catchup.c: sync DB cluster + * + * Copyright (c) 2021-2022, Postgres Professional + * + *------------------------------------------------------------------------- + */ + +#include "pg_probackup.h" + +#if PG_VERSION_NUM < 110000 +#include "catalog/catalog.h" +#endif +#include "catalog/pg_tablespace.h" +#include "access/timeline.h" +#include "pgtar.h" +#include "streamutil.h" + +#include +#include +#include + +#include "utils/thread.h" +#include "utils/file.h" + +/* + * Catchup routines + */ +static PGconn *catchup_init_state(PGNodeInfo *source_node_info, const char *source_pgdata, const char *dest_pgdata); +static void catchup_preflight_checks(PGNodeInfo *source_node_info, PGconn *source_conn, const char *source_pgdata, + const char *dest_pgdata); +static void catchup_check_tablespaces_existance_in_tbsmapping(PGconn *conn); +static parray* catchup_get_tli_history(ConnectionOptions *conn_opt, TimeLineID tli); + +//REVIEW I'd also suggest to wrap all these fields into some CatchupState, but it isn't urgent. +//REVIEW_ANSWER what for? +/* + * Prepare for work: fill some globals, open connection to source database + */ +static PGconn * +catchup_init_state(PGNodeInfo *source_node_info, const char *source_pgdata, const char *dest_pgdata) +{ + PGconn *source_conn; + + /* Initialize PGInfonode */ + pgNodeInit(source_node_info); + + /* Get WAL segments size and system ID of source PG instance */ + instance_config.xlog_seg_size = get_xlog_seg_size(source_pgdata); + instance_config.system_identifier = get_system_identifier(source_pgdata, FIO_DB_HOST, false); + current.start_time = time(NULL); + + strlcpy(current.program_version, PROGRAM_VERSION, sizeof(current.program_version)); + + /* Do some compatibility checks and fill basic info about PG instance */ + source_conn = pgdata_basic_setup(instance_config.conn_opt, source_node_info); + +#if PG_VERSION_NUM >= 110000 + if (!RetrieveWalSegSize(source_conn)) + elog(ERROR, "Failed to retrieve wal_segment_size"); +#endif + + get_ptrack_version(source_conn, source_node_info); + if (source_node_info->ptrack_version_num > 0) + source_node_info->is_ptrack_enabled = pg_is_ptrack_enabled(source_conn, source_node_info->ptrack_version_num); + + /* Obtain current timeline */ +#if PG_VERSION_NUM >= 90600 + current.tli = get_current_timeline(source_conn); +#else + instance_config.pgdata = source_pgdata; + current.tli = get_current_timeline_from_control(source_pgdata, FIO_DB_HOST, false); +#endif + + elog(INFO, "Catchup start, pg_probackup version: %s, " + "PostgreSQL version: %s, " + "remote: %s, source-pgdata: %s, destination-pgdata: %s", + PROGRAM_VERSION, source_node_info->server_version_str, + IsSshProtocol() ? "true" : "false", + source_pgdata, dest_pgdata); + + if (current.from_replica) + elog(INFO, "Running catchup from standby"); + + return source_conn; +} + +/* + * Check that catchup can be performed on source and dest + * this function is for checks, that can be performed without modification of data on disk + */ +static void +catchup_preflight_checks(PGNodeInfo *source_node_info, PGconn *source_conn, + const char *source_pgdata, const char *dest_pgdata) +{ + /* TODO + * gsmol - fallback to FULL mode if dest PGDATA is empty + * kulaginm -- I think this is a harmful feature. If user requested an incremental catchup, then + * he expects that this will be done quickly and efficiently. If, for example, he made a mistake + * with dest_dir, then he will receive a second full copy instead of an error message, and I think + * that in some cases he would prefer the error. + * I propose in future versions to offer a backup_mode auto, in which we will look to the dest_dir + * and decide which of the modes will be the most effective. + * I.e.: + * if(requested_backup_mode == BACKUP_MODE_DIFF_AUTO) + * { + * if(dest_pgdata_is_empty) + * backup_mode = BACKUP_MODE_FULL; + * else + * if(ptrack supported and applicable) + * backup_mode = BACKUP_MODE_DIFF_PTRACK; + * else + * backup_mode = BACKUP_MODE_DIFF_DELTA; + * } + */ + + if (dir_is_empty(dest_pgdata, FIO_LOCAL_HOST)) + { + if (current.backup_mode == BACKUP_MODE_DIFF_PTRACK || + current.backup_mode == BACKUP_MODE_DIFF_DELTA) + elog(ERROR, "\"%s\" is empty, but incremental catchup mode requested.", + dest_pgdata); + } + else /* dest dir not empty */ + { + if (current.backup_mode == BACKUP_MODE_FULL) + elog(ERROR, "Can't perform full catchup into non-empty directory \"%s\".", + dest_pgdata); + } + + /* check that postmaster is not running in destination */ + if (current.backup_mode != BACKUP_MODE_FULL) + { + pid_t pid; + pid = fio_check_postmaster(dest_pgdata, FIO_LOCAL_HOST); + if (pid == 1) /* postmaster.pid is mangled */ + { + char pid_filename[MAXPGPATH]; + join_path_components(pid_filename, dest_pgdata, "postmaster.pid"); + elog(ERROR, "Pid file \"%s\" is mangled, cannot determine whether postmaster is running or not", + pid_filename); + } + else if (pid > 1) /* postmaster is up */ + { + elog(ERROR, "Postmaster with pid %u is running in destination directory \"%s\"", + pid, dest_pgdata); + } + } + + /* check backup_label absence in dest */ + if (current.backup_mode != BACKUP_MODE_FULL) + { + char backup_label_filename[MAXPGPATH]; + + join_path_components(backup_label_filename, dest_pgdata, PG_BACKUP_LABEL_FILE); + if (fio_access(backup_label_filename, F_OK, FIO_LOCAL_HOST) == 0) + elog(ERROR, "Destination directory contains \"" PG_BACKUP_LABEL_FILE "\" file"); + } + + /* Check that connected PG instance, source and destination PGDATA are the same */ + { + uint64 source_conn_id, source_id, dest_id; + + source_conn_id = get_remote_system_identifier(source_conn); + source_id = get_system_identifier(source_pgdata, FIO_DB_HOST, false); /* same as instance_config.system_identifier */ + + if (source_conn_id != source_id) + elog(ERROR, "Database identifiers mismatch: we connected to DB id %lu, but in \"%s\" we found id %lu", + source_conn_id, source_pgdata, source_id); + + if (current.backup_mode != BACKUP_MODE_FULL) + { + ControlFileData dst_control; + get_control_file_or_back_file(dest_pgdata, FIO_LOCAL_HOST, &dst_control); + dest_id = dst_control.system_identifier; + + if (source_conn_id != dest_id) + elog(ERROR, "Database identifiers mismatch: we connected to DB id %llu, but in \"%s\" we found id %llu", + (long long)source_conn_id, dest_pgdata, (long long)dest_id); + } + } + + /* check PTRACK version */ + if (current.backup_mode == BACKUP_MODE_DIFF_PTRACK) + { + if (source_node_info->ptrack_version_num == 0) + elog(ERROR, "This PostgreSQL instance does not support ptrack"); + else if (source_node_info->ptrack_version_num < 200) + elog(ERROR, "Ptrack extension is too old.\n" + "Upgrade ptrack to version >= 2"); + else if (!source_node_info->is_ptrack_enabled) + elog(ERROR, "Ptrack is disabled"); + } + + if (current.from_replica && exclusive_backup) + elog(ERROR, "Catchup from standby is only available for PostgreSQL >= 9.6"); + + /* check that we don't overwrite tablespace in source pgdata */ + catchup_check_tablespaces_existance_in_tbsmapping(source_conn); + + /* check timelines */ + if (current.backup_mode != BACKUP_MODE_FULL) + { + RedoParams dest_redo = { 0, InvalidXLogRecPtr, 0 }; + + /* fill dest_redo.lsn and dest_redo.tli */ + get_redo(dest_pgdata, FIO_LOCAL_HOST, &dest_redo); + elog(LOG, "source.tli = %X, dest_redo.lsn = %X/%X, dest_redo.tli = %X", + current.tli, (uint32) (dest_redo.lsn >> 32), (uint32) dest_redo.lsn, dest_redo.tli); + + if (current.tli != 1) + { + parray *source_timelines; /* parray* of TimeLineHistoryEntry* */ + source_timelines = catchup_get_tli_history(&instance_config.conn_opt, current.tli); + + if (source_timelines == NULL) + elog(ERROR, "Cannot get source timeline history"); + + if (!satisfy_timeline(source_timelines, dest_redo.tli, dest_redo.lsn)) + elog(ERROR, "Destination is not in source timeline history"); + + parray_walk(source_timelines, pfree); + parray_free(source_timelines); + } + else /* special case -- no history files in source */ + { + if (dest_redo.tli != 1) + elog(ERROR, "Source is behind destination in timeline history"); + } + } +} + +/* + * Check that all tablespaces exists in tablespace mapping (--tablespace-mapping option) + * Check that all local mapped directories is empty if it is local FULL catchup + * Emit fatal error if that (not existent in map or not empty) tablespace found + */ +static void +catchup_check_tablespaces_existance_in_tbsmapping(PGconn *conn) +{ + PGresult *res; + int i; + char *tablespace_path = NULL; + const char *linked_path = NULL; + char *query = "SELECT pg_catalog.pg_tablespace_location(oid) " + "FROM pg_catalog.pg_tablespace " + "WHERE pg_catalog.pg_tablespace_location(oid) <> '';"; + + res = pgut_execute(conn, query, 0, NULL); + + if (!res) + elog(ERROR, "Failed to get list of tablespaces"); + + for (i = 0; i < res->ntups; i++) + { + tablespace_path = PQgetvalue(res, i, 0); + Assert (strlen(tablespace_path) > 0); + + canonicalize_path(tablespace_path); + linked_path = get_tablespace_mapping(tablespace_path); + + if (strcmp(tablespace_path, linked_path) == 0) + /* same result -> not found in mapping */ + { + if (!fio_is_remote(FIO_DB_HOST)) + elog(ERROR, "Local catchup executed, but source database contains " + "tablespace (\"%s\"), that is not listed in the map", tablespace_path); + else + elog(WARNING, "Remote catchup executed and source database contains " + "tablespace (\"%s\"), that is not listed in the map", tablespace_path); + } + + if (!is_absolute_path(linked_path)) + elog(ERROR, "Tablespace directory path must be an absolute path: \"%s\"", + linked_path); + + if (current.backup_mode == BACKUP_MODE_FULL + && !dir_is_empty(linked_path, FIO_LOCAL_HOST)) + elog(ERROR, "Target mapped tablespace directory (\"%s\") is not empty in FULL catchup", + linked_path); + } + PQclear(res); +} + +/* + * Get timeline history via replication connection + * returns parray* of TimeLineHistoryEntry* + */ +static parray* +catchup_get_tli_history(ConnectionOptions *conn_opt, TimeLineID tli) +{ + PGresult *res; + PGconn *conn; + char *history; + char query[128]; + parray *result = NULL; + TimeLineHistoryEntry *entry = NULL; + + snprintf(query, sizeof(query), "TIMELINE_HISTORY %u", tli); + + /* + * Connect in replication mode to the server. + */ + conn = pgut_connect_replication(conn_opt->pghost, + conn_opt->pgport, + conn_opt->pgdatabase, + conn_opt->pguser, + false); + + if (!conn) + return NULL; + + res = PQexec(conn, query); + PQfinish(conn); + + if (PQresultStatus(res) != PGRES_TUPLES_OK) + { + elog(WARNING, "Could not send replication command \"%s\": %s", + query, PQresultErrorMessage(res)); + PQclear(res); + return NULL; + } + + /* + * The response to TIMELINE_HISTORY is a single row result set + * with two fields: filename and content + */ + if (PQnfields(res) != 2 || PQntuples(res) != 1) + { + elog(ERROR, "Unexpected response to TIMELINE_HISTORY command: " + "got %d rows and %d fields, expected %d rows and %d fields", + PQntuples(res), PQnfields(res), 1, 2); + PQclear(res); + return NULL; + } + + history = pgut_strdup(PQgetvalue(res, 0, 1)); + result = parse_tli_history_buffer(history, tli); + + /* some cleanup */ + pg_free(history); + PQclear(res); + + /* append last timeline entry (as read_timeline_history() do) */ + entry = pgut_new(TimeLineHistoryEntry); + entry->tli = tli; + entry->end = InvalidXLogRecPtr; + parray_insert(result, 0, entry); + + return result; +} + +/* + * catchup multithreaded copy rountine and helper structure and function + */ + +/* parameters for catchup_thread_runner() passed from catchup_multithreaded_copy() */ +typedef struct +{ + PGNodeInfo *nodeInfo; + const char *from_root; + const char *to_root; + parray *source_filelist; + parray *dest_filelist; + XLogRecPtr sync_lsn; + BackupMode backup_mode; + int thread_num; + size_t transfered_bytes; + bool completed; +} catchup_thread_runner_arg; + +/* Catchup file copier executed in separate thread */ +static void * +catchup_thread_runner(void *arg) +{ + int i; + char from_fullpath[MAXPGPATH]; + char to_fullpath[MAXPGPATH]; + + catchup_thread_runner_arg *arguments = (catchup_thread_runner_arg *) arg; + int n_files = parray_num(arguments->source_filelist); + + /* catchup a file */ + for (i = 0; i < n_files; i++) + { + pgFile *file = (pgFile *) parray_get(arguments->source_filelist, i); + pgFile *dest_file = NULL; + + /* We have already copied all directories */ + if (S_ISDIR(file->mode)) + continue; + + if (file->excluded) + continue; + + if (!pg_atomic_test_set_flag(&file->lock)) + continue; + + /* check for interrupt */ + if (interrupted || thread_interrupted) + elog(ERROR, "Interrupted during catchup"); + + elog(progress ? INFO : LOG, "Progress: (%d/%d). Process file \"%s\"", + i + 1, n_files, file->rel_path); + + /* construct destination filepath */ + Assert(file->external_dir_num == 0); + join_path_components(from_fullpath, arguments->from_root, file->rel_path); + join_path_components(to_fullpath, arguments->to_root, file->rel_path); + + /* Encountered some strange beast */ + if (!S_ISREG(file->mode)) + elog(WARNING, "Unexpected type %d of file \"%s\", skipping", + file->mode, from_fullpath); + + /* Check that file exist in dest pgdata */ + if (arguments->backup_mode != BACKUP_MODE_FULL) + { + pgFile **dest_file_tmp = NULL; + dest_file_tmp = (pgFile **) parray_bsearch(arguments->dest_filelist, + file, pgFileCompareRelPathWithExternal); + if (dest_file_tmp) + { + /* File exists in destination PGDATA */ + file->exists_in_prev = true; + dest_file = *dest_file_tmp; + } + } + + /* Do actual work */ + if (file->is_datafile && !file->is_cfs) + { + catchup_data_file(file, from_fullpath, to_fullpath, + arguments->sync_lsn, + arguments->backup_mode, + arguments->nodeInfo->checksum_version, + dest_file != NULL ? dest_file->size : 0); + } + else + { + backup_non_data_file(file, dest_file, from_fullpath, to_fullpath, + arguments->backup_mode, current.parent_backup, true); + } + + /* file went missing during catchup */ + if (file->write_size == FILE_NOT_FOUND) + continue; + + if (file->write_size == BYTES_INVALID) + { + elog(LOG, "Skipping the unchanged file: \"%s\", read %li bytes", from_fullpath, file->read_size); + continue; + } + + arguments->transfered_bytes += file->write_size; + elog(LOG, "File \"%s\". Copied "INT64_FORMAT " bytes", + from_fullpath, file->write_size); + } + + /* ssh connection to longer needed */ + fio_disconnect(); + + /* Data files transferring is successful */ + arguments->completed = true; + + return NULL; +} + +/* + * main multithreaded copier + * returns size of transfered data file + * or -1 in case of error + */ +static ssize_t +catchup_multithreaded_copy(int num_threads, + PGNodeInfo *source_node_info, + const char *source_pgdata_path, + const char *dest_pgdata_path, + parray *source_filelist, + parray *dest_filelist, + XLogRecPtr sync_lsn, + BackupMode backup_mode) +{ + /* arrays with meta info for multi threaded catchup */ + catchup_thread_runner_arg *threads_args; + pthread_t *threads; + + bool all_threads_successful = true; + ssize_t transfered_bytes_result = 0; + int i; + + /* init thread args */ + threads_args = (catchup_thread_runner_arg *) palloc(sizeof(catchup_thread_runner_arg) * num_threads); + for (i = 0; i < num_threads; i++) + threads_args[i] = (catchup_thread_runner_arg){ + .nodeInfo = source_node_info, + .from_root = source_pgdata_path, + .to_root = dest_pgdata_path, + .source_filelist = source_filelist, + .dest_filelist = dest_filelist, + .sync_lsn = sync_lsn, + .backup_mode = backup_mode, + .thread_num = i + 1, + .transfered_bytes = 0, + .completed = false, + }; + + /* Run threads */ + thread_interrupted = false; + threads = (pthread_t *) palloc(sizeof(pthread_t) * num_threads); + if (!dry_run) + { + for (i = 0; i < num_threads; i++) + { + elog(VERBOSE, "Start thread num: %i", i); + pthread_create(&threads[i], NULL, &catchup_thread_runner, &(threads_args[i])); + } + } + + /* Wait threads */ + for (i = 0; i < num_threads; i++) + { + if (!dry_run) + pthread_join(threads[i], NULL); + all_threads_successful &= threads_args[i].completed; + transfered_bytes_result += threads_args[i].transfered_bytes; + } + + free(threads); + free(threads_args); + return all_threads_successful ? transfered_bytes_result : -1; +} + +/* + * Sync every file in destination directory to disk + */ +static void +catchup_sync_destination_files(const char* pgdata_path, fio_location location, parray *filelist, pgFile *pg_control_file) +{ + char fullpath[MAXPGPATH]; + time_t start_time, end_time; + char pretty_time[20]; + int i; + + elog(INFO, "Syncing copied files to disk"); + time(&start_time); + + for (i = 0; i < parray_num(filelist); i++) + { + pgFile *file = (pgFile *) parray_get(filelist, i); + + /* TODO: sync directory ? + * - at first glance we can rely on fs journaling, + * which is enabled by default on most platforms + * - but PG itself is not relying on fs, its durable_sync + * includes directory sync + */ + if (S_ISDIR(file->mode) || file->excluded) + continue; + + Assert(file->external_dir_num == 0); + join_path_components(fullpath, pgdata_path, file->rel_path); + if (fio_sync(fullpath, location) != 0) + elog(ERROR, "Cannot sync file \"%s\": %s", fullpath, strerror(errno)); + } + + /* + * sync pg_control file + */ + join_path_components(fullpath, pgdata_path, pg_control_file->rel_path); + if (fio_sync(fullpath, location) != 0) + elog(ERROR, "Cannot sync file \"%s\": %s", fullpath, strerror(errno)); + + time(&end_time); + pretty_time_interval(difftime(end_time, start_time), + pretty_time, lengthof(pretty_time)); + elog(INFO, "Files are synced, time elapsed: %s", pretty_time); +} + +/* + * Filter filelist helper function (used to process --exclude-path's) + * filelist -- parray of pgFile *, can't be NULL + * exclude_absolute_paths_list -- sorted parray of char * (absolute paths, starting with '/'), can be NULL + * exclude_relative_paths_list -- sorted parray of char * (relative paths), can be NULL + * logging_string -- helper parameter, used for generating verbose log messages ("Source" or "Destination") + */ +static void +filter_filelist(parray *filelist, const char *pgdata, + parray *exclude_absolute_paths_list, parray *exclude_relative_paths_list, + const char *logging_string) +{ + int i; + + if (exclude_absolute_paths_list == NULL && exclude_relative_paths_list == NULL) + return; + + for (i = 0; i < parray_num(filelist); ++i) + { + char full_path[MAXPGPATH]; + pgFile *file = (pgFile *) parray_get(filelist, i); + join_path_components(full_path, pgdata, file->rel_path); + + if ( + (exclude_absolute_paths_list != NULL + && parray_bsearch(exclude_absolute_paths_list, full_path, pgPrefixCompareString)!= NULL + ) || ( + exclude_relative_paths_list != NULL + && parray_bsearch(exclude_relative_paths_list, file->rel_path, pgPrefixCompareString)!= NULL) + ) + { + elog(INFO, "%s file \"%s\" excluded with --exclude-path option", logging_string, full_path); + file->excluded = true; + } + } +} + +/* + * Entry point of pg_probackup CATCHUP subcommand. + * exclude_*_paths_list are parray's of char * + */ +int +do_catchup(const char *source_pgdata, const char *dest_pgdata, int num_threads, bool sync_dest_files, + parray *exclude_absolute_paths_list, parray *exclude_relative_paths_list) +{ + PGconn *source_conn = NULL; + PGNodeInfo source_node_info; + bool backup_logs = false; + parray *source_filelist = NULL; + pgFile *source_pg_control_file = NULL; + parray *dest_filelist = NULL; + char dest_xlog_path[MAXPGPATH]; + + RedoParams dest_redo = { 0, InvalidXLogRecPtr, 0 }; + PGStopBackupResult stop_backup_result; + bool catchup_isok = true; + + int i; + + /* for fancy reporting */ + time_t start_time, end_time; + ssize_t transfered_datafiles_bytes = 0; + ssize_t transfered_walfiles_bytes = 0; + char pretty_source_bytes[20]; + + char dest_pg_control_fullpath[MAXPGPATH]; + char dest_pg_control_bak_fullpath[MAXPGPATH]; + + source_conn = catchup_init_state(&source_node_info, source_pgdata, dest_pgdata); + catchup_preflight_checks(&source_node_info, source_conn, source_pgdata, dest_pgdata); + + /* we need to sort --exclude_path's for future searching */ + if (exclude_absolute_paths_list != NULL) + parray_qsort(exclude_absolute_paths_list, pgCompareString); + if (exclude_relative_paths_list != NULL) + parray_qsort(exclude_relative_paths_list, pgCompareString); + + elog(INFO, "Database catchup start"); + + if (current.backup_mode != BACKUP_MODE_FULL) + { + dest_filelist = parray_new(); + dir_list_file(dest_filelist, dest_pgdata, + true, true, false, backup_logs, true, 0, FIO_LOCAL_HOST); + filter_filelist(dest_filelist, dest_pgdata, exclude_absolute_paths_list, exclude_relative_paths_list, "Destination"); + + // fill dest_redo.lsn and dest_redo.tli + get_redo(dest_pgdata, FIO_LOCAL_HOST, &dest_redo); + elog(INFO, "syncLSN = %X/%X", (uint32) (dest_redo.lsn >> 32), (uint32) dest_redo.lsn); + + /* + * Future improvement to catch partial catchup: + * 1. rename dest pg_control into something like pg_control.pbk + * (so user can't start partial catchup'ed instance from this point) + * 2. try to read by get_redo() pg_control and pg_control.pbk (to detect partial catchup) + * 3. at the end (after copy of correct pg_control), remove pg_control.pbk + */ + } + + /* + * Make sure that sync point is withing ptrack tracking range + * TODO: move to separate function to use in both backup.c and catchup.c + */ + if (current.backup_mode == BACKUP_MODE_DIFF_PTRACK) + { + XLogRecPtr ptrack_lsn = get_last_ptrack_lsn(source_conn, &source_node_info); + + if (ptrack_lsn > dest_redo.lsn || ptrack_lsn == InvalidXLogRecPtr) + elog(ERROR, "LSN from ptrack_control in source %X/%X is greater than checkpoint LSN in destination %X/%X.\n" + "You can perform only FULL catchup.", + (uint32) (ptrack_lsn >> 32), (uint32) (ptrack_lsn), + (uint32) (dest_redo.lsn >> 32), + (uint32) (dest_redo.lsn)); + } + + { + char label[1024]; + /* notify start of backup to PostgreSQL server */ + time2iso(label, lengthof(label), current.start_time, false); + strncat(label, " with pg_probackup", lengthof(label) - + strlen(" with pg_probackup")); + + /* Call pg_start_backup function in PostgreSQL connect */ + pg_start_backup(label, smooth_checkpoint, ¤t, &source_node_info, source_conn); + elog(INFO, "pg_start_backup START LSN %X/%X", (uint32) (current.start_lsn >> 32), (uint32) (current.start_lsn)); + } + + /* Sanity: source cluster must be "in future" relatively to dest cluster */ + if (current.backup_mode != BACKUP_MODE_FULL && + dest_redo.lsn > current.start_lsn) + elog(ERROR, "Current START LSN %X/%X is lower than SYNC LSN %X/%X, " + "it may indicate that we are trying to catchup with PostgreSQL instance from the past", + (uint32) (current.start_lsn >> 32), (uint32) (current.start_lsn), + (uint32) (dest_redo.lsn >> 32), (uint32) (dest_redo.lsn)); + + /* Start stream replication */ + join_path_components(dest_xlog_path, dest_pgdata, PG_XLOG_DIR); + if (!dry_run) + { + fio_mkdir(dest_xlog_path, DIR_PERMISSION, FIO_LOCAL_HOST); + start_WAL_streaming(source_conn, dest_xlog_path, &instance_config.conn_opt, + current.start_lsn, current.tli, false); + } + else + elog(INFO, "WAL streaming skipping with --dry-run option"); + + source_filelist = parray_new(); + + /* list files with the logical path. omit $PGDATA */ + if (fio_is_remote(FIO_DB_HOST)) + fio_list_dir(source_filelist, source_pgdata, + true, true, false, backup_logs, true, 0); + else + dir_list_file(source_filelist, source_pgdata, + true, true, false, backup_logs, true, 0, FIO_LOCAL_HOST); + + //REVIEW FIXME. Let's fix that before release. + // TODO what if wal is not a dir (symlink to a dir)? + // - Currently backup/restore transform pg_wal symlink to directory + // so the problem is not only with catchup. + // if we want to make it right - we must provide the way + // for symlink remapping during restore and catchup. + // By default everything must be left as it is. + + /* close ssh session in main thread */ + fio_disconnect(); + + /* + * Sort pathname ascending. It is necessary to create intermediate + * directories sequentially. + * + * For example: + * 1 - create 'base' + * 2 - create 'base/1' + * + * Sorted array is used at least in parse_filelist_filenames(), + * extractPageMap(), make_pagemap_from_ptrack(). + */ + parray_qsort(source_filelist, pgFileCompareRelPathWithExternal); + + //REVIEW Do we want to do similar calculation for dest? + //REVIEW_ANSWER what for? + { + ssize_t source_bytes = 0; + char pretty_bytes[20]; + + source_bytes += calculate_datasize_of_filelist(source_filelist); + + /* Extract information about files in source_filelist parsing their names:*/ + parse_filelist_filenames(source_filelist, source_pgdata); + filter_filelist(source_filelist, source_pgdata, exclude_absolute_paths_list, exclude_relative_paths_list, "Source"); + + current.pgdata_bytes += calculate_datasize_of_filelist(source_filelist); + + pretty_size(current.pgdata_bytes, pretty_source_bytes, lengthof(pretty_source_bytes)); + pretty_size(source_bytes - current.pgdata_bytes, pretty_bytes, lengthof(pretty_bytes)); + elog(INFO, "Source PGDATA size: %s (excluded %s)", pretty_source_bytes, pretty_bytes); + } + + elog(INFO, "Start LSN (source): %X/%X, TLI: %X", + (uint32) (current.start_lsn >> 32), (uint32) (current.start_lsn), + current.tli); + if (current.backup_mode != BACKUP_MODE_FULL) + elog(INFO, "LSN in destination: %X/%X, TLI: %X", + (uint32) (dest_redo.lsn >> 32), (uint32) (dest_redo.lsn), + dest_redo.tli); + + /* Build page mapping in PTRACK mode */ + if (current.backup_mode == BACKUP_MODE_DIFF_PTRACK) + { + time(&start_time); + elog(INFO, "Extracting pagemap of changed blocks"); + + /* Build the page map from ptrack information */ + make_pagemap_from_ptrack_2(source_filelist, source_conn, + source_node_info.ptrack_schema, + source_node_info.ptrack_version_num, + dest_redo.lsn); + time(&end_time); + elog(INFO, "Pagemap successfully extracted, time elapsed: %.0f sec", + difftime(end_time, start_time)); + } + + /* + * Make directories before catchup + */ + /* + * We iterate over source_filelist and for every directory with parent 'pg_tblspc' + * we must lookup this directory name in tablespace map. + * If we got a match, we treat this directory as tablespace. + * It means that we create directory specified in tablespace map and + * original directory created as symlink to it. + */ + for (i = 0; i < parray_num(source_filelist); i++) + { + pgFile *file = (pgFile *) parray_get(source_filelist, i); + char parent_dir[MAXPGPATH]; + + if (!S_ISDIR(file->mode) || file->excluded) + continue; + + /* + * check if it is fake "directory" and is a tablespace link + * this is because we passed the follow_symlink when building the list + */ + /* get parent dir of rel_path */ + strncpy(parent_dir, file->rel_path, MAXPGPATH); + get_parent_directory(parent_dir); + + /* check if directory is actually link to tablespace */ + if (strcmp(parent_dir, PG_TBLSPC_DIR) != 0) + { + /* if the entry is a regular directory, create it in the destination */ + char dirpath[MAXPGPATH]; + + join_path_components(dirpath, dest_pgdata, file->rel_path); + + elog(LOG, "Create directory '%s'", dirpath); + if (!dry_run) + fio_mkdir(dirpath, DIR_PERMISSION, FIO_LOCAL_HOST); + } + else + { + /* this directory located in pg_tblspc */ + const char *linked_path = NULL; + char to_path[MAXPGPATH]; + + // TODO perform additional check that this is actually symlink? + { /* get full symlink path and map this path to new location */ + char source_full_path[MAXPGPATH]; + char symlink_content[MAXPGPATH]; + join_path_components(source_full_path, source_pgdata, file->rel_path); + fio_readlink(source_full_path, symlink_content, sizeof(symlink_content), FIO_DB_HOST); + /* we checked that mapping exists in preflight_checks for local catchup */ + linked_path = get_tablespace_mapping(symlink_content); + elog(INFO, "Map tablespace full_path: \"%s\" old_symlink_content: \"%s\" new_symlink_content: \"%s\"\n", + source_full_path, + symlink_content, + linked_path); + } + + if (!is_absolute_path(linked_path)) + elog(ERROR, "Tablespace directory path must be an absolute path: %s\n", + linked_path); + + join_path_components(to_path, dest_pgdata, file->rel_path); + + elog(INFO, "Create directory \"%s\" and symbolic link \"%s\"", + linked_path, to_path); + + if (!dry_run) + { + /* create tablespace directory */ + if (fio_mkdir(linked_path, file->mode, FIO_LOCAL_HOST) != 0) + elog(ERROR, "Could not create tablespace directory \"%s\": %s", + linked_path, strerror(errno)); + + /* create link to linked_path */ + if (fio_symlink(linked_path, to_path, true, FIO_LOCAL_HOST) < 0) + elog(ERROR, "Could not create symbolic link \"%s\" -> \"%s\": %s", + linked_path, to_path, strerror(errno)); + } + } + } + + /* + * find pg_control file (in already sorted source_filelist) + * and exclude it from list for future special processing + */ + { + int control_file_elem_index; + pgFile search_key; + MemSet(&search_key, 0, sizeof(pgFile)); + /* pgFileCompareRelPathWithExternal uses only .rel_path and .external_dir_num for comparision */ + search_key.rel_path = XLOG_CONTROL_FILE; + search_key.external_dir_num = 0; + control_file_elem_index = parray_bsearch_index(source_filelist, &search_key, pgFileCompareRelPathWithExternal); + if(control_file_elem_index < 0) + elog(ERROR, "\"%s\" not found in \"%s\"\n", XLOG_CONTROL_FILE, source_pgdata); + source_pg_control_file = parray_remove(source_filelist, control_file_elem_index); + } + + /* TODO before public release: must be more careful with pg_control. + * when running catchup or incremental restore + * cluster is actually in two states + * simultaneously - old and new, so + * it must contain both pg_control files + * describing those states: global/pg_control_old, global/pg_control_new + * 1. This approach will provide us with means of + * robust detection of previos failures and thus correct operation retrying (or forbidding). + * 2. We will have the ability of preventing instance from starting + * in the middle of our operations. + */ + + /* + * remove absent source files in dest (dropped tables, etc...) + * note: global/pg_control will also be deleted here + * mark dest files (that excluded with source --exclude-path) also for exclusion + */ + if (current.backup_mode != BACKUP_MODE_FULL) + { + elog(INFO, "Removing redundant files in destination directory"); + parray_qsort(dest_filelist, pgFileCompareRelPathWithExternalDesc); + for (i = 0; i < parray_num(dest_filelist); i++) + { + bool redundant = true; + pgFile *file = (pgFile *) parray_get(dest_filelist, i); + pgFile **src_file = NULL; + + //TODO optimize it and use some merge-like algorithm + //instead of bsearch for each file. + src_file = (pgFile **) parray_bsearch(source_filelist, file, pgFileCompareRelPathWithExternal); + + if (src_file!= NULL && !(*src_file)->excluded && file->excluded) + (*src_file)->excluded = true; + + if (src_file!= NULL || file->excluded) + redundant = false; + + /* pg_filenode.map are always copied, because it's crc cannot be trusted */ + Assert(file->external_dir_num == 0); + if (pg_strcasecmp(file->name, RELMAPPER_FILENAME) == 0) + redundant = true; + /* global/pg_control.pbk.bak is always keeped, because it's needed for restart failed incremental restore */ + if (pg_strcasecmp(file->rel_path, XLOG_CONTROL_BAK_FILE) == 0) + redundant = false; + + /* if file does not exists in destination list, then we can safely unlink it */ + if (redundant) + { + char fullpath[MAXPGPATH]; + + join_path_components(fullpath, dest_pgdata, file->rel_path); + if (!dry_run) + { + fio_delete(file->mode, fullpath, FIO_LOCAL_HOST); + } + elog(LOG, "Deleted file \"%s\"", fullpath); + + /* shrink dest pgdata list */ + pgFileFree(file); + parray_remove(dest_filelist, i); + i--; + } + } + } + + /* clear file locks */ + pfilearray_clear_locks(source_filelist); + + /* Sort by size for load balancing */ + parray_qsort(source_filelist, pgFileCompareSizeDesc); + + /* Sort the array for binary search */ + if (dest_filelist) + parray_qsort(dest_filelist, pgFileCompareRelPathWithExternal); + + join_path_components(dest_pg_control_fullpath, dest_pgdata, XLOG_CONTROL_FILE); + join_path_components(dest_pg_control_bak_fullpath, dest_pgdata, XLOG_CONTROL_BAK_FILE); + /* + * rename (if it exist) dest control file before restoring + * if it doesn't exist, that mean, that we already restoring in a previously failed + * pgdata, where XLOG_CONTROL_BAK_FILE exist + */ + if (current.backup_mode != BACKUP_MODE_FULL && !dry_run) + { + if (!fio_access(dest_pg_control_fullpath, F_OK, FIO_LOCAL_HOST)) + { + pgFile *dst_control; + dst_control = pgFileNew(dest_pg_control_bak_fullpath, XLOG_CONTROL_BAK_FILE, + true,0, FIO_BACKUP_HOST); + + if(!fio_access(dest_pg_control_bak_fullpath, F_OK, FIO_LOCAL_HOST)) + fio_delete(dst_control->mode, dest_pg_control_bak_fullpath, FIO_LOCAL_HOST); + fio_rename(dest_pg_control_fullpath, dest_pg_control_bak_fullpath, FIO_LOCAL_HOST); + pgFileFree(dst_control); + } + } + + /* run copy threads */ + elog(INFO, "Start transferring data files"); + time(&start_time); + transfered_datafiles_bytes = catchup_multithreaded_copy(num_threads, &source_node_info, + source_pgdata, dest_pgdata, + source_filelist, dest_filelist, + dest_redo.lsn, current.backup_mode); + catchup_isok = transfered_datafiles_bytes != -1; + + /* at last copy control file */ + if (catchup_isok && !dry_run) + { + char from_fullpath[MAXPGPATH]; + char to_fullpath[MAXPGPATH]; + join_path_components(from_fullpath, source_pgdata, source_pg_control_file->rel_path); + join_path_components(to_fullpath, dest_pgdata, source_pg_control_file->rel_path); + copy_pgcontrol_file(from_fullpath, FIO_DB_HOST, + to_fullpath, FIO_LOCAL_HOST, source_pg_control_file); + transfered_datafiles_bytes += source_pg_control_file->size; + + /* Now backup control file can be deled */ + if (current.backup_mode != BACKUP_MODE_FULL && !fio_access(dest_pg_control_bak_fullpath, F_OK, FIO_LOCAL_HOST)){ + pgFile *dst_control; + dst_control = pgFileNew(dest_pg_control_bak_fullpath, XLOG_CONTROL_BAK_FILE, + true,0, FIO_BACKUP_HOST); + fio_delete(dst_control->mode, dest_pg_control_bak_fullpath, FIO_LOCAL_HOST); + pgFileFree(dst_control); + } + } + + if (!catchup_isok && !dry_run) + { + char pretty_time[20]; + char pretty_transfered_data_bytes[20]; + + time(&end_time); + pretty_time_interval(difftime(end_time, start_time), + pretty_time, lengthof(pretty_time)); + pretty_size(transfered_datafiles_bytes, pretty_transfered_data_bytes, lengthof(pretty_transfered_data_bytes)); + + elog(ERROR, "Catchup failed. Transfered: %s, time elapsed: %s", + pretty_transfered_data_bytes, pretty_time); + } + + /* Notify end of backup */ + { + //REVIEW Is it relevant to catchup? I suppose it isn't, since catchup is a new code. + //If we do need it, please write a comment explaining that. + /* kludge against some old bug in archive_timeout. TODO: remove in 3.0.0 */ + int timeout = (instance_config.archive_timeout > 0) ? + instance_config.archive_timeout : ARCHIVE_TIMEOUT_DEFAULT; + char *stop_backup_query_text = NULL; + + pg_silent_client_messages(source_conn); + + /* Execute pg_stop_backup using PostgreSQL connection */ + pg_stop_backup_send(source_conn, source_node_info.server_version, current.from_replica, exclusive_backup, &stop_backup_query_text); + + /* + * Wait for the result of pg_stop_backup(), but no longer than + * archive_timeout seconds + */ + pg_stop_backup_consume(source_conn, source_node_info.server_version, exclusive_backup, timeout, stop_backup_query_text, &stop_backup_result); + + /* Cleanup */ + pg_free(stop_backup_query_text); + } + + if (!dry_run) + wait_wal_and_calculate_stop_lsn(dest_xlog_path, stop_backup_result.lsn, ¤t); + +#if PG_VERSION_NUM >= 90600 + /* Write backup_label */ + Assert(stop_backup_result.backup_label_content != NULL); + if (!dry_run) + { + pg_stop_backup_write_file_helper(dest_pgdata, PG_BACKUP_LABEL_FILE, "backup label", + stop_backup_result.backup_label_content, stop_backup_result.backup_label_content_len, + NULL); + } + free(stop_backup_result.backup_label_content); + stop_backup_result.backup_label_content = NULL; + stop_backup_result.backup_label_content_len = 0; + + /* tablespace_map */ + if (stop_backup_result.tablespace_map_content != NULL) + { + // TODO what if tablespace is created during catchup? + /* Because we have already created symlinks in pg_tblspc earlier, + * we do not need to write the tablespace_map file. + * So this call is unnecessary: + * pg_stop_backup_write_file_helper(dest_pgdata, PG_TABLESPACE_MAP_FILE, "tablespace map", + * stop_backup_result.tablespace_map_content, stop_backup_result.tablespace_map_content_len, + * NULL); + */ + free(stop_backup_result.tablespace_map_content); + stop_backup_result.tablespace_map_content = NULL; + stop_backup_result.tablespace_map_content_len = 0; + } +#endif + + /* wait for end of wal streaming and calculate wal size transfered */ + if (!dry_run) + { + parray *wal_files_list = NULL; + wal_files_list = parray_new(); + + if (wait_WAL_streaming_end(wal_files_list)) + elog(ERROR, "WAL streaming failed"); + + for (i = 0; i < parray_num(wal_files_list); i++) + { + pgFile *file = (pgFile *) parray_get(wal_files_list, i); + transfered_walfiles_bytes += file->size; + } + + parray_walk(wal_files_list, pgFileFree); + parray_free(wal_files_list); + wal_files_list = NULL; + } + + /* + * In case of backup from replica >= 9.6 we must fix minRecPoint + */ + if (current.from_replica && !exclusive_backup) + { + set_min_recovery_point(source_pg_control_file, dest_pgdata, current.stop_lsn); + } + + /* close ssh session in main thread */ + fio_disconnect(); + + /* fancy reporting */ + { + char pretty_transfered_data_bytes[20]; + char pretty_transfered_wal_bytes[20]; + char pretty_time[20]; + + time(&end_time); + pretty_time_interval(difftime(end_time, start_time), + pretty_time, lengthof(pretty_time)); + pretty_size(transfered_datafiles_bytes, pretty_transfered_data_bytes, lengthof(pretty_transfered_data_bytes)); + pretty_size(transfered_walfiles_bytes, pretty_transfered_wal_bytes, lengthof(pretty_transfered_wal_bytes)); + + elog(INFO, "Databases synchronized. Transfered datafiles size: %s, transfered wal size: %s, time elapsed: %s", + pretty_transfered_data_bytes, pretty_transfered_wal_bytes, pretty_time); + + if (current.backup_mode != BACKUP_MODE_FULL) + elog(INFO, "Catchup incremental ratio (less is better): %.f%% (%s/%s)", + ((float) transfered_datafiles_bytes / current.pgdata_bytes) * 100, + pretty_transfered_data_bytes, pretty_source_bytes); + } + + /* Sync all copied files unless '--no-sync' flag is used */ + if (sync_dest_files && !dry_run) + catchup_sync_destination_files(dest_pgdata, FIO_LOCAL_HOST, source_filelist, source_pg_control_file); + else + elog(WARNING, "Files are not synced to disk"); + + /* Cleanup */ + if (dest_filelist && !dry_run) + { + parray_walk(dest_filelist, pgFileFree); + } + parray_free(dest_filelist); + parray_walk(source_filelist, pgFileFree); + parray_free(source_filelist); + pgFileFree(source_pg_control_file); + + return 0; +} diff --git a/src/checkdb.c b/src/checkdb.c index e878d688f..2a7d4e9eb 100644 --- a/src/checkdb.c +++ b/src/checkdb.c @@ -39,6 +39,8 @@ typedef struct ConnectionArgs conn_arg; /* number of thread for debugging */ int thread_num; + /* pgdata path */ + const char *from_root; /* * Return value from the thread: * 0 everything is ok @@ -81,6 +83,7 @@ typedef struct pg_indexEntry char *name; char *namespace; bool heapallindexed_is_supported; + bool checkunique_is_supported; /* schema where amcheck extension is located */ char *amcheck_nspname; /* lock for synchronization of parallel threads */ @@ -130,6 +133,7 @@ check_files(void *arg) int i; check_files_arg *arguments = (check_files_arg *) arg; int n_files_list = 0; + char from_fullpath[MAXPGPATH]; if (arguments->files_list) n_files_list = parray_num(arguments->files_list); @@ -137,49 +141,28 @@ check_files(void *arg) /* check a file */ for (i = 0; i < n_files_list; i++) { - int ret; - struct stat buf; pgFile *file = (pgFile *) parray_get(arguments->files_list, i); + /* check for interrupt */ + if (interrupted || thread_interrupted) + elog(ERROR, "Interrupted during checkdb"); + + /* No need to check directories */ + if (S_ISDIR(file->mode)) + continue; + if (!pg_atomic_test_set_flag(&file->lock)) continue; - elog(VERBOSE, "Checking file: \"%s\" ", file->path); + join_path_components(from_fullpath, arguments->from_root, file->rel_path); - /* check for interrupt */ - if (interrupted || thread_interrupted) - elog(ERROR, "interrupted during checkdb"); + elog(VERBOSE, "Checking file: \"%s\" ", from_fullpath); if (progress) elog(INFO, "Progress: (%d/%d). Process file \"%s\"", - i + 1, n_files_list, file->path); + i + 1, n_files_list, from_fullpath); - /* stat file to check its current state */ - ret = stat(file->path, &buf); - if (ret == -1) - { - if (errno == ENOENT) - { - /* - * If file is not found, this is not en error. - * It could have been deleted by concurrent postgres transaction. - */ - elog(LOG, "File \"%s\" is not found", file->path); - continue; - } - else - { - elog(ERROR, - "can't stat file to check \"%s\": %s", - file->path, strerror(errno)); - } - } - - /* No need to check directories */ - if (S_ISDIR(buf.st_mode)) - continue; - - if (S_ISREG(buf.st_mode)) + if (S_ISREG(file->mode)) { /* check only uncompressed by cfs datafiles */ if (file->is_datafile && !file->is_cfs) @@ -189,13 +172,14 @@ check_files(void *arg) * uses global variables to set connections. * Need refactoring. */ - if (!check_data_file(&(arguments->conn_arg), file, + if (!check_data_file(&(arguments->conn_arg), + file, from_fullpath, arguments->checksum_version)) arguments->ret = 2; /* corruption found */ } } else - elog(WARNING, "unexpected file type %d", buf.st_mode); + elog(WARNING, "unexpected file type %d", file->mode); } /* Ret values: @@ -224,8 +208,8 @@ do_block_validation(char *pgdata, uint32 checksum_version) files_list = parray_new(); /* list files with the logical path. omit $PGDATA */ - dir_list_file(files_list, pgdata, - true, true, false, 0, FIO_DB_HOST); + dir_list_file(files_list, pgdata, true, true, + false, false, true, 0, FIO_DB_HOST); /* * Sort pathname ascending. @@ -234,7 +218,7 @@ do_block_validation(char *pgdata, uint32 checksum_version) * 1 - create 'base' * 2 - create 'base/1' */ - parray_qsort(files_list, pgFileComparePath); + parray_qsort(files_list, pgFileCompareRelPathWithExternal); /* Extract information about files in pgdata parsing their names:*/ parse_filelist_filenames(files_list, pgdata); @@ -258,6 +242,7 @@ do_block_validation(char *pgdata, uint32 checksum_version) arg->files_list = files_list; arg->checksum_version = checksum_version; + arg->from_root = pgdata; arg->conn_arg.conn = NULL; arg->conn_arg.cancel_conn = NULL; @@ -308,6 +293,7 @@ check_indexes(void *arg) int i; check_indexes_arg *arguments = (check_indexes_arg *) arg; int n_indexes = 0; + my_thread_num = arguments->thread_num; if (arguments->index_list) n_indexes = parray_num(arguments->index_list); @@ -366,16 +352,20 @@ get_index_list(const char *dbname, bool first_db_with_amcheck, { PGresult *res; char *amcheck_nspname = NULL; + char *amcheck_extname = NULL; + char *amcheck_extversion = NULL; int i; bool heapallindexed_is_supported = false; + bool checkunique_is_supported = false; parray *index_list = NULL; + /* Check amcheck extension version */ res = pgut_execute(db_conn, "SELECT " "extname, nspname, extversion " - "FROM pg_namespace n " - "JOIN pg_extension e " + "FROM pg_catalog.pg_namespace n " + "JOIN pg_catalog.pg_extension e " "ON n.oid=e.extnamespace " - "WHERE e.extname IN ('amcheck', 'amcheck_next') " + "WHERE e.extname IN ('amcheck'::name, 'amcheck_next'::name) " "ORDER BY extversion DESC " "LIMIT 1", 0, NULL); @@ -394,24 +384,68 @@ get_index_list(const char *dbname, bool first_db_with_amcheck, return NULL; } + amcheck_extname = pgut_malloc(strlen(PQgetvalue(res, 0, 0)) + 1); + strcpy(amcheck_extname, PQgetvalue(res, 0, 0)); amcheck_nspname = pgut_malloc(strlen(PQgetvalue(res, 0, 1)) + 1); strcpy(amcheck_nspname, PQgetvalue(res, 0, 1)); + amcheck_extversion = pgut_malloc(strlen(PQgetvalue(res, 0, 2)) + 1); + strcpy(amcheck_extversion, PQgetvalue(res, 0, 2)); + PQclear(res); /* heapallindexed_is_supported is database specific */ - if (strcmp(PQgetvalue(res, 0, 2), "1.0") != 0 && - strcmp(PQgetvalue(res, 0, 2), "1") != 0) + /* TODO this is wrong check, heapallindexed supported also in 1.1.1, 1.2 and 1.2.1... */ + if (strcmp(amcheck_extversion, "1.0") != 0 && + strcmp(amcheck_extversion, "1") != 0) heapallindexed_is_supported = true; elog(INFO, "Amchecking database '%s' using extension '%s' " "version %s from schema '%s'", - dbname, PQgetvalue(res, 0, 0), - PQgetvalue(res, 0, 2), PQgetvalue(res, 0, 1)); + dbname, amcheck_extname, + amcheck_extversion, amcheck_nspname); if (!heapallindexed_is_supported && heapallindexed) elog(WARNING, "Extension '%s' version %s in schema '%s'" "do not support 'heapallindexed' option", - PQgetvalue(res, 0, 0), PQgetvalue(res, 0, 2), - PQgetvalue(res, 0, 1)); + amcheck_extname, amcheck_extversion, + amcheck_nspname); + +#ifndef PGPRO_EE + /* + * Will support when the vanilla patch will commited https://commitfest.postgresql.org/32/2976/ + */ + checkunique_is_supported = false; +#else + /* + * Check bt_index_check function signature to determine support of checkunique parameter + * This can't be exactly checked by checking extension version, + * For example, 1.1.1 and 1.2.1 supports this parameter, but 1.2 doesn't (PGPROEE-12.4.1) + */ + res = pgut_execute(db_conn, "SELECT " + " oid " + "FROM pg_catalog.pg_proc " + "WHERE " + " pronamespace = $1::regnamespace " + "AND proname = 'bt_index_check' " + "AND 'checkunique' = ANY(proargnames) " + "AND (pg_catalog.string_to_array(proargtypes::text, ' ')::regtype[])[pg_catalog.array_position(proargnames, 'checkunique')] = 'bool'::regtype", + 1, (const char **) &amcheck_nspname); + + if (PQresultStatus(res) != PGRES_TUPLES_OK) + { + PQclear(res); + elog(ERROR, "Cannot check 'checkunique' option is supported in bt_index_check function %s: %s", + dbname, PQerrorMessage(db_conn)); + } + + checkunique_is_supported = PQntuples(res) >= 1; + PQclear(res); +#endif + + if (!checkunique_is_supported && checkunique) + elog(WARNING, "Extension '%s' version %s in schema '%s' " + "do not support 'checkunique' parameter", + amcheck_extname, amcheck_extversion, + amcheck_nspname); /* * In order to avoid duplicates, select global indexes @@ -427,7 +461,9 @@ get_index_list(const char *dbname, bool first_db_with_amcheck, "LEFT JOIN pg_catalog.pg_class cls ON idx.indexrelid=cls.oid " "LEFT JOIN pg_catalog.pg_namespace nmspc ON cls.relnamespace=nmspc.oid " "LEFT JOIN pg_catalog.pg_am am ON cls.relam=am.oid " - "WHERE am.amname='btree' AND cls.relpersistence != 't' " + "WHERE am.amname='btree' " + "AND cls.relpersistence != 't' " + "AND cls.relkind != 'I' " "ORDER BY nmspc.nspname DESC", 0, NULL); } @@ -439,8 +475,10 @@ get_index_list(const char *dbname, bool first_db_with_amcheck, "LEFT JOIN pg_catalog.pg_class cls ON idx.indexrelid=cls.oid " "LEFT JOIN pg_catalog.pg_namespace nmspc ON cls.relnamespace=nmspc.oid " "LEFT JOIN pg_catalog.pg_am am ON cls.relam=am.oid " - "WHERE am.amname='btree' AND cls.relpersistence != 't' AND " - "(cls.reltablespace IN " + "WHERE am.amname='btree' " + "AND cls.relpersistence != 't' " + "AND cls.relkind != 'I' " + "AND (cls.reltablespace IN " "(SELECT oid from pg_catalog.pg_tablespace where spcname <> 'pg_global') " "OR cls.reltablespace = 0) " "ORDER BY nmspc.nspname DESC", @@ -455,7 +493,7 @@ get_index_list(const char *dbname, bool first_db_with_amcheck, char *namespace = NULL; /* index oid */ - ind->indexrelid = atoi(PQgetvalue(res, i, 0)); + ind->indexrelid = atoll(PQgetvalue(res, i, 0)); /* index relname */ name = PQgetvalue(res, i, 1); @@ -468,6 +506,7 @@ get_index_list(const char *dbname, bool first_db_with_amcheck, strcpy(ind->namespace, namespace); /* enough buffer size guaranteed */ ind->heapallindexed_is_supported = heapallindexed_is_supported; + ind->checkunique_is_supported = checkunique_is_supported; ind->amcheck_nspname = pgut_malloc(strlen(amcheck_nspname) + 1); strcpy(ind->amcheck_nspname, amcheck_nspname); pg_atomic_clear_flag(&ind->lock); @@ -479,6 +518,9 @@ get_index_list(const char *dbname, bool first_db_with_amcheck, } PQclear(res); + free(amcheck_extversion); + free(amcheck_nspname); + free(amcheck_extname); return index_list; } @@ -488,38 +530,46 @@ static bool amcheck_one_index(check_indexes_arg *arguments, pg_indexEntry *ind) { - PGresult *res; - char *params[2]; + PGresult *res; + char *params[3]; + static const char *queries[] = { + "SELECT %s.bt_index_check(index => $1)", + "SELECT %s.bt_index_check(index => $1, heapallindexed => $2)", + "SELECT %s.bt_index_check(index => $1, heapallindexed => $2, checkunique => $3)", + }; + int params_count; char *query = NULL; - params[0] = palloc(64); + if (interrupted) + elog(ERROR, "Interrupted"); +#define INDEXRELID 0 +#define HEAPALLINDEXED 1 +#define CHECKUNIQUE 2 /* first argument is index oid */ - sprintf(params[0], "%i", ind->indexrelid); + params[INDEXRELID] = palloc(64); + sprintf(params[INDEXRELID], "%u", ind->indexrelid); /* second argument is heapallindexed */ - params[1] = heapallindexed ? "true" : "false"; - - if (interrupted) - elog(ERROR, "Interrupted"); + params[HEAPALLINDEXED] = heapallindexed ? "true" : "false"; + /* third optional argument is checkunique */ + params[CHECKUNIQUE] = checkunique ? "true" : "false"; +#undef CHECKUNIQUE +#undef HEAPALLINDEXED - if (ind->heapallindexed_is_supported) - { - query = palloc(strlen(ind->amcheck_nspname)+strlen("SELECT .bt_index_check($1, $2)")+1); - sprintf(query, "SELECT %s.bt_index_check($1, $2)", ind->amcheck_nspname); + params_count = ind->checkunique_is_supported ? + 3 : + ( ind->heapallindexed_is_supported ? 2 : 1 ); - res = pgut_execute_parallel(arguments->conn_arg.conn, - arguments->conn_arg.cancel_conn, - query, 2, (const char **)params, true, true, true); - } - else - { - query = palloc(strlen(ind->amcheck_nspname)+strlen("SELECT .bt_index_check($1)")+1); - sprintf(query, "SELECT %s.bt_index_check($1)", ind->amcheck_nspname); + /* + * Prepare query text with schema name + * +1 for \0 and -2 for %s + */ + query = palloc(strlen(ind->amcheck_nspname) + strlen(queries[params_count - 1]) + 1 - 2); + sprintf(query, queries[params_count - 1], ind->amcheck_nspname); - res = pgut_execute_parallel(arguments->conn_arg.conn, + res = pgut_execute_parallel(arguments->conn_arg.conn, arguments->conn_arg.cancel_conn, - query, 1, (const char **)params, true, true, true); - } + query, params_count, (const char **)params, true, true, true); if (PQresultStatus(res) != PGRES_TUPLES_OK) { @@ -527,7 +577,7 @@ amcheck_one_index(check_indexes_arg *arguments, arguments->thread_num, arguments->conn_opt.pgdatabase, ind->namespace, ind->name, PQresultErrorMessage(res)); - pfree(params[0]); + pfree(params[INDEXRELID]); pfree(query); PQclear(res); return false; @@ -537,7 +587,8 @@ amcheck_one_index(check_indexes_arg *arguments, arguments->thread_num, arguments->conn_opt.pgdatabase, ind->namespace, ind->name); - pfree(params[0]); + pfree(params[INDEXRELID]); +#undef INDEXRELID pfree(query); PQclear(res); return true; @@ -571,8 +622,8 @@ do_amcheck(ConnectionOptions conn_opt, PGconn *conn) res_db = pgut_execute(conn, "SELECT datname, oid, dattablespace " - "FROM pg_database " - "WHERE datname NOT IN ('template0', 'template1')", + "FROM pg_catalog.pg_database " + "WHERE datname NOT IN ('template0'::name, 'template1'::name)", 0, NULL); /* we don't need this connection anymore */ @@ -699,7 +750,7 @@ do_checkdb(bool need_amcheck, if (!skip_block_validation) { if (!pgdata) - elog(ERROR, "required parameter not specified: PGDATA " + elog(ERROR, "Required parameter not specified: PGDATA " "(-D, --pgdata)"); /* get node info */ diff --git a/src/configure.c b/src/configure.c index f9d92fc1f..964548343 100644 --- a/src/configure.c +++ b/src/configure.c @@ -17,10 +17,14 @@ static void assign_log_level_console(ConfigOption *opt, const char *arg); static void assign_log_level_file(ConfigOption *opt, const char *arg); +static void assign_log_format_console(ConfigOption *opt, const char *arg); +static void assign_log_format_file(ConfigOption *opt, const char *arg); static void assign_compress_alg(ConfigOption *opt, const char *arg); static char *get_log_level_console(ConfigOption *opt); static char *get_log_level_file(ConfigOption *opt); +static char *get_log_format_console(ConfigOption *opt); +static char *get_log_format_file(ConfigOption *opt); static char *get_compress_alg(ConfigOption *opt); static void show_configure_start(void); @@ -49,7 +53,7 @@ ConfigOption instance_options[] = /* Instance options */ { 's', 'D', "pgdata", - &instance_config.pgdata, SOURCE_CMD, 0, + &instance_config.pgdata, SOURCE_CMD, SOURCE_DEFAULT, OPTION_INSTANCE_GROUP, 0, option_get_value }, { @@ -66,49 +70,49 @@ ConfigOption instance_options[] = #endif { 's', 'E', "external-dirs", - &instance_config.external_dir_str, SOURCE_CMD, 0, + &instance_config.external_dir_str, SOURCE_CMD, SOURCE_DEFAULT, OPTION_INSTANCE_GROUP, 0, option_get_value }, /* Connection options */ { 's', 'd', "pgdatabase", - &instance_config.conn_opt.pgdatabase, SOURCE_CMD, 0, + &instance_config.conn_opt.pgdatabase, SOURCE_CMD, SOURCE_DEFAULT, OPTION_CONN_GROUP, 0, option_get_value }, { 's', 'h', "pghost", - &instance_config.conn_opt.pghost, SOURCE_CMD, 0, + &instance_config.conn_opt.pghost, SOURCE_CMD, SOURCE_DEFAULT, OPTION_CONN_GROUP, 0, option_get_value }, { 's', 'p', "pgport", - &instance_config.conn_opt.pgport, SOURCE_CMD, 0, + &instance_config.conn_opt.pgport, SOURCE_CMD, SOURCE_DEFAULT, OPTION_CONN_GROUP, 0, option_get_value }, { 's', 'U', "pguser", - &instance_config.conn_opt.pguser, SOURCE_CMD, 0, + &instance_config.conn_opt.pguser, SOURCE_CMD, SOURCE_DEFAULT, OPTION_CONN_GROUP, 0, option_get_value }, /* Replica options */ { 's', 202, "master-db", - &instance_config.master_conn_opt.pgdatabase, SOURCE_CMD, 0, + &instance_config.master_conn_opt.pgdatabase, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REPLICA_GROUP, 0, option_get_value }, { 's', 203, "master-host", - &instance_config.master_conn_opt.pghost, SOURCE_CMD, 0, + &instance_config.master_conn_opt.pghost, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REPLICA_GROUP, 0, option_get_value }, { 's', 204, "master-port", - &instance_config.master_conn_opt.pgport, SOURCE_CMD, 0, + &instance_config.master_conn_opt.pgport, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REPLICA_GROUP, 0, option_get_value }, { 's', 205, "master-user", - &instance_config.master_conn_opt.pguser, SOURCE_CMD, 0, + &instance_config.master_conn_opt.pguser, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REPLICA_GROUP, 0, option_get_value }, { @@ -122,98 +126,133 @@ ConfigOption instance_options[] = &instance_config.archive_timeout, SOURCE_CMD, SOURCE_DEFAULT, OPTION_ARCHIVE_GROUP, OPTION_UNIT_S, option_get_value }, + { + 's', 208, "archive-host", + &instance_config.archive.host, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_ARCHIVE_GROUP, 0, option_get_value + }, + { + 's', 209, "archive-port", + &instance_config.archive.port, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_ARCHIVE_GROUP, 0, option_get_value + }, + { + 's', 210, "archive-user", + &instance_config.archive.user, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_ARCHIVE_GROUP, 0, option_get_value + }, + { + 's', 211, "restore-command", + &instance_config.restore_command, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_ARCHIVE_GROUP, 0, option_get_value + }, /* Logging options */ { - 'f', 208, "log-level-console", - assign_log_level_console, SOURCE_CMD, 0, + 'f', 212, "log-level-console", + assign_log_level_console, SOURCE_CMD, SOURCE_DEFAULT, OPTION_LOG_GROUP, 0, get_log_level_console }, { - 'f', 209, "log-level-file", - assign_log_level_file, SOURCE_CMD, 0, + 'f', 213, "log-level-file", + assign_log_level_file, SOURCE_CMD, SOURCE_DEFAULT, OPTION_LOG_GROUP, 0, get_log_level_file }, { - 's', 210, "log-filename", - &instance_config.logger.log_filename, SOURCE_CMD, 0, + 'f', 214, "log-format-console", + assign_log_format_console, SOURCE_CMD_STRICT, SOURCE_DEFAULT, + OPTION_LOG_GROUP, 0, get_log_format_console + }, + { + 'f', 215, "log-format-file", + assign_log_format_file, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_LOG_GROUP, 0, get_log_format_file + }, + { + 's', 216, "log-filename", + &instance_config.logger.log_filename, SOURCE_CMD, SOURCE_DEFAULT, OPTION_LOG_GROUP, 0, option_get_value }, { - 's', 211, "error-log-filename", - &instance_config.logger.error_log_filename, SOURCE_CMD, 0, + 's', 217, "error-log-filename", + &instance_config.logger.error_log_filename, SOURCE_CMD, SOURCE_DEFAULT, OPTION_LOG_GROUP, 0, option_get_value }, { - 's', 212, "log-directory", - &instance_config.logger.log_directory, SOURCE_CMD, 0, + 's', 218, "log-directory", + &instance_config.logger.log_directory, SOURCE_CMD, SOURCE_DEFAULT, OPTION_LOG_GROUP, 0, option_get_value }, { - 'U', 213, "log-rotation-size", + 'U', 219, "log-rotation-size", &instance_config.logger.log_rotation_size, SOURCE_CMD, SOURCE_DEFAULT, OPTION_LOG_GROUP, OPTION_UNIT_KB, option_get_value }, { - 'U', 214, "log-rotation-age", + 'U', 220, "log-rotation-age", &instance_config.logger.log_rotation_age, SOURCE_CMD, SOURCE_DEFAULT, OPTION_LOG_GROUP, OPTION_UNIT_MS, option_get_value }, /* Retention options */ { - 'u', 215, "retention-redundancy", - &instance_config.retention_redundancy, SOURCE_CMD, 0, + 'u', 221, "retention-redundancy", + &instance_config.retention_redundancy, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_RETENTION_GROUP, 0, option_get_value + }, + { + 'u', 222, "retention-window", + &instance_config.retention_window, SOURCE_CMD, SOURCE_DEFAULT, OPTION_RETENTION_GROUP, 0, option_get_value }, { - 'u', 216, "retention-window", - &instance_config.retention_window, SOURCE_CMD, 0, + 'u', 223, "wal-depth", + &instance_config.wal_depth, SOURCE_CMD, SOURCE_DEFAULT, OPTION_RETENTION_GROUP, 0, option_get_value }, /* Compression options */ { - 'f', 217, "compress-algorithm", - assign_compress_alg, SOURCE_CMD, 0, + 'f', 224, "compress-algorithm", + assign_compress_alg, SOURCE_CMD, SOURCE_DEFAULT, OPTION_COMPRESS_GROUP, 0, get_compress_alg }, { - 'u', 218, "compress-level", - &instance_config.compress_level, SOURCE_CMD, 0, + 'u', 225, "compress-level", + &instance_config.compress_level, SOURCE_CMD, SOURCE_DEFAULT, OPTION_COMPRESS_GROUP, 0, option_get_value }, /* Remote backup options */ { - 's', 219, "remote-proto", - &instance_config.remote.proto, SOURCE_CMD, 0, + 's', 226, "remote-proto", + &instance_config.remote.proto, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REMOTE_GROUP, 0, option_get_value }, { - 's', 220, "remote-host", - &instance_config.remote.host, SOURCE_CMD, 0, + 's', 227, "remote-host", + &instance_config.remote.host, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REMOTE_GROUP, 0, option_get_value }, { - 's', 221, "remote-port", - &instance_config.remote.port, SOURCE_CMD, 0, + 's', 228, "remote-port", + &instance_config.remote.port, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REMOTE_GROUP, 0, option_get_value }, { - 's', 222, "remote-path", - &instance_config.remote.path, SOURCE_CMD, 0, + 's', 229, "remote-path", + &instance_config.remote.path, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REMOTE_GROUP, 0, option_get_value }, { - 's', 223, "remote-user", - &instance_config.remote.user, SOURCE_CMD, 0, + 's', 230, "remote-user", + &instance_config.remote.user, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REMOTE_GROUP, 0, option_get_value }, { - 's', 224, "ssh-options", - &instance_config.remote.ssh_options, SOURCE_CMD, 0, + 's', 231, "ssh-options", + &instance_config.remote.ssh_options, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REMOTE_GROUP, 0, option_get_value }, { - 's', 225, "ssh-config", - &instance_config.remote.ssh_config, SOURCE_CMD, 0, + 's', 232, "ssh-config", + &instance_config.remote.ssh_config, SOURCE_CMD, SOURCE_DEFAULT, OPTION_REMOTE_GROUP, 0, option_get_value }, { 0 } @@ -230,7 +269,7 @@ static const char *current_group = NULL; * Show configure options including default values. */ void -do_show_config(void) +do_show_config(bool show_base_units) { int i; @@ -238,10 +277,13 @@ do_show_config(void) for (i = 0; instance_options[i].type; i++) { + if (show_base_units && strchr("bBiIuU", instance_options[i].type) && instance_options[i].get_value == *option_get_value) + instance_options[i].flags |= GET_VAL_IN_BASE_UNITS; /* Set flag */ if (show_format == SHOW_PLAIN) show_configure_plain(&instance_options[i]); else show_configure_json(&instance_options[i]); + instance_options[i].flags &= ~(GET_VAL_IN_BASE_UNITS); /* Reset flag. It was resetted in option_get_value(). Probably this reset isn't needed */ } show_configure_end(); @@ -252,18 +294,16 @@ do_show_config(void) * values into the file. */ void -do_set_config(bool missing_ok) +do_set_config(InstanceState *instanceState, bool missing_ok) { - char path[MAXPGPATH]; char path_temp[MAXPGPATH]; FILE *fp; int i; - join_path_components(path, backup_instance_path, BACKUP_CATALOG_CONF_FILE); - snprintf(path_temp, sizeof(path_temp), "%s.tmp", path); + snprintf(path_temp, sizeof(path_temp), "%s.tmp", instanceState->instance_config_path); - if (!missing_ok && !fileExists(path, FIO_LOCAL_HOST)) - elog(ERROR, "Configuration file \"%s\" doesn't exist", path); + if (!missing_ok && !fileExists(instanceState->instance_config_path, FIO_LOCAL_HOST)) + elog(ERROR, "Configuration file \"%s\" doesn't exist", instanceState->instance_config_path); fp = fopen(path_temp, "wt"); if (fp == NULL) @@ -274,6 +314,7 @@ do_set_config(bool missing_ok) for (i = 0; instance_options[i].type; i++) { + int rc = 0; ConfigOption *opt = &instance_options[i]; char *value; @@ -294,25 +335,37 @@ do_set_config(bool missing_ok) } if (strchr(value, ' ')) - fprintf(fp, "%s = '%s'\n", opt->lname, value); + rc = fprintf(fp, "%s = '%s'\n", opt->lname, value); else - fprintf(fp, "%s = %s\n", opt->lname, value); + rc = fprintf(fp, "%s = %s\n", opt->lname, value); + + if (rc < 0) + elog(ERROR, "Cannot write to configuration file: \"%s\"", path_temp); + pfree(value); } - fclose(fp); + if (ferror(fp) || fflush(fp)) + elog(ERROR, "Cannot write to configuration file: \"%s\"", path_temp); - if (rename(path_temp, path) < 0) + if (fclose(fp)) + elog(ERROR, "Cannot close configuration file: \"%s\"", path_temp); + + if (fio_sync(path_temp, FIO_LOCAL_HOST) != 0) + elog(ERROR, "Failed to sync temp configuration file \"%s\": %s", + path_temp, strerror(errno)); + + if (rename(path_temp, instanceState->instance_config_path) < 0) { int errno_temp = errno; unlink(path_temp); elog(ERROR, "Cannot rename configuration file \"%s\" to \"%s\": %s", - path_temp, path, strerror(errno_temp)); + path_temp, instanceState->instance_config_path, strerror(errno_temp)); } } void -init_config(InstanceConfig *config) +init_config(InstanceConfig *config, const char *instance_name) { MemSet(config, 0, sizeof(InstanceConfig)); @@ -335,6 +388,7 @@ init_config(InstanceConfig *config) config->retention_redundancy = RETENTION_REDUNDANCY_DEFAULT; config->retention_window = RETENTION_WINDOW_DEFAULT; + config->wal_depth = 0; config->compress_alg = COMPRESS_ALG_DEFAULT; config->compress_level = COMPRESS_LEVEL_DEFAULT; @@ -342,6 +396,282 @@ init_config(InstanceConfig *config) config->remote.proto = (char*)"ssh"; } +/* + * read instance config from file + */ +InstanceConfig * +readInstanceConfigFile(InstanceState *instanceState) +{ + InstanceConfig *instance = pgut_new(InstanceConfig); + char *log_level_console = NULL; + char *log_level_file = NULL; + char *log_format_console = NULL; + char *log_format_file = NULL; + char *compress_alg = NULL; + int parsed_options; + + ConfigOption instance_options[] = + { + /* Instance options */ + { + 's', 'D', "pgdata", + &instance->pgdata, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_INSTANCE_GROUP, 0, option_get_value + }, + { + 'U', 200, "system-identifier", + &instance->system_identifier, SOURCE_FILE_STRICT, 0, + OPTION_INSTANCE_GROUP, 0, option_get_value + }, + #if PG_VERSION_NUM >= 110000 + { + 'u', 201, "xlog-seg-size", + &instance->xlog_seg_size, SOURCE_FILE_STRICT, 0, + OPTION_INSTANCE_GROUP, 0, option_get_value + }, + #endif + { + 's', 'E', "external-dirs", + &instance->external_dir_str, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_INSTANCE_GROUP, 0, option_get_value + }, + /* Connection options */ + { + 's', 'd', "pgdatabase", + &instance->conn_opt.pgdatabase, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_CONN_GROUP, 0, option_get_value + }, + { + 's', 'h', "pghost", + &instance->conn_opt.pghost, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_CONN_GROUP, 0, option_get_value + }, + { + 's', 'p', "pgport", + &instance->conn_opt.pgport, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_CONN_GROUP, 0, option_get_value + }, + { + 's', 'U', "pguser", + &instance->conn_opt.pguser, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_CONN_GROUP, 0, option_get_value + }, + /* Replica options */ + { + 's', 202, "master-db", + &instance->master_conn_opt.pgdatabase, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REPLICA_GROUP, 0, option_get_value + }, + { + 's', 203, "master-host", + &instance->master_conn_opt.pghost, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REPLICA_GROUP, 0, option_get_value + }, + { + 's', 204, "master-port", + &instance->master_conn_opt.pgport, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REPLICA_GROUP, 0, option_get_value + }, + { + 's', 205, "master-user", + &instance->master_conn_opt.pguser, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REPLICA_GROUP, 0, option_get_value + }, + { + 'u', 206, "replica-timeout", + &instance->replica_timeout, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REPLICA_GROUP, OPTION_UNIT_S, option_get_value + }, + /* Archive options */ + { + 'u', 207, "archive-timeout", + &instance->archive_timeout, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_ARCHIVE_GROUP, OPTION_UNIT_S, option_get_value + }, + { + 's', 208, "archive-host", + &instance_config.archive.host, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_ARCHIVE_GROUP, 0, option_get_value + }, + { + 's', 209, "archive-port", + &instance_config.archive.port, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_ARCHIVE_GROUP, 0, option_get_value + }, + { + 's', 210, "archive-user", + &instance_config.archive.user, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_ARCHIVE_GROUP, 0, option_get_value + }, + { + 's', 211, "restore-command", + &instance->restore_command, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_ARCHIVE_GROUP, 0, option_get_value + }, + + /* Instance options */ + { + 's', 'D', "pgdata", + &instance->pgdata, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_INSTANCE_GROUP, 0, option_get_value + }, + + /* Logging options */ + { + 's', 212, "log-level-console", + &log_level_console, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_LOG_GROUP, 0, option_get_value + }, + { + 's', 213, "log-level-file", + &log_level_file, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_LOG_GROUP, 0, option_get_value + }, + { + 's', 214, "log-format-console", + &log_format_console, SOURCE_CMD_STRICT, SOURCE_DEFAULT, + OPTION_LOG_GROUP, 0, option_get_value + }, + { + 's', 215, "log-format-file", + &log_format_file, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_LOG_GROUP, 0, option_get_value + }, + { + 's', 216, "log-filename", + &instance->logger.log_filename, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_LOG_GROUP, 0, option_get_value + }, + { + 's', 217, "error-log-filename", + &instance->logger.error_log_filename, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_LOG_GROUP, 0, option_get_value + }, + { + 's', 218, "log-directory", + &instance->logger.log_directory, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_LOG_GROUP, 0, option_get_value + }, + { + 'U', 219, "log-rotation-size", + &instance->logger.log_rotation_size, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_LOG_GROUP, OPTION_UNIT_KB, option_get_value + }, + { + 'U', 220, "log-rotation-age", + &instance->logger.log_rotation_age, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_LOG_GROUP, OPTION_UNIT_MS, option_get_value + }, + /* Retention options */ + { + 'u', 221, "retention-redundancy", + &instance->retention_redundancy, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_RETENTION_GROUP, 0, option_get_value + }, + { + 'u', 222, "retention-window", + &instance->retention_window, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_RETENTION_GROUP, 0, option_get_value + }, + { + 'u', 223, "wal-depth", + &instance->wal_depth, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_RETENTION_GROUP, 0, option_get_value + }, + /* Compression options */ + { + 's', 224, "compress-algorithm", + &compress_alg, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_LOG_GROUP, 0, option_get_value + }, + { + 'u', 225, "compress-level", + &instance->compress_level, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_COMPRESS_GROUP, 0, option_get_value + }, + /* Remote backup options */ + { + 's', 226, "remote-proto", + &instance->remote.proto, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REMOTE_GROUP, 0, option_get_value + }, + { + 's', 227, "remote-host", + &instance->remote.host, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REMOTE_GROUP, 0, option_get_value + }, + { + 's', 228, "remote-port", + &instance->remote.port, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REMOTE_GROUP, 0, option_get_value + }, + { + 's', 229, "remote-path", + &instance->remote.path, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REMOTE_GROUP, 0, option_get_value + }, + { + 's', 230, "remote-user", + &instance->remote.user, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REMOTE_GROUP, 0, option_get_value + }, + { + 's', 231, "ssh-options", + &instance->remote.ssh_options, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REMOTE_GROUP, 0, option_get_value + }, + { + 's', 232, "ssh-config", + &instance->remote.ssh_config, SOURCE_CMD, SOURCE_DEFAULT, + OPTION_REMOTE_GROUP, 0, option_get_value + }, + { 0 } + }; + + + init_config(instance, instanceState->instance_name); + + if (fio_access(instanceState->instance_config_path, F_OK, FIO_BACKUP_HOST) != 0) + { + elog(WARNING, "Control file \"%s\" doesn't exist", instanceState->instance_config_path); + pfree(instance); + return NULL; + } + + parsed_options = config_read_opt(instanceState->instance_config_path, + instance_options, WARNING, true, true); + + if (parsed_options == 0) + { + elog(WARNING, "Control file \"%s\" is empty", instanceState->instance_config_path); + pfree(instance); + return NULL; + } + + if (log_level_console) + instance->logger.log_level_console = parse_log_level(log_level_console); + + if (log_level_file) + instance->logger.log_level_file = parse_log_level(log_level_file); + + if (log_format_console) + instance->logger.log_format_console = parse_log_format(log_format_console); + + if (log_format_file) + instance->logger.log_format_file = parse_log_format(log_format_file); + + if (compress_alg) + instance->compress_alg = parse_compress_alg(compress_alg); + +#if PG_VERSION_NUM >= 110000 + /* If for some reason xlog-seg-size is missing, then set it to 16MB */ + if (!instance->xlog_seg_size) + instance->xlog_seg_size = DEFAULT_XLOG_SEG_SIZE; +#endif + + return instance; +} + static void assign_log_level_console(ConfigOption *opt, const char *arg) { @@ -354,6 +684,18 @@ assign_log_level_file(ConfigOption *opt, const char *arg) instance_config.logger.log_level_file = parse_log_level(arg); } +static void +assign_log_format_console(ConfigOption *opt, const char *arg) +{ + instance_config.logger.log_format_console = parse_log_format(arg); +} + +static void +assign_log_format_file(ConfigOption *opt, const char *arg) +{ + instance_config.logger.log_format_file = parse_log_format(arg); +} + static void assign_compress_alg(ConfigOption *opt, const char *arg) { @@ -372,6 +714,18 @@ get_log_level_file(ConfigOption *opt) return pstrdup(deparse_log_level(instance_config.logger.log_level_file)); } +static char * +get_log_format_console(ConfigOption *opt) +{ + return pstrdup(deparse_log_format(instance_config.logger.log_format_console)); +} + +static char * +get_log_format_file(ConfigOption *opt) +{ + return pstrdup(deparse_log_format(instance_config.logger.log_format_file)); +} + static char * get_compress_alg(ConfigOption *opt) { @@ -450,6 +804,6 @@ show_configure_json(ConfigOption *opt) return; json_add_value(&show_buf, opt->lname, value, json_level, - true); + !(opt->flags & GET_VAL_IN_BASE_UNITS)); pfree(value); } diff --git a/src/data.c b/src/data.c index 2b9121ede..1a9616bae 100644 --- a/src/data.c +++ b/src/data.c @@ -3,7 +3,7 @@ * data.c: utils to parse and backup data pages * * Portions Copyright (c) 2009-2013, NIPPON TELEGRAPH AND TELEPHONE CORPORATION - * Portions Copyright (c) 2015-2019, Postgres Professional + * Portions Copyright (c) 2015-2022, Postgres Professional * *------------------------------------------------------------------------- */ @@ -25,21 +25,24 @@ #include "utils/thread.h" /* Union to ease operations on relation pages */ -typedef union DataPage +typedef struct DataPage { - PageHeaderData page_data; - char data[BLCKSZ]; + BackupPageHeader bph; + char data[BLCKSZ]; } DataPage; +static bool get_page_header(FILE *in, const char *fullpath, BackupPageHeader *bph, + pg_crc32 *crc, bool use_crc32c); + #ifdef HAVE_LIBZ /* Implementation of zlib compression method */ static int32 zlib_compress(void *dst, size_t dst_size, void const *src, size_t src_size, int level) { - uLongf compressed_size = dst_size; - int rc = compress2(dst, &compressed_size, src, src_size, - level); + uLongf compressed_size = dst_size; + int rc = compress2(dst, &compressed_size, src, src_size, + level); return rc == Z_OK ? compressed_size : rc; } @@ -48,8 +51,8 @@ zlib_compress(void *dst, size_t dst_size, void const *src, size_t src_size, static int32 zlib_decompress(void *dst, size_t dst_size, void const *src, size_t src_size) { - uLongf dest_len = dst_size; - int rc = uncompress(dst, &dest_len, src, src_size); + uLongf dest_len = dst_size; + int rc = uncompress(dst, &dest_len, src, src_size); return rc == Z_OK ? dest_len : rc; } @@ -60,7 +63,7 @@ zlib_decompress(void *dst, size_t dst_size, void const *src, size_t src_size) * written in the destination buffer, or -1 if compression fails. */ int32 -do_compress(void* dst, size_t dst_size, void const* src, size_t src_size, +do_compress(void *dst, size_t dst_size, void const *src, size_t src_size, CompressAlg alg, int level, const char **errormsg) { switch (alg) @@ -70,13 +73,13 @@ do_compress(void* dst, size_t dst_size, void const* src, size_t src_size, return -1; #ifdef HAVE_LIBZ case ZLIB_COMPRESS: - { - int32 ret; - ret = zlib_compress(dst, dst_size, src, src_size, level); - if (ret < Z_OK && errormsg) - *errormsg = zError(ret); - return ret; - } + { + int32 ret; + ret = zlib_compress(dst, dst_size, src, src_size, level); + if (ret < Z_OK && errormsg) + *errormsg = zError(ret); + return ret; + } #endif case PGLZ_COMPRESS: return pglz_compress(src, src_size, dst, PGLZ_strategy_always); @@ -89,35 +92,39 @@ do_compress(void* dst, size_t dst_size, void const* src, size_t src_size, * Decompresses source into dest using algorithm. Returns the number of bytes * decompressed in the destination buffer, or -1 if decompression fails. */ -static int32 -do_decompress(void* dst, size_t dst_size, void const* src, size_t src_size, +int32 +do_decompress(void *dst, size_t dst_size, void const *src, size_t src_size, CompressAlg alg, const char **errormsg) { switch (alg) { case NONE_COMPRESS: case NOT_DEFINED_COMPRESS: - if (errormsg) + if (errormsg) *errormsg = "Invalid compression algorithm"; return -1; #ifdef HAVE_LIBZ case ZLIB_COMPRESS: - { - int32 ret; - ret = zlib_decompress(dst, dst_size, src, src_size); - if (ret < Z_OK && errormsg) - *errormsg = zError(ret); - return ret; - } + { + int32 ret; + ret = zlib_decompress(dst, dst_size, src, src_size); + if (ret < Z_OK && errormsg) + *errormsg = zError(ret); + return ret; + } #endif case PGLZ_COMPRESS: + +#if PG_VERSION_NUM >= 120000 + return pglz_decompress(src, src_size, dst, dst_size, true); +#else return pglz_decompress(src, src_size, dst, dst_size); +#endif } return -1; } - #define ZLIB_MAGIC 0x78 /* @@ -135,7 +142,7 @@ page_may_be_compressed(Page page, CompressAlg alg, uint32 backup_version) phdr = (PageHeader) page; /* First check if page header is valid (it seems to be fast enough check) */ - if (!(PageGetPageSize(phdr) == BLCKSZ && + if (!(PageGetPageSize(page) == BLCKSZ && // PageGetPageLayoutVersion(phdr) == PG_PAGE_LAYOUT_VERSION && (phdr->pd_flags & ~PD_VALID_FLAG_BITS) == 0 && phdr->pd_lower >= SizeOfPageHeaderData && @@ -154,7 +161,7 @@ page_may_be_compressed(Page page, CompressAlg alg, uint32 backup_version) /* For zlib we can check page magic: * https://stackoverflow.com/questions/9050260/what-does-a-zlib-header-look-like */ - if (alg == ZLIB_COMPRESS && *(char*)page != ZLIB_MAGIC) + if (alg == ZLIB_COMPRESS && *(char *)page != ZLIB_MAGIC) { return false; } @@ -174,7 +181,7 @@ parse_page(Page page, XLogRecPtr *lsn) /* Get lsn from page header */ *lsn = PageXLogRecPtrGet(phdr->pd_lsn); - if (PageGetPageSize(phdr) == BLCKSZ && + if (PageGetPageSize(page) == BLCKSZ && // PageGetPageLayoutVersion(phdr) == PG_PAGE_LAYOUT_VERSION && (phdr->pd_flags & ~PD_VALID_FLAG_BITS) == 0 && phdr->pd_lower >= SizeOfPageHeaderData && @@ -187,95 +194,67 @@ parse_page(Page page, XLogRecPtr *lsn) return false; } -/* Read one page from file directly accessing disk - * return value: - * 0 - if the page is not found - * 1 - if the page is found and valid - * -1 - if the page is found but invalid +/* We know that header is invalid, store specific + * details in errormsg. */ -static int -read_page_from_file(pgFile *file, BlockNumber blknum, - FILE *in, Page page, XLogRecPtr *page_lsn, - uint32 checksum_version) +void +get_header_errormsg(Page page, char **errormsg) { - off_t offset = blknum * BLCKSZ; - ssize_t read_len = 0; - - /* read the block */ - read_len = fio_pread(in, page, offset); - - if (read_len != BLCKSZ) - { - /* The block could have been truncated. It is fine. */ - if (read_len == 0) - { - elog(LOG, "File %s, block %u, file was truncated", - file->path, blknum); - return 0; - } - else - { - elog(WARNING, "File: %s, block %u, expected block size %u," - "but read %zu, try again", - file->path, blknum, BLCKSZ, read_len); - return -1; - } - } - - /* - * If we found page with invalid header, at first check if it is zeroed, - * which is a valid state for page. If it is not, read it and check header - * again, because it's possible that we've read a partly flushed page. - * If after several attempts page header is still invalid, throw an error. - * The same idea is applied to checksum verification. - */ - if (!parse_page(page, page_lsn)) - { - int i; - /* Check if the page is zeroed. */ - for(i = 0; i < BLCKSZ && page[i] == 0; i++); + PageHeader phdr = (PageHeader) page; + *errormsg = pgut_malloc(ERRMSG_MAX_LEN); + + if (PageGetPageSize(page) != BLCKSZ) + snprintf(*errormsg, ERRMSG_MAX_LEN, "page header invalid, " + "page size %lu is not equal to block size %u", + PageGetPageSize(page), BLCKSZ); + + else if (phdr->pd_lower < SizeOfPageHeaderData) + snprintf(*errormsg, ERRMSG_MAX_LEN, "page header invalid, " + "pd_lower %i is less than page header size %lu", + phdr->pd_lower, SizeOfPageHeaderData); + + else if (phdr->pd_lower > phdr->pd_upper) + snprintf(*errormsg, ERRMSG_MAX_LEN, "page header invalid, " + "pd_lower %u is greater than pd_upper %u", + phdr->pd_lower, phdr->pd_upper); + + else if (phdr->pd_upper > phdr->pd_special) + snprintf(*errormsg, ERRMSG_MAX_LEN, "page header invalid, " + "pd_upper %u is greater than pd_special %u", + phdr->pd_upper, phdr->pd_special); + + else if (phdr->pd_special > BLCKSZ) + snprintf(*errormsg, ERRMSG_MAX_LEN, "page header invalid, " + "pd_special %u is greater than block size %u", + phdr->pd_special, BLCKSZ); + + else if (phdr->pd_special != MAXALIGN(phdr->pd_special)) + snprintf(*errormsg, ERRMSG_MAX_LEN, "page header invalid, " + "pd_special %i is misaligned, expected %lu", + phdr->pd_special, MAXALIGN(phdr->pd_special)); + + else if (phdr->pd_flags & ~PD_VALID_FLAG_BITS) + snprintf(*errormsg, ERRMSG_MAX_LEN, "page header invalid, " + "pd_flags mask contain illegal bits"); - /* Page is zeroed. No need to check header and checksum. */ - if (i == BLCKSZ) - { - elog(LOG, "File: %s blknum %u, empty page", file->path, blknum); - return 1; - } + else + snprintf(*errormsg, ERRMSG_MAX_LEN, "page header invalid"); +} - /* - * If page is not completely empty and we couldn't parse it, - * try again several times. If it didn't help, throw error - */ - elog(LOG, "File: %s blknum %u have wrong page header, try again", - file->path, blknum); - return -1; - } +/* We know that checksumms are mismatched, store specific + * details in errormsg. + */ +void +get_checksum_errormsg(Page page, char **errormsg, BlockNumber absolute_blkno) +{ + PageHeader phdr = (PageHeader) page; + *errormsg = pgut_malloc(ERRMSG_MAX_LEN); - /* Verify checksum */ - if (checksum_version) - { - BlockNumber blkno = file->segno * RELSEG_SIZE + blknum; - /* - * If checksum is wrong, sleep a bit and then try again - * several times. If it didn't help, throw error - */ - if (pg_checksum_page(page, blkno) != ((PageHeader) page)->pd_checksum) - { - elog(LOG, "File: %s blknum %u have wrong checksum, try again", - file->path, blknum); - return -1; - } - else - { - /* page header and checksum are correct */ - return 1; - } - } - else - { - /* page header is correct and checksum check is disabled */ - return 1; - } + snprintf(*errormsg, ERRMSG_MAX_LEN, + "page verification failed, " + "calculated checksum %u but expected %u", + phdr->pd_checksum, + pg_checksum_page(page, absolute_blkno)); } /* @@ -284,26 +263,30 @@ read_page_from_file(pgFile *file, BlockNumber blknum, * should be a pointer to allocated BLCKSZ of bytes. * * Prints appropriate warnings/errors/etc into log. - * Returns 0 if page was successfully retrieved - * SkipCurrentPage(-3) if we need to skip this page - * PageIsTruncated(-2) if the page was truncated - * PageIsCorrupted(-4) if the page check mismatch + * Returns: + * PageIsOk(0) if page was successfully retrieved + * PageIsTruncated(-1) if the page was truncated + * SkipCurrentPage(-2) if we need to skip this page, + * only used for DELTA and PTRACK backup + * PageIsCorrupted(-3) if the page checksum mismatch + * or header corruption, + * only used for checkdb + * TODO: probably we should always + * return it to the caller */ static int32 -prepare_page(ConnectionArgs *arguments, - pgFile *file, XLogRecPtr prev_backup_start_lsn, - BlockNumber blknum, BlockNumber nblocks, - FILE *in, BlockNumber *n_skipped, +prepare_page(pgFile *file, XLogRecPtr prev_backup_start_lsn, + BlockNumber blknum, FILE *in, BackupMode backup_mode, - Page page, - bool strict, - uint32 checksum_version) + Page page, bool strict, + uint32 checksum_version, + const char *from_fullpath, + PageState *page_st) { - XLogRecPtr page_lsn = 0; - int try_again = 100; + int try_again = PAGE_READ_ATTEMPTS; bool page_is_valid = false; - bool page_is_truncated = false; BlockNumber absolute_blknum = file->segno * RELSEG_SIZE + blknum; + int rc = 0; /* check for interrupt */ if (interrupted || thread_interrupted) @@ -314,202 +297,182 @@ prepare_page(ConnectionArgs *arguments, * Under high write load it's possible that we've read partly * flushed page, so try several times before throwing an error. */ - if (backup_mode != BACKUP_MODE_DIFF_PTRACK) + while (!page_is_valid && try_again--) { - while(!page_is_valid && try_again) - { - int result = read_page_from_file(file, blknum, in, page, - &page_lsn, checksum_version); + /* read the block */ + int read_len = fio_pread(in, page, blknum * BLCKSZ); - try_again--; - if (result == 0) - { - /* This block was truncated.*/ - page_is_truncated = true; - /* Page is not actually valid, but it is absent - * and we're not going to reread it or validate */ - page_is_valid = true; - } - - if (result == 1) - page_is_valid = true; - - /* - * If ptrack support is available use it to get invalid block - * instead of rereading it 99 times - */ - //elog(WARNING, "Checksum_Version: %i", checksum_version ? 1 : 0); - - if (result == -1 && is_ptrack_support && strict) - { - elog(WARNING, "File %s, block %u, try to fetch via SQL", - file->path, blknum); - break; - } - } - /* - * If page is not valid after 100 attempts to read it - * throw an error. - */ - - if (!page_is_valid && - ((strict && !is_ptrack_support) || !strict)) + /* The block could have been truncated. It is fine. */ + if (read_len == 0) { - /* show this message for checkdb or backup without ptrack support */ - elog(WARNING, "Corruption detected in file \"%s\", block %u", - file->path, blknum); + elog(VERBOSE, "Cannot read block %u of \"%s\": " + "block truncated", blknum, from_fullpath); + return PageIsTruncated; } + else if (read_len < 0) + elog(ERROR, "Cannot read block %u of \"%s\": %s", + blknum, from_fullpath, strerror(errno)); + else if (read_len != BLCKSZ) + elog(WARNING, "Cannot read block %u of \"%s\": " + "read %i of %d, try again", + blknum, from_fullpath, read_len, BLCKSZ); + else + { + /* We have BLCKSZ of raw data, validate it */ + rc = validate_one_page(page, absolute_blknum, + InvalidXLogRecPtr, page_st, + checksum_version); + switch (rc) + { + case PAGE_IS_ZEROED: + elog(VERBOSE, "File: \"%s\" blknum %u, empty page", from_fullpath, blknum); + return PageIsOk; + + case PAGE_IS_VALID: + /* in DELTA or PTRACK modes we must compare lsn */ + if (backup_mode == BACKUP_MODE_DIFF_DELTA || backup_mode == BACKUP_MODE_DIFF_PTRACK) + page_is_valid = true; + else + return PageIsOk; + break; - /* Backup with invalid block and without ptrack support must throw error */ - if (!page_is_valid && strict && !is_ptrack_support) - elog(ERROR, "Data file corruption. Canceling backup"); + case PAGE_HEADER_IS_INVALID: + elog(VERBOSE, "File: \"%s\" blknum %u have wrong page header, try again", + from_fullpath, blknum); + break; - /* Checkdb not going futher */ - if (!strict) - { - if (page_is_valid) - return 0; - else - return PageIsCorrupted; + case PAGE_CHECKSUM_MISMATCH: + elog(VERBOSE, "File: \"%s\" blknum %u have wrong checksum, try again", + from_fullpath, blknum); + break; + default: + Assert(false); + } } + /* avoid re-reading once buffered data, flushing on further attempts, see PBCKP-150 */ + fflush(in); } - if (backup_mode == BACKUP_MODE_DIFF_PTRACK || (!page_is_valid && is_ptrack_support)) + /* + * If page is not valid after PAGE_READ_ATTEMPTS attempts to read it + * throw an error. + */ + if (!page_is_valid) { - size_t page_size = 0; - Page ptrack_page = NULL; - ptrack_page = (Page) pg_ptrack_get_block(arguments, file->dbOid, file->tblspcOid, - file->relOid, absolute_blknum, &page_size); + int elevel = ERROR; + char *errormsg = NULL; + + /* Get the details of corruption */ + if (rc == PAGE_HEADER_IS_INVALID) + get_header_errormsg(page, &errormsg); + else if (rc == PAGE_CHECKSUM_MISMATCH) + get_checksum_errormsg(page, &errormsg, + file->segno * RELSEG_SIZE + blknum); + + /* Error out in case of merge or backup without ptrack support; + * issue warning in case of checkdb or backup with ptrack support + */ + if (!strict) + elevel = WARNING; - if (ptrack_page == NULL) - { - /* This block was truncated.*/ - page_is_truncated = true; - } - else if (page_size != BLCKSZ) - { - free(ptrack_page); - elog(ERROR, "File: %s, block %u, expected block size %d, but read %zu", - file->path, absolute_blknum, BLCKSZ, page_size); - } + if (errormsg) + elog(elevel, "Corruption detected in file \"%s\", block %u: %s", + from_fullpath, blknum, errormsg); else - { - /* - * We need to copy the page that was successfully - * retrieved from ptrack into our output "page" parameter. - * We must set checksum here, because it is outdated - * in the block recieved from shared buffers. - */ - memcpy(page, ptrack_page, BLCKSZ); - free(ptrack_page); - if (checksum_version) - ((PageHeader) page)->pd_checksum = pg_checksum_page(page, absolute_blknum); - } - /* get lsn from page, provided by pg_ptrack_get_block() */ - if (backup_mode == BACKUP_MODE_DIFF_DELTA && - file->exists_in_prev && - !page_is_truncated && - !parse_page(page, &page_lsn)) - elog(ERROR, "Cannot parse page after pg_ptrack_get_block. " - "Possible risk of a memory corruption"); + elog(elevel, "Corruption detected in file \"%s\", block %u", + from_fullpath, blknum); + pg_free(errormsg); + return PageIsCorrupted; } - /* Nullified pages must be copied by DELTA backup, just to be safe */ - if (backup_mode == BACKUP_MODE_DIFF_DELTA && + /* Checkdb not going futher */ + if (!strict) + return PageIsOk; + + /* + * Skip page if page lsn is less than START_LSN of parent backup. + * Nullified pages must be copied by DELTA backup, just to be safe. + */ + if ((backup_mode == BACKUP_MODE_DIFF_DELTA || backup_mode == BACKUP_MODE_DIFF_PTRACK) && file->exists_in_prev && - !page_is_truncated && - page_lsn && - page_lsn < prev_backup_start_lsn) + page_st->lsn > 0 && + page_st->lsn < prev_backup_start_lsn) { - elog(VERBOSE, "Skipping blknum: %u in file: %s", blknum, file->path); - (*n_skipped)++; + elog(VERBOSE, "Skipping blknum %u in file: \"%s\", file->exists_in_prev: %s, page_st->lsn: %X/%X, prev_backup_start_lsn: %X/%X", + blknum, from_fullpath, + file->exists_in_prev ? "true" : "false", + (uint32) (page_st->lsn >> 32), (uint32) page_st->lsn, + (uint32) (prev_backup_start_lsn >> 32), (uint32) prev_backup_start_lsn); return SkipCurrentPage; } - if (page_is_truncated) - return PageIsTruncated; - - return 0; + return PageIsOk; } -static void +/* split this function in two: compress() and backup() */ +static int compress_and_backup_page(pgFile *file, BlockNumber blknum, FILE *in, FILE *out, pg_crc32 *crc, int page_state, Page page, - CompressAlg calg, int clevel) + CompressAlg calg, int clevel, + const char *from_fullpath, const char *to_fullpath) { - BackupPageHeader header; - size_t write_buffer_size = sizeof(header); - char write_buffer[BLCKSZ+sizeof(header)]; - char compressed_page[BLCKSZ*2]; /* compressed page may require more space than uncompressed */ - - if(page_state == SkipCurrentPage) - return; - - header.block = blknum; - header.compressed_size = page_state; - - if(page_state == PageIsTruncated) + int compressed_size = 0; + size_t write_buffer_size = 0; + char write_buffer[BLCKSZ*2]; /* compressed page may require more space than uncompressed */ + BackupPageHeader* bph = (BackupPageHeader*)write_buffer; + const char *errormsg = NULL; + + /* Compress the page */ + compressed_size = do_compress(write_buffer + sizeof(BackupPageHeader), + sizeof(write_buffer) - sizeof(BackupPageHeader), + page, BLCKSZ, calg, clevel, + &errormsg); + /* Something went wrong and errormsg was assigned, throw a warning */ + if (compressed_size < 0 && errormsg != NULL) + elog(WARNING, "An error occured during compressing block %u of file \"%s\": %s", + blknum, from_fullpath, errormsg); + + file->compress_alg = calg; /* TODO: wtf? why here? */ + + /* compression didn`t worked */ + if (compressed_size <= 0 || compressed_size >= BLCKSZ) { - /* - * The page was truncated. Write only header - * to know that we must truncate restored file - */ - memcpy(write_buffer, &header, sizeof(header)); - } - else - { - const char *errormsg = NULL; - - /* The page was not truncated, so we need to compress it */ - header.compressed_size = do_compress(compressed_page, sizeof(compressed_page), - page, BLCKSZ, calg, clevel, - &errormsg); - /* Something went wrong and errormsg was assigned, throw a warning */ - if (header.compressed_size < 0 && errormsg != NULL) - elog(WARNING, "An error occured during compressing block %u of file \"%s\": %s", - blknum, file->path, errormsg); - - file->compress_alg = calg; - file->read_size += BLCKSZ; - - /* The page was successfully compressed. */ - if (header.compressed_size > 0 && header.compressed_size < BLCKSZ) - { - memcpy(write_buffer, &header, sizeof(header)); - memcpy(write_buffer + sizeof(header), - compressed_page, header.compressed_size); - write_buffer_size += MAXALIGN(header.compressed_size); - } - /* Non-positive value means that compression failed. Write it as is. */ - else - { - header.compressed_size = BLCKSZ; - memcpy(write_buffer, &header, sizeof(header)); - memcpy(write_buffer + sizeof(header), page, BLCKSZ); - write_buffer_size += header.compressed_size; - } + /* Do not compress page */ + memcpy(write_buffer + sizeof(BackupPageHeader), page, BLCKSZ); + compressed_size = BLCKSZ; } - - /* elog(VERBOSE, "backup blkno %u, compressed_size %d write_buffer_size %ld", - blknum, header.compressed_size, write_buffer_size); */ + bph->block = blknum; + bph->compressed_size = compressed_size; + write_buffer_size = compressed_size + sizeof(BackupPageHeader); /* Update CRC */ COMP_FILE_CRC32(true, *crc, write_buffer, write_buffer_size); /* write data page */ if (fio_fwrite(out, write_buffer, write_buffer_size) != write_buffer_size) - { - int errno_tmp = errno; - - fclose(in); - fio_fclose(out); - elog(ERROR, "File: %s, cannot write backup at block %u: %s", - file->path, blknum, strerror(errno_tmp)); - } + elog(ERROR, "File: \"%s\", cannot write at block %u: %s", + to_fullpath, blknum, strerror(errno)); file->write_size += write_buffer_size; + file->uncompressed_size += BLCKSZ; + + return compressed_size; +} + +/* Write page as-is. TODO: make it fastpath option in compress_and_backup_page() */ +static int +write_page(pgFile *file, FILE *out, Page page) +{ + /* write data page */ + if (fio_fwrite(out, page, BLCKSZ) != BLCKSZ) + return -1; + + file->write_size += BLCKSZ; + file->uncompressed_size += BLCKSZ; + + return BLCKSZ; } /* @@ -520,20 +483,29 @@ compress_and_backup_page(pgFile *file, BlockNumber blknum, * incremental backup), validate checksum, optionally compress and write to * backup with special header. */ -bool -backup_data_file(backup_files_arg* arguments, - const char *to_path, pgFile *file, +void +backup_data_file(pgFile *file, const char *from_fullpath, const char *to_fullpath, XLogRecPtr prev_backup_start_lsn, BackupMode backup_mode, - CompressAlg calg, int clevel, bool missing_ok) + CompressAlg calg, int clevel, uint32 checksum_version, + HeaderMap *hdr_map, bool is_merge) { - FILE *in; - FILE *out; - BlockNumber blknum = 0; - BlockNumber nblocks = 0; - BlockNumber n_blocks_skipped = 0; - BlockNumber n_blocks_read = 0; - int page_state; - char curr_page[BLCKSZ]; + int rc; + bool use_pagemap; + char *errmsg = NULL; + BlockNumber err_blknum = 0; + /* page headers */ + BackupPageHeader2 *headers = NULL; + + /* sanity */ + if (file->size % BLCKSZ != 0) + elog(WARNING, "File: \"%s\", invalid file size %zu", from_fullpath, file->size); + + /* + * Compute expected number of blocks in the file. + * NOTE This is a normal situation, if the file size has changed + * since the moment we computed it. + */ + file->n_blocks = file->size/BLCKSZ; /* * Skip unchanged file only if it exists in previous backup. @@ -546,280 +518,637 @@ backup_data_file(backup_files_arg* arguments, file->exists_in_prev && !file->pagemap_isabsent) { /* - * There are no changed blocks since last backup. We want make + * There are no changed blocks since last backup. We want to make * incremental backup, so we should exit. */ - elog(VERBOSE, "Skipping the unchanged file: %s", file->path); - return false; + file->write_size = BYTES_INVALID; + return; } /* reset size summary */ file->read_size = 0; file->write_size = 0; + file->uncompressed_size = 0; INIT_FILE_CRC32(true, file->crc); - /* open backup mode file for read */ - in = fio_fopen(file->path, PG_BINARY_R, FIO_DB_HOST); - if (in == NULL) - { - FIN_FILE_CRC32(true, file->crc); - - /* - * If file is not found, this is not en error. - * It could have been deleted by concurrent postgres transaction. - */ - if (errno == ENOENT) - { - if (missing_ok) - { - elog(LOG, "File \"%s\" is not found", file->path); - file->write_size = FILE_NOT_FOUND; - return false; - } - else - elog(ERROR, "File \"%s\" is not found", file->path); - } - - elog(ERROR, "cannot open file \"%s\": %s", - file->path, strerror(errno)); - } - - if (file->size % BLCKSZ != 0) - elog(WARNING, "File: \"%s\", invalid file size %zu", file->path, file->size); - - /* - * Compute expected number of blocks in the file. - * NOTE This is a normal situation, if the file size has changed - * since the moment we computed it. - */ - nblocks = file->size/BLCKSZ; - - /* open backup file for write */ - out = fio_fopen(to_path, PG_BINARY_W, FIO_BACKUP_HOST); - if (out == NULL) - { - int errno_tmp = errno; - fio_fclose(in); - elog(ERROR, "cannot open backup file \"%s\": %s", - to_path, strerror(errno_tmp)); - } - /* * Read each page, verify checksum and write it to backup. * If page map is empty or file is not present in previous backup * backup all pages of the relation. * - * We will enter here if backup_mode is FULL or DELTA. + * In PTRACK 1.x there was a problem + * of data files with missing _ptrack map. + * Such files should be fully copied. */ + if (file->pagemap.bitmapsize == PageBitmapIsEmpty || - file->pagemap_isabsent || !file->exists_in_prev) + file->pagemap_isabsent || !file->exists_in_prev || + !file->pagemap.bitmap) + use_pagemap = false; + else + use_pagemap = true; + + /* Remote mode */ + if (fio_is_remote(FIO_DB_HOST)) { - if (backup_mode != BACKUP_MODE_DIFF_PTRACK && fio_is_remote_file(in)) - { - int rc = fio_send_pages(in, out, file, - backup_mode == BACKUP_MODE_DIFF_DELTA && file->exists_in_prev ? prev_backup_start_lsn : InvalidXLogRecPtr, - &n_blocks_skipped, calg, clevel); - if (rc == PAGE_CHECKSUM_MISMATCH && is_ptrack_support) - goto RetryUsingPtrack; - if (rc < 0) - elog(ERROR, "Failed to read file \"%s\": %s", - file->path, rc == PAGE_CHECKSUM_MISMATCH ? "data file checksum mismatch" : strerror(-rc)); - n_blocks_read = rc; - } - else - { - RetryUsingPtrack: - for (blknum = 0; blknum < nblocks; blknum++) - { - page_state = prepare_page(&(arguments->conn_arg), file, prev_backup_start_lsn, - blknum, nblocks, in, &n_blocks_skipped, - backup_mode, curr_page, true, current.checksum_version); - compress_and_backup_page(file, blknum, in, out, &(file->crc), - page_state, curr_page, calg, clevel); - n_blocks_read++; - if (page_state == PageIsTruncated) - break; - } - } - if (backup_mode == BACKUP_MODE_DIFF_DELTA) - file->n_blocks = n_blocks_read; + rc = fio_send_pages(to_fullpath, from_fullpath, file, + /* send prev backup START_LSN */ + (backup_mode == BACKUP_MODE_DIFF_DELTA || backup_mode == BACKUP_MODE_DIFF_PTRACK) && + file->exists_in_prev ? prev_backup_start_lsn : InvalidXLogRecPtr, + calg, clevel, checksum_version, + /* send pagemap if any */ + use_pagemap, + /* variables for error reporting */ + &err_blknum, &errmsg, &headers); } - /* - * If page map is not empty we scan only changed blocks. - * - * We will enter here if backup_mode is PAGE or PTRACK. - */ else { - datapagemap_iterator_t *iter; - iter = datapagemap_iterate(&file->pagemap); - while (datapagemap_next(iter, &blknum)) - { - page_state = prepare_page(&(arguments->conn_arg), file, prev_backup_start_lsn, - blknum, nblocks, in, &n_blocks_skipped, - backup_mode, curr_page, true, current.checksum_version); - compress_and_backup_page(file, blknum, in, out, &(file->crc), - page_state, curr_page, calg, clevel); - n_blocks_read++; - if (page_state == PageIsTruncated) - break; - } + /* TODO: stop handling errors internally */ + rc = send_pages(to_fullpath, from_fullpath, file, + /* send prev backup START_LSN */ + (backup_mode == BACKUP_MODE_DIFF_DELTA || backup_mode == BACKUP_MODE_DIFF_PTRACK) && + file->exists_in_prev ? prev_backup_start_lsn : InvalidXLogRecPtr, + calg, clevel, checksum_version, use_pagemap, + &headers, backup_mode); + } - pg_free(file->pagemap.bitmap); - pg_free(iter); + /* check for errors */ + if (rc == FILE_MISSING) + { + elog(is_merge ? ERROR : LOG, "File not found: \"%s\"", from_fullpath); + file->write_size = FILE_NOT_FOUND; + goto cleanup; } - /* update file permission */ - if (fio_chmod(to_path, FILE_PERMISSION, FIO_BACKUP_HOST) == -1) + else if (rc == WRITE_FAILED) + elog(ERROR, "Cannot write block %u of \"%s\": %s", + err_blknum, to_fullpath, strerror(errno)); + + else if (rc == PAGE_CORRUPTION) + { + if (errmsg) + elog(ERROR, "Corruption detected in file \"%s\", block %u: %s", + from_fullpath, err_blknum, errmsg); + else + elog(ERROR, "Corruption detected in file \"%s\", block %u", + from_fullpath, err_blknum); + } + /* OPEN_FAILED and READ_FAILED */ + else if (rc == OPEN_FAILED) + { + if (errmsg) + elog(ERROR, "%s", errmsg); + else + elog(ERROR, "Cannot open file \"%s\"", from_fullpath); + } + else if (rc == READ_FAILED) { - int errno_tmp = errno; - fio_fclose(in); - fio_fclose(out); - elog(ERROR, "cannot change mode of \"%s\": %s", file->path, - strerror(errno_tmp)); + if (errmsg) + elog(ERROR, "%s", errmsg); + else + elog(ERROR, "Cannot read file \"%s\"", from_fullpath); } - if (fio_fflush(out) != 0 || - fio_fclose(out)) - elog(ERROR, "cannot write backup file \"%s\": %s", - to_path, strerror(errno)); - fio_fclose(in); + file->read_size = rc * BLCKSZ; - FIN_FILE_CRC32(true, file->crc); + /* refresh n_blocks for FULL and DELTA */ + if (backup_mode == BACKUP_MODE_FULL || + backup_mode == BACKUP_MODE_DIFF_DELTA) + file->n_blocks = file->read_size / BLCKSZ; - /* - * If we have pagemap then file in the backup can't be a zero size. - * Otherwise, we will clear the last file. - */ - if (n_blocks_read != 0 && n_blocks_read == n_blocks_skipped) + /* Determine that file didn`t changed in case of incremental backup */ + if (backup_mode != BACKUP_MODE_FULL && + file->exists_in_prev && + file->write_size == 0 && + file->n_blocks > 0) { - if (fio_unlink(to_path, FIO_BACKUP_HOST) == -1) - elog(ERROR, "cannot remove file \"%s\": %s", to_path, - strerror(errno)); - return false; + file->write_size = BYTES_INVALID; } - return true; +cleanup: + + /* finish CRC calculation */ + FIN_FILE_CRC32(true, file->crc); + + /* dump page headers */ + write_page_headers(headers, file, hdr_map, is_merge); + + pg_free(errmsg); + pg_free(file->pagemap.bitmap); + pg_free(headers); } /* - * Restore files in the from_root directory to the to_root directory with - * same relative path. - * - * If write_header is true then we add header to each restored block, currently - * it is used for MERGE command. + * Catchup data file in the from_root directory to the to_root directory with + * same relative path. If sync_lsn is not NULL, only pages with equal or + * higher lsn will be copied. + * Not just copy file, but read it block by block (use bitmap in case of + * incremental catchup), validate page checksum. */ void -restore_data_file(const char *to_path, pgFile *file, bool allow_truncate, - bool write_header, uint32 backup_version) +catchup_data_file(pgFile *file, const char *from_fullpath, const char *to_fullpath, + XLogRecPtr sync_lsn, BackupMode backup_mode, + uint32 checksum_version, size_t prev_size) { - FILE *in = NULL; - FILE *out = NULL; - BackupPageHeader header; - BlockNumber blknum = 0, - truncate_from = 0; - bool need_truncate = false; - - /* BYTES_INVALID allowed only in case of restoring file from DELTA backup */ - if (file->write_size != BYTES_INVALID) - { - /* open backup mode file for read */ - in = fopen(file->path, PG_BINARY_R); - if (in == NULL) - { - elog(ERROR, "Cannot open backup file \"%s\": %s", file->path, - strerror(errno)); - } - } + int rc; + bool use_pagemap; + char *errmsg = NULL; + BlockNumber err_blknum = 0; /* - * Open backup file for write. We use "r+" at first to overwrite only - * modified pages for differential restore. If the file does not exist, - * re-open it with "w" to create an empty file. + * Compute expected number of blocks in the file. + * NOTE This is a normal situation, if the file size has changed + * since the moment we computed it. */ - out = fio_fopen(to_path, PG_BINARY_R "+", FIO_DB_HOST); - if (out == NULL) - { - int errno_tmp = errno; - fclose(in); - elog(ERROR, "Cannot open restore target file \"%s\": %s", - to_path, strerror(errno_tmp)); - } + file->n_blocks = file->size/BLCKSZ; - while (true) + /* + * Skip unchanged file only if it exists in destination directory. + * This way we can correctly handle null-sized files which are + * not tracked by pagemap and thus always marked as unchanged. + */ + if (backup_mode == BACKUP_MODE_DIFF_PTRACK && + file->pagemap.bitmapsize == PageBitmapIsEmpty && + file->exists_in_prev && file->size == prev_size && !file->pagemap_isabsent) { - off_t write_pos; - size_t read_len; - DataPage compressed_page; /* used as read buffer */ - DataPage page; - int32 uncompressed_size = 0; - - /* File didn`t changed. Nothing to copy */ - if (file->write_size == BYTES_INVALID) - break; - /* - * We need to truncate result file if data file in an incremental backup - * less than data file in a full backup. We know it thanks to n_blocks. - * - * It may be equal to -1, then we don't want to truncate the result - * file. + * There are none changed pages. */ - if (file->n_blocks != BLOCKNUM_INVALID && - (blknum + 1) > file->n_blocks) - { - truncate_from = blknum; - need_truncate = true; - break; + file->write_size = BYTES_INVALID; + return; + } + + /* reset size summary */ + file->read_size = 0; + file->write_size = 0; + file->uncompressed_size = 0; + + /* + * If page map is empty or file is not present in destination directory, + * then copy backup all pages of the relation. + */ + + if (file->pagemap.bitmapsize == PageBitmapIsEmpty || + file->pagemap_isabsent || !file->exists_in_prev || + !file->pagemap.bitmap) + use_pagemap = false; + else + use_pagemap = true; + + if (use_pagemap) + elog(LOG, "Using pagemap for file \"%s\"", file->rel_path); + + /* Remote mode */ + if (fio_is_remote(FIO_DB_HOST)) + { + rc = fio_copy_pages(to_fullpath, from_fullpath, file, + /* send prev backup START_LSN */ + ((backup_mode == BACKUP_MODE_DIFF_DELTA || backup_mode == BACKUP_MODE_DIFF_PTRACK) && + file->exists_in_prev) ? sync_lsn : InvalidXLogRecPtr, + NONE_COMPRESS, 1, checksum_version, + /* send pagemap if any */ + use_pagemap, + /* variables for error reporting */ + &err_blknum, &errmsg); + } + else + { + /* TODO: stop handling errors internally */ + rc = copy_pages(to_fullpath, from_fullpath, file, + /* send prev backup START_LSN */ + ((backup_mode == BACKUP_MODE_DIFF_DELTA || backup_mode == BACKUP_MODE_DIFF_PTRACK) && + file->exists_in_prev) ? sync_lsn : InvalidXLogRecPtr, + checksum_version, use_pagemap, backup_mode); + } + + /* check for errors */ + if (rc == FILE_MISSING) + { + elog(LOG, "File not found: \"%s\"", from_fullpath); + file->write_size = FILE_NOT_FOUND; + goto cleanup; + } + + else if (rc == WRITE_FAILED) + elog(ERROR, "Cannot write block %u of \"%s\": %s", + err_blknum, to_fullpath, strerror(errno)); + + else if (rc == PAGE_CORRUPTION) + { + if (errmsg) + elog(ERROR, "Corruption detected in file \"%s\", block %u: %s", + from_fullpath, err_blknum, errmsg); + else + elog(ERROR, "Corruption detected in file \"%s\", block %u", + from_fullpath, err_blknum); + } + /* OPEN_FAILED and READ_FAILED */ + else if (rc == OPEN_FAILED) + { + if (errmsg) + elog(ERROR, "%s", errmsg); + else + elog(ERROR, "Cannot open file \"%s\"", from_fullpath); + } + else if (rc == READ_FAILED) + { + if (errmsg) + elog(ERROR, "%s", errmsg); + else + elog(ERROR, "Cannot read file \"%s\"", from_fullpath); + } + + file->read_size = rc * BLCKSZ; + + /* Determine that file didn`t changed in case of incremental catchup */ + if (backup_mode != BACKUP_MODE_FULL && + file->exists_in_prev && + file->write_size == 0 && + file->n_blocks > 0) + { + file->write_size = BYTES_INVALID; + } + +cleanup: + pg_free(errmsg); + pg_free(file->pagemap.bitmap); +} + +/* + * Backup non data file + * We do not apply compression to this file. + * If file exists in previous backup, then compare checksums + * and make a decision about copying or skiping the file. + */ +void +backup_non_data_file(pgFile *file, pgFile *prev_file, + const char *from_fullpath, const char *to_fullpath, + BackupMode backup_mode, time_t parent_backup_time, + bool missing_ok) +{ + /* special treatment for global/pg_control */ + if (file->external_dir_num == 0 && strcmp(file->rel_path, XLOG_CONTROL_FILE) == 0) + { + copy_pgcontrol_file(from_fullpath, FIO_DB_HOST, + to_fullpath, FIO_BACKUP_HOST, file); + return; + } + + /* + * If non-data file exists in previous backup + * and its mtime is less than parent backup start time ... */ + if ((pg_strcasecmp(file->name, RELMAPPER_FILENAME) != 0) && + (prev_file && file->exists_in_prev && + file->size == prev_file->size && + file->mtime <= parent_backup_time)) + { + /* + * file could be deleted under our feets. + * But then backup_non_data_file_internal will handle it safely + */ + if (file->forkName != cfm) + file->crc = fio_get_crc32(from_fullpath, FIO_DB_HOST, false, true); + else + file->crc = fio_get_crc32_truncated(from_fullpath, FIO_DB_HOST, true); + + /* ...and checksum is the same... */ + if (EQ_TRADITIONAL_CRC32(file->crc, prev_file->crc)) + { + file->write_size = BYTES_INVALID; + /* get full size from previous backup for unchanged file */ + file->uncompressed_size = prev_file->uncompressed_size; + return; /* ...skip copying file. */ } + } + + backup_non_data_file_internal(from_fullpath, FIO_DB_HOST, + to_fullpath, file, missing_ok); +} + +/* + * Iterate over parent backup chain and lookup given destination file in + * filelist of every chain member starting with FULL backup. + * Apply changed blocks to destination file from every backup in parent chain. + */ +size_t +restore_data_file(parray *parent_chain, pgFile *dest_file, FILE *out, + const char *to_fullpath, bool use_bitmap, PageState *checksum_map, + XLogRecPtr shift_lsn, datapagemap_t *lsn_map, bool use_headers) +{ + size_t total_write_len = 0; + char *in_buf = pgut_malloc(STDIO_BUFSIZE); + int backup_seq = 0; - /* read BackupPageHeader */ - read_len = fread(&header, 1, sizeof(header), in); - if (read_len != sizeof(header)) + /* + * FULL -> INCR -> DEST + * 2 1 0 + * Restore of backups of older versions cannot be optimized with bitmap + * because of n_blocks + */ + if (use_bitmap) + /* start with dest backup */ + backup_seq = 0; + else + /* start with full backup */ + backup_seq = parray_num(parent_chain) - 1; + +// for (i = parray_num(parent_chain) - 1; i >= 0; i--) +// for (i = 0; i < parray_num(parent_chain); i++) + while (backup_seq >= 0 && backup_seq < parray_num(parent_chain)) + { + char from_root[MAXPGPATH]; + char from_fullpath[MAXPGPATH]; + FILE *in = NULL; + + pgFile **res_file = NULL; + pgFile *tmp_file = NULL; + + /* page headers */ + BackupPageHeader2 *headers = NULL; + + pgBackup *backup = (pgBackup *) parray_get(parent_chain, backup_seq); + + if (use_bitmap) + backup_seq++; + else + backup_seq--; + + /* lookup file in intermediate backup */ + res_file = parray_bsearch(backup->files, dest_file, pgFileCompareRelPathWithExternal); + tmp_file = (res_file) ? *res_file : NULL; + + /* Destination file is not exists yet at this moment */ + if (tmp_file == NULL) + continue; + + /* + * Skip file if it haven't changed since previous backup + * and thus was not backed up. + */ + if (tmp_file->write_size == BYTES_INVALID) + continue; + + /* If file was truncated in intermediate backup, + * it is ok not to truncate it now, because old blocks will be + * overwritten by new blocks from next backup. + */ + if (tmp_file->write_size == 0) + continue; + + /* + * At this point we are sure, that something is going to be copied + * Open source file. + */ + join_path_components(from_root, backup->root_dir, DATABASE_DIR); + join_path_components(from_fullpath, from_root, tmp_file->rel_path); + + in = fopen(from_fullpath, PG_BINARY_R); + if (in == NULL) + elog(ERROR, "Cannot open backup file \"%s\": %s", from_fullpath, + strerror(errno)); + + /* set stdio buffering for input data file */ + setvbuf(in, in_buf, _IOFBF, STDIO_BUFSIZE); + + /* get headers for this file */ + if (use_headers && tmp_file->n_headers > 0) + headers = get_data_file_headers(&(backup->hdr_map), tmp_file, + parse_program_version(backup->program_version), + true); + + if (use_headers && !headers && tmp_file->n_headers > 0) + elog(ERROR, "Failed to get page headers for file \"%s\"", from_fullpath); + + /* + * Restore the file. + * Datafiles are backed up block by block and every block + * have BackupPageHeader with meta information, so we cannot just + * copy the file from backup. + */ + total_write_len += restore_data_file_internal(in, out, tmp_file, + parse_program_version(backup->program_version), + from_fullpath, to_fullpath, dest_file->n_blocks, + use_bitmap ? &(dest_file)->pagemap : NULL, + checksum_map, backup->checksum_version, + /* shiftmap can be used only if backup state precedes the shift */ + backup->stop_lsn <= shift_lsn ? lsn_map : NULL, + headers); + + if (fclose(in) != 0) + elog(ERROR, "Cannot close file \"%s\": %s", from_fullpath, + strerror(errno)); + + pg_free(headers); + +// datapagemap_print_debug(&(dest_file)->pagemap); + } + pg_free(in_buf); + + return total_write_len; +} + +/* Restore block from "in" file to "out" file. + * If "nblocks" is greater than zero, then skip restoring blocks, + * whose position if greater than "nblocks". + * If map is NULL, then page bitmap cannot be used for restore optimization + * Page bitmap optimize restore of incremental chains, consisting of more than one + * backup. We restoring from newest to oldest and page, once restored, marked in map. + * When the same page, but in older backup, encountered, we check the map, if it is + * marked as already restored, then page is skipped. + */ +size_t +restore_data_file_internal(FILE *in, FILE *out, pgFile *file, uint32 backup_version, + const char *from_fullpath, const char *to_fullpath, int nblocks, + datapagemap_t *map, PageState *checksum_map, int checksum_version, + datapagemap_t *lsn_map, BackupPageHeader2 *headers) +{ + BlockNumber blknum = 0; + int n_hdr = -1; + size_t write_len = 0; + off_t cur_pos_out = 0; + off_t cur_pos_in = 0; + + /* should not be possible */ + Assert(!(backup_version >= 20400 && file->n_headers <= 0)); + + /* + * We rely on stdio buffering of input and output. + * For buffering to be efficient, we try to minimize the + * number of lseek syscalls, because it forces buffer flush. + * For that, we track current write position in + * output file and issue fseek only when offset of block to be + * written not equal to current write position, which happens + * a lot when blocks from incremental backup are restored, + * but should never happen in case of blocks from FULL backup. + */ + if (fio_fseek(out, cur_pos_out) < 0) + elog(ERROR, "Cannot seek block %u of \"%s\": %s", + blknum, to_fullpath, strerror(errno)); + + for (;;) + { + off_t write_pos; + size_t len; + size_t read_len; + DataPage page; + int32 compressed_size = 0; + bool is_compressed = false; + + /* incremental restore vars */ + uint16 page_crc = 0; + XLogRecPtr page_lsn = InvalidXLogRecPtr; + + /* check for interrupt */ + if (interrupted || thread_interrupted) + elog(ERROR, "Interrupted during data file restore"); + + /* newer backups have headers in separate storage */ + if (headers) + { + n_hdr++; + if (n_hdr >= file->n_headers) + break; + + blknum = headers[n_hdr].block; + page_lsn = headers[n_hdr].lsn; + page_crc = headers[n_hdr].checksum; + /* calculate payload size by comparing current and next page positions, + * page header is not included */ + compressed_size = headers[n_hdr+1].pos - headers[n_hdr].pos - sizeof(BackupPageHeader); + + Assert(compressed_size > 0); + Assert(compressed_size <= BLCKSZ); + + read_len = compressed_size + sizeof(BackupPageHeader); + } + else { - int errno_tmp = errno; - if (read_len == 0 && feof(in)) - break; /* EOF found */ - else if (read_len != 0 && feof(in)) - elog(ERROR, - "Odd size page found at block %u of \"%s\"", - blknum, file->path); + /* We get into this function either when restoring old backup + * or when merging something. Align read_len only when restoring + * or merging old backups. + */ + if (get_page_header(in, from_fullpath, &(page).bph, NULL, false)) + { + cur_pos_in += sizeof(BackupPageHeader); + + /* backward compatibility kludge TODO: remove in 3.0 */ + blknum = page.bph.block; + compressed_size = page.bph.compressed_size; + + /* this has a potential to backfire when retrying merge of old backups, + * so we just forbid the retrying of failed merges between versions >= 2.4.0 and + * version < 2.4.0 + */ + if (backup_version >= 20400) + read_len = compressed_size; + else + /* For some unknown and possibly dump reason I/O operations + * in versions < 2.4.0 were always aligned to 8 bytes. + * Now we have to deal with backward compatibility. + */ + read_len = MAXALIGN(compressed_size); + } else - elog(ERROR, "Cannot read header of block %u of \"%s\": %s", - blknum, file->path, strerror(errno_tmp)); + break; } - if (header.block == 0 && header.compressed_size == 0) + /* + * Backward compatibility kludge: in the good old days + * n_blocks attribute was available only in DELTA backups. + * File truncate in PAGE and PTRACK happened on the fly when + * special value PageIsTruncated is encountered. + * It was inefficient. + * + * Nowadays every backup type has n_blocks, so instead of + * writing and then truncating redundant data, writing + * is not happening in the first place. + * TODO: remove in 3.0.0 + */ + if (compressed_size == PageIsTruncated) { - elog(VERBOSE, "Skip empty block of \"%s\"", file->path); - continue; + /* + * Block header contains information that this block was truncated. + * We need to truncate file to this length. + */ + + elog(VERBOSE, "Truncate file \"%s\" to block %u", to_fullpath, blknum); + + /* To correctly truncate file, we must first flush STDIO buffers */ + if (fio_fflush(out) != 0) + elog(ERROR, "Cannot flush file \"%s\": %s", to_fullpath, strerror(errno)); + + /* Set position to the start of file */ + if (fio_fseek(out, 0) < 0) + elog(ERROR, "Cannot seek to the start of file \"%s\": %s", to_fullpath, strerror(errno)); + + if (fio_ftruncate(out, blknum * BLCKSZ) != 0) + elog(ERROR, "Cannot truncate file \"%s\": %s", to_fullpath, strerror(errno)); + + break; } - if (header.block < blknum) - elog(ERROR, "Backup is broken at block %u of \"%s\"", - blknum, file->path); + Assert(compressed_size > 0); + Assert(compressed_size <= BLCKSZ); + + /* no point in writing redundant data */ + if (nblocks > 0 && blknum >= nblocks) + break; + + if (compressed_size > BLCKSZ) + elog(ERROR, "Size of a blknum %i exceed BLCKSZ: %i", blknum, compressed_size); - blknum = header.block; + /* Incremental restore in LSN mode */ + if (map && lsn_map && datapagemap_is_set(lsn_map, blknum)) + datapagemap_add(map, blknum); - if (header.compressed_size == PageIsTruncated) + if (map && checksum_map && checksum_map[blknum].checksum != 0) { + //elog(INFO, "HDR CRC: %u, MAP CRC: %u", page_crc, checksum_map[blknum].checksum); /* - * Backup contains information that this block was truncated. - * We need to truncate file to this length. + * The heart of incremental restore in CHECKSUM mode + * If page in backup has the same checksum and lsn as + * page in backup, then page can be skipped. */ - truncate_from = blknum; - need_truncate = true; - break; + if (page_crc == checksum_map[blknum].checksum && + page_lsn == checksum_map[blknum].lsn) + { + datapagemap_add(map, blknum); + } + } + + /* if this page is marked as already restored, then skip it */ + if (map && datapagemap_is_set(map, blknum)) + { + /* Backward compatibility kludge TODO: remove in 3.0 + * go to the next page. + */ + if (!headers && fseek(in, read_len, SEEK_CUR) != 0) + elog(ERROR, "Cannot seek block %u of \"%s\": %s", + blknum, from_fullpath, strerror(errno)); + continue; } - Assert(header.compressed_size <= BLCKSZ); + if (headers && + cur_pos_in != headers[n_hdr].pos) + { + if (fseek(in, headers[n_hdr].pos, SEEK_SET) != 0) + elog(ERROR, "Cannot seek to offset %u of \"%s\": %s", + headers[n_hdr].pos, from_fullpath, strerror(errno)); + + cur_pos_in = headers[n_hdr].pos; + } /* read a page from file */ - read_len = fread(compressed_page.data, 1, - MAXALIGN(header.compressed_size), in); - if (read_len != MAXALIGN(header.compressed_size)) - elog(ERROR, "Cannot read block %u of \"%s\" read %zu of %d", - blknum, file->path, read_len, header.compressed_size); + if (headers) + len = fread(&page, 1, read_len, in); + else + len = fread(page.data, 1, read_len, in); + + if (len != read_len) + elog(ERROR, "Cannot read block %u file \"%s\": %s", + blknum, from_fullpath, strerror(errno)); + + cur_pos_in += read_len; /* * if page size is smaller than BLCKSZ, decompress the page. @@ -827,370 +1156,1042 @@ restore_data_file(const char *to_path, pgFile *file, bool allow_truncate, * we have to check, whether it is compressed or not using * page_may_be_compressed() function. */ - if (header.compressed_size != BLCKSZ - || page_may_be_compressed(compressed_page.data, file->compress_alg, - backup_version)) + if (compressed_size != BLCKSZ + || page_may_be_compressed(page.data, file->compress_alg, backup_version)) { - const char *errormsg = NULL; - - uncompressed_size = do_decompress(page.data, BLCKSZ, - compressed_page.data, - header.compressed_size, - file->compress_alg, &errormsg); - if (uncompressed_size < 0 && errormsg != NULL) - elog(WARNING, "An error occured during decompressing block %u of file \"%s\": %s", - blknum, file->path, errormsg); - - if (uncompressed_size != BLCKSZ) - elog(ERROR, "Page of file \"%s\" uncompressed to %d bytes. != BLCKSZ", - file->path, uncompressed_size); + is_compressed = true; } - write_pos = (write_header) ? blknum * (BLCKSZ + sizeof(header)) : - blknum * BLCKSZ; - /* * Seek and write the restored page. + * When restoring file from FULL backup, pages are written sequentially, + * so there is no need to issue fseek for every page. */ - if (fio_fseek(out, write_pos) < 0) - elog(ERROR, "Cannot seek block %u of \"%s\": %s", - blknum, to_path, strerror(errno)); + write_pos = blknum * BLCKSZ; - if (write_header) + if (cur_pos_out != write_pos) { - /* We uncompressed the page, so its size is BLCKSZ */ - header.compressed_size = BLCKSZ; - if (fio_fwrite(out, &header, sizeof(header)) != sizeof(header)) - elog(ERROR, "Cannot write header of block %u of \"%s\": %s", - blknum, file->path, strerror(errno)); + if (fio_fseek(out, write_pos) < 0) + elog(ERROR, "Cannot seek block %u of \"%s\": %s", + blknum, to_fullpath, strerror(errno)); + + cur_pos_out = write_pos; } - /* if we uncompressed the page - write page.data, - * if page wasn't compressed - - * write what we've read - compressed_page.data + /* + * If page is compressed and restore is in remote mode, + * send compressed page to the remote side. */ - if (uncompressed_size == BLCKSZ) + if (is_compressed) { - if (fio_fwrite(out, page.data, BLCKSZ) != BLCKSZ) - elog(ERROR, "Cannot write block %u of \"%s\": %s", - blknum, file->path, strerror(errno)); + ssize_t rc; + rc = fio_fwrite_async_compressed(out, page.data, compressed_size, file->compress_alg); + + if (!fio_is_remote_file(out) && rc != BLCKSZ) + elog(ERROR, "Cannot write block %u of \"%s\": %s, size: %u", + blknum, to_fullpath, strerror(errno), compressed_size); } else { - if (fio_fwrite(out, compressed_page.data, BLCKSZ) != BLCKSZ) + if (fio_fwrite_async(out, page.data, BLCKSZ) != BLCKSZ) elog(ERROR, "Cannot write block %u of \"%s\": %s", - blknum, file->path, strerror(errno)); + blknum, to_fullpath, strerror(errno)); } + + write_len += BLCKSZ; + cur_pos_out += BLCKSZ; /* update current write position */ + + /* Mark page as restored to avoid reading this page when restoring parent backups */ + if (map) + datapagemap_add(map, blknum); } - /* - * DELTA backup have no knowledge about truncated blocks as PAGE or PTRACK do - * But during DELTA backup we read every file in PGDATA and thus DELTA backup - * knows exact size of every file at the time of backup. - * So when restoring file from DELTA backup we, knowing it`s size at - * a time of a backup, can truncate file to this size. - */ - if (allow_truncate && file->n_blocks != BLOCKNUM_INVALID && !need_truncate) + elog(LOG, "Copied file \"%s\": %lu bytes", from_fullpath, write_len); + return write_len; +} + +/* + * Copy file to backup. + * We do not apply compression to these files, because + * it is either small control file or already compressed cfs file. + */ +void +restore_non_data_file_internal(FILE *in, FILE *out, pgFile *file, + const char *from_fullpath, const char *to_fullpath) +{ + size_t read_len = 0; + char *buf = pgut_malloc(STDIO_BUFSIZE); /* 64kB buffer */ + + /* copy content */ + for (;;) { - struct stat st; - if (fio_ffstat(out, &st) == 0 && st.st_size > file->n_blocks * BLCKSZ) + read_len = 0; + + /* check for interrupt */ + if (interrupted || thread_interrupted) + elog(ERROR, "Interrupted during non-data file restore"); + + read_len = fread(buf, 1, STDIO_BUFSIZE, in); + + if (ferror(in)) + elog(ERROR, "Cannot read backup file \"%s\": %s", + from_fullpath, strerror(errno)); + + if (read_len > 0) { - truncate_from = file->n_blocks; - need_truncate = true; + if (fio_fwrite_async(out, buf, read_len) != read_len) + elog(ERROR, "Cannot write to \"%s\": %s", to_fullpath, + strerror(errno)); } + + if (feof(in)) + break; } - if (need_truncate) - { - off_t write_pos; + pg_free(buf); + + elog(LOG, "Copied file \"%s\": %lu bytes", from_fullpath, file->write_size); +} + +size_t +restore_non_data_file(parray *parent_chain, pgBackup *dest_backup, + pgFile *dest_file, FILE *out, const char *to_fullpath, + bool already_exists) +{ + char from_root[MAXPGPATH]; + char from_fullpath[MAXPGPATH]; + FILE *in = NULL; - write_pos = (write_header) ? truncate_from * (BLCKSZ + sizeof(header)) : - truncate_from * BLCKSZ; + pgFile *tmp_file = NULL; + pgBackup *tmp_backup = NULL; + /* Check if full copy of destination file is available in destination backup */ + if (dest_file->write_size > 0) + { + tmp_file = dest_file; + tmp_backup = dest_backup; + } + else + { /* - * Truncate file to this length. + * Iterate over parent chain starting from direct parent of destination + * backup to oldest backup in chain, and look for the first + * full copy of destination file. + * Full copy is latest possible destination file with size equal or + * greater than zero. */ - if (fio_ftruncate(out, write_pos) != 0) - elog(ERROR, "Cannot truncate \"%s\": %s", - file->path, strerror(errno)); - elog(VERBOSE, "Delta truncate file %s to block %u", - file->path, truncate_from); + tmp_backup = dest_backup->parent_backup_link; + while (tmp_backup) + { + pgFile **res_file = NULL; + + /* lookup file in intermediate backup */ + res_file = parray_bsearch(tmp_backup->files, dest_file, pgFileCompareRelPathWithExternal); + tmp_file = (res_file) ? *res_file : NULL; + + /* + * It should not be possible not to find destination file in intermediate + * backup, without encountering full copy first. + */ + if (!tmp_file) + { + elog(ERROR, "Failed to locate non-data file \"%s\" in backup %s", + dest_file->rel_path, backup_id_of(tmp_backup)); + continue; + } + + /* Full copy is found and it is null sized, nothing to do here */ + if (tmp_file->write_size == 0) + { + /* In case of incremental restore truncate file just to be safe */ + if (already_exists && fio_ftruncate(out, 0)) + elog(ERROR, "Cannot truncate file \"%s\": %s", + to_fullpath, strerror(errno)); + return 0; + } + + /* Full copy is found */ + if (tmp_file->write_size > 0) + break; + + tmp_backup = tmp_backup->parent_backup_link; + } } - /* update file permission */ - if (fio_chmod(to_path, file->mode, FIO_DB_HOST) == -1) + /* sanity */ + if (!tmp_backup) + elog(ERROR, "Failed to locate a backup containing full copy of non-data file \"%s\"", + to_fullpath); + + if (!tmp_file) + elog(ERROR, "Failed to locate a full copy of non-data file \"%s\"", to_fullpath); + + if (tmp_file->write_size <= 0) + elog(ERROR, "Full copy of non-data file has invalid size: %li. " + "Metadata corruption in backup %s in file: \"%s\"", + tmp_file->write_size, backup_id_of(tmp_backup), + to_fullpath); + + /* incremental restore */ + if (already_exists) { - int errno_tmp = errno; + /* compare checksums of already existing file and backup file */ + pg_crc32 file_crc; + if (tmp_file->forkName == cfm && + tmp_file->uncompressed_size > tmp_file->write_size) + file_crc = fio_get_crc32_truncated(to_fullpath, FIO_DB_HOST, false); + else + file_crc = fio_get_crc32(to_fullpath, FIO_DB_HOST, false, false); - if (in) - fclose(in); - fio_fclose(out); - elog(ERROR, "Cannot change mode of \"%s\": %s", to_path, - strerror(errno_tmp)); + if (file_crc == tmp_file->crc) + { + elog(LOG, "Already existing non-data file \"%s\" has the same checksum, skip restore", + to_fullpath); + return 0; + } + + /* Checksum mismatch, truncate file and overwrite it */ + if (fio_ftruncate(out, 0)) + elog(ERROR, "Cannot truncate file \"%s\": %s", + to_fullpath, strerror(errno)); } - if (fio_fflush(out) != 0 || - fio_fclose(out)) - elog(ERROR, "Cannot write \"%s\": %s", to_path, strerror(errno)); + if (tmp_file->external_dir_num == 0) + join_path_components(from_root, tmp_backup->root_dir, DATABASE_DIR); + else + { + char external_prefix[MAXPGPATH]; - if (in) - fclose(in); + join_path_components(external_prefix, tmp_backup->root_dir, EXTERNAL_DIR); + makeExternalDirPathByNum(from_root, external_prefix, tmp_file->external_dir_num); + } + + join_path_components(from_fullpath, from_root, dest_file->rel_path); + + in = fopen(from_fullpath, PG_BINARY_R); + if (in == NULL) + elog(ERROR, "Cannot open backup file \"%s\": %s", from_fullpath, + strerror(errno)); + + /* disable stdio buffering for non-data files */ + setvbuf(in, NULL, _IONBF, BUFSIZ); + + /* do actual work */ + restore_non_data_file_internal(in, out, tmp_file, from_fullpath, to_fullpath); + + if (fclose(in) != 0) + elog(ERROR, "Cannot close file \"%s\": %s", from_fullpath, + strerror(errno)); + + return tmp_file->write_size; } /* * Copy file to backup. * We do not apply compression to these files, because * it is either small control file or already compressed cfs file. + * TODO: optimize remote copying */ -bool -copy_file(fio_location from_location, const char *to_root, - fio_location to_location, pgFile *file, bool missing_ok) +void +backup_non_data_file_internal(const char *from_fullpath, + fio_location from_location, + const char *to_fullpath, pgFile *file, + bool missing_ok) { - char to_path[MAXPGPATH]; - FILE *in; - FILE *out; - size_t read_len = 0; - int errno_tmp; - char buf[BLCKSZ]; - pg_crc32 crc; + FILE *out = NULL; + char *errmsg = NULL; + int rc; + bool cut_zero_tail; + + cut_zero_tail = file->forkName == cfm; - INIT_FILE_CRC32(true, crc); + INIT_FILE_CRC32(true, file->crc); /* reset size summary */ file->read_size = 0; file->write_size = 0; + file->uncompressed_size = 0; - /* open backup mode file for read */ - in = fio_fopen(file->path, PG_BINARY_R, from_location); - if (in == NULL) - { - FIN_FILE_CRC32(true, crc); - file->crc = crc; - + /* open backup file for write */ + out = fopen(to_fullpath, PG_BINARY_W); + if (out == NULL) + elog(ERROR, "Cannot open destination file \"%s\": %s", + to_fullpath, strerror(errno)); + + /* update file permission */ + if (chmod(to_fullpath, file->mode) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", to_fullpath, + strerror(errno)); + + /* backup remote file */ + if (fio_is_remote(FIO_DB_HOST)) + rc = fio_send_file(from_fullpath, out, cut_zero_tail, file, &errmsg); + else + rc = fio_send_file_local(from_fullpath, out, cut_zero_tail, file, &errmsg); + + /* handle errors */ + if (rc == FILE_MISSING) + { /* maybe deleted, it's not error in case of backup */ + if (missing_ok) + { + elog(LOG, "File \"%s\" is not found", from_fullpath); + file->write_size = FILE_NOT_FOUND; + goto cleanup; + } + else + elog(ERROR, "File \"%s\" is not found", from_fullpath); + } + else if (rc == WRITE_FAILED) + elog(ERROR, "Cannot write to \"%s\": %s", to_fullpath, strerror(errno)); + else if (rc != SEND_OK) + { + if (errmsg) + elog(ERROR, "%s", errmsg); + else + elog(ERROR, "Cannot access remote file \"%s\"", from_fullpath); + } + + file->uncompressed_size = file->read_size; + +cleanup: + if (errmsg != NULL) + pg_free(errmsg); + + /* finish CRC calculation and store into pgFile */ + FIN_FILE_CRC32(true, file->crc); + + if (out && fclose(out)) + elog(ERROR, "Cannot close the file \"%s\": %s", to_fullpath, strerror(errno)); +} + +/* + * Create empty file, used for partial restore + */ +bool +create_empty_file(fio_location from_location, const char *to_root, + fio_location to_location, pgFile *file) +{ + char to_path[MAXPGPATH]; + FILE *out; + + /* open file for write */ + join_path_components(to_path, to_root, file->rel_path); + out = fio_fopen(to_path, PG_BINARY_W, to_location); + + if (out == NULL) + elog(ERROR, "Cannot open destination file \"%s\": %s", + to_path, strerror(errno)); + + /* update file permission */ + if (fio_chmod(to_path, file->mode, to_location) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", to_path, + strerror(errno)); + + if (fio_fclose(out)) + elog(ERROR, "Cannot close \"%s\": %s", to_path, strerror(errno)); + + return true; +} + +/* + * Validate given page. + * This function is expected to be executed multiple times, + * so avoid using elog within it. + * lsn from page is assigned to page_lsn pointer. + * TODO: switch to enum for return codes. + */ +int +validate_one_page(Page page, BlockNumber absolute_blkno, + XLogRecPtr stop_lsn, PageState *page_st, + uint32 checksum_version) +{ + page_st->lsn = InvalidXLogRecPtr; + page_st->checksum = 0; + + /* new level of paranoia */ + if (page == NULL) + return PAGE_IS_NOT_FOUND; + + /* check that page header is ok */ + if (!parse_page(page, &(page_st)->lsn)) + { + int i; + /* Check if the page is zeroed. */ + for (i = 0; i < BLCKSZ && page[i] == 0; i++); + + /* Page is zeroed. No need to verify checksums */ + if (i == BLCKSZ) + return PAGE_IS_ZEROED; + + /* Page does not looking good */ + return PAGE_HEADER_IS_INVALID; + } + + /* Verify checksum */ + page_st->checksum = pg_checksum_page(page, absolute_blkno); + + if (checksum_version) + { + /* Checksums are enabled, so check them. */ + if (page_st->checksum != ((PageHeader) page)->pd_checksum) + return PAGE_CHECKSUM_MISMATCH; + } + + /* At this point page header is sane, if checksums are enabled - the`re ok. + * Check that page is not from future. + * Note, this check should be used only by validate command. + */ + if (stop_lsn > 0) + { + /* Get lsn from page header. Ensure that page is from our time. */ + if (page_st->lsn > stop_lsn) + return PAGE_LSN_FROM_FUTURE; + } + + return PAGE_IS_VALID; +} + +/* + * Validate pages of datafile in PGDATA one by one. + * + * returns true if the file is valid + * also returns true if the file was not found + */ +bool +check_data_file(ConnectionArgs *arguments, pgFile *file, + const char *from_fullpath, uint32 checksum_version) +{ + FILE *in; + BlockNumber blknum = 0; + BlockNumber nblocks = 0; + int page_state; + char curr_page[BLCKSZ]; + bool is_valid = true; + + in = fopen(from_fullpath, PG_BINARY_R); + if (in == NULL) + { + /* + * If file is not found, this is not en error. + * It could have been deleted by concurrent postgres transaction. + */ if (errno == ENOENT) { - if (missing_ok) + elog(LOG, "File \"%s\" is not found", from_fullpath); + return true; + } + + elog(WARNING, "Cannot open file \"%s\": %s", + from_fullpath, strerror(errno)); + return false; + } + + if (file->size % BLCKSZ != 0) + elog(WARNING, "File: \"%s\", invalid file size %zu", from_fullpath, file->size); + + /* + * Compute expected number of blocks in the file. + * NOTE This is a normal situation, if the file size has changed + * since the moment we computed it. + */ + nblocks = file->size/BLCKSZ; + + for (blknum = 0; blknum < nblocks; blknum++) + { + PageState page_st; + page_state = prepare_page(file, InvalidXLogRecPtr, + blknum, in, BACKUP_MODE_FULL, + curr_page, false, checksum_version, + from_fullpath, &page_st); + + if (page_state == PageIsTruncated) + break; + + if (page_state == PageIsCorrupted) + { + /* Page is corrupted, no need to elog about it, + * prepare_page() already done that + */ + is_valid = false; + continue; + } + } + + fclose(in); + return is_valid; +} + +/* Valiate pages of datafile in backup one by one */ +bool +validate_file_pages(pgFile *file, const char *fullpath, XLogRecPtr stop_lsn, + uint32 checksum_version, uint32 backup_version, HeaderMap *hdr_map) +{ + size_t read_len = 0; + bool is_valid = true; + FILE *in; + pg_crc32 crc; + bool use_crc32c = backup_version <= 20021 || backup_version >= 20025; + BackupPageHeader2 *headers = NULL; + int n_hdr = -1; + off_t cur_pos_in = 0; + + elog(LOG, "Validate relation blocks for file \"%s\"", fullpath); + + /* should not be possible */ + Assert(!(backup_version >= 20400 && file->n_headers <= 0)); + + in = fopen(fullpath, PG_BINARY_R); + if (in == NULL) + elog(ERROR, "Cannot open file \"%s\": %s", + fullpath, strerror(errno)); + + headers = get_data_file_headers(hdr_map, file, backup_version, false); + + if (!headers && file->n_headers > 0) + { + elog(WARNING, "Cannot get page headers for file \"%s\"", fullpath); + return false; + } + + /* calc CRC of backup file */ + INIT_FILE_CRC32(use_crc32c, crc); + + /* read and validate pages one by one */ + while (true) + { + int rc = 0; + size_t len = 0; + DataPage compressed_page; /* used as read buffer */ + int compressed_size = 0; + DataPage page; + BlockNumber blknum = 0; + PageState page_st; + + if (interrupted || thread_interrupted) + elog(ERROR, "Interrupted during data file validation"); + + /* newer backups have page headers in separate storage */ + if (headers) + { + n_hdr++; + if (n_hdr >= file->n_headers) + break; + + blknum = headers[n_hdr].block; + /* calculate payload size by comparing current and next page positions, + * page header is not included. + */ + compressed_size = headers[n_hdr+1].pos - headers[n_hdr].pos - sizeof(BackupPageHeader); + + Assert(compressed_size > 0); + Assert(compressed_size <= BLCKSZ); + + read_len = sizeof(BackupPageHeader) + compressed_size; + + if (cur_pos_in != headers[n_hdr].pos) { - elog(LOG, "File \"%s\" is not found", file->path); - file->write_size = FILE_NOT_FOUND; - return false; + if (fio_fseek(in, headers[n_hdr].pos) < 0) + elog(ERROR, "Cannot seek block %u of \"%s\": %s", + blknum, fullpath, strerror(errno)); + else + elog(VERBOSE, "Seek to %u", headers[n_hdr].pos); + + cur_pos_in = headers[n_hdr].pos; + } + } + /* old backups rely on header located directly in data file */ + else + { + if (get_page_header(in, fullpath, &(compressed_page).bph, &crc, use_crc32c)) + { + /* Backward compatibility kludge, TODO: remove in 3.0 + * for some reason we padded compressed pages in old versions + */ + blknum = compressed_page.bph.block; + compressed_size = compressed_page.bph.compressed_size; + read_len = MAXALIGN(compressed_size); } else - elog(ERROR, "File \"%s\" is not found", file->path); + break; } - elog(ERROR, "cannot open source file \"%s\": %s", file->path, - strerror(errno)); + /* backward compatibility kludge TODO: remove in 3.0 */ + if (compressed_size == PageIsTruncated) + { + elog(VERBOSE, "Block %u of \"%s\" is truncated", + blknum, fullpath); + continue; + } + + Assert(compressed_size <= BLCKSZ); + Assert(compressed_size > 0); + + if (headers) + len = fread(&compressed_page, 1, read_len, in); + else + len = fread(compressed_page.data, 1, read_len, in); + + if (len != read_len) + { + elog(WARNING, "Cannot read block %u file \"%s\": %s", + blknum, fullpath, strerror(errno)); + return false; + } + + /* update current position */ + cur_pos_in += read_len; + + if (headers) + COMP_FILE_CRC32(use_crc32c, crc, &compressed_page, read_len); + else + COMP_FILE_CRC32(use_crc32c, crc, compressed_page.data, read_len); + + if (compressed_size != BLCKSZ + || page_may_be_compressed(compressed_page.data, file->compress_alg, + backup_version)) + { + int32 uncompressed_size = 0; + const char *errormsg = NULL; + + uncompressed_size = do_decompress(page.data, BLCKSZ, + compressed_page.data, + compressed_size, + file->compress_alg, + &errormsg); + if (uncompressed_size < 0 && errormsg != NULL) + { + elog(WARNING, "An error occured during decompressing block %u of file \"%s\": %s", + blknum, fullpath, errormsg); + return false; + } + + if (uncompressed_size != BLCKSZ) + { + if (compressed_size == BLCKSZ) + { + is_valid = false; + continue; + } + elog(WARNING, "Page %u of file \"%s\" uncompressed to %d bytes. != BLCKSZ", + blknum, fullpath, uncompressed_size); + return false; + } + + rc = validate_one_page(page.data, + file->segno * RELSEG_SIZE + blknum, + stop_lsn, &page_st, checksum_version); + } + else + rc = validate_one_page(compressed_page.data, + file->segno * RELSEG_SIZE + blknum, + stop_lsn, &page_st, checksum_version); + + switch (rc) + { + case PAGE_IS_NOT_FOUND: + elog(VERBOSE, "File \"%s\", block %u, page is NULL", file->rel_path, blknum); + break; + case PAGE_IS_ZEROED: + elog(VERBOSE, "File: %s blknum %u, empty zeroed page", file->rel_path, blknum); + break; + case PAGE_HEADER_IS_INVALID: + elog(WARNING, "Page header is looking insane: %s, block %i", file->rel_path, blknum); + is_valid = false; + break; + case PAGE_CHECKSUM_MISMATCH: + elog(WARNING, "File: %s blknum %u have wrong checksum: %u", file->rel_path, blknum, page_st.checksum); + is_valid = false; + break; + case PAGE_LSN_FROM_FUTURE: + elog(WARNING, "File: %s, block %u, checksum is %s. " + "Page is from future: pageLSN %X/%X stopLSN %X/%X", + file->rel_path, blknum, + checksum_version ? "correct" : "not enabled", + (uint32) (page_st.lsn >> 32), (uint32) page_st.lsn, + (uint32) (stop_lsn >> 32), (uint32) stop_lsn); + break; + } } - /* open backup file for write */ - join_path_components(to_path, to_root, file->rel_path); - out = fio_fopen(to_path, PG_BINARY_W, to_location); - if (out == NULL) + FIN_FILE_CRC32(use_crc32c, crc); + fclose(in); + + if (crc != file->crc) { - int errno_tmp = errno; - fio_fclose(in); - elog(ERROR, "cannot open destination file \"%s\": %s", - to_path, strerror(errno_tmp)); + elog(WARNING, "Invalid CRC of backup file \"%s\": %X. Expected %X", + fullpath, crc, file->crc); + is_valid = false; } - /* copy content and calc CRC */ - for (;;) + pg_free(headers); + + return is_valid; +} + +/* read local data file and construct map with block checksums */ +PageState* +get_checksum_map(const char *fullpath, uint32 checksum_version, + int n_blocks, XLogRecPtr dest_stop_lsn, BlockNumber segmentno) +{ + PageState *checksum_map = NULL; + FILE *in = NULL; + BlockNumber blknum = 0; + char read_buffer[BLCKSZ]; + char in_buf[STDIO_BUFSIZE]; + + /* open file */ + in = fopen(fullpath, "r+b"); + if (!in) + elog(ERROR, "Cannot open source file \"%s\": %s", fullpath, strerror(errno)); + + /* truncate up to blocks */ + if (ftruncate(fileno(in), n_blocks * BLCKSZ) != 0) + elog(ERROR, "Cannot truncate file to blknum %u \"%s\": %s", + n_blocks, fullpath, strerror(errno)); + + setvbuf(in, in_buf, _IOFBF, STDIO_BUFSIZE); + + /* initialize array of checksums */ + checksum_map = pgut_malloc(n_blocks * sizeof(PageState)); + memset(checksum_map, 0, n_blocks * sizeof(PageState)); + + for (blknum = 0; blknum < n_blocks; blknum++) { - read_len = 0; + size_t read_len = fread(read_buffer, 1, BLCKSZ, in); + PageState page_st; + + /* report error */ + if (ferror(in)) + elog(ERROR, "Cannot read block %u of \"%s\": %s", + blknum, fullpath, strerror(errno)); + + if (read_len == BLCKSZ) + { + int rc = validate_one_page(read_buffer, segmentno + blknum, + dest_stop_lsn, &page_st, + checksum_version); + + if (rc == PAGE_IS_VALID) + { +// if (checksum_version) +// checksum_map[blknum].checksum = ((PageHeader) read_buffer)->pd_checksum; +// else +// checksum_map[blknum].checksum = page_st.checksum; + checksum_map[blknum].checksum = page_st.checksum; + checksum_map[blknum].lsn = page_st.lsn; + } + } + else + elog(ERROR, "Failed to read blknum %u from file \"%s\"", blknum, fullpath); + + if (feof(in)) + break; + + if (interrupted || thread_interrupted) + elog(ERROR, "Interrupted during page reading"); + } + + if (in) + fclose(in); + + return checksum_map; +} + +/* return bitmap of valid blocks, bitmap is empty, then NULL is returned */ +datapagemap_t * +get_lsn_map(const char *fullpath, uint32 checksum_version, + int n_blocks, XLogRecPtr shift_lsn, BlockNumber segmentno) +{ + FILE *in = NULL; + BlockNumber blknum = 0; + char read_buffer[BLCKSZ]; + char in_buf[STDIO_BUFSIZE]; + datapagemap_t *lsn_map = NULL; + + Assert(shift_lsn > 0); + + /* open file */ + in = fopen(fullpath, "r+b"); + if (!in) + elog(ERROR, "Cannot open source file \"%s\": %s", fullpath, strerror(errno)); + + /* truncate up to blocks */ + if (ftruncate(fileno(in), n_blocks * BLCKSZ) != 0) + elog(ERROR, "Cannot truncate file to blknum %u \"%s\": %s", + n_blocks, fullpath, strerror(errno)); + + setvbuf(in, in_buf, _IOFBF, STDIO_BUFSIZE); + + lsn_map = pgut_malloc(sizeof(datapagemap_t)); + memset(lsn_map, 0, sizeof(datapagemap_t)); + + for (blknum = 0; blknum < n_blocks; blknum++) + { + size_t read_len = fread(read_buffer, 1, BLCKSZ, in); + PageState page_st; + + /* report error */ + if (ferror(in)) + elog(ERROR, "Cannot read block %u of \"%s\": %s", + blknum, fullpath, strerror(errno)); + + if (read_len == BLCKSZ) + { + int rc = validate_one_page(read_buffer, segmentno + blknum, + shift_lsn, &page_st, checksum_version); + + if (rc == PAGE_IS_VALID) + datapagemap_add(lsn_map, blknum); + } + else + elog(ERROR, "Cannot read block %u from file \"%s\": %s", + blknum, fullpath, strerror(errno)); - if ((read_len = fio_fread(in, buf, sizeof(buf))) != sizeof(buf)) + if (feof(in)) break; - if (fio_fwrite(out, buf, read_len) != read_len) - { - errno_tmp = errno; - /* oops */ - fio_fclose(in); - fio_fclose(out); - elog(ERROR, "cannot write to \"%s\": %s", to_path, - strerror(errno_tmp)); - } - /* update CRC */ - COMP_FILE_CRC32(true, crc, buf, read_len); - - file->read_size += read_len; + if (interrupted || thread_interrupted) + elog(ERROR, "Interrupted during page reading"); } - errno_tmp = errno; - if (read_len < 0) + if (in) + fclose(in); + + if (lsn_map->bitmapsize == 0) { - fio_fclose(in); - fio_fclose(out); - elog(ERROR, "cannot read backup mode file \"%s\": %s", - file->path, strerror(errno_tmp)); + pg_free(lsn_map); + lsn_map = NULL; } - /* copy odd part. */ - if (read_len > 0) - { - if (fio_fwrite(out, buf, read_len) != read_len) - { - errno_tmp = errno; - /* oops */ - fio_fclose(in); - fio_fclose(out); - elog(ERROR, "cannot write to \"%s\": %s", to_path, - strerror(errno_tmp)); - } - /* update CRC */ - COMP_FILE_CRC32(true, crc, buf, read_len); + return lsn_map; +} - file->read_size += read_len; - } +/* Every page in data file contains BackupPageHeader, extract it */ +bool +get_page_header(FILE *in, const char *fullpath, BackupPageHeader* bph, + pg_crc32 *crc, bool use_crc32c) +{ + /* read BackupPageHeader */ + size_t read_len = fread(bph, 1, sizeof(BackupPageHeader), in); - file->write_size = (int64) file->read_size; - /* finish CRC calculation and store into pgFile */ - FIN_FILE_CRC32(true, crc); - file->crc = crc; + if (ferror(in)) + elog(ERROR, "Cannot read file \"%s\": %s", + fullpath, strerror(errno)); - /* update file permission */ - if (fio_chmod(to_path, file->mode, to_location) == -1) + if (read_len != sizeof(BackupPageHeader)) { - errno_tmp = errno; - fio_fclose(in); - fio_fclose(out); - elog(ERROR, "cannot change mode of \"%s\": %s", to_path, - strerror(errno_tmp)); + if (read_len == 0 && feof(in)) + return false; /* EOF found */ + else if (read_len != 0 && feof(in)) + elog(ERROR, + "Odd size page found at offset %ld of \"%s\"", + ftello(in), fullpath); + else + elog(ERROR, "Cannot read header at offset %ld of \"%s\": %s", + ftello(in), fullpath, strerror(errno)); } - if (fio_fflush(out) != 0 || - fio_fclose(out)) - elog(ERROR, "cannot write \"%s\": %s", to_path, strerror(errno)); - fio_fclose(in); + /* In older versions < 2.4.0, when crc for file was calculated, header was + * not included in crc calculations. Now it is. And now we have + * the problem of backward compatibility for backups of old versions + */ + if (crc) + COMP_FILE_CRC32(use_crc32c, *crc, bph, read_len); + + if (bph->block == 0 && bph->compressed_size == 0) + elog(ERROR, "Empty block in file \"%s\"", fullpath); + Assert(bph->compressed_size != 0); return true; } -/* - * Validate given page. - * - * Returns value: - * 0 - if the page is not found - * 1 - if the page is found and valid - * -1 - if the page is found but invalid - */ -#define PAGE_IS_NOT_FOUND 0 -#define PAGE_IS_FOUND_AND_VALID 1 -#define PAGE_IS_FOUND_AND_NOT_VALID -1 -static int -validate_one_page(Page page, pgFile *file, - BlockNumber blknum, XLogRecPtr stop_lsn, - uint32 checksum_version) +/* Open local backup file for writing, set permissions and buffering */ +FILE* +open_local_file_rw(const char *to_fullpath, char **out_buf, uint32 buf_size) { - PageHeader phdr; - XLogRecPtr lsn; + FILE *out = NULL; + /* open backup file for write */ + out = fopen(to_fullpath, PG_BINARY_W); + if (out == NULL) + elog(ERROR, "Cannot open backup file \"%s\": %s", + to_fullpath, strerror(errno)); - /* new level of paranoia */ - if (page == NULL) + /* update file permission */ + if (chmod(to_fullpath, FILE_PERMISSION) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", to_fullpath, + strerror(errno)); + + /* enable stdio buffering for output file */ + *out_buf = pgut_malloc(buf_size); + setvbuf(out, *out_buf, _IOFBF, buf_size); + + return out; +} + +/* backup local file */ +int +send_pages(const char *to_fullpath, const char *from_fullpath, + pgFile *file, XLogRecPtr prev_backup_start_lsn, CompressAlg calg, int clevel, + uint32 checksum_version, bool use_pagemap, BackupPageHeader2 **headers, + BackupMode backup_mode) +{ + FILE *in = NULL; + FILE *out = NULL; + off_t cur_pos_out = 0; + char curr_page[BLCKSZ]; + int n_blocks_read = 0; + BlockNumber blknum = 0; + datapagemap_iterator_t *iter = NULL; + int compressed_size = 0; + BackupPageHeader2 *header = NULL; + parray *harray = NULL; + + /* stdio buffers */ + char *in_buf = NULL; + char *out_buf = NULL; + + /* open source file for read */ + in = fopen(from_fullpath, PG_BINARY_R); + if (in == NULL) { - elog(LOG, "File \"%s\", block %u, page is NULL", file->path, blknum); - return PAGE_IS_NOT_FOUND; + /* + * If file is not found, this is not en error. + * It could have been deleted by concurrent postgres transaction. + */ + if (errno == ENOENT) + return FILE_MISSING; + + elog(ERROR, "Cannot open file \"%s\": %s", from_fullpath, strerror(errno)); } - phdr = (PageHeader) page; + /* + * Enable stdio buffering for local input file, + * unless the pagemap is involved, which + * imply a lot of random access. + */ - if (PageIsNew(page)) + if (use_pagemap) { - int i; + iter = datapagemap_iterate(&file->pagemap); + datapagemap_next(iter, &blknum); /* set first block */ - /* Check if the page is zeroed. */ - for(i = 0; i < BLCKSZ && page[i] == 0; i++); + setvbuf(in, NULL, _IONBF, BUFSIZ); + } + else + { + in_buf = pgut_malloc(STDIO_BUFSIZE); + setvbuf(in, in_buf, _IOFBF, STDIO_BUFSIZE); + } - if (i == BLCKSZ) - { - elog(LOG, "File: %s blknum %u, page is New, empty zeroed page", - file->path, blknum); - return PAGE_IS_FOUND_AND_VALID; - } - else + harray = parray_new(); + + while (blknum < file->n_blocks) + { + PageState page_st; + int rc = prepare_page(file, prev_backup_start_lsn, + blknum, in, backup_mode, curr_page, + true, checksum_version, + from_fullpath, &page_st); + + if (rc == PageIsTruncated) + break; + + else if (rc == PageIsOk) { - elog(WARNING, "File: %s blknum %u, page is New, but not zeroed", - file->path, blknum); + /* lazily open backup file (useful for s3) */ + if (!out) + out = open_local_file_rw(to_fullpath, &out_buf, STDIO_BUFSIZE); + + header = pgut_new0(BackupPageHeader2); + *header = (BackupPageHeader2){ + .block = blknum, + .pos = cur_pos_out, + .lsn = page_st.lsn, + .checksum = page_st.checksum, + }; + + parray_append(harray, header); + + compressed_size = compress_and_backup_page(file, blknum, in, out, &(file->crc), + rc, curr_page, calg, clevel, + from_fullpath, to_fullpath); + cur_pos_out += compressed_size + sizeof(BackupPageHeader); } - /* Page is zeroed. No sense in checking header and checksum. */ - return PAGE_IS_FOUND_AND_VALID; - } + n_blocks_read++; - /* Verify checksum */ - if (checksum_version) - { - /* Checksums are enabled, so check them. */ - if (!(pg_checksum_page(page, file->segno * RELSEG_SIZE + blknum) - == ((PageHeader) page)->pd_checksum)) + /* next block */ + if (use_pagemap) { - elog(WARNING, "File: %s blknum %u have wrong checksum", - file->path, blknum); - return PAGE_IS_FOUND_AND_NOT_VALID; + /* exit if pagemap is exhausted */ + if (!datapagemap_next(iter, &blknum)) + break; } + else + blknum++; } - /* Check page for the sights of insanity. - * TODO: We should give more information about what exactly is looking "wrong" - */ - if (!(PageGetPageSize(phdr) == BLCKSZ && - // PageGetPageLayoutVersion(phdr) == PG_PAGE_LAYOUT_VERSION && - (phdr->pd_flags & ~PD_VALID_FLAG_BITS) == 0 && - phdr->pd_lower >= SizeOfPageHeaderData && - phdr->pd_lower <= phdr->pd_upper && - phdr->pd_upper <= phdr->pd_special && - phdr->pd_special <= BLCKSZ && - phdr->pd_special == MAXALIGN(phdr->pd_special))) - { - /* Page does not looking good */ - elog(WARNING, "Page header is looking insane: %s, block %i", - file->path, blknum); - return PAGE_IS_FOUND_AND_NOT_VALID; - } - - /* At this point page header is sane, if checksums are enabled - the`re ok. - * Check that page is not from future. + /* + * Add dummy header, so we can later extract the length of last header + * as difference between their offsets. */ - if (stop_lsn > 0) + if (parray_num(harray) > 0) { - /* Get lsn from page header. Ensure that page is from our time. */ - lsn = PageXLogRecPtrGet(phdr->pd_lsn); + size_t hdr_num = parray_num(harray); + size_t i; - if (lsn > stop_lsn) + file->n_headers = (int) hdr_num; /* is it valid? */ + *headers = (BackupPageHeader2 *) pgut_malloc0((hdr_num + 1) * sizeof(BackupPageHeader2)); + for (i = 0; i < hdr_num; i++) { - elog(WARNING, "File: %s, block %u, checksum is %s. " - "Page is from future: pageLSN %X/%X stopLSN %X/%X", - file->path, blknum, checksum_version ? "correct" : "not enabled", - (uint32) (lsn >> 32), (uint32) lsn, - (uint32) (stop_lsn >> 32), (uint32) stop_lsn); - return PAGE_IS_FOUND_AND_NOT_VALID; + header = (BackupPageHeader2 *)parray_get(harray, i); + (*headers)[i] = *header; + pg_free(header); } + (*headers)[hdr_num] = (BackupPageHeader2){.pos=cur_pos_out}; } + parray_free(harray); + + /* cleanup */ + if (in && fclose(in)) + elog(ERROR, "Cannot close the source file \"%s\": %s", + to_fullpath, strerror(errno)); + + /* close local output file */ + if (out && fclose(out)) + elog(ERROR, "Cannot close the backup file \"%s\": %s", + to_fullpath, strerror(errno)); - return PAGE_IS_FOUND_AND_VALID; + pg_free(iter); + pg_free(in_buf); + pg_free(out_buf); + + return n_blocks_read; } /* - * Valiate pages of datafile in PGDATA one by one. - * - * returns true if the file is valid - * also returns true if the file was not found + * Copy local data file just as send_pages but without attaching additional header and compression */ -bool -check_data_file(ConnectionArgs *arguments, - pgFile *file, uint32 checksum_version) +int +copy_pages(const char *to_fullpath, const char *from_fullpath, + pgFile *file, XLogRecPtr sync_lsn, + uint32 checksum_version, bool use_pagemap, + BackupMode backup_mode) { - FILE *in; - BlockNumber blknum = 0; - BlockNumber nblocks = 0; - BlockNumber n_blocks_skipped = 0; - int page_state; - char curr_page[BLCKSZ]; - bool is_valid = true; - - in = fopen(file->path, PG_BINARY_R); + FILE *in = NULL; + FILE *out = NULL; + char curr_page[BLCKSZ]; + int n_blocks_read = 0; + BlockNumber blknum = 0; + datapagemap_iterator_t *iter = NULL; + + /* stdio buffers */ + char *in_buf = NULL; + char *out_buf = NULL; + + /* open source file for read */ + in = fopen(from_fullpath, PG_BINARY_R); if (in == NULL) { /* @@ -1198,205 +2199,329 @@ check_data_file(ConnectionArgs *arguments, * It could have been deleted by concurrent postgres transaction. */ if (errno == ENOENT) - { - elog(LOG, "File \"%s\" is not found", file->path); - return true; - } + return FILE_MISSING; - elog(WARNING, "cannot open file \"%s\": %s", - file->path, strerror(errno)); - return false; + elog(ERROR, "Cannot open file \"%s\": %s", from_fullpath, strerror(errno)); } - if (file->size % BLCKSZ != 0) - elog(WARNING, "File: \"%s\", invalid file size %zu", file->path, file->size); - /* - * Compute expected number of blocks in the file. - * NOTE This is a normal situation, if the file size has changed - * since the moment we computed it. + * Enable stdio buffering for local input file, + * unless the pagemap is involved, which + * imply a lot of random access. */ - nblocks = file->size/BLCKSZ; - for (blknum = 0; blknum < nblocks; blknum++) + if (use_pagemap) { - page_state = prepare_page(arguments, file, InvalidXLogRecPtr, - blknum, nblocks, in, &n_blocks_skipped, - BACKUP_MODE_FULL, curr_page, false, checksum_version); + iter = datapagemap_iterate(&file->pagemap); + datapagemap_next(iter, &blknum); /* set first block */ - if (page_state == PageIsTruncated) + setvbuf(in, NULL, _IONBF, BUFSIZ); + } + else + { + in_buf = pgut_malloc(STDIO_BUFSIZE); + setvbuf(in, in_buf, _IOFBF, STDIO_BUFSIZE); + } + + out = fio_fopen(to_fullpath, PG_BINARY_R "+", FIO_BACKUP_HOST); + if (out == NULL) + elog(ERROR, "Cannot open destination file \"%s\": %s", + to_fullpath, strerror(errno)); + + /* update file permission */ + if (chmod(to_fullpath, file->mode) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", to_fullpath, + strerror(errno)); + + /* Enable buffering for output file */ + out_buf = pgut_malloc(STDIO_BUFSIZE); + setvbuf(out, out_buf, _IOFBF, STDIO_BUFSIZE); + + while (blknum < file->n_blocks) + { + PageState page_st; + int rc = prepare_page(file, sync_lsn, + blknum, in, backup_mode, curr_page, + true, checksum_version, + from_fullpath, &page_st); + if (rc == PageIsTruncated) break; - if (page_state == PageIsCorrupted) + else if (rc == PageIsOk) { - /* Page is corrupted, no need to elog about it, - * prepare_page() already done that - */ - is_valid = false; - continue; + if (fseek(out, blknum * BLCKSZ, SEEK_SET) != 0) + elog(ERROR, "Cannot seek to position %u in destination file \"%s\": %s", + blknum * BLCKSZ, to_fullpath, strerror(errno)); + + if (write_page(file, out, curr_page) != BLCKSZ) + elog(ERROR, "File: \"%s\", cannot write at block %u: %s", + to_fullpath, blknum, strerror(errno)); } - /* At this point page is found and its checksum is ok, if any - * but could be 'insane' - * TODO: between prepare_page and validate_one_page we - * compute and compare checksum twice, it`s ineffective - */ - if (validate_one_page(curr_page, file, blknum, - InvalidXLogRecPtr, - 0) == PAGE_IS_FOUND_AND_NOT_VALID) + n_blocks_read++; + + /* next block */ + if (use_pagemap) { - /* Page is corrupted */ - is_valid = false; + /* exit if pagemap is exhausted */ + if (!datapagemap_next(iter, &blknum)) + break; } + else + blknum++; } - fclose(in); - return is_valid; + /* truncate output file if required */ + if (fseek(out, 0, SEEK_END) != 0) + elog(ERROR, "Cannot seek to end of file position in destination file \"%s\": %s", + to_fullpath, strerror(errno)); + { + long pos = ftell(out); + + if (pos < 0) + elog(ERROR, "Cannot get position in destination file \"%s\": %s", + to_fullpath, strerror(errno)); + + if (pos != file->size) + { + if (fflush(out) != 0) + elog(ERROR, "Cannot flush destination file \"%s\": %s", + to_fullpath, strerror(errno)); + + if (ftruncate(fileno(out), file->size) == -1) + elog(ERROR, "Cannot ftruncate file \"%s\" to size %lu: %s", + to_fullpath, file->size, strerror(errno)); + } + } + + /* cleanup */ + if (fclose(in)) + elog(ERROR, "Cannot close the source file \"%s\": %s", + to_fullpath, strerror(errno)); + + /* close output file */ + if (fclose(out)) + elog(ERROR, "Cannot close the destination file \"%s\": %s", + to_fullpath, strerror(errno)); + + pg_free(iter); + pg_free(in_buf); + pg_free(out_buf); + + return n_blocks_read; } -/* Valiate pages of datafile in backup one by one */ -bool -check_file_pages(pgFile *file, XLogRecPtr stop_lsn, uint32 checksum_version, - uint32 backup_version) +/* + * Attempt to open header file, read content and return as + * array of headers. + * TODO: some access optimizations would be great here: + * less fseeks, buffering, descriptor sharing, etc. + */ +BackupPageHeader2* +get_data_file_headers(HeaderMap *hdr_map, pgFile *file, uint32 backup_version, bool strict) { - size_t read_len = 0; - bool is_valid = true; - FILE *in; - pg_crc32 crc; - bool use_crc32c = backup_version <= 20021 || backup_version >= 20025; + bool success = false; + FILE *in = NULL; + size_t read_len = 0; + pg_crc32 hdr_crc; + BackupPageHeader2 *headers = NULL; + /* header decompression */ + int z_len = 0; + char *zheaders = NULL; + const char *errormsg = NULL; + + if (backup_version < 20400) + return NULL; + + if (file->n_headers <= 0) + return NULL; + + /* TODO: consider to make this descriptor thread-specific */ + in = fopen(hdr_map->path, PG_BINARY_R); + + if (!in) + { + elog(strict ? ERROR : WARNING, "Cannot open header file \"%s\": %s", hdr_map->path, strerror(errno)); + return NULL; + } + /* disable buffering for header file */ + setvbuf(in, NULL, _IONBF, 0); - elog(VERBOSE, "Validate relation blocks for file %s", file->path); + if (fseeko(in, file->hdr_off, SEEK_SET)) + { + elog(strict ? ERROR : WARNING, "Cannot seek to position %llu in page header map \"%s\": %s", + file->hdr_off, hdr_map->path, strerror(errno)); + goto cleanup; + } - in = fopen(file->path, PG_BINARY_R); - if (in == NULL) + /* + * The actual number of headers in header file is n+1, last one is a dummy header, + * used for calculation of read_len for actual last header. + */ + read_len = (file->n_headers+1) * sizeof(BackupPageHeader2); + + /* allocate memory for compressed headers */ + zheaders = pgut_malloc(file->hdr_size); + memset(zheaders, 0, file->hdr_size); + + if (fread(zheaders, 1, file->hdr_size, in) != file->hdr_size) { - if (errno == ENOENT) - { - elog(WARNING, "File \"%s\" is not found", file->path); - return false; - } + elog(strict ? ERROR : WARNING, "Cannot read header file at offset: %llu len: %i \"%s\": %s", + file->hdr_off, file->hdr_size, hdr_map->path, strerror(errno)); + goto cleanup; + } - elog(ERROR, "Cannot open file \"%s\": %s", - file->path, strerror(errno)); + /* allocate memory for uncompressed headers */ + headers = pgut_malloc(read_len); + memset(headers, 0, read_len); + + z_len = do_decompress(headers, read_len, zheaders, file->hdr_size, + ZLIB_COMPRESS, &errormsg); + if (z_len <= 0) + { + if (errormsg) + elog(strict ? ERROR : WARNING, "An error occured during metadata decompression for file \"%s\": %s", + file->rel_path, errormsg); + else + elog(strict ? ERROR : WARNING, "An error occured during metadata decompression for file \"%s\": %i", + file->rel_path, z_len); + + goto cleanup; } - /* calc CRC of backup file */ - INIT_FILE_CRC32(use_crc32c, crc); + /* validate checksum */ + INIT_FILE_CRC32(true, hdr_crc); + COMP_FILE_CRC32(true, hdr_crc, headers, read_len); + FIN_FILE_CRC32(true, hdr_crc); - /* read and validate pages one by one */ - while (true) + if (hdr_crc != file->hdr_crc) { - DataPage compressed_page; /* used as read buffer */ - DataPage page; - BackupPageHeader header; - BlockNumber blknum = 0; + elog(strict ? ERROR : WARNING, "Header map for file \"%s\" crc mismatch \"%s\" " + "offset: %llu, len: %lu, current: %u, expected: %u", + file->rel_path, hdr_map->path, file->hdr_off, read_len, hdr_crc, file->hdr_crc); + goto cleanup; + } - if (interrupted || thread_interrupted) - elog(ERROR, "Interrupted during data file validation"); + success = true; - /* read BackupPageHeader */ - read_len = fread(&header, 1, sizeof(header), in); - if (read_len != sizeof(header)) - { - int errno_tmp = errno; - if (read_len == 0 && feof(in)) - break; /* EOF found */ - else if (read_len != 0 && feof(in)) - elog(WARNING, - "Odd size page found at block %u of \"%s\"", - blknum, file->path); - else - elog(WARNING, "Cannot read header of block %u of \"%s\": %s", - blknum, file->path, strerror(errno_tmp)); - return false; - } +cleanup: - COMP_FILE_CRC32(use_crc32c, crc, &header, read_len); + pg_free(zheaders); + if (in && fclose(in)) + elog(ERROR, "Cannot close file \"%s\"", hdr_map->path); - if (header.block == 0 && header.compressed_size == 0) - { - elog(VERBOSE, "Skip empty block of \"%s\"", file->path); - continue; - } + if (!success) + { + pg_free(headers); + headers = NULL; + } - if (header.block < blknum) - { - elog(WARNING, "Backup is broken at block %u of \"%s\"", - blknum, file->path); - return false; - } + return headers; +} - blknum = header.block; +/* write headers of all blocks belonging to file to header map and + * save its offset and size */ +void +write_page_headers(BackupPageHeader2 *headers, pgFile *file, HeaderMap *hdr_map, bool is_merge) +{ + size_t read_len = 0; + char *map_path = NULL; + /* header compression */ + int z_len = 0; + char *zheaders = NULL; + const char *errormsg = NULL; + + if (file->n_headers <= 0) + return; - if (header.compressed_size == PageIsTruncated) - { - elog(LOG, "Block %u of \"%s\" is truncated", - blknum, file->path); - continue; - } + /* when running merge we must write headers into temp map */ + map_path = (is_merge) ? hdr_map->path_tmp : hdr_map->path; + read_len = (file->n_headers + 1) * sizeof(BackupPageHeader2); - Assert(header.compressed_size <= BLCKSZ); + /* calculate checksums */ + INIT_FILE_CRC32(true, file->hdr_crc); + COMP_FILE_CRC32(true, file->hdr_crc, headers, read_len); + FIN_FILE_CRC32(true, file->hdr_crc); - read_len = fread(compressed_page.data, 1, - MAXALIGN(header.compressed_size), in); - if (read_len != MAXALIGN(header.compressed_size)) - { - elog(WARNING, "Cannot read block %u of \"%s\" read %zu of %d", - blknum, file->path, read_len, header.compressed_size); - return false; - } + zheaders = pgut_malloc(read_len * 2); + memset(zheaders, 0, read_len * 2); - COMP_FILE_CRC32(use_crc32c, crc, compressed_page.data, read_len); + /* compress headers */ + z_len = do_compress(zheaders, read_len * 2, headers, + read_len, ZLIB_COMPRESS, 1, &errormsg); - if (header.compressed_size != BLCKSZ - || page_may_be_compressed(compressed_page.data, file->compress_alg, - backup_version)) - { - int32 uncompressed_size = 0; - const char *errormsg = NULL; + /* writing to header map must be serialized */ + pthread_lock(&(hdr_map->mutex)); /* what if we crash while trying to obtain mutex? */ - uncompressed_size = do_decompress(page.data, BLCKSZ, - compressed_page.data, - header.compressed_size, - file->compress_alg, - &errormsg); - if (uncompressed_size < 0 && errormsg != NULL) - elog(WARNING, "An error occured during decompressing block %u of file \"%s\": %s", - blknum, file->path, errormsg); + if (!hdr_map->fp) + { + elog(LOG, "Creating page header map \"%s\"", map_path); - if (uncompressed_size != BLCKSZ) - { - if (header.compressed_size == BLCKSZ) - { - is_valid = false; - continue; - } - elog(WARNING, "Page of file \"%s\" uncompressed to %d bytes. != BLCKSZ", - file->path, uncompressed_size); - return false; - } + hdr_map->fp = fopen(map_path, PG_BINARY_A); + if (hdr_map->fp == NULL) + elog(ERROR, "Cannot open header file \"%s\": %s", + map_path, strerror(errno)); - if (validate_one_page(page.data, file, blknum, - stop_lsn, checksum_version) == PAGE_IS_FOUND_AND_NOT_VALID) - is_valid = false; - } + /* enable buffering for header file */ + hdr_map->buf = pgut_malloc(LARGE_CHUNK_SIZE); + setvbuf(hdr_map->fp, hdr_map->buf, _IOFBF, LARGE_CHUNK_SIZE); + + /* update file permission */ + if (chmod(map_path, FILE_PERMISSION) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", map_path, + strerror(errno)); + + file->hdr_off = 0; + } + else + file->hdr_off = hdr_map->offset; + + if (z_len <= 0) + { + if (errormsg) + elog(ERROR, "An error occured during compressing metadata for file \"%s\": %s", + file->rel_path, errormsg); else - { - if (validate_one_page(compressed_page.data, file, blknum, - stop_lsn, checksum_version) == PAGE_IS_FOUND_AND_NOT_VALID) - is_valid = false; - } + elog(ERROR, "An error occured during compressing metadata for file \"%s\": %i", + file->rel_path, z_len); } - FIN_FILE_CRC32(use_crc32c, crc); - fclose(in); + elog(VERBOSE, "Writing headers for file \"%s\" offset: %llu, len: %i, crc: %u", + file->rel_path, file->hdr_off, z_len, file->hdr_crc); - if (crc != file->crc) + if (fwrite(zheaders, 1, z_len, hdr_map->fp) != z_len) { - elog(WARNING, "Invalid CRC of backup file \"%s\": %X. Expected %X", - file->path, crc, file->crc); - is_valid = false; + pthread_mutex_unlock(&(hdr_map->mutex)); + elog(ERROR, "Cannot write to file \"%s\": %s", map_path, strerror(errno)); } - return is_valid; + file->hdr_size = z_len; /* save the length of compressed headers */ + hdr_map->offset += z_len; /* update current offset in map */ + + /* End critical section */ + pthread_mutex_unlock(&(hdr_map->mutex)); + + pg_free(zheaders); +} + +void +init_header_map(pgBackup *backup) +{ + backup->hdr_map.fp = NULL; + backup->hdr_map.buf = NULL; + join_path_components(backup->hdr_map.path, backup->root_dir, HEADER_MAP); + join_path_components(backup->hdr_map.path_tmp, backup->root_dir, HEADER_MAP_TMP); + backup->hdr_map.mutex = (pthread_mutex_t)PTHREAD_MUTEX_INITIALIZER; +} + +void +cleanup_header_map(HeaderMap *hdr_map) +{ + /* cleanup descriptor */ + if (hdr_map->fp && fclose(hdr_map->fp)) + elog(ERROR, "Cannot close file \"%s\"", hdr_map->path); + hdr_map->fp = NULL; + hdr_map->offset = 0; + pg_free(hdr_map->buf); + hdr_map->buf = NULL; } diff --git a/src/datapagemap.c b/src/datapagemap.c new file mode 100644 index 000000000..7e4202a72 --- /dev/null +++ b/src/datapagemap.c @@ -0,0 +1,113 @@ +/*------------------------------------------------------------------------- + * + * datapagemap.c + * A data structure for keeping track of data pages that have changed. + * + * This is a fairly simple bitmap. + * + * Copyright (c) 2013-2019, PostgreSQL Global Development Group + * + *------------------------------------------------------------------------- + */ + +#include "postgres_fe.h" + +#include "datapagemap.h" + +struct datapagemap_iterator +{ + datapagemap_t *map; + BlockNumber nextblkno; +}; + +/***** + * Public functions + */ + +/* + * Add a block to the bitmap. + */ +void +datapagemap_add(datapagemap_t *map, BlockNumber blkno) +{ + int offset; + int bitno; + int oldsize = map->bitmapsize; + + offset = blkno / 8; + bitno = blkno % 8; + + /* enlarge or create bitmap if needed */ + if (oldsize <= offset) + { + int newsize; + + /* + * The minimum to hold the new bit is offset + 1. But add some + * headroom, so that we don't need to repeatedly enlarge the bitmap in + * the common case that blocks are modified in order, from beginning + * of a relation to the end. + */ + newsize = (oldsize == 0) ? 16 : oldsize; + while (newsize <= offset) { + newsize <<= 1; + } + + map->bitmap = pg_realloc(map->bitmap, newsize); + + /* zero out the newly allocated region */ + memset(&map->bitmap[oldsize], 0, newsize - oldsize); + + map->bitmapsize = newsize; + } + + /* Set the bit */ + map->bitmap[offset] |= (1 << bitno); +} + +/* + * Start iterating through all entries in the page map. + * + * After datapagemap_iterate, call datapagemap_next to return the entries, + * until it returns false. After you're done, use pg_free() to destroy the + * iterator. + */ +datapagemap_iterator_t * +datapagemap_iterate(datapagemap_t *map) +{ + datapagemap_iterator_t *iter; + + iter = pg_malloc(sizeof(datapagemap_iterator_t)); + iter->map = map; + iter->nextblkno = 0; + + return iter; +} + +bool +datapagemap_next(datapagemap_iterator_t *iter, BlockNumber *blkno) +{ + datapagemap_t *map = iter->map; + + for (;;) + { + BlockNumber blk = iter->nextblkno; + int nextoff = blk / 8; + int bitno = blk % 8; + + if (nextoff >= map->bitmapsize) + break; + + iter->nextblkno++; + + if (map->bitmap[nextoff] & (1 << bitno)) + { + *blkno = blk; + return true; + } + } + + /* no more set bits in this bitmap. */ + return false; +} + diff --git a/src/datapagemap.h b/src/datapagemap.h new file mode 100644 index 000000000..6ad7a6204 --- /dev/null +++ b/src/datapagemap.h @@ -0,0 +1,34 @@ +/*------------------------------------------------------------------------- + * + * datapagemap.h + * + * Copyright (c) 2013-2019, PostgreSQL Global Development Group + * + *------------------------------------------------------------------------- + */ +#ifndef DATAPAGEMAP_H +#define DATAPAGEMAP_H + +#if PG_VERSION_NUM < 160000 +#include "storage/relfilenode.h" +#else +#include "storage/relfilelocator.h" +#define RelFileNode RelFileLocator +#endif +#include "storage/block.h" + + +struct datapagemap +{ + char *bitmap; + int bitmapsize; +}; + +typedef struct datapagemap datapagemap_t; +typedef struct datapagemap_iterator datapagemap_iterator_t; + +extern void datapagemap_add(datapagemap_t *map, BlockNumber blkno); +extern datapagemap_iterator_t *datapagemap_iterate(datapagemap_t *map); +extern bool datapagemap_next(datapagemap_iterator_t *iter, BlockNumber *blkno); + +#endif /* DATAPAGEMAP_H */ diff --git a/src/delete.c b/src/delete.c index 83be5d522..f48ecc95f 100644 --- a/src/delete.c +++ b/src/delete.c @@ -14,30 +14,33 @@ #include #include -static void delete_walfiles(XLogRecPtr oldest_lsn, TimeLineID oldest_tli, - uint32 xlog_seg_size); +static void delete_walfiles_in_tli(InstanceState *instanceState, XLogRecPtr keep_lsn, timelineInfo *tli, + uint32 xlog_seg_size, bool dry_run); static void do_retention_internal(parray *backup_list, parray *to_keep_list, parray *to_purge_list); -static void do_retention_merge(parray *backup_list, parray *to_keep_list, - parray *to_purge_list); +static void do_retention_merge(InstanceState *instanceState, parray *backup_list, + parray *to_keep_list, parray *to_purge_list, + bool no_validate, bool no_sync); static void do_retention_purge(parray *to_keep_list, parray *to_purge_list); -static void do_retention_wal(void); +static void do_retention_wal(InstanceState *instanceState, bool dry_run); +// TODO: more useful messages for dry run. static bool backup_deleted = false; /* At least one backup was deleted */ static bool backup_merged = false; /* At least one merge was enacted */ +static bool wal_deleted = false; /* At least one WAL segments was deleted */ void -do_delete(time_t backup_id) +do_delete(InstanceState *instanceState, time_t backup_id) { int i; parray *backup_list, *delete_list; pgBackup *target_backup = NULL; - XLogRecPtr oldest_lsn = InvalidXLogRecPtr; - TimeLineID oldest_tli = 0; + int64 size_to_delete = 0; + char size_to_delete_pretty[20]; /* Get complete list of backups */ - backup_list = catalog_get_backup_list(INVALID_BACKUP_ID); + backup_list = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); delete_list = parray_new(); @@ -63,49 +66,50 @@ do_delete(time_t backup_id) pgBackup *backup = (pgBackup *) parray_get(backup_list, i); /* check if backup is descendant of delete target */ - if (is_parent(target_backup->start_time, backup, false)) + if (is_parent(target_backup->start_time, backup, true)) + { parray_append(delete_list, backup); - } - parray_append(delete_list, target_backup); - - /* Lock marked for delete backups */ - catalog_lock_backup_list(delete_list, parray_num(delete_list) - 1, 0); - /* Delete backups from the end of list */ - for (i = (int) parray_num(delete_list) - 1; i >= 0; i--) - { - pgBackup *backup = (pgBackup *) parray_get(delete_list, (size_t) i); - - if (interrupted) - elog(ERROR, "interrupted during delete backup"); + elog(LOG, "Backup %s %s be deleted", + backup_id_of(backup), dry_run? "can":"will"); - delete_backup_files(backup); + size_to_delete += backup->data_bytes; + if (backup->stream) + size_to_delete += backup->wal_bytes; + } } - parray_free(delete_list); + /* Report the resident size to delete */ + if (size_to_delete >= 0) + { + pretty_size(size_to_delete, size_to_delete_pretty, lengthof(size_to_delete_pretty)); + elog(INFO, "Resident data size to free by delete of backup %s : %s", + backup_id_of(target_backup), size_to_delete_pretty); + } - /* Clean WAL segments */ - if (delete_wal) + if (!dry_run) { - Assert(target_backup); + /* Lock marked for delete backups */ + catalog_lock_backup_list(delete_list, parray_num(delete_list) - 1, 0, false, true); - /* Find oldest LSN, used by backups */ - for (i = (int) parray_num(backup_list) - 1; i >= 0; i--) + /* Delete backups from the end of list */ + for (i = (int) parray_num(delete_list) - 1; i >= 0; i--) { - pgBackup *backup = (pgBackup *) parray_get(backup_list, (size_t) i); + pgBackup *backup = (pgBackup *) parray_get(delete_list, (size_t) i); - if (backup->status == BACKUP_STATUS_OK || backup->status == BACKUP_STATUS_DONE) - { - oldest_lsn = backup->start_lsn; - oldest_tli = backup->tli; - break; - } - } + if (interrupted) + elog(ERROR, "interrupted during delete backup"); - delete_walfiles(oldest_lsn, oldest_tli, instance_config.xlog_seg_size); + delete_backup_files(backup); + } } + /* Clean WAL segments */ + if (delete_wal) + do_retention_wal(instanceState, dry_run); + /* cleanup */ + parray_free(delete_list); parray_walk(backup_list, pgBackupFree); parray_free(backup_list); } @@ -120,7 +124,7 @@ do_delete(time_t backup_id) * which FULL backup should be keeped for redundancy obligation(only valid do), * but if invalid backup is not guarded by retention - it is removed */ -int do_retention(void) +void do_retention(InstanceState *instanceState, bool no_validate, bool no_sync) { parray *backup_list = NULL; parray *to_keep_list = parray_new(); @@ -132,8 +136,11 @@ int do_retention(void) backup_deleted = false; backup_merged = false; + /* For now retention is possible only locally */ + MyLocation = FIO_LOCAL_HOST; + /* Get a complete list of backups. */ - backup_list = catalog_get_backup_list(INVALID_BACKUP_ID); + backup_list = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); if (parray_num(backup_list) == 0) backup_list_is_empty = true; @@ -151,7 +158,13 @@ int do_retention(void) /* Retention is disabled but we still can cleanup wal */ elog(WARNING, "Retention policy is not set"); if (!delete_wal) - return 0; + { + parray_walk(backup_list, pgBackupFree); + parray_free(backup_list); + parray_free(to_keep_list); + parray_free(to_purge_list); + return; + } } else /* At least one retention policy is active */ @@ -166,14 +179,14 @@ int do_retention(void) do_retention_internal(backup_list, to_keep_list, to_purge_list); if (merge_expired && !dry_run && !backup_list_is_empty) - do_retention_merge(backup_list, to_keep_list, to_purge_list); + do_retention_merge(instanceState, backup_list, to_keep_list, to_purge_list, no_validate, no_sync); if (delete_expired && !dry_run && !backup_list_is_empty) do_retention_purge(to_keep_list, to_purge_list); /* TODO: some sort of dry run for delete_wal */ - if (delete_wal && !dry_run) - do_retention_wal(); + if (delete_wal) + do_retention_wal(instanceState, dry_run); /* TODO: consider dry-run flag */ @@ -185,14 +198,14 @@ int do_retention(void) else elog(INFO, "There are no backups to delete by retention policy"); + if (!wal_deleted) + elog(INFO, "There is no WAL to purge by retention policy"); + /* Cleanup */ parray_walk(backup_list, pgBackupFree); parray_free(backup_list); parray_free(to_keep_list); parray_free(to_purge_list); - - return 0; - } /* Evaluate every backup by retention policies and populate purge and keep lists. @@ -203,7 +216,6 @@ static void do_retention_internal(parray *backup_list, parray *to_keep_list, parray *to_purge_list) { int i; - time_t current_time; parray *redundancy_full_backup_list = NULL; @@ -215,9 +227,6 @@ do_retention_internal(parray *backup_list, parray *to_keep_list, parray *to_purg /* For fancy reporting */ uint32 actual_window = 0; - /* Get current time */ - current_time = time(NULL); - /* Calculate n_full_backups and days_threshold */ if (instance_config.retention_redundancy > 0) { @@ -225,22 +234,23 @@ do_retention_internal(parray *backup_list, parray *to_keep_list, parray *to_purg { pgBackup *backup = (pgBackup *) parray_get(backup_list, i); - /* Consider only valid FULL backups for Redundancy */ - if (instance_config.retention_redundancy > 0 && - backup->backup_mode == BACKUP_MODE_FULL && - (backup->status == BACKUP_STATUS_OK || - backup->status == BACKUP_STATUS_DONE)) + if (backup->backup_mode == BACKUP_MODE_FULL) { - n_full_backups++; - /* Add every FULL backup that satisfy Redundancy policy to separate list */ - if (n_full_backups <= instance_config.retention_redundancy) + if (n_full_backups < instance_config.retention_redundancy) { if (!redundancy_full_backup_list) redundancy_full_backup_list = parray_new(); parray_append(redundancy_full_backup_list, backup); } + + /* Consider only valid FULL backups for Redundancy fulfillment */ + if (backup->status == BACKUP_STATUS_OK || + backup->status == BACKUP_STATUS_DONE) + { + n_full_backups++; + } } } /* Sort list of full backups to keep */ @@ -259,6 +269,7 @@ do_retention_internal(parray *backup_list, parray *to_keep_list, parray *to_purg { bool redundancy_keep = false; + time_t backup_time = 0; pgBackup *backup = (pgBackup *) parray_get(backup_list, (size_t) i); /* check if backup`s FULL ancestor is in redundancy list */ @@ -280,10 +291,16 @@ do_retention_internal(parray *backup_list, parray *to_keep_list, parray *to_purg cur_full_backup_num++; } - /* Check if backup in needed by retention policy - * TODO: consider that ERROR backup most likely to have recovery_time == 0 + /* Invalid and running backups most likely to have recovery_time == 0, + * so in this case use start_time instead. */ - if ((days_threshold == 0 || (days_threshold > backup->recovery_time)) && + if (backup->recovery_time) + backup_time = backup->recovery_time; + else + backup_time = backup->start_time; + + /* Check if backup in needed by retention policy */ + if ((days_threshold == 0 || (days_threshold > backup_time)) && (instance_config.retention_redundancy == 0 || !redundancy_keep)) { /* This backup is not guarded by retention @@ -300,8 +317,22 @@ do_retention_internal(parray *backup_list, parray *to_keep_list, parray *to_purg * FULL CORRUPT out of retention */ + /* Save backup from purge if backup is pinned and + * expire date is not yet due. + */ + if ((backup->expire_time > 0) && + (backup->expire_time > current_time)) + { + char expire_timestamp[100]; + time2iso(expire_timestamp, lengthof(expire_timestamp), backup->expire_time, false); + + elog(LOG, "Backup %s is pinned until '%s', retain", + backup_id_of(backup), expire_timestamp); + continue; + } + /* Add backup to purge_list */ - elog(VERBOSE, "Mark backup %s for purge.", base36enc(backup->start_time)); + elog(VERBOSE, "Mark backup %s for purge.", backup_id_of(backup)); parray_append(to_purge_list, backup); continue; } @@ -325,10 +356,12 @@ do_retention_internal(parray *backup_list, parray *to_keep_list, parray *to_purg { pgBackup *backup = (pgBackup *) parray_get(backup_list, i); - /* Do not keep invalid backups by retention */ - if (backup->status != BACKUP_STATUS_OK && - backup->status != BACKUP_STATUS_DONE) - continue; + /* Do not keep invalid backups by retention + * Turns out it was not a very good idea - [Issue #114] + */ + //if (backup->status != BACKUP_STATUS_OK && + // backup->status != BACKUP_STATUS_DONE) + // continue; /* only incremental backups should be in keep list */ if (backup->backup_mode == BACKUP_MODE_FULL) @@ -361,6 +394,7 @@ do_retention_internal(parray *backup_list, parray *to_keep_list, parray *to_purg for (i = 0; i < parray_num(backup_list); i++) { char *action = "Active"; + uint32 pinning_window = 0; pgBackup *backup = (pgBackup *) parray_get(backup_list, i); @@ -370,26 +404,36 @@ do_retention_internal(parray *backup_list, parray *to_keep_list, parray *to_purg if (backup->recovery_time == 0) actual_window = 0; else - actual_window = (current_time - backup->recovery_time)/(60 * 60 * 24); + actual_window = (current_time - backup->recovery_time)/(3600 * 24); + + /* For pinned backups show expire date */ + if (backup->expire_time > 0 && backup->expire_time > backup->recovery_time) + pinning_window = (backup->expire_time - backup->recovery_time)/(3600 * 24); /* TODO: add ancestor(chain full backup) ID */ elog(INFO, "Backup %s, mode: %s, status: %s. Redundancy: %i/%i, Time Window: %ud/%ud. %s", - base36enc(backup->start_time), - pgBackupGetBackupMode(backup), + backup_id_of(backup), + pgBackupGetBackupMode(backup, false), status2str(backup->status), cur_full_backup_num, instance_config.retention_redundancy, - actual_window, instance_config.retention_window, + actual_window, + pinning_window ? pinning_window : instance_config.retention_window, action); - if (backup->backup_mode == BACKUP_MODE_FULL) + /* Only valid full backups are count to something */ + if (backup->backup_mode == BACKUP_MODE_FULL && + (backup->status == BACKUP_STATUS_OK || + backup->status == BACKUP_STATUS_DONE)) cur_full_backup_num++; } } /* Merge partially expired incremental chains */ static void -do_retention_merge(parray *backup_list, parray *to_keep_list, parray *to_purge_list) +do_retention_merge(InstanceState *instanceState, parray *backup_list, + parray *to_keep_list, parray *to_purge_list, + bool no_validate, bool no_sync) { int i; int j; @@ -413,7 +457,6 @@ do_retention_merge(parray *backup_list, parray *to_keep_list, parray *to_purge_l /* Merging happens here */ for (i = 0; i < parray_num(to_keep_list); i++) { - char *keep_backup_id = NULL; pgBackup *full_backup = NULL; parray *merge_list = NULL; @@ -423,7 +466,7 @@ do_retention_merge(parray *backup_list, parray *to_keep_list, parray *to_purge_l if (!keep_backup) continue; - elog(INFO, "Consider backup %s for merge", base36enc(keep_backup->start_time)); + elog(INFO, "Consider backup %s for merge", backup_id_of(keep_backup)); /* Got valid incremental backup, find its FULL ancestor */ full_backup = find_parent_full_backup(keep_backup); @@ -431,7 +474,7 @@ do_retention_merge(parray *backup_list, parray *to_keep_list, parray *to_purge_l /* Failed to find parent */ if (!full_backup) { - elog(WARNING, "Failed to find FULL parent for %s", base36enc(keep_backup->start_time)); + elog(WARNING, "Failed to find FULL parent for %s", backup_id_of(keep_backup)); continue; } @@ -441,7 +484,7 @@ do_retention_merge(parray *backup_list, parray *to_keep_list, parray *to_purge_l pgBackupCompareIdDesc)) { elog(WARNING, "Skip backup %s for merging, " - "because his FULL parent is not marked for purge", base36enc(keep_backup->start_time)); + "because his FULL parent is not marked for purge", backup_id_of(keep_backup)); continue; } @@ -450,15 +493,14 @@ do_retention_merge(parray *backup_list, parray *to_keep_list, parray *to_purge_l * backups from purge_list. */ - keep_backup_id = base36enc_dup(keep_backup->start_time); - elog(INFO, "Merge incremental chain between FULL backup %s and backup %s", - base36enc(full_backup->start_time), keep_backup_id); - pg_free(keep_backup_id); + elog(INFO, "Merge incremental chain between full backup %s and backup %s", + backup_id_of(full_backup), + backup_id_of(keep_backup)); merge_list = parray_new(); /* Form up a merge list */ - while(keep_backup->parent_backup_link) + while (keep_backup->parent_backup_link) { parray_append(merge_list, keep_backup); keep_backup = keep_backup->parent_backup_link; @@ -482,8 +524,21 @@ do_retention_merge(parray *backup_list, parray *to_keep_list, parray *to_purge_l parray_rm(to_purge_list, full_backup, pgBackupCompareId); /* Lock merge chain */ - catalog_lock_backup_list(merge_list, parray_num(merge_list) - 1, 0); + catalog_lock_backup_list(merge_list, parray_num(merge_list) - 1, 0, true, true); + + /* Consider this extreme case */ + // PAGEa1 PAGEb1 both valid + // \ / + // FULL + /* Check that FULL backup do not has multiple descendants + * full_backup always point to current full_backup after merge + */ +// if (is_prolific(backup_list, full_backup)) +// { +// elog(WARNING, "Backup %s has multiple valid descendants. " +// "Automatic merge is not possible.", backup_id_of(full_backup)); +// } /* Merge list example: * 0 PAGE3 @@ -491,37 +546,29 @@ do_retention_merge(parray *backup_list, parray *to_keep_list, parray *to_purge_l * 2 PAGE1 * 3 FULL * - * Consequentially merge incremental backups from PAGE1 to PAGE3 - * into FULL. + * Merge incremental chain from PAGE3 into FULL. */ + keep_backup = parray_get(merge_list, 0); + merge_chain(instanceState, merge_list, full_backup, keep_backup, no_validate, no_sync); + backup_merged = true; for (j = parray_num(merge_list) - 2; j >= 0; j--) { - pgBackup *from_backup = (pgBackup *) parray_get(merge_list, j); - - - /* Consider this extreme case */ - // PAGEa1 PAGEb1 both valid - // \ / - // FULL - - /* Check that FULL backup do not has multiple descendants - * full_backup always point to current full_backup after merge - */ - if (is_prolific(backup_list, full_backup)) - { - elog(WARNING, "Backup %s has multiple valid descendants. " - "Automatic merge is not possible.", base36enc(full_backup->start_time)); - break; - } - - merge_backups(full_backup, from_backup); - backup_merged = true; + pgBackup *tmp_backup = (pgBackup *) parray_get(merge_list, j); /* Try to remove merged incremental backup from both keep and purge lists */ - parray_rm(to_purge_list, from_backup, pgBackupCompareId); - parray_set(to_keep_list, i, NULL); + parray_rm(to_purge_list, tmp_backup, pgBackupCompareId); + for (i = 0; i < parray_num(to_keep_list); i++) + if (parray_get(to_keep_list, i) == tmp_backup) + { + parray_set(to_keep_list, i, NULL); + break; + } } + if (!no_validate) + pgBackupValidate(full_backup, NULL); + if (full_backup->status == BACKUP_STATUS_CORRUPT) + elog(ERROR, "Merging of backup %s failed", backup_id_of(full_backup)); /* Cleanup */ parray_free(merge_list); @@ -553,7 +600,7 @@ do_retention_purge(parray *to_keep_list, parray *to_purge_list) pgBackup *delete_backup = (pgBackup *) parray_get(to_purge_list, j); elog(LOG, "Consider backup %s for purge", - base36enc(delete_backup->start_time)); + backup_id_of(delete_backup)); /* Evaluate marked for delete backup against every backup in keep list. * If marked for delete backup is recognized as parent of one of those, @@ -561,8 +608,6 @@ do_retention_purge(parray *to_keep_list, parray *to_purge_list) */ for (i = 0; i < parray_num(to_keep_list); i++) { - char *keeped_backup_id; - pgBackup *keep_backup = (pgBackup *) parray_get(to_keep_list, i); /* item could have been nullified in merge */ @@ -573,24 +618,22 @@ do_retention_purge(parray *to_keep_list, parray *to_purge_list) if (keep_backup->backup_mode == BACKUP_MODE_FULL) continue; - keeped_backup_id = base36enc_dup(keep_backup->start_time); - elog(LOG, "Check if backup %s is parent of backup %s", - base36enc(delete_backup->start_time), keeped_backup_id); + backup_id_of(delete_backup), + backup_id_of(keep_backup)); if (is_parent(delete_backup->start_time, keep_backup, true)) { /* We must not delete this backup, evict it from purge list */ - elog(LOG, "Retain backup %s from purge because his " + elog(LOG, "Retain backup %s because his " "descendant %s is guarded by retention", - base36enc(delete_backup->start_time), keeped_backup_id); + backup_id_of(delete_backup), + backup_id_of(keep_backup)); purge = false; - pg_free(keeped_backup_id); break; } - pg_free(keeped_backup_id); } /* Retain backup */ @@ -598,11 +641,11 @@ do_retention_purge(parray *to_keep_list, parray *to_purge_list) continue; /* Actual purge */ - if (!lock_backup(delete_backup)) + if (!lock_backup(delete_backup, false, true)) { /* If the backup still is used, do not interrupt and go to the next */ elog(WARNING, "Cannot lock backup %s directory, skip purging", - base36enc(delete_backup->start_time)); + backup_id_of(delete_backup)); continue; } @@ -613,57 +656,91 @@ do_retention_purge(parray *to_keep_list, parray *to_purge_list) } } -/* Purge WAL */ +/* + * Purge WAL + * Iterate over timelines + * Look for WAL segment not reachable from existing backups + * and delete them. + */ static void -do_retention_wal(void) +do_retention_wal(InstanceState *instanceState, bool dry_run) { - parray *backup_list = NULL; - - XLogRecPtr oldest_lsn = InvalidXLogRecPtr; - TimeLineID oldest_tli = 0; - bool backup_list_is_empty = false; + parray *tli_list; + int i; - /* Get list of backups. */ - backup_list = catalog_get_backup_list(INVALID_BACKUP_ID); + //TODO check that instanceState is not NULL + tli_list = catalog_get_timelines(instanceState, &instance_config); - if (parray_num(backup_list) == 0) - backup_list_is_empty = true; - - /* Save LSN and Timeline to remove unnecessary WAL segments */ - if (!backup_list_is_empty) + for (i = 0; i < parray_num(tli_list); i++) { - pgBackup *backup = NULL; - /* Get LSN and TLI of oldest alive backup */ - backup = (pgBackup *) parray_get(backup_list, parray_num(backup_list) -1); + timelineInfo *tlinfo = (timelineInfo *) parray_get(tli_list, i); - oldest_tli = backup->tli; - oldest_lsn = backup->start_lsn; - } + /* + * Empty timeline (only mentioned in timeline history file) + * has nothing to cleanup. + */ + if (tlinfo->n_xlog_files == 0 && parray_num(tlinfo->xlog_filelist) == 0) + continue; - /* Be paranoid */ - if (!backup_list_is_empty && XLogRecPtrIsInvalid(oldest_lsn)) - elog(ERROR, "Not going to purge WAL because LSN is invalid"); + /* + * If closest backup exists, then timeline is reachable from + * at least one backup and no file should be removed. + * Unless wal-depth is enabled. + */ + if ((tlinfo->closest_backup) && instance_config.wal_depth == 0) + continue; - /* Purge WAL files */ - delete_walfiles(oldest_lsn, oldest_tli, instance_config.xlog_seg_size); + /* WAL retention keeps this timeline from purge */ + if (tlinfo->anchor_tli > 0 && tlinfo->anchor_tli != tlinfo->tli) + continue; - /* Cleanup */ - parray_walk(backup_list, pgBackupFree); - parray_free(backup_list); + /* + * Purge all WAL segments before START LSN of oldest backup. + * If timeline doesn't have a backup, then whole timeline + * can be safely purged. + * Note, that oldest_backup is not necessarily valid here, + * but still we keep wal for it. + * If wal-depth is enabled then use anchor_lsn instead + * of oldest_backup. + */ + if (tlinfo->oldest_backup) + { + if (!(XLogRecPtrIsInvalid(tlinfo->anchor_lsn))) + { + delete_walfiles_in_tli(instanceState, tlinfo->anchor_lsn, + tlinfo, instance_config.xlog_seg_size, dry_run); + } + else + { + delete_walfiles_in_tli(instanceState, tlinfo->oldest_backup->start_lsn, + tlinfo, instance_config.xlog_seg_size, dry_run); + } + } + else + { + if (!(XLogRecPtrIsInvalid(tlinfo->anchor_lsn))) + delete_walfiles_in_tli(instanceState, tlinfo->anchor_lsn, + tlinfo, instance_config.xlog_seg_size, dry_run); + else + delete_walfiles_in_tli(instanceState, InvalidXLogRecPtr, + tlinfo, instance_config.xlog_seg_size, dry_run); + } + } } /* * Delete backup files of the backup and update the status of the backup to * BACKUP_STATUS_DELETED. + * TODO: delete files on multiple threads */ void delete_backup_files(pgBackup *backup) { size_t i; - char path[MAXPGPATH]; char timestamp[100]; - parray *files; + parray *files; size_t num_files; + char full_path[MAXPGPATH]; /* * If the backup was deleted already, there is nothing to do. @@ -671,41 +748,45 @@ delete_backup_files(pgBackup *backup) if (backup->status == BACKUP_STATUS_DELETED) { elog(WARNING, "Backup %s already deleted", - base36enc(backup->start_time)); + backup_id_of(backup)); return; } - time2iso(timestamp, lengthof(timestamp), backup->recovery_time); + if (backup->recovery_time) + time2iso(timestamp, lengthof(timestamp), backup->recovery_time, false); + else + time2iso(timestamp, lengthof(timestamp), backup->start_time, false); elog(INFO, "Delete: %s %s", - base36enc(backup->start_time), timestamp); + backup_id_of(backup), timestamp); /* * Update STATUS to BACKUP_STATUS_DELETING in preparation for the case which * the error occurs before deleting all backup files. */ - write_backup_status(backup, BACKUP_STATUS_DELETING); + write_backup_status(backup, BACKUP_STATUS_DELETING, false); /* list files to be deleted */ files = parray_new(); - pgBackupGetPath(backup, path, lengthof(path), NULL); - dir_list_file(files, path, false, true, true, 0, FIO_BACKUP_HOST); + dir_list_file(files, backup->root_dir, false, false, true, false, false, 0, FIO_BACKUP_HOST); /* delete leaf node first */ - parray_qsort(files, pgFileComparePathDesc); + parray_qsort(files, pgFileCompareRelPathWithExternalDesc); num_files = parray_num(files); for (i = 0; i < num_files; i++) { pgFile *file = (pgFile *) parray_get(files, i); - if (progress) - elog(INFO, "Progress: (%zd/%zd). Process file \"%s\"", - i + 1, num_files, file->path); + join_path_components(full_path, backup->root_dir, file->rel_path); if (interrupted) elog(ERROR, "interrupted during delete backup"); - pgFileDelete(file); + if (progress) + elog(INFO, "Progress: (%zd/%zd). Delete file \"%s\"", + i + 1, num_files, full_path); + + pgFileDelete(file->mode, full_path); } parray_walk(files, pgFileFree); @@ -716,133 +797,198 @@ delete_backup_files(pgBackup *backup) } /* - * Deletes WAL segments up to oldest_lsn or all WAL segments (if all backups - * was deleted and so oldest_lsn is invalid). + * Purge WAL archive. One timeline at a time. + * If 'keep_lsn' is InvalidXLogRecPtr, then whole timeline can be purged + * If 'keep_lsn' is valid LSN, then every lesser segment can be purged. + * If 'dry_run' is set, then don`t actually delete anything. * - * oldest_lsn - if valid, function deletes WAL segments, which contain lsn - * older than oldest_lsn. If it is invalid function deletes all WAL segments. - * oldest_tli - is used to construct oldest WAL segment in addition to - * oldest_lsn. + * Case 1: + * archive is not empty, 'keep_lsn' is valid and we can delete something. + * Case 2: + * archive is not empty, 'keep_lsn' is valid and prevening us from deleting anything. + * Case 3: + * archive is not empty, 'keep_lsn' is invalid, drop all WAL files in archive, + * belonging to the timeline. + * Case 4: + * archive is empty, 'keep_lsn' is valid, assume corruption of WAL archive. + * Case 5: + * archive is empty, 'keep_lsn' is invalid, drop backup history files + * and partial WAL segments in archive. + * + * Q: Maybe we should stop treating partial WAL segments as second-class citizens? */ static void -delete_walfiles(XLogRecPtr oldest_lsn, TimeLineID oldest_tli, - uint32 xlog_seg_size) +delete_walfiles_in_tli(InstanceState *instanceState, XLogRecPtr keep_lsn, timelineInfo *tlinfo, + uint32 xlog_seg_size, bool dry_run) { - XLogSegNo targetSegNo; - char oldestSegmentNeeded[MAXFNAMELEN]; - DIR *arcdir; - struct dirent *arcde; - char wal_file[MAXPGPATH]; - char max_wal_file[MAXPGPATH]; - char min_wal_file[MAXPGPATH]; - int rc; + XLogSegNo FirstToDeleteSegNo; + XLogSegNo OldestToKeepSegNo = 0; + char first_to_del_str[MAXFNAMELEN]; + char oldest_to_keep_str[MAXFNAMELEN]; + int i; + size_t wal_size_logical = 0; + size_t wal_size_actual = 0; + char wal_pretty_size[20]; + bool purge_all = false; - max_wal_file[0] = '\0'; - min_wal_file[0] = '\0'; - if (!XLogRecPtrIsInvalid(oldest_lsn)) + /* Timeline is completely empty */ + if (parray_num(tlinfo->xlog_filelist) == 0) { - GetXLogSegNo(oldest_lsn, targetSegNo, xlog_seg_size); - GetXLogFileName(oldestSegmentNeeded, oldest_tli, targetSegNo, - xlog_seg_size); + elog(INFO, "Timeline %i is empty, nothing to remove", tlinfo->tli); + return; + } - elog(LOG, "removing WAL segments older than %s", oldestSegmentNeeded); + if (XLogRecPtrIsInvalid(keep_lsn)) + { + /* Drop all files in timeline */ + elog(INFO, "On timeline %i all files %s be removed", + tlinfo->tli, dry_run?"can":"will"); + FirstToDeleteSegNo = tlinfo->begin_segno; + OldestToKeepSegNo = tlinfo->end_segno; + purge_all = true; } else - elog(LOG, "removing all WAL segments"); + { + /* Drop all segments between begin_segno and segment with keep_lsn (excluding) */ + FirstToDeleteSegNo = tlinfo->begin_segno; + GetXLogSegNo(keep_lsn, OldestToKeepSegNo, xlog_seg_size); + } - /* - * Now it is time to do the actual work and to remove all the segments - * not needed anymore. - */ - if ((arcdir = opendir(arclog_path)) != NULL) + if (OldestToKeepSegNo > 0 && OldestToKeepSegNo > FirstToDeleteSegNo) { - while (errno = 0, (arcde = readdir(arcdir)) != NULL) + /* translate segno number into human readable format */ + GetXLogFileName(first_to_del_str, tlinfo->tli, FirstToDeleteSegNo, xlog_seg_size); + GetXLogFileName(oldest_to_keep_str, tlinfo->tli, OldestToKeepSegNo, xlog_seg_size); + + elog(INFO, "On timeline %i WAL segments between %s and %s %s be removed", + tlinfo->tli, first_to_del_str, + oldest_to_keep_str, dry_run?"can":"will"); + } + + /* sanity */ + if (OldestToKeepSegNo > FirstToDeleteSegNo) + { + wal_size_logical = (OldestToKeepSegNo - FirstToDeleteSegNo) * xlog_seg_size; + + /* In case of 'purge all' scenario OldestToKeepSegNo will be deleted too */ + if (purge_all) + wal_size_logical += xlog_seg_size; + } + else if (OldestToKeepSegNo < FirstToDeleteSegNo) + { + /* It is actually possible for OldestToKeepSegNo to be less than FirstToDeleteSegNo + * in case of : + * 1. WAL archive corruption. + * 2. There is no actual WAL archive to speak of and + * 'keep_lsn' is coming from STREAM backup. + */ + + if (FirstToDeleteSegNo > 0 && OldestToKeepSegNo > 0) { - /* - * We ignore the timeline part of the WAL segment identifiers in - * deciding whether a segment is still needed. This ensures that - * we won't prematurely remove a segment from a parent timeline. - * We could probably be a little more proactive about removing - * segments of non-parent timelines, but that would be a whole lot - * more complicated. - * - * We use the alphanumeric sorting property of the filenames to - * decide which ones are earlier than the exclusiveCleanupFileName - * file. Note that this means files are not removed in the order - * they were originally written, in case this worries you. - * - * We also should not forget that WAL segment can be compressed. - */ - if (IsXLogFileName(arcde->d_name) || - IsPartialXLogFileName(arcde->d_name) || - IsBackupHistoryFileName(arcde->d_name) || - IsCompressedXLogFileName(arcde->d_name)) + GetXLogFileName(first_to_del_str, tlinfo->tli, FirstToDeleteSegNo, xlog_seg_size); + GetXLogFileName(oldest_to_keep_str, tlinfo->tli, OldestToKeepSegNo, xlog_seg_size); + + elog(LOG, "On timeline %i first segment %s is greater than oldest segment to keep %s", + tlinfo->tli, first_to_del_str, oldest_to_keep_str); + } + } + else if (OldestToKeepSegNo == FirstToDeleteSegNo && !purge_all) + { + /* 'Nothing to delete' scenario because of 'keep_lsn' + * with possible exception of partial and backup history files. + */ + elog(INFO, "Nothing to remove on timeline %i", tlinfo->tli); + } + + /* Report the logical size to delete */ + if (wal_size_logical > 0) + { + pretty_size(wal_size_logical, wal_pretty_size, lengthof(wal_pretty_size)); + elog(INFO, "Logical WAL size to remove on timeline %i : %s", + tlinfo->tli, wal_pretty_size); + } + + /* Calculate the actual size to delete */ + for (i = 0; i < parray_num(tlinfo->xlog_filelist); i++) + { + xlogFile *wal_file = (xlogFile *) parray_get(tlinfo->xlog_filelist, i); + + if (purge_all || wal_file->segno < OldestToKeepSegNo) + wal_size_actual += wal_file->file.size; + } + + /* Report the actual size to delete */ + if (wal_size_actual > 0) + { + pretty_size(wal_size_actual, wal_pretty_size, lengthof(wal_pretty_size)); + elog(INFO, "Resident WAL size to free on timeline %i : %s", + tlinfo->tli, wal_pretty_size); + } + + if (dry_run) + return; + + for (i = 0; i < parray_num(tlinfo->xlog_filelist); i++) + { + xlogFile *wal_file = (xlogFile *) parray_get(tlinfo->xlog_filelist, i); + + if (interrupted) + elog(ERROR, "interrupted during WAL archive purge"); + + /* Any segment equal or greater than EndSegNo must be kept + * unless it`s a 'purge all' scenario. + */ + if (purge_all || wal_file->segno < OldestToKeepSegNo) + { + char wal_fullpath[MAXPGPATH]; + + join_path_components(wal_fullpath, instanceState->instance_wal_subdir_path, wal_file->file.name); + + /* save segment from purging */ + if (wal_file->keep) { - if (XLogRecPtrIsInvalid(oldest_lsn) || - strncmp(arcde->d_name + 8, oldestSegmentNeeded + 8, 16) < 0) - { - /* - * Use the original file name again now, including any - * extension that might have been chopped off before testing - * the sequence. - */ - snprintf(wal_file, MAXPGPATH, "%s/%s", - arclog_path, arcde->d_name); - - rc = unlink(wal_file); - if (rc != 0) - { - elog(WARNING, "could not remove file \"%s\": %s", - wal_file, strerror(errno)); - break; - } - elog(LOG, "removed WAL segment \"%s\"", wal_file); - - if (max_wal_file[0] == '\0' || - strcmp(max_wal_file + 8, arcde->d_name + 8) < 0) - strcpy(max_wal_file, arcde->d_name); - - if (min_wal_file[0] == '\0' || - strcmp(min_wal_file + 8, arcde->d_name + 8) > 0) - strcpy(min_wal_file, arcde->d_name); - } + elog(VERBOSE, "Retain WAL segment \"%s\"", wal_fullpath); + continue; } - } - if (min_wal_file[0] != '\0') - elog(INFO, "removed min WAL segment \"%s\"", min_wal_file); - if (max_wal_file[0] != '\0') - elog(INFO, "removed max WAL segment \"%s\"", max_wal_file); + /* unlink segment */ + if (fio_unlink(wal_fullpath, FIO_BACKUP_HOST) < 0) + { + /* Missing file is not considered as error condition */ + if (errno != ENOENT) + elog(ERROR, "Could not remove file \"%s\": %s", + wal_fullpath, strerror(errno)); + } + else + { + if (wal_file->type == SEGMENT) + elog(VERBOSE, "Removed WAL segment \"%s\"", wal_fullpath); + else if (wal_file->type == TEMP_SEGMENT) + elog(VERBOSE, "Removed temp WAL segment \"%s\"", wal_fullpath); + else if (wal_file->type == PARTIAL_SEGMENT) + elog(VERBOSE, "Removed partial WAL segment \"%s\"", wal_fullpath); + else if (wal_file->type == BACKUP_HISTORY_FILE) + elog(VERBOSE, "Removed backup history file \"%s\"", wal_fullpath); + } - if (errno) - elog(WARNING, "could not read archive location \"%s\": %s", - arclog_path, strerror(errno)); - if (closedir(arcdir)) - elog(WARNING, "could not close archive location \"%s\": %s", - arclog_path, strerror(errno)); + wal_deleted = true; + } } - else - elog(WARNING, "could not open archive location \"%s\": %s", - arclog_path, strerror(errno)); } /* Delete all backup files and wal files of given instance. */ int -do_delete_instance(void) +do_delete_instance(InstanceState *instanceState) { parray *backup_list; - parray *xlog_files_list; int i; - int rc; - char instance_config_path[MAXPGPATH]; - /* Delete all backups. */ - backup_list = catalog_get_backup_list(INVALID_BACKUP_ID); + backup_list = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); - catalog_lock_backup_list(backup_list, 0, parray_num(backup_list) - 1); + catalog_lock_backup_list(backup_list, 0, parray_num(backup_list) - 1, true, true); for (i = 0; i < parray_num(backup_list); i++) { @@ -855,42 +1001,130 @@ do_delete_instance(void) parray_free(backup_list); /* Delete all wal files. */ - xlog_files_list = parray_new(); - dir_list_file(xlog_files_list, arclog_path, false, false, false, 0, FIO_BACKUP_HOST); - - for (i = 0; i < parray_num(xlog_files_list); i++) - { - pgFile *wal_file = (pgFile *) parray_get(xlog_files_list, i); - if (S_ISREG(wal_file->mode)) - { - rc = unlink(wal_file->path); - if (rc != 0) - elog(WARNING, "Failed to remove file \"%s\": %s", - wal_file->path, strerror(errno)); - } - } - - /* Cleanup */ - parray_walk(xlog_files_list, pgFileFree); - parray_free(xlog_files_list); + pgut_rmtree(instanceState->instance_wal_subdir_path, false, true); /* Delete backup instance config file */ - join_path_components(instance_config_path, backup_instance_path, BACKUP_CATALOG_CONF_FILE); - if (remove(instance_config_path)) + if (remove(instanceState->instance_config_path)) { - elog(ERROR, "Can't remove \"%s\": %s", instance_config_path, + elog(ERROR, "Can't remove \"%s\": %s", instanceState->instance_config_path, strerror(errno)); } /* Delete instance root directories */ - if (rmdir(backup_instance_path) != 0) - elog(ERROR, "Can't remove \"%s\": %s", backup_instance_path, + if (rmdir(instanceState->instance_backup_subdir_path) != 0) + elog(ERROR, "Can't remove \"%s\": %s", instanceState->instance_backup_subdir_path, strerror(errno)); - if (rmdir(arclog_path) != 0) - elog(ERROR, "Can't remove \"%s\": %s", arclog_path, + if (rmdir(instanceState->instance_wal_subdir_path) != 0) + elog(ERROR, "Can't remove \"%s\": %s", instanceState->instance_wal_subdir_path, strerror(errno)); - elog(INFO, "Instance '%s' successfully deleted", instance_name); + elog(INFO, "Instance '%s' successfully deleted", instanceState->instance_name); return 0; } + +/* Delete all backups of given status in instance */ +void +do_delete_status(InstanceState *instanceState, InstanceConfig *instance_config, const char *status) +{ + int i; + parray *backup_list, *delete_list; + const char *pretty_status; + int n_deleted = 0, n_found = 0; + int64 size_to_delete = 0; + char size_to_delete_pretty[20]; + pgBackup *backup; + + BackupStatus status_for_delete = str2status(status); + delete_list = parray_new(); + + if (status_for_delete == BACKUP_STATUS_INVALID) + elog(ERROR, "Unknown value for '--status' option: '%s'", status); + + /* + * User may have provided status string in lower case, but + * we should print backup statuses consistently with show command, + * so convert it. + */ + pretty_status = status2str(status_for_delete); + + backup_list = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); + + if (parray_num(backup_list) == 0) + { + elog(WARNING, "Instance '%s' has no backups", instanceState->instance_name); + parray_free(delete_list); + parray_free(backup_list); + return; + } + + if (dry_run) + elog(INFO, "Deleting all backups with status '%s' in dry run mode", pretty_status); + else + elog(INFO, "Deleting all backups with status '%s'", pretty_status); + + /* Selects backups with specified status and their children into delete_list array. */ + for (i = 0; i < parray_num(backup_list); i++) + { + backup = (pgBackup *) parray_get(backup_list, i); + + if (backup->status == status_for_delete) + { + n_found++; + + /* incremental backup can be already in delete_list due to append_children() */ + if (parray_contains(delete_list, backup)) + continue; + parray_append(delete_list, backup); + + append_children(backup_list, backup, delete_list); + } + } + + parray_qsort(delete_list, pgBackupCompareIdDesc); + + /* delete and calculate free size from delete_list */ + for (i = 0; i < parray_num(delete_list); i++) + { + backup = (pgBackup *)parray_get(delete_list, i); + + elog(INFO, "Backup %s with status %s %s be deleted", + backup_id_of(backup), status2str(backup->status), dry_run ? "can" : "will"); + + size_to_delete += backup->data_bytes; + if (backup->stream) + size_to_delete += backup->wal_bytes; + + if (!dry_run && lock_backup(backup, false, true)) + delete_backup_files(backup); + + n_deleted++; + } + + /* Inform about data size to free */ + if (size_to_delete >= 0) + { + pretty_size(size_to_delete, size_to_delete_pretty, lengthof(size_to_delete_pretty)); + elog(INFO, "Resident data size to free by delete of %i backups: %s", + n_deleted, size_to_delete_pretty); + } + + /* delete selected backups */ + if (!dry_run && n_deleted > 0) + elog(INFO, "Successfully deleted %i %s from instance '%s'", + n_deleted, n_deleted == 1 ? "backup" : "backups", + instanceState->instance_name); + + + if (n_found == 0) + elog(WARNING, "Instance '%s' has no backups with status '%s'", + instanceState->instance_name, pretty_status); + + // we don`t do WAL purge here, because it is impossible to correctly handle + // dry-run case. + + /* Cleanup */ + parray_free(delete_list); + parray_walk(backup_list, pgBackupFree); + parray_free(backup_list); +} diff --git a/src/dir.c b/src/dir.c index 948ca5aeb..4b1bc2816 100644 --- a/src/dir.c +++ b/src/dir.c @@ -3,11 +3,12 @@ * dir.c: directory operation utility. * * Portions Copyright (c) 2009-2013, NIPPON TELEGRAPH AND TELEPHONE CORPORATION - * Portions Copyright (c) 2015-2019, Postgres Professional + * Portions Copyright (c) 2015-2022, Postgres Professional * *------------------------------------------------------------------------- */ +#include #include "pg_probackup.h" #include "utils/file.h" @@ -28,7 +29,7 @@ * start so they are not included in backups. The directories themselves are * kept and included as empty to preserve access permissions. */ -const char *pgdata_exclude_dir[] = +static const char *pgdata_exclude_dir[] = { PG_XLOG_DIR, /* @@ -79,6 +80,9 @@ static char *pgdata_exclude_files[] = "recovery.conf", "postmaster.pid", "postmaster.opts", + "probackup_recovery.conf", + "recovery.signal", + "standby.signal", NULL }; @@ -118,14 +122,17 @@ typedef struct TablespaceCreatedList TablespaceCreatedListCell *tail; } TablespaceCreatedList; -static int BlackListCompare(const void *str1, const void *str2); +static char dir_check_file(pgFile *file, bool backup_logs); -static char dir_check_file(pgFile *file); -static void dir_list_file_internal(parray *files, pgFile *parent, bool exclude, - bool follow_symlink, parray *black_list, - int external_dir_num, fio_location location); +static void dir_list_file_internal(parray *files, pgFile *parent, const char *parent_dir, + bool exclude, bool follow_symlink, bool backup_logs, + bool skip_hidden, int external_dir_num, fio_location location); static void opt_path_map(ConfigOption *opt, const char *arg, TablespaceList *list, const char *type); +static void cleanup_tablespace(const char *path); + +static void control_string_bad_format(const char* str); + /* Tablespace mapping */ static TablespaceList tablespace_dirs = {NULL, NULL}; @@ -134,25 +141,29 @@ static TablespaceList external_remap_list = {NULL, NULL}; /* * Create directory, also create parent directories if necessary. + * In strict mode treat already existing directory as error. + * Return values: + * 0 - ok + * -1 - error (check errno) */ int -dir_create_dir(const char *dir, mode_t mode) +dir_create_dir(const char *dir, mode_t mode, bool strict) { char parent[MAXPGPATH]; - strncpy(parent, dir, MAXPGPATH); + strlcpy(parent, dir, MAXPGPATH); get_parent_directory(parent); /* Create parent first */ - if (access(parent, F_OK) == -1) - dir_create_dir(parent, mode); + if (strlen(parent) > 0 && access(parent, F_OK) == -1) + dir_create_dir(parent, mode, false); /* Create directory */ if (mkdir(dir, mode) == -1) { - if (errno == EEXIST) /* already exist */ + if (errno == EEXIST && !strict) /* already exist */ return 0; - elog(ERROR, "cannot create directory \"%s\": %s", dir, strerror(errno)); + return -1; } return 0; @@ -171,40 +182,36 @@ pgFileNew(const char *path, const char *rel_path, bool follow_symlink, /* file not found is not an error case */ if (errno == ENOENT) return NULL; - elog(ERROR, "cannot stat file \"%s\": %s", path, + elog(ERROR, "Cannot stat file \"%s\": %s", path, strerror(errno)); } - file = pgFileInit(path, rel_path); + file = pgFileInit(rel_path); file->size = st.st_size; file->mode = st.st_mode; + file->mtime = st.st_mtime; file->external_dir_num = external_dir_num; return file; } pgFile * -pgFileInit(const char *path, const char *rel_path) +pgFileInit(const char *rel_path) { pgFile *file; - char *file_name; + char *file_name = NULL; file = (pgFile *) pgut_malloc(sizeof(pgFile)); MemSet(file, 0, sizeof(pgFile)); - file->forkName = pgut_malloc(MAXPGPATH); - file->forkName[0] = '\0'; - - file->path = pgut_strdup(path); - canonicalize_path(file->path); - file->rel_path = pgut_strdup(rel_path); canonicalize_path(file->rel_path); /* Get file name from the path */ - file_name = last_dir_separator(file->path); + file_name = last_dir_separator(file->rel_path); + if (file_name == NULL) - file->name = file->path; + file->name = file->rel_path; else { file_name++; @@ -214,6 +221,12 @@ pgFileInit(const char *path, const char *rel_path) /* Number of blocks readed during backup */ file->n_blocks = BLOCKNUM_INVALID; + /* Number of blocks backed up during backup */ + file->n_headers = 0; + + // May be add? + // pg_atomic_clear_flag(file->lock); + file->excluded = false; return file; } @@ -222,88 +235,33 @@ pgFileInit(const char *path, const char *rel_path) * If the pgFile points directory, the directory must be empty. */ void -pgFileDelete(pgFile *file) +pgFileDelete(mode_t mode, const char *full_path) { - if (S_ISDIR(file->mode)) + if (S_ISDIR(mode)) { - if (rmdir(file->path) == -1) + if (rmdir(full_path) == -1) { if (errno == ENOENT) return; else if (errno == ENOTDIR) /* could be symbolic link */ goto delete_file; - elog(ERROR, "cannot remove directory \"%s\": %s", - file->path, strerror(errno)); + elog(ERROR, "Cannot remove directory \"%s\": %s", + full_path, strerror(errno)); } return; } delete_file: - if (remove(file->path) == -1) + if (remove(full_path) == -1) { if (errno == ENOENT) return; - elog(ERROR, "cannot remove file \"%s\": %s", file->path, + elog(ERROR, "Cannot remove file \"%s\": %s", full_path, strerror(errno)); } } -pg_crc32 -pgFileGetCRC(const char *file_path, bool use_crc32c, bool raise_on_deleted, - size_t *bytes_read, fio_location location) -{ - FILE *fp; - pg_crc32 crc = 0; - char buf[1024]; - size_t len = 0; - size_t total = 0; - int errno_tmp; - - INIT_FILE_CRC32(use_crc32c, crc); - - /* open file in binary read mode */ - fp = fio_fopen(file_path, PG_BINARY_R, location); - if (fp == NULL) - { - if (!raise_on_deleted && errno == ENOENT) - { - FIN_FILE_CRC32(use_crc32c, crc); - return crc; - } - else - elog(ERROR, "cannot open file \"%s\": %s", - file_path, strerror(errno)); - } - - /* calc CRC of file */ - for (;;) - { - if (interrupted) - elog(ERROR, "interrupted during CRC calculation"); - - len = fio_fread(fp, buf, sizeof(buf)); - if(len == 0) - break; - /* update CRC */ - COMP_FILE_CRC32(use_crc32c, crc, buf, len); - total += len; - } - - if (bytes_read) - *bytes_read = total; - - errno_tmp = errno; - if (len < 0) - elog(WARNING, "cannot read \"%s\": %s", file_path, - strerror(errno_tmp)); - - FIN_FILE_CRC32(use_crc32c, crc); - fio_fclose(fp); - - return crc; -} - void pgFileFree(void *file) { @@ -314,23 +272,18 @@ pgFileFree(void *file) file_ptr = (pgFile *) file; - if (file_ptr->linked) - free(file_ptr->linked); - - if (file_ptr->forkName) - free(file_ptr->forkName); - - pfree(file_ptr->path); + pfree(file_ptr->linked); pfree(file_ptr->rel_path); + pfree(file); } /* Compare two pgFile with their path in ascending order of ASCII code. */ int -pgFileComparePath(const void *f1, const void *f2) +pgFileMapComparePath(const void *f1, const void *f2) { - pgFile *f1p = *(pgFile **)f1; - pgFile *f2p = *(pgFile **)f2; + page_map_entry *f1p = *(page_map_entry **)f1; + page_map_entry *f2p = *(page_map_entry **)f2; return strcmp(f1p->path, f2p->path); } @@ -345,28 +298,24 @@ pgFileCompareName(const void *f1, const void *f2) return strcmp(f1p->name, f2p->name); } -/* - * Compare two pgFile with their path and external_dir_num - * in ascending order of ASCII code. - */ +/* Compare pgFile->name with string in ascending order of ASCII code. */ int -pgFileComparePathWithExternal(const void *f1, const void *f2) +pgFileCompareNameWithString(const void *f1, const void *f2) { pgFile *f1p = *(pgFile **)f1; - pgFile *f2p = *(pgFile **)f2; - int res; + char *f2s = *(char **)f2; - res = strcmp(f1p->path, f2p->path); - if (!res) - { - if (f1p->external_dir_num > f2p->external_dir_num) - return 1; - else if (f1p->external_dir_num < f2p->external_dir_num) - return -1; - else - return 0; - } - return res; + return strcmp(f1p->name, f2s); +} + +/* Compare pgFile->rel_path with string in ascending order of ASCII code. */ +int +pgFileCompareRelPathWithString(const void *f1, const void *f2) +{ + pgFile *f1p = *(pgFile **)f1; + char *f2s = *(char **)f2; + + return strcmp(f1p->rel_path, f2s); } /* @@ -393,21 +342,14 @@ pgFileCompareRelPathWithExternal(const void *f1, const void *f2) return res; } -/* Compare two pgFile with their path in descending order of ASCII code. */ -int -pgFileComparePathDesc(const void *f1, const void *f2) -{ - return -pgFileComparePath(f1, f2); -} - /* - * Compare two pgFile with their path and external_dir_num + * Compare two pgFile with their rel_path and external_dir_num * in descending order of ASCII code. */ int -pgFileComparePathWithExternalDesc(const void *f1, const void *f2) +pgFileCompareRelPathWithExternalDesc(const void *f1, const void *f2) { - return -pgFileComparePathWithExternal(f1, f2); + return -pgFileCompareRelPathWithExternal(f1, f2); } /* Compare two pgFile with their linked directory path. */ @@ -435,60 +377,72 @@ pgFileCompareSize(const void *f1, const void *f2) return 0; } -static int -BlackListCompare(const void *str1, const void *str2) +/* Compare two pgFile with their size in descending order */ +int +pgFileCompareSizeDesc(const void *f1, const void *f2) +{ + return -1 * pgFileCompareSize(f1, f2); +} + +int +pgCompareString(const void *str1, const void *str2) { return strcmp(*(char **) str1, *(char **) str2); } +/* + * From bsearch(3): "The compar routine is expected to have two argu‐ + * ments which point to the key object and to an array member, in that order" + * But in practice this is opposite, so we took strlen from second string (search key) + * This is checked by tests.catchup.CatchupTest.test_catchup_with_exclude_path + */ +int +pgPrefixCompareString(const void *str1, const void *str2) +{ + const char *s1 = *(char **) str1; + const char *s2 = *(char **) str2; + return strncmp(s1, s2, strlen(s2)); +} + +/* Compare two Oids */ +int +pgCompareOid(const void *f1, const void *f2) +{ + Oid *v1 = *(Oid **) f1; + Oid *v2 = *(Oid **) f2; + + if (*v1 > *v2) + return 1; + else if (*v1 < *v2) + return -1; + else + return 0;} + + +void +db_map_entry_free(void *entry) +{ + db_map_entry *m = (db_map_entry *) entry; + + free(m->datname); + free(entry); +} + /* * List files, symbolic links and directories in the directory "root" and add * pgFile objects to "files". We add "root" to "files" if add_root is true. * * When follow_symlink is true, symbolic link is ignored and only file or * directory linked to will be listed. + * + * TODO: make it strictly local */ void dir_list_file(parray *files, const char *root, bool exclude, bool follow_symlink, - bool add_root, int external_dir_num, fio_location location) + bool add_root, bool backup_logs, bool skip_hidden, int external_dir_num, + fio_location location) { pgFile *file; - parray *black_list = NULL; - char path[MAXPGPATH]; - - join_path_components(path, backup_instance_path, PG_BLACK_LIST); - /* List files with black list */ - if (root && instance_config.pgdata && - strcmp(root, instance_config.pgdata) == 0 && - fileExists(path, FIO_BACKUP_HOST)) - { - FILE *black_list_file = NULL; - char buf[MAXPGPATH * 2]; - char black_item[MAXPGPATH * 2]; - - black_list = parray_new(); - black_list_file = fio_open_stream(path, FIO_BACKUP_HOST); - - if (black_list_file == NULL) - elog(ERROR, "cannot open black_list: %s", strerror(errno)); - - while (fgets(buf, lengthof(buf), black_list_file) != NULL) - { - black_item[0] = '\0'; - join_path_components(black_item, instance_config.pgdata, buf); - - if (black_item[strlen(black_item) - 1] == '\n') - black_item[strlen(black_item) - 1] = '\0'; - - if (black_item[0] == '#' || black_item[0] == '\0') - continue; - - parray_append(black_list, pgut_strdup(black_item)); - } - - fio_close_stream(black_list_file); - parray_qsort(black_list, BlackListCompare); - } file = pgFileNew(root, "", follow_symlink, external_dir_num, location); if (file == NULL) @@ -504,25 +458,19 @@ dir_list_file(parray *files, const char *root, bool exclude, bool follow_symlink { if (external_dir_num > 0) elog(ERROR, " --external-dirs option \"%s\": directory or symbolic link expected", - file->path); + root); else - elog(WARNING, "Skip \"%s\": unexpected file format", file->path); + elog(WARNING, "Skip \"%s\": unexpected file format", root); return; } if (add_root) parray_append(files, file); - dir_list_file_internal(files, file, exclude, follow_symlink, black_list, - external_dir_num, location); + dir_list_file_internal(files, file, root, exclude, follow_symlink, + backup_logs, skip_hidden, external_dir_num, location); if (!add_root) pgFileFree(file); - - if (black_list) - { - parray_walk(black_list, pfree); - parray_free(black_list); - } } #define CHECK_FALSE 0 @@ -543,7 +491,7 @@ dir_list_file(parray *files, const char *root, bool exclude, bool follow_symlink * - datafiles */ static char -dir_check_file(pgFile *file) +dir_check_file(pgFile *file, bool backup_logs) { int i; int sscanf_res; @@ -557,20 +505,20 @@ dir_check_file(pgFile *file) if (!exclusive_backup) { for (i = 0; pgdata_exclude_files_non_exclusive[i]; i++) - if (strcmp(file->name, + if (strcmp(file->rel_path, pgdata_exclude_files_non_exclusive[i]) == 0) { /* Skip */ - elog(VERBOSE, "Excluding file: %s", file->name); + elog(LOG, "Excluding file: %s", file->name); return CHECK_FALSE; } } for (i = 0; pgdata_exclude_files[i]; i++) - if (strcmp(file->name, pgdata_exclude_files[i]) == 0) + if (strcmp(file->rel_path, pgdata_exclude_files[i]) == 0) { /* Skip */ - elog(VERBOSE, "Excluding file: %s", file->name); + elog(LOG, "Excluding file: %s", file->name); return CHECK_FALSE; } } @@ -578,7 +526,7 @@ dir_check_file(pgFile *file) * If the directory name is in the exclude list, do not list the * contents. */ - else if (S_ISDIR(file->mode) && !in_tablespace) + else if (S_ISDIR(file->mode) && !in_tablespace && file->external_dir_num == 0) { /* * If the item in the exclude list starts with '/', compare to @@ -587,20 +535,20 @@ dir_check_file(pgFile *file) */ for (i = 0; pgdata_exclude_dir[i]; i++) { - /* Full-path exclude*/ - if (pgdata_exclude_dir[i][0] == '/') + /* exclude by dirname */ + if (strcmp(file->name, pgdata_exclude_dir[i]) == 0) { - if (strcmp(file->path, pgdata_exclude_dir[i]) == 0) - { - elog(VERBOSE, "Excluding directory content: %s", - file->name); - return CHECK_EXCLUDE_FALSE; - } + elog(LOG, "Excluding directory content: %s", file->rel_path); + return CHECK_EXCLUDE_FALSE; } - else if (strcmp(file->name, pgdata_exclude_dir[i]) == 0) + } + + if (!backup_logs) + { + if (strcmp(file->rel_path, PG_LOG_DIR) == 0) { - elog(VERBOSE, "Excluding directory content: %s", - file->name); + /* Skip */ + elog(LOG, "Excluding directory content: %s", file->rel_path); return CHECK_EXCLUDE_FALSE; } } @@ -642,26 +590,16 @@ dir_check_file(pgFile *file) */ if (sscanf_res == 2 && strcmp(tmp_rel_path, TABLESPACE_VERSION_DIRECTORY) != 0) return CHECK_FALSE; - - if (sscanf_res == 3 && S_ISDIR(file->mode) && - strcmp(tmp_rel_path, TABLESPACE_VERSION_DIRECTORY) == 0) - file->is_database = true; } else if (path_is_prefix_of_path("global", file->rel_path)) { file->tblspcOid = GLOBALTABLESPACE_OID; - - if (S_ISDIR(file->mode) && strcmp(file->name, "global") == 0) - file->is_database = true; } else if (path_is_prefix_of_path("base", file->rel_path)) { file->tblspcOid = DEFAULTTABLESPACE_OID; sscanf(file->rel_path, "base/%u/", &(file->dbOid)); - - if (S_ISDIR(file->mode) && strcmp(file->name, "base") != 0) - file->is_database = true; } /* Do not backup ptrack_init files */ @@ -677,47 +615,22 @@ dir_check_file(pgFile *file) { if (strcmp(file->name, "pg_internal.init") == 0) return CHECK_FALSE; + /* Do not backup ptrack2.x temp map files */ +// else if (strcmp(file->name, "ptrack.map") == 0) +// return CHECK_FALSE; + else if (strcmp(file->name, "ptrack.map.mmap") == 0) + return CHECK_FALSE; + else if (strcmp(file->name, "ptrack.map.tmp") == 0) + return CHECK_FALSE; /* Do not backup temp files */ else if (file->name[0] == 't' && isdigit(file->name[1])) return CHECK_FALSE; else if (isdigit(file->name[0])) { - char *fork_name; - int len; - char suffix[MAXPGPATH]; - - fork_name = strstr(file->name, "_"); - if (fork_name) - { - /* Auxiliary fork of the relfile */ - sscanf(file->name, "%u_%s", &(file->relOid), file->forkName); + set_forkname(file); - /* Do not backup ptrack files */ - if (strcmp(file->forkName, "ptrack") == 0) - return CHECK_FALSE; - } - else - { - /* - * snapfs files: - * RELFILENODE.BLOCKNO.snapmap.SNAPID - * RELFILENODE.BLOCKNO.snap.SNAPID - */ - if (strstr(file->name, "snap") != NULL) - return true; - - len = strlen(file->name); - /* reloid.cfm */ - if (len > 3 && strcmp(file->name + len - 3, "cfm") == 0) - return CHECK_TRUE; - - sscanf_res = sscanf(file->name, "%u.%d.%s", &(file->relOid), - &(file->segno), suffix); - if (sscanf_res == 0) - elog(ERROR, "Cannot parse file name \"%s\"", file->name); - else if (sscanf_res == 1 || sscanf_res == 2) - file->is_datafile = true; - } + if (file->forkName == ptrack) /* Compatibility with left-overs from ptrack1 */ + return CHECK_FALSE; } } @@ -728,20 +641,22 @@ dir_check_file(pgFile *file) * List files in parent->path directory. If "exclude" is true do not add into * "files" files from pgdata_exclude_files and directories from * pgdata_exclude_dir. + * + * TODO: should we check for interrupt here ? */ static void -dir_list_file_internal(parray *files, pgFile *parent, bool exclude, - bool follow_symlink, parray *black_list, - int external_dir_num, fio_location location) +dir_list_file_internal(parray *files, pgFile *parent, const char *parent_dir, + bool exclude, bool follow_symlink, bool backup_logs, + bool skip_hidden, int external_dir_num, fio_location location) { - DIR *dir; + DIR *dir; struct dirent *dent; if (!S_ISDIR(parent->mode)) - elog(ERROR, "\"%s\" is not a directory", parent->path); + elog(ERROR, "\"%s\" is not a directory", parent_dir); /* Open directory and list contents */ - dir = fio_opendir(parent->path, location); + dir = fio_opendir(parent_dir, location); if (dir == NULL) { if (errno == ENOENT) @@ -750,7 +665,7 @@ dir_list_file_internal(parray *files, pgFile *parent, bool exclude, return; } elog(ERROR, "Cannot open directory \"%s\": %s", - parent->path, strerror(errno)); + parent_dir, strerror(errno)); } errno = 0; @@ -761,7 +676,7 @@ dir_list_file_internal(parray *files, pgFile *parent, bool exclude, char rel_child[MAXPGPATH]; char check_res; - join_path_components(child, parent->path, dent->d_name); + join_path_components(child, parent_dir, dent->d_name); join_path_components(rel_child, parent->rel_path, dent->d_name); file = pgFileNew(child, rel_child, follow_symlink, external_dir_num, @@ -777,29 +692,28 @@ dir_list_file_internal(parray *files, pgFile *parent, bool exclude, continue; } - /* - * Add only files, directories and links. Skip sockets and other - * unexpected file formats. - */ - if (!S_ISDIR(file->mode) && !S_ISREG(file->mode)) + /* skip hidden files and directories */ + if (skip_hidden && file->name[0] == '.') { - elog(WARNING, "Skip \"%s\": unexpected file format", file->path); + elog(WARNING, "Skip hidden file: '%s'", child); pgFileFree(file); continue; } - /* Skip if the directory is in black_list defined by user */ - if (black_list && parray_bsearch(black_list, file->path, - BlackListCompare)) + /* + * Add only files, directories and links. Skip sockets and other + * unexpected file formats. + */ + if (!S_ISDIR(file->mode) && !S_ISREG(file->mode)) { - elog(LOG, "Skip \"%s\": it is in the user's black list", file->path); + elog(WARNING, "Skip '%s': unexpected file format", child); pgFileFree(file); continue; } if (exclude) { - check_res = dir_check_file(file); + check_res = dir_check_file(file, backup_logs); if (check_res == CHECK_FALSE) { /* Skip */ @@ -821,8 +735,8 @@ dir_list_file_internal(parray *files, pgFile *parent, bool exclude, * recursively. */ if (S_ISDIR(file->mode)) - dir_list_file_internal(files, file, exclude, follow_symlink, - black_list, external_dir_num, location); + dir_list_file_internal(files, file, child, exclude, follow_symlink, + backup_logs, skip_hidden, external_dir_num, location); } if (errno && errno != ENOENT) @@ -830,7 +744,7 @@ dir_list_file_internal(parray *files, pgFile *parent, bool exclude, int errno_tmp = errno; fio_closedir(dir); elog(ERROR, "Cannot read directory \"%s\": %s", - parent->path, strerror(errno_tmp)); + parent_dir, strerror(errno_tmp)); } fio_closedir(dir); } @@ -841,7 +755,7 @@ dir_list_file_internal(parray *files, pgFile *parent, bool exclude, * * Copy of function get_tablespace_mapping() from pg_basebackup.c. */ -static const char * +const char * get_tablespace_mapping(const char *dir) { TablespaceListCell *cell; @@ -873,14 +787,14 @@ opt_path_map(ConfigOption *opt, const char *arg, TablespaceList *list, for (arg_ptr = arg; *arg_ptr; arg_ptr++) { if (dst_ptr - dst >= MAXPGPATH) - elog(ERROR, "directory name too long"); + elog(ERROR, "Directory name too long"); if (*arg_ptr == '\\' && *(arg_ptr + 1) == '=') ; /* skip backslash escaping = */ else if (*arg_ptr == '=' && (arg_ptr == arg || *(arg_ptr - 1) != '\\')) { if (*cell->new_dir) - elog(ERROR, "multiple \"=\" signs in %s mapping\n", type); + elog(ERROR, "Multiple \"=\" signs in %s mapping\n", type); else dst = dst_ptr = cell->new_dir; } @@ -889,7 +803,7 @@ opt_path_map(ConfigOption *opt, const char *arg, TablespaceList *list, } if (!*cell->old_dir || !*cell->new_dir) - elog(ERROR, "invalid %s mapping format \"%s\", " + elog(ERROR, "Invalid %s mapping format \"%s\", " "must be \"OLDDIR=NEWDIR\"", type, arg); canonicalize_path(cell->old_dir); canonicalize_path(cell->new_dir); @@ -901,11 +815,11 @@ opt_path_map(ConfigOption *opt, const char *arg, TablespaceList *list, * consistent with the new_dir check. */ if (!is_absolute_path(cell->old_dir)) - elog(ERROR, "old directory is not an absolute path in %s mapping: %s\n", + elog(ERROR, "Old directory is not an absolute path in %s mapping: %s\n", type, cell->old_dir); if (!is_absolute_path(cell->new_dir)) - elog(ERROR, "new directory is not an absolute path in %s mapping: %s\n", + elog(ERROR, "New directory is not an absolute path in %s mapping: %s\n", type, cell->new_dir); if (list->tail) @@ -946,13 +860,20 @@ opt_externaldir_map(ConfigOption *opt, const char *arg) */ void create_data_directories(parray *dest_files, const char *data_dir, const char *backup_dir, - bool extract_tablespaces, fio_location location) + bool extract_tablespaces, bool incremental, fio_location location, + const char* waldir_path) { int i; parray *links = NULL; mode_t pg_tablespace_mode = DIR_PERMISSION; char to_path[MAXPGPATH]; + if (waldir_path && !dir_is_empty(waldir_path, location)) + { + elog(ERROR, "WAL directory location is not empty: \"%s\"", waldir_path); + } + + /* get tablespace map */ if (extract_tablespaces) { @@ -1017,12 +938,33 @@ create_data_directories(parray *dest_files, const char *data_dir, const char *ba /* skip external directory content */ if (dir->external_dir_num != 0) continue; + /* Create WAL directory and symlink if waldir_path is setting */ + if (waldir_path && strcmp(dir->rel_path, PG_XLOG_DIR) == 0) { + /* get full path to PG_XLOG_DIR */ + + join_path_components(to_path, data_dir, PG_XLOG_DIR); + + elog(VERBOSE, "Create directory \"%s\" and symbolic link \"%s\"", + waldir_path, to_path); + + /* create tablespace directory from waldir_path*/ + fio_mkdir(waldir_path, pg_tablespace_mode, location); + + /* create link to linked_path */ + if (fio_symlink(waldir_path, to_path, incremental, location) < 0) + elog(ERROR, "Could not create symbolic link \"%s\": %s", + to_path, strerror(errno)); + + continue; + + + } /* tablespace_map exists */ if (links) { /* get parent dir of rel_path */ - strncpy(parent_dir, dir->rel_path, MAXPGPATH); + strlcpy(parent_dir, dir->rel_path, MAXPGPATH); get_parent_directory(parent_dir); /* check if directory is actually link to tablespace */ @@ -1039,19 +981,19 @@ create_data_directories(parray *dest_files, const char *data_dir, const char *ba const char *linked_path = get_tablespace_mapping((*link)->linked); if (!is_absolute_path(linked_path)) - elog(ERROR, "Tablespace directory is not an absolute path: %s\n", + elog(ERROR, "Tablespace directory path must be an absolute path: %s\n", linked_path); join_path_components(to_path, data_dir, dir->rel_path); - elog(VERBOSE, "Create directory \"%s\" and symbolic link \"%s\"", + elog(LOG, "Create directory \"%s\" and symbolic link \"%s\"", linked_path, to_path); /* create tablespace directory */ fio_mkdir(linked_path, pg_tablespace_mode, location); /* create link to linked_path */ - if (fio_symlink(linked_path, to_path, location) < 0) + if (fio_symlink(linked_path, to_path, incremental, location) < 0) elog(ERROR, "Could not create symbolic link \"%s\": %s", to_path, strerror(errno)); @@ -1061,9 +1003,11 @@ create_data_directories(parray *dest_files, const char *data_dir, const char *ba } /* This is not symlink, create directory */ - elog(VERBOSE, "Create directory \"%s\"", dir->rel_path); + elog(LOG, "Create directory \"%s\"", dir->rel_path); join_path_components(to_path, data_dir, dir->rel_path); + + // TODO check exit code fio_mkdir(to_path, dir->mode, location); } @@ -1079,7 +1023,7 @@ create_data_directories(parray *dest_files, const char *data_dir, const char *ba * tablespace_map or tablespace_map.txt. */ void -read_tablespace_map(parray *files, const char *backup_dir) +read_tablespace_map(parray *links, const char *backup_dir) { FILE *fp; char db_path[MAXPGPATH], @@ -1089,72 +1033,96 @@ read_tablespace_map(parray *files, const char *backup_dir) join_path_components(db_path, backup_dir, DATABASE_DIR); join_path_components(map_path, db_path, PG_TABLESPACE_MAP_FILE); - /* Exit if database/tablespace_map doesn't exist */ - if (!fileExists(map_path, FIO_BACKUP_HOST)) - { - elog(LOG, "there is no file tablespace_map"); - return; - } - fp = fio_open_stream(map_path, FIO_BACKUP_HOST); if (fp == NULL) - elog(ERROR, "cannot open \"%s\": %s", map_path, strerror(errno)); + elog(ERROR, "Cannot open tablespace map file \"%s\": %s", map_path, strerror(errno)); while (fgets(buf, lengthof(buf), fp)) { - char link_name[MAXPGPATH], - path[MAXPGPATH]; - pgFile *file; + char link_name[MAXPGPATH]; + char *path; + int n = 0; + pgFile *file; + int i = 0; - if (sscanf(buf, "%1023s %1023s", link_name, path) != 2) - elog(ERROR, "invalid format found in \"%s\"", map_path); - - file = pgut_new(pgFile); - memset(file, 0, sizeof(pgFile)); + if (sscanf(buf, "%s %n", link_name, &n) != 1) + elog(ERROR, "Invalid format found in \"%s\"", map_path); - file->path = pgut_malloc(strlen(link_name) + 1); - strcpy(file->path, link_name); + path = buf + n; - file->name = file->path; + /* Remove newline character at the end of string if any */ + i = strcspn(path, "\n"); + if (strlen(path) > i) + path[i] = '\0'; - file->linked = pgut_malloc(strlen(path) + 1); - strcpy(file->linked, path); + file = pgut_new(pgFile); + memset(file, 0, sizeof(pgFile)); + /* follow the convention for pgFileFree */ + file->name = pgut_strdup(link_name); + file->linked = pgut_strdup(path); canonicalize_path(file->linked); - canonicalize_path(file->path); - parray_append(files, file); + parray_append(links, file); } + if (ferror(fp)) + elog(ERROR, "Failed to read from file: \"%s\"", map_path); + fio_close_stream(fp); } /* * Check that all tablespace mapping entries have correct linked directory - * paths. Linked directories must be empty or do not exist. + * paths. Linked directories must be empty or do not exist, unless + * we are running incremental restore, then linked directories can be nonempty. * * If tablespace-mapping option is supplied, all OLDDIR entries must have * entries in tablespace_map file. + * + * When running incremental restore with tablespace remapping, then + * new tablespace directory MUST be empty, because there is no way + * we can be sure, that files laying there belong to our instance. + * But "force" flag allows to ignore this condition, by wiping out + * the current content on the directory. + * + * Exit codes: + * 1. backup has no tablespaces + * 2. backup has tablespaces and they are empty + * 3. backup has tablespaces and some of them are not empty */ -void -check_tablespace_mapping(pgBackup *backup) +int +check_tablespace_mapping(pgBackup *backup, bool incremental, bool force, bool pgdata_is_empty, bool no_validate) { - char this_backup_path[MAXPGPATH]; - parray *links; + parray *links = parray_new(); size_t i; TablespaceListCell *cell; pgFile *tmp_file = pgut_new(pgFile); + bool tblspaces_are_empty = true; - links = parray_new(); + elog(LOG, "Checking tablespace directories of backup %s", + backup_id_of(backup)); - pgBackupGetPath(backup, this_backup_path, lengthof(this_backup_path), NULL); - read_tablespace_map(links, this_backup_path); + /* validate tablespace map, + * if there are no tablespaces, then there is nothing left to do + */ + if (!validate_tablespace_map(backup, no_validate)) + { + /* + * Sanity check + * If there is no tablespaces in backup, + * then using the '--tablespace-mapping' option is a mistake. + */ + if (tablespace_dirs.head != NULL) + elog(ERROR, "Backup %s has no tablespaceses, nothing to remap " + "via \"--tablespace-mapping\" option", backup_id_of(backup)); + return NoTblspc; + } + + read_tablespace_map(links, backup->root_dir); /* Sort links by the path of a linked file*/ parray_qsort(links, pgFileCompareLinked); - elog(LOG, "check tablespace directories of backup %s", - base36enc(backup->start_time)); - /* 1 - each OLDDIR must have an entry in tablespace_map file (links) */ for (cell = tablespace_dirs.head; cell; cell = cell->next) { @@ -1166,43 +1134,122 @@ check_tablespace_mapping(pgBackup *backup) cell->old_dir); } + /* + * There is difference between incremental restore of already existing + * tablespaceses and remapped tablespaceses. + * Former are allowed to be not empty, because we treat them like an + * extension of PGDATA. + * The latter are not, unless "--force" flag is used. + * in which case the remapped directory is nuked - just to be safe, + * because it is hard to be sure that there are no some tricky corner + * cases of pages from different systems having the same crc. + * This is a strict approach. + * + * Why can`t we not nuke it and just let it roll ? + * What if user just wants to rerun failed restore with the same + * parameters? Nuking is bad for this case. + * + * Consider the example of existing PGDATA: + * .... + * pg_tablespace + * 100500-> /somedirectory + * .... + * + * We want to remap it during restore like that: + * .... + * pg_tablespace + * 100500-> /somedirectory1 + * .... + * + * Usually it is required for "/somedirectory1" to be empty, but + * in case of incremental restore with 'force' flag, which required + * of us to drop already existing content of "/somedirectory1". + * + * TODO: Ideally in case of incremental restore we must also + * drop the "/somedirectory" directory first, but currently + * we don`t do that. + */ + /* 2 - all linked directories must be empty */ for (i = 0; i < parray_num(links); i++) { pgFile *link = (pgFile *) parray_get(links, i); const char *linked_path = link->linked; - TablespaceListCell *cell; + bool remapped = false; for (cell = tablespace_dirs.head; cell; cell = cell->next) + { if (strcmp(link->linked, cell->old_dir) == 0) { linked_path = cell->new_dir; + remapped = true; break; } + } + + if (remapped) + elog(INFO, "Tablespace %s will be remapped from \"%s\" to \"%s\"", link->name, cell->old_dir, cell->new_dir); + else + elog(INFO, "Tablespace %s will be restored using old path \"%s\"", link->name, linked_path); if (!is_absolute_path(linked_path)) - elog(ERROR, "tablespace directory is not an absolute path: %s\n", + elog(ERROR, "Tablespace directory path must be an absolute path: %s\n", linked_path); if (!dir_is_empty(linked_path, FIO_DB_HOST)) - elog(ERROR, "restore tablespace destination is not empty: \"%s\"", - linked_path); + { + + if (!incremental) + elog(ERROR, "Restore tablespace destination is not empty: \"%s\"", linked_path); + + else if (remapped && !force) + elog(ERROR, "Remapped tablespace destination is not empty: \"%s\". " + "Use \"--force\" flag if you want to automatically clean up the " + "content of new tablespace destination", + linked_path); + + else if (pgdata_is_empty && !force) + elog(ERROR, "PGDATA is empty, but tablespace destination is not: \"%s\". " + "Use \"--force\" flag is you want to automatically clean up the " + "content of tablespace destination", + linked_path); + + /* + * TODO: compile the list of tblspc Oids to delete later, + * similar to what we do with database_map. + */ + else if (force && (pgdata_is_empty || remapped)) + { + elog(WARNING, "Cleaning up the content of %s directory: \"%s\"", + remapped ? "remapped tablespace" : "tablespace", linked_path); + cleanup_tablespace(linked_path); + continue; + } + + tblspaces_are_empty = false; + } } free(tmp_file); parray_walk(links, pgFileFree); parray_free(links); + + if (tblspaces_are_empty) + return EmptyTblspc; + + return NotEmptyTblspc; } +/* TODO: Make it consistent with check_tablespace_mapping */ void -check_external_dir_mapping(pgBackup *backup) +check_external_dir_mapping(pgBackup *backup, bool incremental) { TablespaceListCell *cell; parray *external_dirs_to_restore; int i; elog(LOG, "check external directories of backup %s", - base36enc(backup->start_time)); + backup_id_of(backup)); if (!backup->external_dir_str) { @@ -1248,7 +1295,7 @@ check_external_dir_mapping(pgBackup *backup) char *external_dir = (char *) parray_get(external_dirs_to_restore, i); - if (!dir_is_empty(external_dir, FIO_DB_HOST)) + if (!incremental && !dir_is_empty(external_dir, FIO_DB_HOST)) elog(ERROR, "External directory is not empty: \"%s\"", external_dir); } @@ -1271,7 +1318,7 @@ get_external_remap(char *current_dir) return current_dir; } -/* Parsing states for get_control_value() */ +/* Parsing states for get_control_value_str() */ #define CONTROL_WAIT_NAME 1 #define CONTROL_INNAME 2 #define CONTROL_WAIT_COLON 3 @@ -1285,26 +1332,62 @@ get_external_remap(char *current_dir) * The line has the following format: * {"name1":"value1", "name2":"value2"} * - * The value will be returned to "value_str" as string if it is not NULL. If it - * is NULL the value will be returned to "value_int64" as int64. + * The value will be returned in "value_int64" as int64. + * + * Returns true if the value was found in the line and parsed. + */ +bool +get_control_value_int64(const char *str, const char *name, int64 *value_int64, bool is_mandatory) +{ + + char buf_int64[32]; + + assert(value_int64); + + /* Set default value */ + *value_int64 = 0; + + if (!get_control_value_str(str, name, buf_int64, sizeof(buf_int64), is_mandatory)) + return false; + + if (!parse_int64(buf_int64, value_int64, 0)) + { + /* We assume that too big value is -1 */ + if (errno == ERANGE) + *value_int64 = BYTES_INVALID; + else + control_string_bad_format(str); + return false; + } + + return true; +} + +/* + * Get value from json-like line "str" of backup_content.control file. + * + * The line has the following format: + * {"name1":"value1", "name2":"value2"} + * + * The value will be returned to "value_str" as string. * * Returns true if the value was found in the line. */ -static bool -get_control_value(const char *str, const char *name, - char *value_str, int64 *value_int64, bool is_mandatory) + +bool +get_control_value_str(const char *str, const char *name, + char *value_str, size_t value_str_size, bool is_mandatory) { int state = CONTROL_WAIT_NAME; char *name_ptr = (char *) name; char *buf = (char *) str; - char buf_int64[32], /* Buffer for "value_int64" */ - *buf_int64_ptr = buf_int64; + char *const value_str_start = value_str; + + assert(value_str); + assert(value_str_size > 0); - /* Set default values */ - if (value_str) - *value_str = '\0'; - else if (value_int64) - *value_int64 = 0; + /* Set default value */ + *value_str = '\0'; while (*buf) { @@ -1314,7 +1397,7 @@ get_control_value(const char *str, const char *name, if (*buf == '"') state = CONTROL_INNAME; else if (IsAlpha(*buf)) - goto bad_format; + control_string_bad_format(str); break; case CONTROL_INNAME: /* Found target field. Parse value. */ @@ -1333,57 +1416,32 @@ get_control_value(const char *str, const char *name, if (*buf == ':') state = CONTROL_WAIT_VALUE; else if (!IsSpace(*buf)) - goto bad_format; + control_string_bad_format(str); break; case CONTROL_WAIT_VALUE: if (*buf == '"') { state = CONTROL_INVALUE; - buf_int64_ptr = buf_int64; } else if (IsAlpha(*buf)) - goto bad_format; + control_string_bad_format(str); break; case CONTROL_INVALUE: /* Value was parsed, exit */ if (*buf == '"') { - if (value_str) - { - *value_str = '\0'; - } - else if (value_int64) - { - /* Length of buf_uint64 should not be greater than 31 */ - if (buf_int64_ptr - buf_int64 >= 32) - elog(ERROR, "field \"%s\" is out of range in the line %s of the file %s", - name, str, DATABASE_FILE_LIST); - - *buf_int64_ptr = '\0'; - if (!parse_int64(buf_int64, value_int64, 0)) - { - /* We assume that too big value is -1 */ - if (errno == ERANGE) - *value_int64 = BYTES_INVALID; - else - goto bad_format; - } - } - + *value_str = '\0'; return true; } else { - if (value_str) - { - *value_str = *buf; - value_str++; - } - else - { - *buf_int64_ptr = *buf; - buf_int64_ptr++; + /* verify if value_str not exceeds value_str_size limits */ + if (value_str - value_str_start >= value_str_size - 1) { + elog(ERROR, "Field \"%s\" is out of range in the line %s of the file %s", + name, str, DATABASE_FILE_LIST); } + *value_str = *buf; + value_str++; } break; case CONTROL_WAIT_NEXT_NAME: @@ -1400,106 +1458,20 @@ get_control_value(const char *str, const char *name, /* There is no close quotes */ if (state == CONTROL_INNAME || state == CONTROL_INVALUE) - goto bad_format; + control_string_bad_format(str); /* Did not find target field */ if (is_mandatory) - elog(ERROR, "field \"%s\" is not found in the line %s of the file %s", + elog(ERROR, "Field \"%s\" is not found in the line %s of the file %s", name, str, DATABASE_FILE_LIST); return false; - -bad_format: - elog(ERROR, "%s file has invalid format in line %s", - DATABASE_FILE_LIST, str); - return false; /* Make compiler happy */ } -/* - * Construct parray of pgFile from the backup content list. - * If root is not NULL, path will be absolute path. - */ -parray * -dir_read_file_list(const char *root, const char *external_prefix, - const char *file_txt, fio_location location) +static void +control_string_bad_format(const char* str) { - FILE *fp; - parray *files; - char buf[MAXPGPATH * 2]; - - fp = fio_open_stream(file_txt, location); - if (fp == NULL) - elog(ERROR, "cannot open \"%s\": %s", file_txt, strerror(errno)); - - files = parray_new(); - - while (fgets(buf, lengthof(buf), fp)) - { - char path[MAXPGPATH]; - char filepath[MAXPGPATH]; - char linked[MAXPGPATH]; - char compress_alg_string[MAXPGPATH]; - int64 write_size, - mode, /* bit length of mode_t depends on platforms */ - is_datafile, - is_cfs, - external_dir_num, - crc, - segno, - n_blocks; - pgFile *file; - - get_control_value(buf, "path", path, NULL, true); - get_control_value(buf, "size", NULL, &write_size, true); - get_control_value(buf, "mode", NULL, &mode, true); - get_control_value(buf, "is_datafile", NULL, &is_datafile, true); - get_control_value(buf, "is_cfs", NULL, &is_cfs, false); - get_control_value(buf, "crc", NULL, &crc, true); - get_control_value(buf, "compress_alg", compress_alg_string, NULL, false); - get_control_value(buf, "external_dir_num", NULL, &external_dir_num, false); - - if (external_dir_num && external_prefix) - { - char temp[MAXPGPATH]; - - makeExternalDirPathByNum(temp, external_prefix, external_dir_num); - join_path_components(filepath, temp, path); - } - else if (root) - join_path_components(filepath, root, path); - else - strcpy(filepath, path); - - file = pgFileInit(filepath, path); - - file->write_size = (int64) write_size; - file->mode = (mode_t) mode; - file->is_datafile = is_datafile ? true : false; - file->is_cfs = is_cfs ? true : false; - file->crc = (pg_crc32) crc; - file->compress_alg = parse_compress_alg(compress_alg_string); - file->external_dir_num = external_dir_num; - - /* - * Optional fields - */ - - if (get_control_value(buf, "linked", linked, NULL, false) && linked[0]) - { - file->linked = pgut_strdup(linked); - canonicalize_path(file->linked); - } - - if (get_control_value(buf, "segno", NULL, &segno, false)) - file->segno = (int) segno; - - if (get_control_value(buf, "n_blocks", NULL, &n_blocks, false)) - file->n_blocks = (int) n_blocks; - - parray_append(files, file); - } - - fio_close_stream(fp); - return files; + elog(ERROR, "%s file has invalid format in line %s", + DATABASE_FILE_LIST, str); } /* @@ -1517,7 +1489,7 @@ dir_is_empty(const char *path, fio_location location) /* Directory in path doesn't exist */ if (errno == ENOENT) return true; - elog(ERROR, "cannot open directory \"%s\": %s", path, strerror(errno)); + elog(ERROR, "Cannot open directory \"%s\": %s", path, strerror(errno)); } errno = 0; @@ -1533,7 +1505,7 @@ dir_is_empty(const char *path, fio_location location) return false; } if (errno) - elog(ERROR, "cannot read directory \"%s\": %s", path, strerror(errno)); + elog(ERROR, "Cannot read directory \"%s\": %s", path, strerror(errno)); fio_closedir(dir); @@ -1612,7 +1584,7 @@ make_external_directory_list(const char *colon_separated_dirs, bool remap) p = strtok(NULL, EXTERNAL_DIRECTORY_DELIMITER); } pfree(tmp); - parray_qsort(list, BlackListCompare); + parray_qsort(list, pgCompareString); return list; } @@ -1626,8 +1598,7 @@ free_dir_list(parray *list) /* Append to string "path_prefix" int "dir_num" */ void -makeExternalDirPathByNum(char *ret_path, const char *path_prefix, - const int dir_num) +makeExternalDirPathByNum(char *ret_path, const char *path_prefix, const int dir_num) { sprintf(ret_path, "%s%d", path_prefix, dir_num); } @@ -1640,6 +1611,260 @@ backup_contains_external(const char *dir, parray *dirs_list) if (!dirs_list) /* There is no external dirs in backup */ return false; - search_result = parray_bsearch(dirs_list, dir, BlackListCompare); + search_result = parray_bsearch(dirs_list, dir, pgCompareString); return search_result != NULL; } + +/* + * Print database_map + */ +void +print_database_map(FILE *out, parray *database_map) +{ + int i; + + for (i = 0; i < parray_num(database_map); i++) + { + db_map_entry *db_entry = (db_map_entry *) parray_get(database_map, i); + + fio_fprintf(out, "{\"dbOid\":\"%u\", \"datname\":\"%s\"}\n", + db_entry->dbOid, db_entry->datname); + } + +} + +/* + * Create file 'database_map' and add its meta to backup_files_list + * NULL check for database_map must be done by the caller. + */ +void +write_database_map(pgBackup *backup, parray *database_map, parray *backup_files_list) +{ + FILE *fp; + pgFile *file; + char database_dir[MAXPGPATH]; + char database_map_path[MAXPGPATH]; + + join_path_components(database_dir, backup->root_dir, DATABASE_DIR); + join_path_components(database_map_path, database_dir, DATABASE_MAP); + + fp = fio_fopen(database_map_path, PG_BINARY_W, FIO_BACKUP_HOST); + if (fp == NULL) + elog(ERROR, "Cannot open database map \"%s\": %s", database_map_path, + strerror(errno)); + + print_database_map(fp, database_map); + if (fio_fflush(fp) || fio_fclose(fp)) + { + fio_unlink(database_map_path, FIO_BACKUP_HOST); + elog(ERROR, "Cannot write database map \"%s\": %s", + database_map_path, strerror(errno)); + } + + /* Add metadata to backup_content.control */ + file = pgFileNew(database_map_path, DATABASE_MAP, true, 0, + FIO_BACKUP_HOST); + file->crc = pgFileGetCRC(database_map_path, true, false); + file->write_size = file->size; + file->uncompressed_size = file->size; + + parray_append(backup_files_list, file); +} + +/* + * read database map, return NULL if database_map in empty or missing + */ +parray * +read_database_map(pgBackup *backup) +{ + FILE *fp; + parray *database_map; + char buf[MAXPGPATH]; + char path[MAXPGPATH]; + char database_map_path[MAXPGPATH]; + + join_path_components(path, backup->root_dir, DATABASE_DIR); + join_path_components(database_map_path, path, DATABASE_MAP); + + fp = fio_open_stream(database_map_path, FIO_BACKUP_HOST); + if (fp == NULL) + { + /* It is NOT ok for database_map to be missing at this point, so + * we should error here. + * It`s a job of the caller to error if database_map is not empty. + */ + elog(ERROR, "Cannot open \"%s\": %s", database_map_path, strerror(errno)); + } + + database_map = parray_new(); + + while (fgets(buf, lengthof(buf), fp)) + { + char datname[MAXPGPATH]; + int64 dbOid; + + db_map_entry *db_entry = (db_map_entry *) pgut_malloc(sizeof(db_map_entry)); + + get_control_value_int64(buf, "dbOid", &dbOid, true); + get_control_value_str(buf, "datname", datname, sizeof(datname), true); + + db_entry->dbOid = dbOid; + db_entry->datname = pgut_strdup(datname); + + parray_append(database_map, db_entry); + } + + if (ferror(fp)) + elog(ERROR, "Failed to read from file: \"%s\"", database_map_path); + + fio_close_stream(fp); + + /* Return NULL if file is empty */ + if (parray_num(database_map) == 0) + { + parray_free(database_map); + return NULL; + } + + return database_map; +} + +/* + * Use it to cleanup tablespaces + * TODO: Current algorihtm is not very efficient in remote mode, + * due to round-trip to delete every file. + */ +void +cleanup_tablespace(const char *path) +{ + int i; + char fullpath[MAXPGPATH]; + parray *files = parray_new(); + + fio_list_dir(files, path, false, false, false, false, false, 0); + + /* delete leaf node first */ + parray_qsort(files, pgFileCompareRelPathWithExternalDesc); + + for (i = 0; i < parray_num(files); i++) + { + pgFile *file = (pgFile *) parray_get(files, i); + + join_path_components(fullpath, path, file->rel_path); + + fio_delete(file->mode, fullpath, FIO_DB_HOST); + elog(LOG, "Deleted file \"%s\"", fullpath); + } + + parray_walk(files, pgFileFree); + parray_free(files); +} + +/* + * Clear the synchronisation locks in a parray of (pgFile *)'s + */ +void +pfilearray_clear_locks(parray *file_list) +{ + int i; + for (i = 0; i < parray_num(file_list); i++) + { + pgFile *file = (pgFile *) parray_get(file_list, i); + pg_atomic_clear_flag(&file->lock); + } +} + +static inline bool +is_forkname(char *name, size_t *pos, const char *forkname) +{ + size_t fnlen = strlen(forkname); + if (strncmp(name + *pos, forkname, fnlen) != 0) + return false; + *pos += fnlen; + return true; +} + +#define OIDCHARS 10 +#define MAXSEGNO (((uint64_t)1<<32)/RELSEG_SIZE-1) +#define SEGNOCHARS 5 /* when BLCKSZ == (1<<15) */ + +/* Set forkName if possible */ +bool +set_forkname(pgFile *file) +{ + size_t i = 0; + uint64_t oid = 0; /* use 64bit to not check for overflow in a loop */ + uint64_t segno = 0; + + /* pretend it is not relation file */ + file->relOid = 0; + file->forkName = none; + file->is_datafile = false; + + for (i = 0; isdigit(file->name[i]); i++) + { + if (i == 0 && file->name[i] == '0') + return false; + oid = oid * 10 + file->name[i] - '0'; + } + if (i == 0 || i > OIDCHARS || oid > UINT32_MAX) + return false; + + /* usual fork name */ + /* /^\d+_(vm|fsm|init|ptrack)$/ */ + if (is_forkname(file->name, &i, "_vm")) + file->forkName = vm; + else if (is_forkname(file->name, &i, "_fsm")) + file->forkName = fsm; + else if (is_forkname(file->name, &i, "_init")) + file->forkName = init; + else if (is_forkname(file->name, &i, "_ptrack")) + file->forkName = ptrack; + + /* segment number */ + /* /^\d+(_(vm|fsm|init|ptrack))?\.\d+$/ */ + if (file->name[i] == '.' && isdigit(file->name[i+1])) + { + size_t start = i+1; + for (i++; isdigit(file->name[i]); i++) + { + if (i == start && file->name[i] == '0') + return false; + segno = segno * 10 + file->name[i] - '0'; + } + if (i - start > SEGNOCHARS || segno > MAXSEGNO) + return false; + } + + /* CFS family fork names */ + if (file->forkName == none && + is_forkname(file->name, &i, ".cfm.bck")) + { + /* /^\d+(\.\d+)?\.cfm\.bck$/ */ + file->forkName = cfm_bck; + } + if (file->forkName == none && + is_forkname(file->name, &i, ".bck")) + { + /* /^\d+(\.\d+)?\.bck$/ */ + file->forkName = cfs_bck; + } + if (file->forkName == none && + is_forkname(file->name, &i, ".cfm")) + { + /* /^\d+(\.\d+)?.cfm$/ */ + file->forkName = cfm; + } + + /* If there are excess characters, it is not relation file */ + if (file->name[i] != 0) + { + file->forkName = none; + return false; + } + + file->relOid = oid; + file->segno = segno; + file->is_datafile = file->forkName == none; + return true; +} \ No newline at end of file diff --git a/src/fetch.c b/src/fetch.c index 4dfd3a700..5401d815e 100644 --- a/src/fetch.c +++ b/src/fetch.c @@ -32,32 +32,24 @@ slurpFile(const char *datadir, const char *path, size_t *filesize, bool safe, fi struct stat statbuf; char fullpath[MAXPGPATH]; int len; - snprintf(fullpath, sizeof(fullpath), "%s/%s", datadir, path); - if (fio_access(fullpath, R_OK, location) != 0) - { - if (safe) - return NULL; - else - elog(ERROR, "could not open file \"%s\" for reading: %s", - fullpath, strerror(errno)); - } + join_path_components(fullpath, datadir, path); if ((fd = fio_open(fullpath, O_RDONLY | PG_BINARY, location)) == -1) { if (safe) return NULL; else - elog(ERROR, "could not open file \"%s\" for reading: %s", + elog(ERROR, "Could not open file \"%s\" for reading: %s", fullpath, strerror(errno)); } - if (fio_fstat(fd, &statbuf) < 0) + if (fio_stat(fullpath, &statbuf, true, location) < 0) { if (safe) return NULL; else - elog(ERROR, "could not open file \"%s\" for reading: %s", + elog(ERROR, "Could not stat file \"%s\": %s", fullpath, strerror(errno)); } @@ -69,7 +61,7 @@ slurpFile(const char *datadir, const char *path, size_t *filesize, bool safe, fi if (safe) return NULL; else - elog(ERROR, "could not read file \"%s\": %s\n", + elog(ERROR, "Could not read file \"%s\": %s\n", fullpath, strerror(errno)); } @@ -100,7 +92,7 @@ fetchFile(PGconn *conn, const char *filename, size_t *filesize) /* sanity check the result set */ if (PQntuples(res) != 1 || PQgetisnull(res, 0, 0)) - elog(ERROR, "unexpected result set while fetching remote file \"%s\"", + elog(ERROR, "Unexpected result set while fetching remote file \"%s\"", filename); /* Read result to local variables */ diff --git a/src/help.c b/src/help.c index 85b1ed858..e18706a13 100644 --- a/src/help.c +++ b/src/help.c @@ -2,13 +2,16 @@ * * help.c * - * Copyright (c) 2017-2019, Postgres Professional + * Copyright (c) 2017-2021, Postgres Professional * *------------------------------------------------------------------------- */ +#include #include "pg_probackup.h" +static void help_nocmd(void); +static void help_internal(void); static void help_init(void); static void help_backup(void); static void help_restore(void); @@ -16,6 +19,7 @@ static void help_validate(void); static void help_show(void); static void help_delete(void); static void help_merge(void); +static void help_set_backup(void); static void help_set_config(void); static void help_show_config(void); static void help_add_instance(void); @@ -23,66 +27,74 @@ static void help_del_instance(void); static void help_archive_push(void); static void help_archive_get(void); static void help_checkdb(void); +static void help_help(void); +static void help_version(void); +static void help_catchup(void); void -help_command(char *command) +help_print_version(void) { - if (strcmp(command, "init") == 0) - help_init(); - else if (strcmp(command, "backup") == 0) - help_backup(); - else if (strcmp(command, "restore") == 0) - help_restore(); - else if (strcmp(command, "validate") == 0) - help_validate(); - else if (strcmp(command, "show") == 0) - help_show(); - else if (strcmp(command, "delete") == 0) - help_delete(); - else if (strcmp(command, "merge") == 0) - help_merge(); - else if (strcmp(command, "set-config") == 0) - help_set_config(); - else if (strcmp(command, "show-config") == 0) - help_show_config(); - else if (strcmp(command, "add-instance") == 0) - help_add_instance(); - else if (strcmp(command, "del-instance") == 0) - help_del_instance(); - else if (strcmp(command, "archive-push") == 0) - help_archive_push(); - else if (strcmp(command, "archive-get") == 0) - help_archive_get(); - else if (strcmp(command, "checkdb") == 0) - help_checkdb(); - else if (strcmp(command, "--help") == 0 - || strcmp(command, "help") == 0 - || strcmp(command, "-?") == 0 - || strcmp(command, "--version") == 0 - || strcmp(command, "version") == 0 - || strcmp(command, "-V") == 0) - printf(_("No help page for \"%s\" command. Try pg_probackup help\n"), command); - else - printf(_("Unknown command \"%s\". Try pg_probackup help\n"), command); - exit(0); +#ifdef PGPRO_VERSION + fprintf(stdout, "%s %s (Postgres Pro %s %s)\n", + PROGRAM_NAME, PROGRAM_VERSION, + PGPRO_VERSION, PGPRO_EDITION); +#else + fprintf(stdout, "%s %s (PostgreSQL %s)\n", + PROGRAM_NAME, PROGRAM_VERSION, PG_VERSION); +#endif +} + +void +help_command(ProbackupSubcmd const subcmd) +{ + typedef void (* help_function_ptr)(void); + /* Order is important, keep it in sync with utils/configuration.h:enum ProbackupSubcmd declaration */ + static help_function_ptr const help_functions[] = + { + &help_nocmd, + &help_init, + &help_add_instance, + &help_del_instance, + &help_archive_push, + &help_archive_get, + &help_backup, + &help_restore, + &help_validate, + &help_delete, + &help_merge, + &help_show, + &help_set_config, + &help_set_backup, + &help_show_config, + &help_checkdb, + &help_internal, // SSH_CMD + &help_internal, // AGENT_CMD + &help_help, + &help_version, + &help_catchup, + }; + + Assert((int)subcmd < sizeof(help_functions) / sizeof(help_functions[0])); + help_functions[(int)subcmd](); } void help_pg_probackup(void) { - printf(_("\n%s - utility to manage backup/recovery of PostgreSQL database.\n\n"), PROGRAM_NAME); + printf(_("\n%s - utility to manage backup/recovery of PostgreSQL database.\n"), PROGRAM_NAME); - printf(_(" %s help [COMMAND]\n"), PROGRAM_NAME); + printf(_("\n %s help [COMMAND]\n"), PROGRAM_NAME); printf(_("\n %s version\n"), PROGRAM_NAME); - printf(_("\n %s init -B backup-path\n"), PROGRAM_NAME); + printf(_("\n %s init -B backup-dir\n"), PROGRAM_NAME); - printf(_("\n %s set-config -B backup-path --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n %s set-config -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" [-D pgdata-path]\n")); printf(_(" [--external-dirs=external-directories-paths]\n")); printf(_(" [--log-level-console=log-level-console]\n")); printf(_(" [--log-level-file=log-level-file]\n")); + printf(_(" [--log-format-file=log-format-file]\n")); printf(_(" [--log-filename=log-filename]\n")); printf(_(" [--error-log-filename=error-log-filename]\n")); printf(_(" [--log-directory=log-directory]\n")); @@ -90,6 +102,7 @@ help_pg_probackup(void) printf(_(" [--log-rotation-age=log-rotation-age]\n")); printf(_(" [--retention-redundancy=retention-redundancy]\n")); printf(_(" [--retention-window=retention-window]\n")); + printf(_(" [--wal-depth=wal-depth]\n")); printf(_(" [--compress-algorithm=compress-algorithm]\n")); printf(_(" [--compress-level=compress-level]\n")); printf(_(" [--archive-timeout=timeout]\n")); @@ -97,28 +110,40 @@ help_pg_probackup(void) printf(_(" [--remote-proto] [--remote-host]\n")); printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); printf(_(" [--ssh-options]\n")); + printf(_(" [--restore-command=cmdline] [--archive-host=destination]\n")); + printf(_(" [--archive-port=port] [--archive-user=username]\n")); + printf(_(" [--help]\n")); + + printf(_("\n %s set-backup -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); + printf(_(" -i backup-id [--ttl=interval] [--expire-time=timestamp]\n")); + printf(_(" [--note=text]\n")); printf(_(" [--help]\n")); - printf(_("\n %s show-config -B backup-path --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n %s show-config -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" [--format=format]\n")); + printf(_(" [--no-scale-units]\n")); printf(_(" [--help]\n")); - printf(_("\n %s backup -B backup-path -b backup-mode --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n %s backup -B backup-dir -b backup-mode --instance=instance-name\n"), PROGRAM_NAME); printf(_(" [-D pgdata-path] [-C]\n")); - printf(_(" [--stream [-S slot-name]] [--temp-slot]\n")); + printf(_(" [--stream [-S slot-name] [--temp-slot]]\n")); printf(_(" [--backup-pg-log] [-j num-threads] [--progress]\n")); printf(_(" [--no-validate] [--skip-block-validation]\n")); printf(_(" [--external-dirs=external-directories-paths]\n")); + printf(_(" [--no-sync]\n")); printf(_(" [--log-level-console=log-level-console]\n")); printf(_(" [--log-level-file=log-level-file]\n")); + printf(_(" [--log-format-console=log-format-console]\n")); + printf(_(" [--log-format-file=log-format-file]\n")); printf(_(" [--log-filename=log-filename]\n")); printf(_(" [--error-log-filename=error-log-filename]\n")); printf(_(" [--log-directory=log-directory]\n")); printf(_(" [--log-rotation-size=log-rotation-size]\n")); - printf(_(" [--log-rotation-age=log-rotation-age]\n")); + printf(_(" [--log-rotation-age=log-rotation-age] [--no-color]\n")); printf(_(" [--delete-expired] [--delete-wal] [--merge-expired]\n")); printf(_(" [--retention-redundancy=retention-redundancy]\n")); printf(_(" [--retention-window=retention-window]\n")); + printf(_(" [--wal-depth=wal-depth]\n")); printf(_(" [--compress]\n")); printf(_(" [--compress-algorithm=compress-algorithm]\n")); printf(_(" [--compress-level=compress-level]\n")); @@ -128,9 +153,11 @@ help_pg_probackup(void) printf(_(" [--remote-proto] [--remote-host]\n")); printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); printf(_(" [--ssh-options]\n")); + printf(_(" [--ttl=interval] [--expire-time=timestamp] [--note=text]\n")); printf(_(" [--help]\n")); - printf(_("\n %s restore -B backup-path --instance=instance_name\n"), PROGRAM_NAME); + + printf(_("\n %s restore -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" [-D pgdata-path] [-i backup-id] [-j num-threads]\n")); printf(_(" [--recovery-target-time=time|--recovery-target-xid=xid\n")); printf(_(" |--recovery-target-lsn=lsn [--recovery-target-inclusive=boolean]]\n")); @@ -138,17 +165,26 @@ help_pg_probackup(void) printf(_(" [--recovery-target=immediate|latest]\n")); printf(_(" [--recovery-target-name=target-name]\n")); printf(_(" [--recovery-target-action=pause|promote|shutdown]\n")); - printf(_(" [--restore-as-replica]\n")); + printf(_(" [--restore-command=cmdline]\n")); + printf(_(" [-R | --restore-as-replica] [--force]\n")); + printf(_(" [--primary-conninfo=primary_conninfo]\n")); + printf(_(" [-S | --primary-slot-name=slotname]\n")); printf(_(" [--no-validate] [--skip-block-validation]\n")); printf(_(" [-T OLDDIR=NEWDIR] [--progress]\n")); printf(_(" [--external-mapping=OLDDIR=NEWDIR]\n")); - printf(_(" [--skip-external-dirs]\n")); + printf(_(" [--skip-external-dirs] [--no-sync]\n")); + printf(_(" [-X WALDIR | --waldir=WALDIR]\n")); + printf(_(" [-I | --incremental-mode=none|checksum|lsn]\n")); + printf(_(" [--db-include | --db-exclude]\n")); + printf(_(" [--destroy-all-other-dbs]\n")); printf(_(" [--remote-proto] [--remote-host]\n")); printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); printf(_(" [--ssh-options]\n")); + printf(_(" [--archive-host=hostname]\n")); + printf(_(" [--archive-port=port] [--archive-user=username]\n")); printf(_(" [--help]\n")); - printf(_("\n %s validate -B backup-path [--instance=instance_name]\n"), PROGRAM_NAME); + printf(_("\n %s validate -B backup-dir [--instance=instance-name]\n"), PROGRAM_NAME); printf(_(" [-i backup-id] [--progress] [-j num-threads]\n")); printf(_(" [--recovery-target-time=time|--recovery-target-xid=xid\n")); printf(_(" |--recovery-target-lsn=lsn [--recovery-target-inclusive=boolean]]\n")); @@ -157,43 +193,51 @@ help_pg_probackup(void) printf(_(" [--skip-block-validation]\n")); printf(_(" [--help]\n")); - printf(_("\n %s checkdb [-B backup-path] [--instance=instance_name]\n"), PROGRAM_NAME); + printf(_("\n %s checkdb [-B backup-dir] [--instance=instance-name]\n"), PROGRAM_NAME); printf(_(" [-D pgdata-path] [--progress] [-j num-threads]\n")); printf(_(" [--amcheck] [--skip-block-validation]\n")); - printf(_(" [--heapallindexed]\n")); + printf(_(" [--heapallindexed] [--checkunique]\n")); printf(_(" [--help]\n")); - printf(_("\n %s show -B backup-path\n"), PROGRAM_NAME); - printf(_(" [--instance=instance_name [-i backup-id]]\n")); - printf(_(" [--format=format]\n")); - printf(_(" [--help]\n")); + printf(_("\n %s show -B backup-dir\n"), PROGRAM_NAME); + printf(_(" [--instance=instance-name [-i backup-id]]\n")); + printf(_(" [--format=format] [--archive]\n")); + printf(_(" [--no-color] [--help]\n")); - printf(_("\n %s delete -B backup-path --instance=instance_name\n"), PROGRAM_NAME); - printf(_(" [--wal] [-i backup-id | --expired | --merge-expired]\n")); - printf(_(" [--dry-run]\n")); + printf(_("\n %s delete -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); + printf(_(" [-j num-threads] [--progress]\n")); + printf(_(" [--retention-redundancy=retention-redundancy]\n")); + printf(_(" [--retention-window=retention-window]\n")); + printf(_(" [--wal-depth=wal-depth]\n")); + printf(_(" [-i backup-id | --delete-expired | --merge-expired | --status=backup_status]\n")); + printf(_(" [--delete-wal]\n")); + printf(_(" [--dry-run] [--no-validate] [--no-sync]\n")); printf(_(" [--help]\n")); - printf(_("\n %s merge -B backup-path --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n %s merge -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" -i backup-id [--progress] [-j num-threads]\n")); + printf(_(" [--no-validate] [--no-sync]\n")); printf(_(" [--help]\n")); - printf(_("\n %s add-instance -B backup-path -D pgdata-path\n"), PROGRAM_NAME); - printf(_(" --instance=instance_name\n")); + printf(_("\n %s add-instance -B backup-dir -D pgdata-path\n"), PROGRAM_NAME); + printf(_(" --instance=instance-name\n")); printf(_(" [--external-dirs=external-directories-paths]\n")); printf(_(" [--remote-proto] [--remote-host]\n")); printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); printf(_(" [--ssh-options]\n")); printf(_(" [--help]\n")); - printf(_("\n %s del-instance -B backup-path\n"), PROGRAM_NAME); - printf(_(" --instance=instance_name\n")); + printf(_("\n %s del-instance -B backup-dir\n"), PROGRAM_NAME); + printf(_(" --instance=instance-name\n")); printf(_(" [--help]\n")); - printf(_("\n %s archive-push -B backup-path --instance=instance_name\n"), PROGRAM_NAME); - printf(_(" --wal-file-path=wal-file-path\n")); + printf(_("\n %s archive-push -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" --wal-file-name=wal-file-name\n")); - printf(_(" [--overwrite]\n")); - printf(_(" [--compress]\n")); + printf(_(" [--wal-file-path=wal-file-path]\n")); + printf(_(" [-j num-threads] [--batch-size=batch_size]\n")); + printf(_(" [--archive-timeout=timeout]\n")); + printf(_(" [--no-ready-rename] [--no-sync]\n")); + printf(_(" [--overwrite] [--compress]\n")); printf(_(" [--compress-algorithm=compress-algorithm]\n")); printf(_(" [--compress-level=compress-level]\n")); printf(_(" [--remote-proto] [--remote-host]\n")); @@ -201,51 +245,83 @@ help_pg_probackup(void) printf(_(" [--ssh-options]\n")); printf(_(" [--help]\n")); - printf(_("\n %s archive-get -B backup-path --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n %s archive-get -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" --wal-file-path=wal-file-path\n")); printf(_(" --wal-file-name=wal-file-name\n")); + printf(_(" [-j num-threads] [--batch-size=batch_size]\n")); + printf(_(" [--no-validate-wal]\n")); + printf(_(" [--remote-proto] [--remote-host]\n")); + printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); + printf(_(" [--ssh-options]\n")); + printf(_(" [--help]\n")); + + printf(_("\n %s catchup -b catchup-mode\n"), PROGRAM_NAME); + printf(_(" --source-pgdata=path_to_pgdata_on_remote_server\n")); + printf(_(" --destination-pgdata=path_to_local_dir\n")); + printf(_(" [--stream [-S slot-name] [--temp-slot | --perm-slot]]\n")); + printf(_(" [-j num-threads]\n")); + printf(_(" [-T OLDDIR=NEWDIR]\n")); + printf(_(" [--exclude-path=path_prefix]\n")); + printf(_(" [-d dbname] [-h host] [-p port] [-U username]\n")); + printf(_(" [-w --no-password] [-W --password]\n")); printf(_(" [--remote-proto] [--remote-host]\n")); printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); printf(_(" [--ssh-options]\n")); + printf(_(" [--dry-run]\n")); printf(_(" [--help]\n")); if ((PROGRAM_URL || PROGRAM_EMAIL)) { printf("\n"); if (PROGRAM_URL) - printf("Read the website for details. <%s>\n", PROGRAM_URL); + printf(_("Read the website for details <%s>.\n"), PROGRAM_URL); if (PROGRAM_EMAIL) - printf("Report bugs to <%s>.\n", PROGRAM_EMAIL); + printf(_("Report bugs to <%s>.\n"), PROGRAM_EMAIL); } - exit(0); +} + +static void +help_nocmd(void) +{ + printf(_("\nUnknown command. Try pg_probackup help\n\n")); +} + +static void +help_internal(void) +{ + printf(_("\nThis command is intended for internal use\n\n")); } static void help_init(void) { - printf(_("\n%s init -B backup-path\n\n"), PROGRAM_NAME); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n\n")); + printf(_("\n%s init -B backup-dir\n\n"), PROGRAM_NAME); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n\n")); } static void help_backup(void) { - printf(_("\n%s backup -B backup-path -b backup-mode --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n%s backup -B backup-dir -b backup-mode --instance=instance-name\n"), PROGRAM_NAME); printf(_(" [-D pgdata-path] [-C]\n")); - printf(_(" [--stream [-S slot-name] [--temp-slot]\n")); + printf(_(" [--stream [-S slot-name] [--temp-slot]]\n")); printf(_(" [--backup-pg-log] [-j num-threads] [--progress]\n")); printf(_(" [--no-validate] [--skip-block-validation]\n")); printf(_(" [-E external-directories-paths]\n")); + printf(_(" [--no-sync]\n")); printf(_(" [--log-level-console=log-level-console]\n")); printf(_(" [--log-level-file=log-level-file]\n")); + printf(_(" [--log-format-console=log-format-console]\n")); + printf(_(" [--log-format-file=log-format-file]\n")); printf(_(" [--log-filename=log-filename]\n")); printf(_(" [--error-log-filename=error-log-filename]\n")); printf(_(" [--log-directory=log-directory]\n")); printf(_(" [--log-rotation-size=log-rotation-size]\n")); - printf(_(" [--log-rotation-age=log-rotation-age]\n")); + printf(_(" [--log-rotation-age=log-rotation-age] [--no-color]\n")); printf(_(" [--delete-expired] [--delete-wal] [--merge-expired]\n")); printf(_(" [--retention-redundancy=retention-redundancy]\n")); printf(_(" [--retention-window=retention-window]\n")); + printf(_(" [--wal-depth=wal-depth]\n")); printf(_(" [--compress]\n")); printf(_(" [--compress-algorithm=compress-algorithm]\n")); printf(_(" [--compress-level=compress-level]\n")); @@ -254,11 +330,12 @@ help_backup(void) printf(_(" [-w --no-password] [-W --password]\n")); printf(_(" [--remote-proto] [--remote-host]\n")); printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); - printf(_(" [--ssh-options]\n\n")); + printf(_(" [--ssh-options]\n")); + printf(_(" [--ttl=interval] [--expire-time=timestamp] [--note=text]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); printf(_(" -b, --backup-mode=backup-mode backup mode=FULL|PAGE|DELTA|PTRACK\n")); - printf(_(" --instance=instance_name name of the instance\n")); + printf(_(" --instance=instance-name name of the instance\n")); printf(_(" -D, --pgdata=pgdata-path location of the database storage area\n")); printf(_(" -C, --smooth-checkpoint do smooth checkpoint before backup\n")); printf(_(" --stream stream the transaction log and include it in the backup\n")); @@ -272,6 +349,9 @@ help_backup(void) printf(_(" -E --external-dirs=external-directories-paths\n")); printf(_(" backup some directories not from pgdata \n")); printf(_(" (example: --external-dirs=/tmp/dir1:/tmp/dir2)\n")); + printf(_(" --no-sync do not sync backed up files to disk\n")); + printf(_(" --note=text add note to backup\n")); + printf(_(" (example: --note='backup before app update to v13.1')\n")); printf(_("\n Logging options:\n")); printf(_(" --log-level-console=log-level-console\n")); @@ -280,6 +360,12 @@ help_backup(void) printf(_(" --log-level-file=log-level-file\n")); printf(_(" level for file logging (default: off)\n")); printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-console=log-format-console\n")); + printf(_(" defines the format of the console log (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); printf(_(" --log-filename=log-filename\n")); printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n")); @@ -293,19 +379,29 @@ help_backup(void) printf(_(" --log-rotation-age=log-rotation-age\n")); printf(_(" rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n")); printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n")); + printf(_(" --no-color disable the coloring of error and warning console messages\n")); printf(_("\n Retention options:\n")); printf(_(" --delete-expired delete backups expired according to current\n")); printf(_(" retention policy after successful backup completion\n")); printf(_(" --merge-expired merge backups expired according to current\n")); printf(_(" retention policy after successful backup completion\n")); - printf(_(" --delete-wal remove redundant archived wal files\n")); + printf(_(" --delete-wal remove redundant files in WAL archive\n")); printf(_(" --retention-redundancy=retention-redundancy\n")); printf(_(" number of full backups to keep; 0 disables; (default: 0)\n")); printf(_(" --retention-window=retention-window\n")); printf(_(" number of days of recoverability; 0 disables; (default: 0)\n")); + printf(_(" --wal-depth=wal-depth number of latest valid backups per timeline that must\n")); + printf(_(" retain the ability to perform PITR; 0 disables; (default: 0)\n")); printf(_(" --dry-run perform a trial run without any changes\n")); + printf(_("\n Pinning options:\n")); + printf(_(" --ttl=interval pin backup for specified amount of time; 0 unpin\n")); + printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: s)\n")); + printf(_(" (example: --ttl=20d)\n")); + printf(_(" --expire-time=time pin backup until specified time stamp\n")); + printf(_(" (example: --expire-time='2024-01-01 00:00:00+03')\n")); + printf(_("\n Compression options:\n")); printf(_(" --compress alias for --compress-algorithm='zlib' and --compress-level=1\n")); printf(_(" --compress-algorithm=compress-algorithm\n")); @@ -314,20 +410,20 @@ help_backup(void) printf(_(" level of compression [0-9] (default: 1)\n")); printf(_("\n Archive options:\n")); - printf(_(" --archive-timeout=timeout wait timeout for WAL segment archiving (default: 5min)\n")); + printf(_(" --archive-timeout=timeout wait timeout for WAL segment archiving (default: 5min)\n")); printf(_("\n Connection options:\n")); - printf(_(" -U, --username=USERNAME user name to connect as (default: current local user)\n")); - printf(_(" -d, --dbname=DBNAME database to connect (default: username)\n")); - printf(_(" -h, --host=HOSTNAME database server host or socket directory(default: 'local socket')\n")); - printf(_(" -p, --port=PORT database server port (default: 5432)\n")); + printf(_(" -U, --pguser=USERNAME user name to connect as (default: current local user)\n")); + printf(_(" -d, --pgdatabase=DBNAME database to connect (default: username)\n")); + printf(_(" -h, --pghost=HOSTNAME database server host or socket directory(default: 'local socket')\n")); + printf(_(" -p, --pgport=PORT database server port (default: 5432)\n")); printf(_(" -w, --no-password never prompt for password\n")); printf(_(" -W, --password force password prompt\n")); printf(_("\n Remote options:\n")); printf(_(" --remote-proto=protocol remote protocol to use\n")); printf(_(" available options: 'ssh', 'none' (default: ssh)\n")); - printf(_(" --remote-host=hostname remote host address or hostname\n")); + printf(_(" --remote-host=destination remote host address or hostname\n")); printf(_(" --remote-port=port remote host port (default: 22)\n")); printf(_(" --remote-path=path path to directory with pg_probackup binary on remote host\n")); printf(_(" (default: current binary path)\n")); @@ -346,31 +442,69 @@ help_backup(void) static void help_restore(void) { - printf(_("\n%s restore -B backup-path --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n%s restore -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" [-D pgdata-path] [-i backup-id] [-j num-threads]\n")); + printf(_(" [--progress] [--force] [--no-sync]\n")); + printf(_(" [--no-validate] [--skip-block-validation]\n")); + printf(_(" [-T OLDDIR=NEWDIR]\n")); + printf(_(" [--external-mapping=OLDDIR=NEWDIR]\n")); + printf(_(" [--skip-external-dirs]\n")); + printf(_(" [-X WALDIR | --waldir=WALDIR]\n")); + printf(_(" [-I | --incremental-mode=none|checksum|lsn]\n")); + printf(_(" [--db-include dbname | --db-exclude dbname]\n")); + printf(_(" [--destroy-all-other-dbs]\n")); printf(_(" [--recovery-target-time=time|--recovery-target-xid=xid\n")); printf(_(" |--recovery-target-lsn=lsn [--recovery-target-inclusive=boolean]]\n")); printf(_(" [--recovery-target-timeline=timeline]\n")); printf(_(" [--recovery-target=immediate|latest]\n")); printf(_(" [--recovery-target-name=target-name]\n")); printf(_(" [--recovery-target-action=pause|promote|shutdown]\n")); - printf(_(" [--restore-as-replica]\n")); - printf(_(" [--no-validate] [--skip-block-validation]\n")); - printf(_(" [-T OLDDIR=NEWDIR] [--progress]\n")); - printf(_(" [--external-mapping=OLDDIR=NEWDIR]\n")); - printf(_(" [--skip-external-dirs]\n")); + printf(_(" [--restore-command=cmdline]\n")); + printf(_(" [-R | --restore-as-replica]\n")); + printf(_(" [--primary-conninfo=primary_conninfo]\n")); + printf(_(" [-S | --primary-slot-name=slotname]\n")); printf(_(" [--remote-proto] [--remote-host]\n")); printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); - printf(_(" [--ssh-options]\n\n")); + printf(_(" [--ssh-options]\n")); + printf(_(" [--archive-host=hostname] [--archive-port=port]\n")); + printf(_(" [--archive-user=username]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name name of the instance\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name name of the instance\n")); printf(_(" -D, --pgdata=pgdata-path location of the database storage area\n")); printf(_(" -i, --backup-id=backup-id backup to restore\n")); printf(_(" -j, --threads=NUM number of parallel threads\n")); printf(_(" --progress show progress\n")); + printf(_(" --force ignore invalid status of the restored backup\n")); + printf(_(" --no-sync do not sync restored files to disk\n")); + printf(_(" --no-validate disable backup validation during restore\n")); + printf(_(" --skip-block-validation set to validate only file-level checksum\n")); + + printf(_(" -T, --tablespace-mapping=OLDDIR=NEWDIR\n")); + printf(_(" relocate the tablespace from directory OLDDIR to NEWDIR\n")); + printf(_(" --external-mapping=OLDDIR=NEWDIR\n")); + printf(_(" relocate the external directory from OLDDIR to NEWDIR\n")); + printf(_(" --skip-external-dirs do not restore all external directories\n")); + + + printf(_(" -X, --waldir=WALDIR location for the write-ahead log directory\n")); + + + printf(_("\n Incremental restore options:\n")); + printf(_(" -I, --incremental-mode=none|checksum|lsn\n")); + printf(_(" reuse valid pages available in PGDATA if they have not changed\n")); + printf(_(" (default: none)\n")); + + printf(_("\n Partial restore options:\n")); + printf(_(" --db-include dbname restore only specified databases\n")); + printf(_(" --db-exclude dbname do not restore specified databases\n")); + printf(_(" --destroy-all-other-dbs\n")); + printf(_(" allows to do partial restore that is prohibited by default,\n")); + printf(_(" because it might remove all other databases.\n")); + + printf(_("\n Recovery options:\n")); printf(_(" --recovery-target-time=time time stamp up to which recovery will proceed\n")); printf(_(" --recovery-target-xid=xid transaction ID up to which recovery will proceed\n")); printf(_(" --recovery-target-lsn=lsn LSN of the write-ahead log location up to which recovery will proceed\n")); @@ -385,17 +519,15 @@ help_restore(void) printf(_(" --recovery-target-action=pause|promote|shutdown\n")); printf(_(" action the server should take once the recovery target is reached\n")); printf(_(" (default: pause)\n")); + printf(_(" --restore-command=cmdline command to use as 'restore_command' in recovery.conf; 'none' disables\n")); + printf(_("\n Standby options:\n")); printf(_(" -R, --restore-as-replica write a minimal recovery.conf in the output directory\n")); printf(_(" to ease setting up a standby server\n")); - printf(_(" --no-validate disable backup validation during restore\n")); - printf(_(" --skip-block-validation set to validate only file-level checksum\n")); - - printf(_(" -T, --tablespace-mapping=OLDDIR=NEWDIR\n")); - printf(_(" relocate the tablespace from directory OLDDIR to NEWDIR\n")); - printf(_(" --external-mapping=OLDDIR=NEWDIR\n")); - printf(_(" relocate the external directory from OLDDIR to NEWDIR\n")); - printf(_(" --skip-external-dirs do not restore all external directories\n")); + printf(_(" --primary-conninfo=primary_conninfo\n")); + printf(_(" connection string to be used for establishing connection\n")); + printf(_(" with the primary server\n")); + printf(_(" -S, --primary-slot-name=slotname replication slot to be used for WAL streaming from the primary server\n")); printf(_("\n Logging options:\n")); printf(_(" --log-level-console=log-level-console\n")); @@ -404,6 +536,12 @@ help_restore(void) printf(_(" --log-level-file=log-level-file\n")); printf(_(" level for file logging (default: off)\n")); printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-console=log-format-console\n")); + printf(_(" defines the format of the console log (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); printf(_(" --log-filename=log-filename\n")); printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n")); @@ -417,23 +555,29 @@ help_restore(void) printf(_(" --log-rotation-age=log-rotation-age\n")); printf(_(" rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n")); printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n")); + printf(_(" --no-color disable the coloring of error and warning console messages\n")); printf(_("\n Remote options:\n")); printf(_(" --remote-proto=protocol remote protocol to use\n")); printf(_(" available options: 'ssh', 'none' (default: ssh)\n")); - printf(_(" --remote-host=hostname remote host address or hostname\n")); + printf(_(" --remote-host=destination remote host address or hostname\n")); printf(_(" --remote-port=port remote host port (default: 22)\n")); printf(_(" --remote-path=path path to directory with pg_probackup binary on remote host\n")); printf(_(" (default: current binary path)\n")); printf(_(" --remote-user=username user name for ssh connection (default: current user)\n")); printf(_(" --ssh-options=ssh_options additional ssh options (default: none)\n")); - printf(_(" (example: --ssh-options='-c cipher_spec -F configfile')\n\n")); + printf(_(" (example: --ssh-options='-c cipher_spec -F configfile')\n")); + + printf(_("\n Remote WAL archive options:\n")); + printf(_(" --archive-host=destination address or hostname for ssh connection to archive host\n")); + printf(_(" --archive-port=port port for ssh connection to archive host (default: 22)\n")); + printf(_(" --archive-user=username user name for ssh connection to archive host (default: PostgreSQL user)\n\n")); } static void help_validate(void) { - printf(_("\n%s validate -B backup-path [--instance=instance_name]\n"), PROGRAM_NAME); + printf(_("\n%s validate -B backup-dir [--instance=instance-name]\n"), PROGRAM_NAME); printf(_(" [-i backup-id] [--progress] [-j num-threads]\n")); printf(_(" [--recovery-target-time=time|--recovery-target-xid=xid\n")); printf(_(" |--recovery-target-lsn=lsn [--recovery-target-inclusive=boolean]]\n")); @@ -441,8 +585,8 @@ help_validate(void) printf(_(" [--recovery-target-name=target-name]\n")); printf(_(" [--skip-block-validation]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name name of the instance\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name name of the instance\n")); printf(_(" -i, --backup-id=backup-id backup to validate\n")); printf(_(" --progress show progress\n")); @@ -465,6 +609,12 @@ help_validate(void) printf(_(" --log-level-file=log-level-file\n")); printf(_(" level for file logging (default: off)\n")); printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-console=log-format-console\n")); + printf(_(" defines the format of the console log (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); printf(_(" --log-filename=log-filename\n")); printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n")); @@ -477,19 +627,20 @@ help_validate(void) printf(_(" available units: 'kB', 'MB', 'GB', 'TB' (default: kB)\n")); printf(_(" --log-rotation-age=log-rotation-age\n")); printf(_(" rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n")); - printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n\n")); + printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n")); + printf(_(" --no-color disable the coloring of error and warning console messages\n\n")); } static void help_checkdb(void) { - printf(_("\n%s checkdb [-B backup-path] [--instance=instance_name]\n"), PROGRAM_NAME); + printf(_("\n%s checkdb [-B backup-dir] [--instance=instance-name]\n"), PROGRAM_NAME); printf(_(" [-D pgdata-path] [-j num-threads] [--progress]\n")); printf(_(" [--amcheck] [--skip-block-validation]\n")); - printf(_(" [--heapallindexed]\n\n")); + printf(_(" [--heapallindexed] [--checkunique]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name name of the instance\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name name of the instance\n")); printf(_(" -D, --pgdata=pgdata-path location of the database storage area\n")); printf(_(" --progress show progress\n")); @@ -501,6 +652,8 @@ help_checkdb(void) printf(_(" using 'amcheck' or 'amcheck_next' extensions\n")); printf(_(" --heapallindexed also check that heap is indexed\n")); printf(_(" can be used only with '--amcheck' option\n")); + printf(_(" --checkunique also check unique constraints\n")); + printf(_(" can be used only with '--amcheck' option\n")); printf(_("\n Logging options:\n")); printf(_(" --log-level-console=log-level-console\n")); @@ -509,6 +662,12 @@ help_checkdb(void) printf(_(" --log-level-file=log-level-file\n")); printf(_(" level for file logging (default: off)\n")); printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-console=log-format-console\n")); + printf(_(" defines the format of the console log (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); printf(_(" --log-filename=log-filename\n")); printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log\n")); @@ -522,12 +681,13 @@ help_checkdb(void) printf(_(" --log-rotation-age=log-rotation-age\n")); printf(_(" rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n")); printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n")); + printf(_(" --no-color disable the coloring of error and warning console messages\n")); printf(_("\n Connection options:\n")); - printf(_(" -U, --username=USERNAME user name to connect as (default: current local user)\n")); - printf(_(" -d, --dbname=DBNAME database to connect (default: username)\n")); - printf(_(" -h, --host=HOSTNAME database server host or socket directory(default: 'local socket')\n")); - printf(_(" -p, --port=PORT database server port (default: 5432)\n")); + printf(_(" -U, --pguser=USERNAME user name to connect as (default: current local user)\n")); + printf(_(" -d, --pgdatabase=DBNAME database to connect (default: username)\n")); + printf(_(" -h, --pghost=HOSTNAME database server host or socket directory(default: 'local socket')\n")); + printf(_(" -p, --pgport=PORT database server port (default: 5432)\n")); printf(_(" -w, --no-password never prompt for password\n")); printf(_(" -W, --password force password prompt\n\n")); } @@ -535,33 +695,51 @@ help_checkdb(void) static void help_show(void) { - printf(_("\n%s show -B backup-path\n"), PROGRAM_NAME); - printf(_(" [--instance=instance_name [-i backup-id]]\n")); - printf(_(" [--format=format]\n\n")); + printf(_("\n%s show -B backup-dir\n"), PROGRAM_NAME); + printf(_(" [--instance=instance-name [-i backup-id]]\n")); + printf(_(" [--format=format] [--archive]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name show info about specific instance\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name show info about specific instance\n")); printf(_(" -i, --backup-id=backup-id show info about specific backups\n")); - printf(_(" --format=format show format=PLAIN|JSON\n\n")); + printf(_(" --archive show WAL archive information\n")); + printf(_(" --format=format show format=PLAIN|JSON\n")); + printf(_(" --no-color disable the coloring for plain format\n\n")); } static void help_delete(void) { - printf(_("\n%s delete -B backup-path --instance=instance_name\n"), PROGRAM_NAME); - printf(_(" [-i backup-id | --expired | --merge-expired] [--wal]\n")); - printf(_(" [-j num-threads] [--dry-run]\n\n")); + printf(_("\n%s delete -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); + printf(_(" [-i backup-id | --delete-expired | --merge-expired] [--delete-wal]\n")); + printf(_(" [-j num-threads] [--progress]\n")); + printf(_(" [--retention-redundancy=retention-redundancy]\n")); + printf(_(" [--retention-window=retention-window]\n")); + printf(_(" [--wal-depth=wal-depth]\n")); + printf(_(" [--no-validate] [--no-sync]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name name of the instance\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name name of the instance\n")); printf(_(" -i, --backup-id=backup-id backup to delete\n")); - printf(_(" --expired delete backups expired according to current\n")); + printf(_(" -j, --threads=NUM number of parallel threads\n")); + printf(_(" --progress show progress\n")); + printf(_(" --no-validate disable validation during retention merge\n")); + printf(_(" --no-sync do not sync merged files to disk\n")); + + printf(_("\n Retention options:\n")); + printf(_(" --delete-expired delete backups expired according to current\n")); printf(_(" retention policy\n")); printf(_(" --merge-expired merge backups expired according to current\n")); printf(_(" retention policy\n")); - printf(_(" --wal remove unnecessary wal files in WAL ARCHIVE\n")); - printf(_(" -j, --threads=NUM number of parallel threads\n")); + printf(_(" --delete-wal remove redundant files in WAL archive\n")); + printf(_(" --retention-redundancy=retention-redundancy\n")); + printf(_(" number of full backups to keep; 0 disables; (default: 0)\n")); + printf(_(" --retention-window=retention-window\n")); + printf(_(" number of days of recoverability; 0 disables; (default: 0)\n")); + printf(_(" --wal-depth=wal-depth number of latest valid backups per timeline that must\n")); + printf(_(" retain the ability to perform PITR; 0 disables; (default: 0)\n")); printf(_(" --dry-run perform a trial run without any changes\n")); + printf(_(" --status=backup_status delete all backups with specified status\n")); printf(_("\n Logging options:\n")); printf(_(" --log-level-console=log-level-console\n")); @@ -570,6 +748,12 @@ help_delete(void) printf(_(" --log-level-file=log-level-file\n")); printf(_(" level for file logging (default: off)\n")); printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-console=log-format-console\n")); + printf(_(" defines the format of the console log (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); printf(_(" --log-filename=log-filename\n")); printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n")); @@ -582,28 +766,34 @@ help_delete(void) printf(_(" available units: 'kB', 'MB', 'GB', 'TB' (default: kB)\n")); printf(_(" --log-rotation-age=log-rotation-age\n")); printf(_(" rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n")); - printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n\n")); + printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n")); + printf(_(" --no-color disable the coloring of error and warning console messages\n\n")); } static void help_merge(void) { - printf(_("\n%s merge -B backup-path --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n%s merge -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" -i backup-id [-j num-threads] [--progress]\n")); + printf(_(" [--no-validate] [--no-sync]\n")); printf(_(" [--log-level-console=log-level-console]\n")); printf(_(" [--log-level-file=log-level-file]\n")); + printf(_(" [--log-format-console=log-format-console]\n")); + printf(_(" [--log-format-file=log-format-file]\n")); printf(_(" [--log-filename=log-filename]\n")); printf(_(" [--error-log-filename=error-log-filename]\n")); printf(_(" [--log-directory=log-directory]\n")); printf(_(" [--log-rotation-size=log-rotation-size]\n")); printf(_(" [--log-rotation-age=log-rotation-age]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name name of the instance\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name name of the instance\n")); printf(_(" -i, --backup-id=backup-id backup to merge\n")); printf(_(" -j, --threads=NUM number of parallel threads\n")); printf(_(" --progress show progress\n")); + printf(_(" --no-validate disable validation during retention merge\n")); + printf(_(" --no-sync do not sync merged files to disk\n")); printf(_("\n Logging options:\n")); printf(_(" --log-level-console=log-level-console\n")); @@ -612,6 +802,12 @@ help_merge(void) printf(_(" --log-level-file=log-level-file\n")); printf(_(" level for file logging (default: off)\n")); printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-console=log-format-console\n")); + printf(_(" defines the format of the console log (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); printf(_(" --log-filename=log-filename\n")); printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n")); @@ -624,17 +820,36 @@ help_merge(void) printf(_(" available units: 'kB', 'MB', 'GB', 'TB' (default: kB)\n")); printf(_(" --log-rotation-age=log-rotation-age\n")); printf(_(" rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n")); - printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n\n")); + printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n")); + printf(_(" --no-color disable the coloring of error and warning console messages\n\n")); +} + +static void +help_set_backup(void) +{ + printf(_("\n%s set-backup -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); + printf(_(" -i backup-id\n")); + printf(_(" [--ttl=interval] [--expire-time=time] [--note=text]\n\n")); + + printf(_(" --ttl=interval pin backup for specified amount of time; 0 unpin\n")); + printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: s)\n")); + printf(_(" (example: --ttl=20d)\n")); + printf(_(" --expire-time=time pin backup until specified time stamp\n")); + printf(_(" (example: --expire-time='2024-01-01 00:00:00+03')\n")); + printf(_(" --note=text add note to backup; 'none' to remove note\n")); + printf(_(" (example: --note='backup before app update to v13.1')\n")); } static void help_set_config(void) { - printf(_("\n%s set-config -B backup-path --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n%s set-config -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" [-D pgdata-path]\n")); printf(_(" [-E external-directories-paths]\n")); + printf(_(" [--restore-command=cmdline]\n")); printf(_(" [--log-level-console=log-level-console]\n")); printf(_(" [--log-level-file=log-level-file]\n")); + printf(_(" [--log-format-file=log-format-file]\n")); printf(_(" [--log-filename=log-filename]\n")); printf(_(" [--error-log-filename=error-log-filename]\n")); printf(_(" [--log-directory=log-directory]\n")); @@ -642,6 +857,7 @@ help_set_config(void) printf(_(" [--log-rotation-age=log-rotation-age]\n")); printf(_(" [--retention-redundancy=retention-redundancy]\n")); printf(_(" [--retention-window=retention-window]\n")); + printf(_(" [--wal-depth=wal-depth]\n")); printf(_(" [--compress-algorithm=compress-algorithm]\n")); printf(_(" [--compress-level=compress-level]\n")); printf(_(" [--archive-timeout=timeout]\n")); @@ -650,12 +866,13 @@ help_set_config(void) printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); printf(_(" [--ssh-options]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name name of the instance\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name name of the instance\n")); printf(_(" -D, --pgdata=pgdata-path location of the database storage area\n")); printf(_(" -E --external-dirs=external-directories-paths\n")); printf(_(" backup some directories not from pgdata \n")); printf(_(" (example: --external-dirs=/tmp/dir1:/tmp/dir2)\n")); + printf(_(" --restore-command=cmdline command to use as 'restore_command' in recovery.conf; 'none' disables\n")); printf(_("\n Logging options:\n")); printf(_(" --log-level-console=log-level-console\n")); @@ -664,6 +881,9 @@ help_set_config(void) printf(_(" --log-level-file=log-level-file\n")); printf(_(" level for file logging (default: off)\n")); printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); printf(_(" --log-filename=log-filename\n")); printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n")); @@ -683,6 +903,8 @@ help_set_config(void) printf(_(" number of full backups to keep; 0 disables; (default: 0)\n")); printf(_(" --retention-window=retention-window\n")); printf(_(" number of days of recoverability; 0 disables; (default: 0)\n")); + printf(_(" --wal-depth=wal-depth number of latest valid backups with ability to perform\n")); + printf(_(" the point in time recovery; disables; (default: 0)\n")); printf(_("\n Compression options:\n")); printf(_(" --compress alias for --compress-algorithm='zlib' and --compress-level=1\n")); @@ -692,18 +914,18 @@ help_set_config(void) printf(_(" level of compression [0-9] (default: 1)\n")); printf(_("\n Archive options:\n")); - printf(_(" --archive-timeout=timeout wait timeout for WAL segment archiving (default: 5min)\n")); + printf(_(" --archive-timeout=timeout wait timeout for WAL segment archiving (default: 5min)\n")); printf(_("\n Connection options:\n")); - printf(_(" -U, --username=USERNAME user name to connect as (default: current local user)\n")); - printf(_(" -d, --dbname=DBNAME database to connect (default: username)\n")); - printf(_(" -h, --host=HOSTNAME database server host or socket directory(default: 'local socket')\n")); - printf(_(" -p, --port=PORT database server port (default: 5432)\n")); + printf(_(" -U, --pguser=USERNAME user name to connect as (default: current local user)\n")); + printf(_(" -d, --pgdatabase=DBNAME database to connect (default: username)\n")); + printf(_(" -h, --pghost=HOSTNAME database server host or socket directory(default: 'local socket')\n")); + printf(_(" -p, --pgport=PORT database server port (default: 5432)\n")); printf(_("\n Remote options:\n")); printf(_(" --remote-proto=protocol remote protocol to use\n")); printf(_(" available options: 'ssh', 'none' (default: ssh)\n")); - printf(_(" --remote-host=hostname remote host address or hostname\n")); + printf(_(" --remote-host=destination remote host address or hostname\n")); printf(_(" --remote-port=port remote host port (default: 22)\n")); printf(_(" --remote-path=path path to directory with pg_probackup binary on remote host\n")); printf(_(" (default: current binary path)\n")); @@ -711,6 +933,11 @@ help_set_config(void) printf(_(" --ssh-options=ssh_options additional ssh options (default: none)\n")); printf(_(" (example: --ssh-options='-c cipher_spec -F configfile')\n")); + printf(_("\n Remote WAL archive options:\n")); + printf(_(" --archive-host=destination address or hostname for ssh connection to archive host\n")); + printf(_(" --archive-port=port port for ssh connection to archive host (default: 22)\n")); + printf(_(" --archive-user=username user name for ssh connection to archive host (default: PostgreSQL user)\n")); + printf(_("\n Replica options:\n")); printf(_(" --master-user=user_name user name to connect to master (deprecated)\n")); printf(_(" --master-db=db_name database to connect to master (deprecated)\n")); @@ -722,27 +949,28 @@ help_set_config(void) static void help_show_config(void) { - printf(_("\n%s show-config -B backup-path --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n%s show-config -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" [--format=format]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name name of the instance\n")); - printf(_(" --format=format show format=PLAIN|JSON\n\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name name of the instance\n")); + printf(_(" --format=format show format=PLAIN|JSON\n")); + printf(_(" --no-scale-units show memory and time values in default units\n\n")); } static void help_add_instance(void) { - printf(_("\n%s add-instance -B backup-path -D pgdata-path\n"), PROGRAM_NAME); - printf(_(" --instance=instance_name\n")); + printf(_("\n%s add-instance -B backup-dir -D pgdata-path\n"), PROGRAM_NAME); + printf(_(" --instance=instance-name\n")); printf(_(" [-E external-directory-path]\n")); printf(_(" [--remote-proto] [--remote-host]\n")); printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); printf(_(" [--ssh-options]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); printf(_(" -D, --pgdata=pgdata-path location of the database storage area\n")); - printf(_(" --instance=instance_name name of the new instance\n")); + printf(_(" --instance=instance-name name of the new instance\n")); printf(_(" -E --external-dirs=external-directories-paths\n")); printf(_(" backup some directories not from pgdata \n")); @@ -750,44 +978,75 @@ help_add_instance(void) printf(_("\n Remote options:\n")); printf(_(" --remote-proto=protocol remote protocol to use\n")); printf(_(" available options: 'ssh', 'none' (default: ssh)\n")); - printf(_(" --remote-host=hostname remote host address or hostname\n")); + printf(_(" --remote-host=destination remote host address or hostname\n")); printf(_(" --remote-port=port remote host port (default: 22)\n")); printf(_(" --remote-path=path path to directory with pg_probackup binary on remote host\n")); printf(_(" (default: current binary path)\n")); printf(_(" --remote-user=username user name for ssh connection (default: current user)\n")); printf(_(" --ssh-options=ssh_options additional ssh options (default: none)\n")); printf(_(" (example: --ssh-options='-c cipher_spec -F configfile')\n\n")); + + printf(_("\n Logging options:\n")); + printf(_(" --log-level-console=log-level-console\n")); + printf(_(" level for console logging (default: info)\n")); + printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-level-file=log-level-file\n")); + printf(_(" level for file logging (default: off)\n")); + printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); + printf(_(" --log-filename=log-filename\n")); + printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); + printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n")); + printf(_(" --error-log-filename=error-log-filename\n")); + printf(_(" filename for error logging (default: none)\n")); + printf(_(" --log-directory=log-directory\n")); + printf(_(" directory for file logging (default: BACKUP_PATH/log)\n")); + printf(_(" --log-rotation-size=log-rotation-size\n")); + printf(_(" rotate logfile if its size exceeds this value; 0 disables; (default: 0)\n")); + printf(_(" available units: 'kB', 'MB', 'GB', 'TB' (default: kB)\n")); + printf(_(" --log-rotation-age=log-rotation-age\n")); + printf(_(" rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n")); + printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n")); } static void help_del_instance(void) { - printf(_("\n%s del-instance -B backup-path --instance=instance_name\n"), PROGRAM_NAME); + printf(_("\n%s del-instance -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name name of the instance to delete\n\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name name of the instance to delete\n\n")); } static void help_archive_push(void) { - printf(_("\n%s archive-push -B backup-path --instance=instance_name\n"), PROGRAM_NAME); - printf(_(" --wal-file-path=wal-file-path\n")); + printf(_("\n%s archive-push -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" --wal-file-name=wal-file-name\n")); - printf(_(" [--overwrite]\n")); - printf(_(" [--compress]\n")); + printf(_(" [--wal-file-path=wal-file-path]\n")); + printf(_(" [-j num-threads] [--batch-size=batch_size]\n")); + printf(_(" [--archive-timeout=timeout]\n")); + printf(_(" [--no-ready-rename] [--no-sync]\n")); + printf(_(" [--overwrite] [--compress]\n")); printf(_(" [--compress-algorithm=compress-algorithm]\n")); printf(_(" [--compress-level=compress-level]\n")); printf(_(" [--remote-proto] [--remote-host]\n")); printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); printf(_(" [--ssh-options]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name name of the instance to delete\n")); - printf(_(" --wal-file-path=wal-file-path\n")); - printf(_(" relative path name of the WAL file on the server\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name name of the instance to delete\n")); printf(_(" --wal-file-name=wal-file-name\n")); - printf(_(" name of the WAL file to retrieve from the server\n")); + printf(_(" name of the file to copy into WAL archive\n")); + printf(_(" --wal-file-path=wal-file-path\n")); + printf(_(" relative destination path of the WAL archive\n")); + printf(_(" -j, --threads=NUM number of parallel threads\n")); + printf(_(" --batch-size=NUM number of files to be copied\n")); + printf(_(" --archive-timeout=timeout wait timeout before discarding stale temp file(default: 5min)\n")); + printf(_(" --no-ready-rename do not rename '.ready' files in 'archive_status' directory\n")); + printf(_(" --no-sync do not sync WAL file to disk\n")); printf(_(" --overwrite overwrite archived WAL file\n")); printf(_("\n Compression options:\n")); @@ -797,6 +1056,30 @@ help_archive_push(void) printf(_(" --compress-level=compress-level\n")); printf(_(" level of compression [0-9] (default: 1)\n")); + printf(_("\n Logging options:\n")); + printf(_(" --log-level-console=log-level-console\n")); + printf(_(" level for console logging (default: info)\n")); + printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-level-file=log-level-file\n")); + printf(_(" level for file logging (default: off)\n")); + printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); + printf(_(" --log-filename=log-filename\n")); + printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); + printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n")); + printf(_(" --error-log-filename=error-log-filename\n")); + printf(_(" filename for error logging (default: none)\n")); + printf(_(" --log-directory=log-directory\n")); + printf(_(" directory for file logging (default: BACKUP_PATH/log)\n")); + printf(_(" --log-rotation-size=log-rotation-size\n")); + printf(_(" rotate logfile if its size exceeds this value; 0 disables; (default: 0)\n")); + printf(_(" available units: 'kB', 'MB', 'GB', 'TB' (default: kB)\n")); + printf(_(" --log-rotation-age=log-rotation-age\n")); + printf(_(" rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n")); + printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n")); + printf(_("\n Remote options:\n")); printf(_(" --remote-proto=protocol remote protocol to use\n")); printf(_(" available options: 'ssh', 'none' (default: ssh)\n")); @@ -812,19 +1095,139 @@ help_archive_push(void) static void help_archive_get(void) { - printf(_("\n%s archive-get -B backup-path --instance=instance_name\n"), PROGRAM_NAME); - printf(_(" --wal-file-path=wal-file-path\n")); + printf(_("\n%s archive-get -B backup-dir --instance=instance-name\n"), PROGRAM_NAME); printf(_(" --wal-file-name=wal-file-name\n")); + printf(_(" [--wal-file-path=wal-file-path]\n")); + printf(_(" [-j num-threads] [--batch-size=batch_size]\n")); + printf(_(" [--no-validate-wal]\n")); printf(_(" [--remote-proto] [--remote-host]\n")); printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); printf(_(" [--ssh-options]\n\n")); - printf(_(" -B, --backup-path=backup-path location of the backup storage area\n")); - printf(_(" --instance=instance_name name of the instance to delete\n")); + printf(_(" -B, --backup-path=backup-dir location of the backup storage area\n")); + printf(_(" --instance=instance-name name of the instance to delete\n")); printf(_(" --wal-file-path=wal-file-path\n")); printf(_(" relative destination path name of the WAL file on the server\n")); printf(_(" --wal-file-name=wal-file-name\n")); printf(_(" name of the WAL file to retrieve from the archive\n")); + printf(_(" -j, --threads=NUM number of parallel threads\n")); + printf(_(" --batch-size=NUM number of files to be prefetched\n")); + printf(_(" --prefetch-dir=path location of the store area for prefetched WAL files\n")); + printf(_(" --no-validate-wal skip validation of prefetched WAL file before using it\n")); + + printf(_("\n Logging options:\n")); + printf(_(" --log-level-console=log-level-console\n")); + printf(_(" level for console logging (default: info)\n")); + printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-level-file=log-level-file\n")); + printf(_(" level for file logging (default: off)\n")); + printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); + printf(_(" --log-filename=log-filename\n")); + printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); + printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n")); + printf(_(" --error-log-filename=error-log-filename\n")); + printf(_(" filename for error logging (default: none)\n")); + printf(_(" --log-directory=log-directory\n")); + printf(_(" directory for file logging (default: BACKUP_PATH/log)\n")); + printf(_(" --log-rotation-size=log-rotation-size\n")); + printf(_(" rotate logfile if its size exceeds this value; 0 disables; (default: 0)\n")); + printf(_(" available units: 'kB', 'MB', 'GB', 'TB' (default: kB)\n")); + printf(_(" --log-rotation-age=log-rotation-age\n")); + printf(_(" rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n")); + printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n")); + + printf(_("\n Remote options:\n")); + printf(_(" --remote-proto=protocol remote protocol to use\n")); + printf(_(" available options: 'ssh', 'none' (default: ssh)\n")); + printf(_(" --remote-host=hostname remote host address or hostname\n")); + printf(_(" --remote-port=port remote host port (default: 22)\n")); + printf(_(" --remote-path=path path to directory with pg_probackup binary on remote host\n")); + printf(_(" (default: current binary path)\n")); + printf(_(" --remote-user=username user name for ssh connection (default: current user)\n")); + printf(_(" --ssh-options=ssh_options additional ssh options (default: none)\n")); + printf(_(" (example: --ssh-options='-c cipher_spec -F configfile')\n\n")); +} + +static void +help_help(void) +{ + printf(_("\n%s help [command]\n"), PROGRAM_NAME); + printf(_("%s command --help\n\n"), PROGRAM_NAME); +} + +static void +help_version(void) +{ + printf(_("\n%s version\n"), PROGRAM_NAME); + printf(_("%s --version\n\n"), PROGRAM_NAME); +} + +static void +help_catchup(void) +{ + printf(_("\n%s catchup -b catchup-mode\n"), PROGRAM_NAME); + printf(_(" --source-pgdata=path_to_pgdata_on_remote_server\n")); + printf(_(" --destination-pgdata=path_to_local_dir\n")); + printf(_(" [--stream [-S slot-name]] [--temp-slot | --perm-slot]\n")); + printf(_(" [-j num-threads]\n")); + printf(_(" [-T OLDDIR=NEWDIR]\n")); + printf(_(" [--exclude-path=path_prefix]\n")); + printf(_(" [-d dbname] [-h host] [-p port] [-U username]\n")); + printf(_(" [-w --no-password] [-W --password]\n")); + printf(_(" [--remote-proto] [--remote-host]\n")); + printf(_(" [--remote-port] [--remote-path] [--remote-user]\n")); + printf(_(" [--ssh-options]\n")); + printf(_(" [--dry-run]\n")); + printf(_(" [--help]\n\n")); + + printf(_(" -b, --backup-mode=catchup-mode catchup mode=FULL|DELTA|PTRACK\n")); + printf(_(" --stream stream the transaction log (only supported mode)\n")); + printf(_(" -S, --slot=SLOTNAME replication slot to use\n")); + printf(_(" --temp-slot use temporary replication slot\n")); + printf(_(" -P --perm-slot create permanent replication slot\n")); + + printf(_(" -j, --threads=NUM number of parallel threads\n")); + + printf(_(" -T, --tablespace-mapping=OLDDIR=NEWDIR\n")); + printf(_(" relocate the tablespace from directory OLDDIR to NEWDIR\n")); + printf(_(" -x, --exclude-path=path_prefix files with path_prefix (relative to pgdata) will be\n")); + printf(_(" excluded from catchup (can be used multiple times)\n")); + printf(_(" Dangerous option! Use at your own risk!\n")); + + printf(_("\n Connection options:\n")); + printf(_(" -U, --pguser=USERNAME user name to connect as (default: current local user)\n")); + printf(_(" -d, --pgdatabase=DBNAME database to connect (default: username)\n")); + printf(_(" -h, --pghost=HOSTNAME database server host or socket directory(default: 'local socket')\n")); + printf(_(" -p, --pgport=PORT database server port (default: 5432)\n")); + printf(_(" -w, --no-password never prompt for password\n")); + printf(_(" -W, --password force password prompt\n\n")); + + printf(_("\n Logging options:\n")); + printf(_(" --log-level-console=log-level-console\n")); + printf(_(" level for console logging (default: info)\n")); + printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-level-file=log-level-file\n")); + printf(_(" level for file logging (default: off)\n")); + printf(_(" available options: 'off', 'error', 'warning', 'info', 'log', 'verbose'\n")); + printf(_(" --log-format-file=log-format-file\n")); + printf(_(" defines the format of log files (default: plain)\n")); + printf(_(" available options: 'plain', 'json'\n")); + printf(_(" --log-filename=log-filename\n")); + printf(_(" filename for file logging (default: 'pg_probackup.log')\n")); + printf(_(" support strftime format (example: pg_probackup-%%Y-%%m-%%d_%%H%%M%%S.log)\n")); + printf(_(" --error-log-filename=error-log-filename\n")); + printf(_(" filename for error logging (default: none)\n")); + printf(_(" --log-directory=log-directory\n")); + printf(_(" directory for file logging (default: BACKUP_PATH/log)\n")); + printf(_(" --log-rotation-size=log-rotation-size\n")); + printf(_(" rotate logfile if its size exceeds this value; 0 disables; (default: 0)\n")); + printf(_(" available units: 'kB', 'MB', 'GB', 'TB' (default: kB)\n")); + printf(_(" --log-rotation-age=log-rotation-age\n")); + printf(_(" rotate logfile if its age exceeds this value; 0 disables; (default: 0)\n")); + printf(_(" available units: 'ms', 's', 'min', 'h', 'd' (default: min)\n")); printf(_("\n Remote options:\n")); printf(_(" --remote-proto=protocol remote protocol to use\n")); @@ -836,4 +1239,6 @@ help_archive_get(void) printf(_(" --remote-user=username user name for ssh connection (default: current user)\n")); printf(_(" --ssh-options=ssh_options additional ssh options (default: none)\n")); printf(_(" (example: --ssh-options='-c cipher_spec -F configfile')\n\n")); + + printf(_(" --dry-run perform a trial run without any changes\n\n")); } diff --git a/src/init.c b/src/init.c index 4e52292aa..837e2bad0 100644 --- a/src/init.c +++ b/src/init.c @@ -17,79 +17,78 @@ * Initialize backup catalog. */ int -do_init(void) +do_init(CatalogState *catalogState) { - char path[MAXPGPATH]; - char arclog_path_dir[MAXPGPATH]; int results; - results = pg_check_dir(backup_path); + results = pg_check_dir(catalogState->catalog_path); + if (results == 4) /* exists and not empty*/ - elog(ERROR, "backup catalog already exist and it's not empty"); + elog(ERROR, "The backup catalog already exists and is not empty"); else if (results == -1) /*trouble accessing directory*/ { int errno_tmp = errno; - elog(ERROR, "cannot open backup catalog directory \"%s\": %s", - backup_path, strerror(errno_tmp)); + elog(ERROR, "Cannot open backup catalog directory \"%s\": %s", + catalogState->catalog_path, strerror(errno_tmp)); } /* create backup catalog root directory */ - dir_create_dir(backup_path, DIR_PERMISSION); + dir_create_dir(catalogState->catalog_path, DIR_PERMISSION, false); /* create backup catalog data directory */ - join_path_components(path, backup_path, BACKUPS_DIR); - dir_create_dir(path, DIR_PERMISSION); + dir_create_dir(catalogState->backup_subdir_path, DIR_PERMISSION, false); /* create backup catalog wal directory */ - join_path_components(arclog_path_dir, backup_path, "wal"); - dir_create_dir(arclog_path_dir, DIR_PERMISSION); + dir_create_dir(catalogState->wal_subdir_path, DIR_PERMISSION, false); - elog(INFO, "Backup catalog '%s' successfully inited", backup_path); + elog(INFO, "Backup catalog '%s' successfully initialized", catalogState->catalog_path); return 0; } int -do_add_instance(void) +do_add_instance(InstanceState *instanceState, InstanceConfig *instance) { - char path[MAXPGPATH]; - char arclog_path_dir[MAXPGPATH]; struct stat st; + CatalogState *catalogState = instanceState->catalog_state; /* PGDATA is always required */ - if (instance_config.pgdata == NULL) - elog(ERROR, "Required parameter not specified: PGDATA " - "(-D, --pgdata)"); + if (instance->pgdata == NULL) + elog(ERROR, "No postgres data directory specified.\n" + "Please specify it either using environment variable PGDATA or\n" + "command line option --pgdata (-D)"); /* Read system_identifier from PGDATA */ - instance_config.system_identifier = get_system_identifier(instance_config.pgdata); + instance->system_identifier = get_system_identifier(instance->pgdata, FIO_DB_HOST, false); /* Starting from PostgreSQL 11 read WAL segment size from PGDATA */ - instance_config.xlog_seg_size = get_xlog_seg_size(instance_config.pgdata); + instance->xlog_seg_size = get_xlog_seg_size(instance->pgdata); /* Ensure that all root directories already exist */ - if (access(backup_path, F_OK) != 0) - elog(ERROR, "%s directory does not exist.", backup_path); + /* TODO maybe call do_init() here instead of error?*/ + if (access(catalogState->catalog_path, F_OK) != 0) + elog(ERROR, "Directory does not exist: '%s'", catalogState->catalog_path); - join_path_components(path, backup_path, BACKUPS_DIR); - if (access(path, F_OK) != 0) - elog(ERROR, "%s directory does not exist.", path); + if (access(catalogState->backup_subdir_path, F_OK) != 0) + elog(ERROR, "Directory does not exist: '%s'", catalogState->backup_subdir_path); - join_path_components(arclog_path_dir, backup_path, "wal"); - if (access(arclog_path_dir, F_OK) != 0) - elog(ERROR, "%s directory does not exist.", arclog_path_dir); + if (access(catalogState->wal_subdir_path, F_OK) != 0) + elog(ERROR, "Directory does not exist: '%s'", catalogState->wal_subdir_path); - /* Create directory for data files of this specific instance */ - if (stat(backup_instance_path, &st) == 0 && S_ISDIR(st.st_mode)) - elog(ERROR, "instance '%s' already exists", backup_instance_path); - dir_create_dir(backup_instance_path, DIR_PERMISSION); + if (stat(instanceState->instance_backup_subdir_path, &st) == 0 && S_ISDIR(st.st_mode)) + elog(ERROR, "Instance '%s' backup directory already exists: '%s'", + instanceState->instance_name, instanceState->instance_backup_subdir_path); /* * Create directory for wal files of this specific instance. * Existence check is extra paranoid because if we don't have such a * directory in data dir, we shouldn't have it in wal as well. */ - if (stat(arclog_path, &st) == 0 && S_ISDIR(st.st_mode)) - elog(ERROR, "arclog_path '%s' already exists", arclog_path); - dir_create_dir(arclog_path, DIR_PERMISSION); + if (stat(instanceState->instance_wal_subdir_path, &st) == 0 && S_ISDIR(st.st_mode)) + elog(ERROR, "Instance '%s' WAL archive directory already exists: '%s'", + instanceState->instance_name, instanceState->instance_wal_subdir_path); + + /* Create directory for data files of this specific instance */ + dir_create_dir(instanceState->instance_backup_subdir_path, DIR_PERMISSION, false); + dir_create_dir(instanceState->instance_wal_subdir_path, DIR_PERMISSION, false); /* * Write initial configuration file. @@ -99,13 +98,30 @@ do_add_instance(void) * We need to manually set options source to save them to the configuration * file. */ - config_set_opt(instance_options, &instance_config.system_identifier, + config_set_opt(instance_options, &instance->system_identifier, SOURCE_FILE); - config_set_opt(instance_options, &instance_config.xlog_seg_size, + config_set_opt(instance_options, &instance->xlog_seg_size, SOURCE_FILE); + + /* Kludge: do not save remote options into config */ + config_set_opt(instance_options, &instance_config.remote.host, + SOURCE_DEFAULT); + config_set_opt(instance_options, &instance_config.remote.proto, + SOURCE_DEFAULT); + config_set_opt(instance_options, &instance_config.remote.port, + SOURCE_DEFAULT); + config_set_opt(instance_options, &instance_config.remote.path, + SOURCE_DEFAULT); + config_set_opt(instance_options, &instance_config.remote.user, + SOURCE_DEFAULT); + config_set_opt(instance_options, &instance_config.remote.ssh_options, + SOURCE_DEFAULT); + config_set_opt(instance_options, &instance_config.remote.ssh_config, + SOURCE_DEFAULT); + /* pgdata was set through command line */ - do_set_config(true); + do_set_config(instanceState, true); - elog(INFO, "Instance '%s' successfully inited", instance_name); + elog(INFO, "Instance '%s' successfully initialized", instanceState->instance_name); return 0; } diff --git a/src/merge.c b/src/merge.c index 832bd90cb..e8f926795 100644 --- a/src/merge.c +++ b/src/merge.c @@ -2,7 +2,7 @@ * * merge.c: merge FULL and incremental backups * - * Copyright (c) 2018-2019, Postgres Professional + * Copyright (c) 2018-2022, Postgres Professional * *------------------------------------------------------------------------- */ @@ -16,16 +16,21 @@ typedef struct { - parray *to_files; - parray *files; - parray *from_external; + parray *merge_filelist; + parray *parent_chain; - pgBackup *to_backup; - pgBackup *from_backup; - const char *to_root; - const char *from_root; - const char *to_external_prefix; - const char *from_external_prefix; + pgBackup *dest_backup; + pgBackup *full_backup; + + const char *full_database_dir; + const char *full_external_prefix; + +// size_t in_place_merge_bytes; + bool compression_match; + bool program_version_match; + bool use_bitmap; + bool is_retry; + bool no_sync; /* * Return value from the thread. @@ -34,6 +39,7 @@ typedef struct int ret; } merge_files_arg; + static void *merge_files(void *arg); static void reorder_external_dirs(pgBackup *to_backup, parray *to_external, @@ -41,6 +47,20 @@ reorder_external_dirs(pgBackup *to_backup, parray *to_external, static int get_external_index(const char *key, const parray *list); +static void +merge_data_file(parray *parent_chain, pgBackup *full_backup, + pgBackup *dest_backup, pgFile *dest_file, + pgFile *tmp_file, const char *to_root, bool use_bitmap, + bool is_retry, bool no_sync); + +static void +merge_non_data_file(parray *parent_chain, pgBackup *full_backup, + pgBackup *dest_backup, pgFile *dest_file, + pgFile *tmp_file, const char *full_database_dir, + const char *full_external_prefix, bool no_sync); + +static bool is_forward_compatible(parray *parent_chain); + /* * Implementation of MERGE command. * @@ -49,24 +69,25 @@ get_external_index(const char *key, const parray *list); * - Remove unnecessary files, which doesn't exist in the target backup anymore */ void -do_merge(time_t backup_id) +do_merge(InstanceState *instanceState, time_t backup_id, bool no_validate, bool no_sync) { parray *backups; parray *merge_list = parray_new(); pgBackup *dest_backup = NULL; + pgBackup *dest_backup_tmp = NULL; pgBackup *full_backup = NULL; int i; if (backup_id == INVALID_BACKUP_ID) - elog(ERROR, "required parameter is not specified: --backup-id"); + elog(ERROR, "Required parameter is not specified: --backup-id"); - if (instance_name == NULL) - elog(ERROR, "required parameter is not specified: --instance"); + if (instanceState == NULL) + elog(ERROR, "Required parameter is not specified: --instance"); elog(INFO, "Merge started"); /* Get list of all backups sorted in order of descending start time */ - backups = catalog_get_backup_list(INVALID_BACKUP_ID); + backups = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); /* Find destination backup first */ for (i = 0; i < parray_num(backups); i++) @@ -81,72 +102,309 @@ do_merge(time_t backup_id) backup->status != BACKUP_STATUS_DONE && /* It is possible that previous merging was interrupted */ backup->status != BACKUP_STATUS_MERGING && + backup->status != BACKUP_STATUS_MERGED && backup->status != BACKUP_STATUS_DELETING) elog(ERROR, "Backup %s has status: %s", - base36enc(backup->start_time), status2str(backup->status)); - - if (backup->backup_mode == BACKUP_MODE_FULL) - elog(ERROR, "Backup %s is full backup", - base36enc(backup->start_time)); + backup_id_of(backup), status2str(backup->status)); dest_backup = backup; break; } } - /* sanity */ + /* + * Handle the case of crash right after deletion of the target + * incremental backup. We still can recover from this. + * Iterate over backups and look for the FULL backup with + * MERGED status, that has merge-target-id eqial to backup_id. + */ + if (dest_backup == NULL) + { + for (i = 0; i < parray_num(backups); i++) + { + pgBackup *backup = (pgBackup *) parray_get(backups, i); + + if (backup->status == BACKUP_STATUS_MERGED && + backup->merge_dest_backup == backup_id) + { + dest_backup = backup; + break; + } + } + } + if (dest_backup == NULL) elog(ERROR, "Target backup %s was not found", base36enc(backup_id)); - /* get full backup */ - full_backup = find_parent_full_backup(dest_backup); + /* It is possible to use FULL backup as target backup for merge. + * There are two possible cases: + * 1. The user want to merge FULL backup with closest incremental backup. + * In this case we must find suitable destination backup and merge them. + * + * 2. Previous merge has failed after destination backup was deleted, + * but before FULL backup was renamed: + * Example A: + * PAGE2_1 OK + * FULL2 OK + * PAGE1_1 MISSING/DELETING <- + * FULL1 MERGED/MERGING + */ + if (dest_backup->backup_mode == BACKUP_MODE_FULL) + { + full_backup = dest_backup; + dest_backup = NULL; + elog(INFO, "Merge target backup %s is full backup", + backup_id_of(full_backup)); + + /* sanity */ + if (full_backup->status == BACKUP_STATUS_DELETING) + elog(ERROR, "Backup %s has status: %s", + backup_id_of(full_backup), + status2str(full_backup->status)); + + /* Case #1 */ + if (full_backup->status == BACKUP_STATUS_OK || + full_backup->status == BACKUP_STATUS_DONE) + { + /* Check the case of FULL backup having more than one direct children */ + if (is_prolific(backups, full_backup)) + elog(ERROR, "Merge target is full backup and has multiple direct children, " + "you must specify child backup id you want to merge with"); + + elog(INFO, "Looking for closest incremental backup to merge with"); + + /* Look for closest child backup */ + for (i = 0; i < parray_num(backups); i++) + { + pgBackup *backup = (pgBackup *) parray_get(backups, i); + + /* skip unsuitable candidates */ + if (backup->status != BACKUP_STATUS_OK && + backup->status != BACKUP_STATUS_DONE) + continue; + + if (backup->parent_backup == full_backup->start_time) + { + dest_backup = backup; + break; + } + } + + /* sanity */ + if (dest_backup == NULL) + elog(ERROR, "Failed to find merge candidate, " + "backup %s has no valid children", + backup_id_of(full_backup)); + + } + /* Case #2 */ + else if (full_backup->status == BACKUP_STATUS_MERGING) + { + /* + * MERGING - merge was ongoing at the moment of crash. + * We must find destination backup and rerun merge. + * If destination backup is missing, then merge must be aborted, + * there is no recovery from this situation. + */ + + if (full_backup->merge_dest_backup == INVALID_BACKUP_ID) + elog(ERROR, "Failed to determine merge destination backup"); + + /* look up destination backup */ + for (i = 0; i < parray_num(backups); i++) + { + pgBackup *backup = (pgBackup *) parray_get(backups, i); + + if (backup->start_time == full_backup->merge_dest_backup) + { + dest_backup = backup; + break; + } + } + if (!dest_backup) + { + elog(ERROR, "Full backup %s has unfinished merge with missing backup %s", + backup_id_of(full_backup), + base36enc(full_backup->merge_dest_backup)); + } + } + else if (full_backup->status == BACKUP_STATUS_MERGED) + { + /* + * MERGED - merge crashed after files were transfered, but + * before rename could take place. + * If destination backup is missing, this is ok. + * If destination backup is present, then it should be deleted. + * After that FULL backup must acquire destination backup ID. + */ + + /* destination backup may or may not exists */ + for (i = 0; i < parray_num(backups); i++) + { + pgBackup *backup = (pgBackup *) parray_get(backups, i); + + if (backup->start_time == full_backup->merge_dest_backup) + { + dest_backup = backup; + break; + } + } + if (!dest_backup) + { + elog(WARNING, "Full backup %s has unfinished merge with missing backup %s", + backup_id_of(full_backup), + base36enc(full_backup->merge_dest_backup)); + } + } + else + elog(ERROR, "Backup %s has status: %s", + backup_id_of(full_backup), + status2str(full_backup->status)); + } + else + { + /* + * Legal Case #1: + * PAGE2 OK <- target + * PAGE1 OK + * FULL OK + * Legal Case #2: + * PAGE2 MERGING <- target + * PAGE1 MERGING + * FULL MERGING + * Legal Case #3: + * PAGE2 MERGING <- target + * PAGE1 DELETING + * FULL MERGED + * Legal Case #4: + * PAGE2 MERGING <- target + * PAGE1 missing + * FULL MERGED + * Legal Case #5: + * PAGE2 DELETING <- target + * FULL MERGED + * Legal Case #6: + * PAGE2 MERGING <- target + * PAGE1 missing + * FULL MERGED + * Illegal Case #7: + * PAGE2 MERGING <- target + * PAGE1 missing + * FULL MERGING + */ + + if (dest_backup->status == BACKUP_STATUS_MERGING || + dest_backup->status == BACKUP_STATUS_DELETING) + elog(WARNING, "Rerun unfinished merge for backup %s", + backup_id_of(dest_backup)); + + /* First we should try to find parent FULL backup */ + full_backup = find_parent_full_backup(dest_backup); + + /* Chain is broken, one or more member of parent chain is missing */ + if (full_backup == NULL) + { + /* It is the legal state of affairs in Case #4, but + * only for MERGING incremental target backup and only + * if FULL backup has MERGED status. + */ + if (dest_backup->status != BACKUP_STATUS_MERGING) + elog(ERROR, "Failed to find parent full backup for %s", + backup_id_of(dest_backup)); + + /* Find FULL backup that has unfinished merge with dest backup */ + for (i = 0; i < parray_num(backups); i++) + { + pgBackup *backup = (pgBackup *) parray_get(backups, i); + + if (backup->merge_dest_backup == dest_backup->start_time) + { + full_backup = backup; + break; + } + } + + if (!full_backup) + elog(ERROR, "Failed to find full backup that has unfinished merge" + "with backup %s, cannot rerun merge", + backup_id_of(dest_backup)); + + if (full_backup->status == BACKUP_STATUS_MERGED) + elog(WARNING, "Incremental chain is broken, try to recover unfinished merge"); + else + elog(ERROR, "Incremental chain is broken, merge is impossible to finish"); + } + else + { + if ((full_backup->status == BACKUP_STATUS_MERGED || + full_backup->status == BACKUP_STATUS_MERGING) && + dest_backup->start_time != full_backup->merge_dest_backup) + { + elog(ERROR, "Full backup %s has unfinished merge with backup %s", + backup_id_of(full_backup), + base36enc(full_backup->merge_dest_backup)); + } + + } + } /* sanity */ if (full_backup == NULL) elog(ERROR, "Parent full backup for the given backup %s was not found", base36enc(backup_id)); + /* At this point NULL as dest_backup is allowed only in case of full backup + * having status MERGED */ + if (dest_backup == NULL && full_backup->status != BACKUP_STATUS_MERGED) + elog(ERROR, "Cannot run merge for full backup %s", + backup_id_of(full_backup)); + /* sanity */ if (full_backup->status != BACKUP_STATUS_OK && full_backup->status != BACKUP_STATUS_DONE && /* It is possible that previous merging was interrupted */ + full_backup->status != BACKUP_STATUS_MERGED && full_backup->status != BACKUP_STATUS_MERGING) elog(ERROR, "Backup %s has status: %s", - base36enc(full_backup->start_time), status2str(full_backup->status)); + backup_id_of(full_backup), status2str(full_backup->status)); + + /* Form merge list */ + dest_backup_tmp = dest_backup; - /* form merge list */ - while(dest_backup->parent_backup_link) + /* While loop below may looks strange, it is done so on purpose + * to handle both whole and broken incremental chains. + */ + while (dest_backup_tmp) { /* sanity */ - if (dest_backup->status != BACKUP_STATUS_OK && - dest_backup->status != BACKUP_STATUS_DONE && + if (dest_backup_tmp->status != BACKUP_STATUS_OK && + dest_backup_tmp->status != BACKUP_STATUS_DONE && /* It is possible that previous merging was interrupted */ - dest_backup->status != BACKUP_STATUS_MERGING && - dest_backup->status != BACKUP_STATUS_DELETING) + dest_backup_tmp->status != BACKUP_STATUS_MERGING && + dest_backup_tmp->status != BACKUP_STATUS_MERGED && + dest_backup_tmp->status != BACKUP_STATUS_DELETING) elog(ERROR, "Backup %s has status: %s", - base36enc(dest_backup->start_time), status2str(dest_backup->status)); + backup_id_of(dest_backup_tmp), + status2str(dest_backup_tmp->status)); + + if (dest_backup_tmp->backup_mode == BACKUP_MODE_FULL) + break; - parray_append(merge_list, dest_backup); - dest_backup = dest_backup->parent_backup_link; + parray_append(merge_list, dest_backup_tmp); + dest_backup_tmp = dest_backup_tmp->parent_backup_link; } - /* Add FULL backup for easy locking */ + /* Add FULL backup */ parray_append(merge_list, full_backup); /* Lock merge chain */ - catalog_lock_backup_list(merge_list, parray_num(merge_list) - 1, 0); + catalog_lock_backup_list(merge_list, parray_num(merge_list) - 1, 0, true, true); - /* - * Found target and full backups, merge them and intermediate backups - */ - for (i = parray_num(merge_list) - 2; i >= 0; i--) - { - pgBackup *from_backup = (pgBackup *) parray_get(merge_list, i); + /* do actual merge */ + merge_chain(instanceState, merge_list, full_backup, dest_backup, no_validate, no_sync); - merge_backups(full_backup, from_backup); - } - - pgBackupValidate(full_backup); + if (!no_validate) + pgBackupValidate(full_backup, NULL); if (full_backup->status == BACKUP_STATUS_CORRUPT) elog(ERROR, "Merging of backup %s failed", base36enc(backup_id)); @@ -159,127 +417,213 @@ do_merge(time_t backup_id) } /* - * Merge two backups data files using threads. - * - to_backup - FULL, from_backup - incremental. - * - move instance files from from_backup to to_backup - * - remove unnecessary directories and files from to_backup - * - update metadata of from_backup, it becames FULL backup + * Merge backup chain. + * dest_backup - incremental backup. + * parent_chain - array of backups starting with dest_backup and + * ending with full_backup. + * + * Copy backup files from incremental backups from parent_chain into + * full backup directory. + * Remove unnecessary directories and files from full backup directory. + * Update metadata of full backup to represent destination backup. + * + * TODO: stop relying on caller to provide valid parent_chain, make sure + * that chain is ok. */ void -merge_backups(pgBackup *to_backup, pgBackup *from_backup) +merge_chain(InstanceState *instanceState, + parray *parent_chain, pgBackup *full_backup, pgBackup *dest_backup, + bool no_validate, bool no_sync) { - char *to_backup_id = base36enc_dup(to_backup->start_time), - *from_backup_id = base36enc_dup(from_backup->start_time); - char to_backup_path[MAXPGPATH], - to_database_path[MAXPGPATH], - to_external_prefix[MAXPGPATH], - from_backup_path[MAXPGPATH], - from_database_path[MAXPGPATH], - from_external_prefix[MAXPGPATH], - control_file[MAXPGPATH]; - parray *files, - *to_files; - parray *to_external = NULL, - *from_external = NULL; - pthread_t *threads = NULL; - merge_files_arg *threads_args = NULL; int i; + char full_external_prefix[MAXPGPATH]; + char full_database_dir[MAXPGPATH]; + parray *full_externals = NULL, + *dest_externals = NULL; + + parray *result_filelist = NULL; + bool use_bitmap = true; + bool is_retry = false; +// size_t total_in_place_merge_bytes = 0; + + pthread_t *threads = NULL; + merge_files_arg *threads_args = NULL; time_t merge_time; bool merge_isok = true; + /* for fancy reporting */ + time_t end_time; + char pretty_time[20]; + /* in-place merge flags */ + bool compression_match = false; + bool program_version_match = false; + /* It's redundant to check block checksumms during merge */ + skip_block_validation = true; + + /* Handle corner cases of missing destination backup */ + if (dest_backup == NULL && + full_backup->status == BACKUP_STATUS_MERGED) + goto merge_rename; + + if (!dest_backup) + elog(ERROR, "Destination backup is missing, cannot continue merge"); + + if (dest_backup->status == BACKUP_STATUS_MERGING || + full_backup->status == BACKUP_STATUS_MERGING || + full_backup->status == BACKUP_STATUS_MERGED) + { + is_retry = true; + elog(INFO, "Retry failed merge of backup %s with parent chain", backup_id_of(dest_backup)); + } + else + elog(INFO, "Merging backup %s with parent chain", backup_id_of(dest_backup)); - merge_time = time(NULL); - elog(INFO, "Merging backup %s with backup %s", from_backup_id, to_backup_id); + /* sanity */ + if (full_backup->merge_dest_backup != INVALID_BACKUP_ID && + full_backup->merge_dest_backup != dest_backup->start_time) + { + elog(ERROR, "Cannot run merge for %s, because full backup %s has " + "unfinished merge with backup %s", + backup_id_of(dest_backup), + backup_id_of(full_backup), + base36enc(full_backup->merge_dest_backup)); + } /* - * Validate to_backup only if it is BACKUP_STATUS_OK. If it has - * BACKUP_STATUS_MERGING status then it isn't valid backup until merging - * finished. + * Previous merging was interrupted during deleting source backup. It is + * safe just to delete it again. */ - if (to_backup->status == BACKUP_STATUS_OK || - to_backup->status == BACKUP_STATUS_DONE) + if (full_backup->status == BACKUP_STATUS_MERGED) + goto merge_delete; + + /* Forward compatibility is not supported */ + for (i = parray_num(parent_chain) - 1; i >= 0; i--) { - pgBackupValidate(to_backup); - if (to_backup->status == BACKUP_STATUS_CORRUPT) - elog(ERROR, "Interrupt merging"); + pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); + + if (parse_program_version(backup->program_version) > + parse_program_version(PROGRAM_VERSION)) + { + elog(ERROR, "Backup %s has been produced by pg_probackup version %s, " + "but current program version is %s. Forward compatibility " + "is not supported.", + backup_id_of(backup), + backup->program_version, + PROGRAM_VERSION); + } } - /* - * It is OK to validate from_backup if it has BACKUP_STATUS_OK or - * BACKUP_STATUS_MERGING status. + /* If destination backup compression algorithm differs from + * full backup compression algorithm, then in-place merge is + * not possible. */ - Assert(from_backup->status == BACKUP_STATUS_OK || - from_backup->status == BACKUP_STATUS_DONE || - from_backup->status == BACKUP_STATUS_MERGING || - from_backup->status == BACKUP_STATUS_DELETING); - pgBackupValidate(from_backup); - if (from_backup->status == BACKUP_STATUS_CORRUPT) - elog(ERROR, "Interrupt merging"); + if (full_backup->compress_alg == dest_backup->compress_alg) + compression_match = true; + else + elog(WARNING, "In-place merge is disabled because of compression " + "algorithms mismatch"); /* - * Make backup paths. + * If current program version differs from destination backup version, + * then in-place merge is not possible. */ - pgBackupGetPath(to_backup, to_backup_path, lengthof(to_backup_path), NULL); - pgBackupGetPath(to_backup, to_database_path, lengthof(to_database_path), - DATABASE_DIR); - pgBackupGetPath(to_backup, to_external_prefix, lengthof(to_database_path), - EXTERNAL_DIR); - pgBackupGetPath(from_backup, from_backup_path, lengthof(from_backup_path), NULL); - pgBackupGetPath(from_backup, from_database_path, lengthof(from_database_path), - DATABASE_DIR); - pgBackupGetPath(from_backup, from_external_prefix, lengthof(from_database_path), - EXTERNAL_DIR); + program_version_match = is_forward_compatible(parent_chain); - /* - * Get list of files which will be modified or removed. + /* Forbid merge retry for failed merges between 2.4.0 and any + * older version. Several format changes makes it impossible + * to determine the exact format any speific file is got. */ - pgBackupGetPath(to_backup, control_file, lengthof(control_file), - DATABASE_FILE_LIST); - to_files = dir_read_file_list(NULL, NULL, control_file, FIO_BACKUP_HOST); - /* To delete from leaf, sort in reversed order */ - parray_qsort(to_files, pgFileComparePathWithExternalDesc); + if (is_retry && + parse_program_version(dest_backup->program_version) >= 20400 && + parse_program_version(full_backup->program_version) < 20400) + { + elog(ERROR, "Retry of failed merge for backups with different between minor " + "versions is forbidden to avoid data corruption because of storage format " + "changes introduced in 2.4.0 version, please take a new full backup"); + } + /* - * Get list of files which need to be moved. + * Validate or revalidate all members of parent chain + * with sole exception of FULL backup. If it has MERGING status + * then it isn't valid backup until merging is finished. */ - pgBackupGetPath(from_backup, control_file, lengthof(control_file), - DATABASE_FILE_LIST); - files = dir_read_file_list(NULL, NULL, control_file, FIO_BACKUP_HOST); - /* sort by size for load balancing */ - parray_qsort(files, pgFileCompareSize); + if (!no_validate) + { + elog(INFO, "Validate parent chain for backup %s", + backup_id_of(dest_backup)); + + for (i = parray_num(parent_chain) - 1; i >= 0; i--) + { + pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); + + /* FULL backup is not to be validated if its status is MERGING */ + if (backup->backup_mode == BACKUP_MODE_FULL && + backup->status == BACKUP_STATUS_MERGING) + { + continue; + } + + pgBackupValidate(backup, NULL); + + if (backup->status != BACKUP_STATUS_OK) + elog(ERROR, "Backup %s has status %s, merge is aborted", + backup_id_of(backup), status2str(backup->status)); + } + } /* - * Previous merging was interrupted during deleting source backup. It is - * safe just to delete it again. + * Get backup files. */ - if (from_backup->status == BACKUP_STATUS_DELETING) - goto delete_source_backup; + for (i = parray_num(parent_chain) - 1; i >= 0; i--) + { + pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); - write_backup_status(to_backup, BACKUP_STATUS_MERGING); - write_backup_status(from_backup, BACKUP_STATUS_MERGING); + backup->files = get_backup_filelist(backup, true); + parray_qsort(backup->files, pgFileCompareRelPathWithExternal); - create_data_directories(files, to_database_path, from_backup_path, false, FIO_BACKUP_HOST); + /* Set MERGING status for every member of the chain */ + if (backup->backup_mode == BACKUP_MODE_FULL) + { + /* In case of FULL backup also remember backup_id of + * of destination backup we are merging with, so + * we can safely allow rerun merge in case of failure. + */ + backup->merge_dest_backup = dest_backup->start_time; + backup->status = BACKUP_STATUS_MERGING; + write_backup(backup, true); + } + else + write_backup_status(backup, BACKUP_STATUS_MERGING, true); + } - threads = (pthread_t *) palloc(sizeof(pthread_t) * num_threads); - threads_args = (merge_files_arg *) palloc(sizeof(merge_files_arg) * num_threads); + /* Construct path to database dir: /backup_dir/instance_name/FULL/database */ + join_path_components(full_database_dir, full_backup->root_dir, DATABASE_DIR); + /* Construct path to external dir: /backup_dir/instance_name/FULL/external */ + join_path_components(full_external_prefix, full_backup->root_dir, EXTERNAL_DIR); - /* Create external directories lists */ - if (to_backup->external_dir_str) - to_external = make_external_directory_list(to_backup->external_dir_str, - false); - if (from_backup->external_dir_str) - from_external = make_external_directory_list(from_backup->external_dir_str, - false); + /* Create directories */ + create_data_directories(dest_backup->files, full_database_dir, + dest_backup->root_dir, false, false, FIO_BACKUP_HOST, NULL); + /* External directories stuff */ + if (dest_backup->external_dir_str) + dest_externals = make_external_directory_list(dest_backup->external_dir_str, false); + if (full_backup->external_dir_str) + full_externals = make_external_directory_list(full_backup->external_dir_str, false); /* - * Rename external directories in to_backup (if exists) - * according to numeration of external dirs in from_backup. + * Rename external directories in FULL backup (if exists) + * according to numeration of external dirs in destionation backup. */ - if (to_external) - reorder_external_dirs(to_backup, to_external, from_external); + if (full_externals && dest_externals) + reorder_external_dirs(full_backup, full_externals, dest_externals); + + /* bitmap optimization rely on n_blocks, which is generally available since 2.3.0 */ + if (parse_program_version(dest_backup->program_version) < 20300) + use_bitmap = false; /* Setup threads */ - for (i = 0; i < parray_num(files); i++) + for (i = 0; i < parray_num(dest_backup->files); i++) { - pgFile *file = (pgFile *) parray_get(files, i); + pgFile *file = (pgFile *) parray_get(dest_backup->files, i); /* if the entry was an external directory, create it in the backup */ if (file->external_dir_num && S_ISDIR(file->mode)) @@ -287,28 +631,36 @@ merge_backups(pgBackup *to_backup, pgBackup *from_backup) char dirpath[MAXPGPATH]; char new_container[MAXPGPATH]; - makeExternalDirPathByNum(new_container, to_external_prefix, + makeExternalDirPathByNum(new_container, full_external_prefix, file->external_dir_num); - join_path_components(dirpath, new_container, file->path); - dir_create_dir(dirpath, DIR_PERMISSION); + join_path_components(dirpath, new_container, file->rel_path); + dir_create_dir(dirpath, DIR_PERMISSION, false); } + pg_atomic_init_flag(&file->lock); } + threads = (pthread_t *) palloc(sizeof(pthread_t) * num_threads); + threads_args = (merge_files_arg *) palloc(sizeof(merge_files_arg) * num_threads); + thread_interrupted = false; + merge_time = time(NULL); + elog(INFO, "Start merging backup files"); for (i = 0; i < num_threads; i++) { merge_files_arg *arg = &(threads_args[i]); - - arg->to_files = to_files; - arg->files = files; - arg->to_backup = to_backup; - arg->from_backup = from_backup; - arg->to_root = to_database_path; - arg->from_root = from_database_path; - arg->from_external = from_external; - arg->to_external_prefix = to_external_prefix; - arg->from_external_prefix = from_external_prefix; + arg->merge_filelist = parray_new(); + arg->parent_chain = parent_chain; + arg->dest_backup = dest_backup; + arg->full_backup = full_backup; + arg->full_database_dir = full_database_dir; + arg->full_external_prefix = full_external_prefix; + + arg->compression_match = compression_match; + arg->program_version_match = program_version_match; + arg->use_bitmap = use_bitmap; + arg->is_retry = is_retry; + arg->no_sync = no_sync; /* By default there are some error */ arg->ret = 1; @@ -318,121 +670,215 @@ merge_backups(pgBackup *to_backup, pgBackup *from_backup) } /* Wait threads */ + result_filelist = parray_new(); for (i = 0; i < num_threads; i++) { pthread_join(threads[i], NULL); if (threads_args[i].ret == 1) merge_isok = false; + + /* Compile final filelist */ + parray_concat(result_filelist, threads_args[i].merge_filelist); + + /* cleanup */ + parray_free(threads_args[i].merge_filelist); + //total_in_place_merge_bytes += threads_args[i].in_place_merge_bytes; + } + + time(&end_time); + pretty_time_interval(difftime(end_time, merge_time), + pretty_time, lengthof(pretty_time)); + + if (merge_isok) + elog(INFO, "Backup files are successfully merged, time elapsed: %s", + pretty_time); + else + elog(ERROR, "Backup files merging failed, time elapsed: %s", + pretty_time); + + /* If temp header map is open, then close it and make rename */ + if (full_backup->hdr_map.fp) + { + cleanup_header_map(&(full_backup->hdr_map)); + + /* sync new header map to disk */ + if (fio_sync(full_backup->hdr_map.path_tmp, FIO_BACKUP_HOST) != 0) + elog(ERROR, "Cannot sync temp header map \"%s\": %s", + full_backup->hdr_map.path_tmp, strerror(errno)); + + /* Replace old header map with new one */ + if (rename(full_backup->hdr_map.path_tmp, full_backup->hdr_map.path)) + elog(ERROR, "Could not rename file \"%s\" to \"%s\": %s", + full_backup->hdr_map.path_tmp, full_backup->hdr_map.path, strerror(errno)); + } + + /* Close page header maps */ + for (i = parray_num(parent_chain) - 1; i >= 0; i--) + { + pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); + cleanup_header_map(&(backup->hdr_map)); } - if (!merge_isok) - elog(ERROR, "Data files merging failed"); /* - * Update to_backup metadata. + * Update FULL backup metadata. * We cannot set backup status to OK just yet, * because it still has old start_time. */ - StrNCpy(to_backup->program_version, PROGRAM_VERSION, - sizeof(to_backup->program_version)); - to_backup->parent_backup = INVALID_BACKUP_ID; - to_backup->start_lsn = from_backup->start_lsn; - to_backup->stop_lsn = from_backup->stop_lsn; - to_backup->recovery_time = from_backup->recovery_time; - to_backup->recovery_xid = from_backup->recovery_xid; - pfree(to_backup->external_dir_str); - to_backup->external_dir_str = from_backup->external_dir_str; - from_backup->external_dir_str = NULL; /* For safe pgBackupFree() */ - to_backup->merge_time = merge_time; - to_backup->end_time = time(NULL); - - /* - * Target backup must inherit wal mode too. + strlcpy(full_backup->program_version, PROGRAM_VERSION, + sizeof(full_backup->program_version)); + full_backup->parent_backup = INVALID_BACKUP_ID; + full_backup->start_lsn = dest_backup->start_lsn; + full_backup->stop_lsn = dest_backup->stop_lsn; + full_backup->recovery_time = dest_backup->recovery_time; + full_backup->recovery_xid = dest_backup->recovery_xid; + full_backup->tli = dest_backup->tli; + full_backup->from_replica = dest_backup->from_replica; + + pfree(full_backup->external_dir_str); + full_backup->external_dir_str = pgut_strdup(dest_backup->external_dir_str); + pfree(full_backup->primary_conninfo); + full_backup->primary_conninfo = pgut_strdup(dest_backup->primary_conninfo); + + full_backup->merge_time = merge_time; + full_backup->end_time = time(NULL); + + full_backup->compress_alg = dest_backup->compress_alg; + full_backup->compress_level = dest_backup->compress_level; + + /* If incremental backup is pinned, + * then result FULL backup must also be pinned. + * And reverse, if FULL backup was pinned and dest was not, + * then pinning is no more. */ - to_backup->stream = from_backup->stream; - /* Compute summary of size of regular files in the backup */ - to_backup->data_bytes = 0; - for (i = 0; i < parray_num(files); i++) - { - pgFile *file = (pgFile *) parray_get(files, i); + full_backup->expire_time = dest_backup->expire_time; - if (S_ISDIR(file->mode)) - { - to_backup->data_bytes += 4096; - continue; - } - /* Count the amount of the data actually copied */ - if (file->write_size > 0) - to_backup->data_bytes += file->write_size; - } - /* compute size of wal files of this backup stored in the archive */ - if (!to_backup->stream) - to_backup->wal_bytes = instance_config.xlog_seg_size * - (to_backup->stop_lsn / instance_config.xlog_seg_size - - to_backup->start_lsn / instance_config.xlog_seg_size + 1); - else - to_backup->wal_bytes = BYTES_INVALID; + pg_free(full_backup->note); + full_backup->note = NULL; - write_backup_filelist(to_backup, files, from_database_path, NULL); - write_backup(to_backup); + if (dest_backup->note) + full_backup->note = pgut_strdup(dest_backup->note); -delete_source_backup: - /* - * Files were copied into to_backup. It is time to remove source backup - * entirely. + /* FULL backup must inherit wal mode. */ + full_backup->stream = dest_backup->stream; + + /* ARCHIVE backup must inherit wal_bytes too. + * STREAM backup will have its wal_bytes calculated by + * write_backup_filelist(). */ - delete_backup_files(from_backup); + if (!dest_backup->stream) + full_backup->wal_bytes = dest_backup->wal_bytes; - /* - * Delete files which are not in from_backup file list. + parray_qsort(result_filelist, pgFileCompareRelPathWithExternal); + + write_backup_filelist(full_backup, result_filelist, full_database_dir, NULL, true); + write_backup(full_backup, true); + + /* Delete FULL backup files, that do not exists in destination backup + * Both arrays must be sorted in in reversed order to delete from leaf */ - parray_qsort(files, pgFileComparePathWithExternalDesc); - for (i = 0; i < parray_num(to_files); i++) + parray_qsort(dest_backup->files, pgFileCompareRelPathWithExternalDesc); + parray_qsort(full_backup->files, pgFileCompareRelPathWithExternalDesc); + for (i = 0; i < parray_num(full_backup->files); i++) { - pgFile *file = (pgFile *) parray_get(to_files, i); + pgFile *full_file = (pgFile *) parray_get(full_backup->files, i); - if (file->external_dir_num && to_external) + if (full_file->external_dir_num && full_externals) { - char *dir_name = parray_get(to_external, file->external_dir_num - 1); - if (backup_contains_external(dir_name, from_external)) + char *dir_name = parray_get(full_externals, full_file->external_dir_num - 1); + if (backup_contains_external(dir_name, full_externals)) /* Dir already removed*/ continue; } - if (parray_bsearch(files, file, pgFileComparePathWithExternalDesc) == NULL) + if (parray_bsearch(dest_backup->files, full_file, pgFileCompareRelPathWithExternalDesc) == NULL) { - char to_file_path[MAXPGPATH]; - char *prev_path; + char full_file_path[MAXPGPATH]; /* We need full path, file object has relative path */ - join_path_components(to_file_path, to_database_path, file->path); - prev_path = file->path; - file->path = to_file_path; - - pgFileDelete(file); - elog(VERBOSE, "Deleted \"%s\"", file->path); + join_path_components(full_file_path, full_database_dir, full_file->rel_path); - file->path = prev_path; + pgFileDelete(full_file->mode, full_file_path); + elog(LOG, "Deleted \"%s\"", full_file_path); } } + /* Critical section starts. + * Change status of FULL backup. + * Files are merged into FULL backup. It is time to remove incremental chain. + */ + full_backup->status = BACKUP_STATUS_MERGED; + write_backup(full_backup, true); + +merge_delete: + for (i = parray_num(parent_chain) - 2; i >= 0; i--) + { + pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); + delete_backup_files(backup); + } + /* - * Rename FULL backup directory. + * PAGE2 DELETED + * PAGE1 DELETED + * FULL MERGED + * If we crash now, automatic rerun of failed merge is still possible: + * The user should start merge with full backup ID as an argument to option '-i'. + */ + +merge_rename: + /* + * Rename FULL backup directory to destination backup directory. + */ + if (dest_backup) + { + elog(LOG, "Rename %s to %s", full_backup->root_dir, dest_backup->root_dir); + if (rename(full_backup->root_dir, dest_backup->root_dir) == -1) + elog(ERROR, "Could not rename directory \"%s\" to \"%s\": %s", + full_backup->root_dir, dest_backup->root_dir, strerror(errno)); + + /* update root_dir after rename */ + pg_free(full_backup->root_dir); + full_backup->root_dir = pgut_strdup(dest_backup->root_dir); + } + else + { + /* Ugly */ + char destination_path[MAXPGPATH]; + + join_path_components(destination_path, instanceState->instance_backup_subdir_path, + base36enc(full_backup->merge_dest_backup)); + + elog(LOG, "Rename %s to %s", full_backup->root_dir, destination_path); + if (rename(full_backup->root_dir, destination_path) == -1) + elog(ERROR, "Could not rename directory \"%s\" to \"%s\": %s", + full_backup->root_dir, destination_path, strerror(errno)); + + /* update root_dir after rename */ + pg_free(full_backup->root_dir); + full_backup->root_dir = pgut_strdup(destination_path); + } + + /* Reinit path to database_dir */ + join_path_components(full_backup->database_dir, full_backup->root_dir, DATABASE_DIR); + + /* If we crash here, it will produce full backup in MERGED + * status, located in directory with wrong backup id. + * It should not be a problem. */ - elog(INFO, "Rename %s to %s", to_backup_id, from_backup_id); - if (rename(to_backup_path, from_backup_path) == -1) - elog(ERROR, "Could not rename directory \"%s\" to \"%s\": %s", - to_backup_path, from_backup_path, strerror(errno)); /* - * Merging finished, now we can safely update ID of the destination backup. - * TODO: for this critical section we must save incremental backup start_tome - * to FULL backup meta, so even if crash happens after incremental backup removal - * but before full backup obtaining new start_time we could safely continue - * this failed backup. + * Merging finished, now we can safely update ID of the FULL backup */ - to_backup->status = BACKUP_STATUS_OK; - to_backup->start_time = from_backup->start_time; - write_backup(to_backup); + elog(INFO, "Rename merged full backup %s to %s", + backup_id_of(full_backup), + base36enc(full_backup->merge_dest_backup)); + + full_backup->status = BACKUP_STATUS_OK; + full_backup->start_time = full_backup->merge_dest_backup; + /* XXX BACKUP_ID change it when backup_id wouldn't match start_time */ + full_backup->backup_id = full_backup->start_time; + full_backup->merge_dest_backup = INVALID_BACKUP_ID; + write_backup(full_backup, true); + /* Critical section end */ /* Cleanup */ if (threads) @@ -441,270 +887,230 @@ merge_backups(pgBackup *to_backup, pgBackup *from_backup) pfree(threads); } - parray_walk(to_files, pgFileFree); - parray_free(to_files); + if (result_filelist) + { + parray_walk(result_filelist, pgFileFree); + parray_free(result_filelist); + } - parray_walk(files, pgFileFree); - parray_free(files); + if (dest_externals != NULL) + free_dir_list(dest_externals); + + if (full_externals != NULL) + free_dir_list(full_externals); + + for (i = parray_num(parent_chain) - 1; i >= 0; i--) + { + pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); - pfree(to_backup_id); - pfree(from_backup_id); + if (backup->files) + { + parray_walk(backup->files, pgFileFree); + parray_free(backup->files); + } + } } /* - * Thread worker of merge_backups(). + * Thread worker of merge_chain(). */ static void * merge_files(void *arg) { - merge_files_arg *argument = (merge_files_arg *) arg; - pgBackup *to_backup = argument->to_backup; - pgBackup *from_backup = argument->from_backup; - int i, - num_files = parray_num(argument->files); + int i; + merge_files_arg *arguments = (merge_files_arg *) arg; + size_t n_files = parray_num(arguments->dest_backup->files); - for (i = 0; i < num_files; i++) + for (i = 0; i < n_files; i++) { - pgFile *file = (pgFile *) parray_get(argument->files, i); - pgFile *to_file; - pgFile **res_file; - char to_file_path[MAXPGPATH]; /* Path of target file */ - char from_file_path[MAXPGPATH]; - char *prev_file_path; + pgFile *dest_file = (pgFile *) parray_get(arguments->dest_backup->files, i); + pgFile *tmp_file; + bool in_place = false; /* keep file as it is */ /* check for interrupt */ if (interrupted || thread_interrupted) - elog(ERROR, "Interrupted during merging backups"); + elog(ERROR, "Interrupted during merge"); - /* Directories were created before */ - if (S_ISDIR(file->mode)) + if (!pg_atomic_test_set_flag(&dest_file->lock)) continue; - if (!pg_atomic_test_set_flag(&file->lock)) - continue; + tmp_file = pgFileInit(dest_file->rel_path); + tmp_file->mode = dest_file->mode; + tmp_file->is_datafile = dest_file->is_datafile; + tmp_file->is_cfs = dest_file->is_cfs; + tmp_file->external_dir_num = dest_file->external_dir_num; + tmp_file->dbOid = dest_file->dbOid; + + /* Directories were created before */ + if (S_ISDIR(dest_file->mode)) + goto done; - if (progress) - elog(INFO, "Progress: (%d/%d). Process file \"%s\"", - i + 1, num_files, file->path); + elog(progress ? INFO : LOG, "Progress: (%d/%lu). Merging file \"%s\"", + i + 1, n_files, dest_file->rel_path); - res_file = parray_bsearch(argument->to_files, file, - pgFileComparePathWithExternalDesc); - to_file = (res_file) ? *res_file : NULL; + if (dest_file->is_datafile && !dest_file->is_cfs) + tmp_file->segno = dest_file->segno; - join_path_components(to_file_path, argument->to_root, file->path); + // If destination file is 0 sized, then go for the next + if (dest_file->write_size == 0) + { + if (!dest_file->is_datafile || dest_file->is_cfs) + tmp_file->crc = dest_file->crc; + + tmp_file->write_size = 0; + goto done; + } /* - * Skip files which haven't changed since previous backup. But in case - * of DELTA backup we must truncate the target file to n_blocks. - * Unless it is a non data file, in this case truncation is not needed. + * If file didn`t changed over the course of all incremental chain, + * then do in-place merge, unless destination backup has + * different compression algorithm. + * In-place merge is also impossible, if program version of destination + * backup differs from PROGRAM_VERSION */ - if (file->write_size == BYTES_INVALID) + if (arguments->program_version_match && arguments->compression_match && + !arguments->is_retry) { - /* sanity */ - if (!to_file) - elog(ERROR, "The file \"%s\" is missing in FULL backup %s", - file->rel_path, base36enc(to_backup->start_time)); - - /* for not changed files of all types in PAGE and PTRACK */ - if (from_backup->backup_mode != BACKUP_MODE_DIFF_DELTA || - /* and not changed non-data files in DELTA */ - (!file->is_datafile || file->is_cfs)) - { - elog(VERBOSE, "Skip merging file \"%s\", the file didn't change", - file->rel_path); + /* + * Case 1: + * in this case in place merge is possible: + * 0 PAGE; file, size BYTES_INVALID + * 1 PAGE; file, size BYTES_INVALID + * 2 FULL; file, size 100500 + * + * Case 2: + * in this case in place merge is possible: + * 0 PAGE; file, size 0 + * 1 PAGE; file, size 0 + * 2 FULL; file, size 100500 + * + * Case 3: + * in this case in place merge is impossible: + * 0 PAGE; file, size BYTES_INVALID + * 1 PAGE; file, size 100501 + * 2 FULL; file, size 100500 + * + * Case 4 (good candidate for future optimization): + * in this case in place merge is impossible: + * 0 PAGE; file, size BYTES_INVALID + * 1 PAGE; file, size 100501 + * 2 FULL; file, not exists yet + */ - /* - * If the file wasn't changed, retreive its - * write_size and compression algorihtm from previous FULL backup. - */ - file->compress_alg = to_file->compress_alg; - file->write_size = to_file->write_size; + in_place = true; - /* - * Recalculate crc for backup prior to 2.0.25. - */ - if (parse_program_version(from_backup->program_version) < 20025) - file->crc = pgFileGetCRC(to_file_path, true, true, NULL, FIO_LOCAL_HOST); - /* Otherwise just get it from the previous file */ - else - file->crc = to_file->crc; + for (i = parray_num(arguments->parent_chain) - 1; i >= 0; i--) + { + pgFile **res_file = NULL; + pgFile *file = NULL; - continue; - } - } + pgBackup *backup = (pgBackup *) parray_get(arguments->parent_chain, i); - /* TODO optimization: file from incremental backup has size 0, then - * just truncate the file from FULL backup - */ + /* lookup file in intermediate backup */ + res_file = parray_bsearch(backup->files, dest_file, pgFileCompareRelPathWithExternal); + file = (res_file) ? *res_file : NULL; - /* We need to make full path, file object has relative path */ - if (file->external_dir_num) - { - char temp[MAXPGPATH]; - makeExternalDirPathByNum(temp, argument->from_external_prefix, - file->external_dir_num); + /* Destination file is not exists yet, + * in-place merge is impossible + */ + if (file == NULL) + { + in_place = false; + break; + } - join_path_components(from_file_path, temp, file->path); + /* Skip file from FULL backup */ + if (backup->backup_mode == BACKUP_MODE_FULL) + continue; + + if (file->write_size != BYTES_INVALID) + { + in_place = false; + break; + } + } } - else - join_path_components(from_file_path, argument->from_root, - file->path); - prev_file_path = file->path; - file->path = from_file_path; /* - * Move the file. We need to decompress it and compress again if - * necessary. + * In-place merge means that file in FULL backup stays as it is, + * no additional actions are required. + * page header map cannot be trusted when retrying, so no + * in place merge for retry. */ - elog(VERBOSE, "Merging file \"%s\", is_datafile %d, is_cfs %d", - file->path, file->is_database, file->is_cfs); - - if (file->is_datafile && !file->is_cfs) + if (in_place) { - /* - * We need more complicate algorithm if target file should be - * compressed. - */ - if (to_backup->compress_alg == PGLZ_COMPRESS || - to_backup->compress_alg == ZLIB_COMPRESS) + pgFile **res_file = NULL; + pgFile *file = NULL; + res_file = parray_bsearch(arguments->full_backup->files, dest_file, + pgFileCompareRelPathWithExternal); + file = (res_file) ? *res_file : NULL; + + /* If file didn`t changed in any way, then in-place merge is possible */ + if (file && + file->n_blocks == dest_file->n_blocks) { - char merge_to_file_path[MAXPGPATH]; - char tmp_file_path[MAXPGPATH]; - char *prev_path; - - snprintf(merge_to_file_path, MAXPGPATH, "%s_merge", to_file_path); - snprintf(tmp_file_path, MAXPGPATH, "%s_tmp", to_file_path); - - /* Start the magic */ - - /* - * Merge files: - * - if to_file in FULL backup exists, restore and decompress it to to_file_merge - * - decompress source file if necessary and merge it with the - * target decompressed file in to_file_merge. - * - compress result file to to_file_tmp - * - rename to_file_tmp to to_file - */ + BackupPageHeader2 *headers = NULL; - /* - * We need to decompress target file in FULL backup if it exists. - */ - if (to_file) + elog(LOG, "The file didn`t changed since FULL backup, skip merge: \"%s\"", + file->rel_path); + + tmp_file->crc = file->crc; + tmp_file->write_size = file->write_size; + + if (dest_file->is_datafile && !dest_file->is_cfs) { - elog(VERBOSE, "Merge target and source files into the temporary path \"%s\"", - merge_to_file_path); - - // TODO: truncate merge_to_file_path just in case? - - /* - * file->path is relative, to_file_path - is absolute. - * Substitute them. - */ - prev_path = to_file->path; - to_file->path = to_file_path; - /* Decompress target file into temporary one */ - restore_data_file(merge_to_file_path, to_file, false, false, - parse_program_version(to_backup->program_version)); - to_file->path = prev_path; + tmp_file->n_blocks = file->n_blocks; + tmp_file->compress_alg = file->compress_alg; + tmp_file->uncompressed_size = file->n_blocks * BLCKSZ; + + tmp_file->n_headers = file->n_headers; + tmp_file->hdr_crc = file->hdr_crc; } else - elog(VERBOSE, "Restore source file into the temporary path \"%s\"", - merge_to_file_path); - - /* TODO: Optimize merge of new files */ - - /* Merge source file with target file */ - restore_data_file(merge_to_file_path, file, - from_backup->backup_mode == BACKUP_MODE_DIFF_DELTA, - false, - parse_program_version(from_backup->program_version)); - - elog(VERBOSE, "Compress file and save it into the directory \"%s\"", - argument->to_root); - - /* Again we need to change path */ - prev_path = file->path; - file->path = merge_to_file_path; - /* backup_data_file() requires file size to calculate nblocks */ - file->size = pgFileSize(file->path); - /* Now we can compress the file */ - backup_data_file(NULL, /* We shouldn't need 'arguments' here */ - tmp_file_path, file, - to_backup->start_lsn, - to_backup->backup_mode, - to_backup->compress_alg, - to_backup->compress_level, - false); - - file->path = prev_path; - - /* rename temp file */ - if (rename(tmp_file_path, to_file_path) == -1) - elog(ERROR, "Could not rename file \"%s\" to \"%s\": %s", - file->path, tmp_file_path, strerror(errno)); - - /* We can remove temporary file */ - if (unlink(merge_to_file_path)) - elog(ERROR, "Could not remove temporary file \"%s\": %s", - merge_to_file_path, strerror(errno)); - } - /* - * Otherwise merging algorithm is simpler. - */ - else - { - /* We can merge in-place here */ - restore_data_file(to_file_path, file, - from_backup->backup_mode == BACKUP_MODE_DIFF_DELTA, - true, - parse_program_version(from_backup->program_version)); - - /* - * We need to calculate write_size, restore_data_file() doesn't - * do that. - */ - file->write_size = pgFileSize(to_file_path); - file->crc = pgFileGetCRC(to_file_path, true, true, NULL, FIO_LOCAL_HOST); + tmp_file->uncompressed_size = file->uncompressed_size; + + /* Copy header metadata from old map into a new one */ + tmp_file->n_headers = file->n_headers; + headers = get_data_file_headers(&(arguments->full_backup->hdr_map), file, + parse_program_version(arguments->full_backup->program_version), + true); + + /* sanity */ + if (!headers && file->n_headers > 0) + elog(ERROR, "Failed to get headers for file \"%s\"", file->rel_path); + + write_page_headers(headers, tmp_file, &(arguments->full_backup->hdr_map), true); + pg_free(headers); + + //TODO: report in_place merge bytes. + goto done; } } - else if (file->external_dir_num) - { - char to_root[MAXPGPATH]; - int new_dir_num; - char *file_external_path = parray_get(argument->from_external, - file->external_dir_num - 1); - - Assert(argument->from_external); - new_dir_num = get_external_index(file_external_path, - argument->from_external); - makeExternalDirPathByNum(to_root, argument->to_external_prefix, - new_dir_num); - copy_file(FIO_LOCAL_HOST, to_root, FIO_LOCAL_HOST, file, false); - } - else if (strcmp(file->name, "pg_control") == 0) - copy_pgcontrol_file(argument->from_root, FIO_LOCAL_HOST, argument->to_root, FIO_LOCAL_HOST, file); - else - copy_file(FIO_LOCAL_HOST, argument->to_root, FIO_LOCAL_HOST, file, false); - - /* - * We need to save compression algorithm type of the target backup to be - * able to restore in the future. - */ - file->compress_alg = to_backup->compress_alg; - if (file->write_size != BYTES_INVALID) - elog(VERBOSE, "Merged file \"%s\": " INT64_FORMAT " bytes", - file->path, file->write_size); + if (dest_file->is_datafile && !dest_file->is_cfs) + merge_data_file(arguments->parent_chain, + arguments->full_backup, + arguments->dest_backup, + dest_file, tmp_file, + arguments->full_database_dir, + arguments->use_bitmap, + arguments->is_retry, + arguments->no_sync); else - elog(ERROR, "Merge of file \"%s\" failed. Invalid size: %i", - file->path, BYTES_INVALID); - - /* Restore relative path */ - file->path = prev_file_path; + merge_non_data_file(arguments->parent_chain, + arguments->full_backup, + arguments->dest_backup, + dest_file, tmp_file, + arguments->full_database_dir, + arguments->full_external_prefix, + arguments->no_sync); + +done: + parray_append(arguments->merge_filelist, tmp_file); } /* Data files merging is successful */ - argument->ret = 0; + arguments->ret = 0; return NULL; } @@ -715,16 +1121,23 @@ remove_dir_with_files(const char *path) { parray *files = parray_new(); int i; + char full_path[MAXPGPATH]; - dir_list_file(files, path, true, true, true, 0, FIO_LOCAL_HOST); - parray_qsort(files, pgFileComparePathDesc); + dir_list_file(files, path, false, false, true, false, false, 0, FIO_LOCAL_HOST); + parray_qsort(files, pgFileCompareRelPathWithExternalDesc); for (i = 0; i < parray_num(files); i++) { pgFile *file = (pgFile *) parray_get(files, i); - pgFileDelete(file); - elog(VERBOSE, "Deleted \"%s\"", file->path); + join_path_components(full_path, path, file->rel_path); + + pgFileDelete(file->mode, full_path); + elog(LOG, "Deleted \"%s\"", full_path); } + + /* cleanup */ + parray_walk(files, pgFileFree); + parray_free(files); } /* Get index of external directory */ @@ -751,8 +1164,7 @@ reorder_external_dirs(pgBackup *to_backup, parray *to_external, char externaldir_template[MAXPGPATH]; int i; - pgBackupGetPath(to_backup, externaldir_template, - lengthof(externaldir_template), EXTERNAL_DIR); + join_path_components(externaldir_template, to_backup->root_dir, EXTERNAL_DIR); for (i = 0; i < parray_num(to_external); i++) { int from_num = get_external_index(parray_get(to_external, i), @@ -769,10 +1181,261 @@ reorder_external_dirs(pgBackup *to_backup, parray *to_external, char new_path[MAXPGPATH]; makeExternalDirPathByNum(old_path, externaldir_template, i + 1); makeExternalDirPathByNum(new_path, externaldir_template, from_num); - elog(VERBOSE, "Rename %s to %s", old_path, new_path); + elog(LOG, "Rename %s to %s", old_path, new_path); if (rename (old_path, new_path) == -1) elog(ERROR, "Could not rename directory \"%s\" to \"%s\": %s", old_path, new_path, strerror(errno)); } } } + +/* Merge is usually happens as usual backup/restore via temp files, unless + * file didn`t changed since FULL backup AND full a dest backup have the + * same compression algorithm. In this case file can be left as it is. + */ +void +merge_data_file(parray *parent_chain, pgBackup *full_backup, + pgBackup *dest_backup, pgFile *dest_file, pgFile *tmp_file, + const char *full_database_dir, bool use_bitmap, bool is_retry, + bool no_sync) +{ + FILE *out = NULL; + char *buffer = pgut_malloc(STDIO_BUFSIZE); + char to_fullpath[MAXPGPATH]; + char to_fullpath_tmp1[MAXPGPATH]; /* used for restore */ + char to_fullpath_tmp2[MAXPGPATH]; /* used for backup */ + + /* The next possible optimization is copying "as is" the file + * from intermediate incremental backup, that didn`t changed in + * subsequent incremental backups. TODO. + */ + + /* set fullpath of destination file and temp files */ + join_path_components(to_fullpath, full_database_dir, tmp_file->rel_path); + snprintf(to_fullpath_tmp1, MAXPGPATH, "%s_tmp1", to_fullpath); + snprintf(to_fullpath_tmp2, MAXPGPATH, "%s_tmp2", to_fullpath); + + /* open temp file */ + out = fopen(to_fullpath_tmp1, PG_BINARY_W); + if (out == NULL) + elog(ERROR, "Cannot open merge target file \"%s\": %s", + to_fullpath_tmp1, strerror(errno)); + setvbuf(out, buffer, _IOFBF, STDIO_BUFSIZE); + + /* restore file into temp file */ + tmp_file->size = restore_data_file(parent_chain, dest_file, out, to_fullpath_tmp1, + use_bitmap, NULL, InvalidXLogRecPtr, NULL, + /* when retrying merge header map cannot be trusted */ + is_retry ? false : true); + if (fclose(out) != 0) + elog(ERROR, "Cannot close file \"%s\": %s", + to_fullpath_tmp1, strerror(errno)); + + pg_free(buffer); + + /* tmp_file->size is greedy, even if there is single 8KB block in file, + * that was overwritten twice during restore_data_file, we would assume that its size is + * 16KB. + * TODO: maybe we should just trust dest_file->n_blocks? + * No, we can`t, because current binary can be used to merge + * 2 backups of old versions, where n_blocks is missing. + */ + + backup_data_file(tmp_file, to_fullpath_tmp1, to_fullpath_tmp2, + InvalidXLogRecPtr, BACKUP_MODE_FULL, + dest_backup->compress_alg, dest_backup->compress_level, + dest_backup->checksum_version, + &(full_backup->hdr_map), true); + + /* drop restored temp file */ + if (unlink(to_fullpath_tmp1) == -1) + elog(ERROR, "Cannot remove file \"%s\": %s", to_fullpath_tmp1, + strerror(errno)); + + /* + * In old (=<2.2.7) versions of pg_probackup n_blocks attribute of files + * in PAGE and PTRACK wasn`t filled. + */ + //Assert(tmp_file->n_blocks == dest_file->n_blocks); + + /* Backward compatibility kludge: + * When merging old backups, it is possible that + * to_fullpath_tmp2 size will be 0, and so it will be + * truncated in backup_data_file(). + * TODO: remove in 3.0.0 + */ + if (tmp_file->write_size == 0) + return; + + /* sync second temp file to disk */ + if (!no_sync && fio_sync(to_fullpath_tmp2, FIO_BACKUP_HOST) != 0) + elog(ERROR, "Cannot sync merge temp file \"%s\": %s", + to_fullpath_tmp2, strerror(errno)); + + /* Do atomic rename from second temp file to destination file */ + if (rename(to_fullpath_tmp2, to_fullpath) == -1) + elog(ERROR, "Could not rename file \"%s\" to \"%s\": %s", + to_fullpath_tmp2, to_fullpath, strerror(errno)); + + /* drop temp file */ + unlink(to_fullpath_tmp1); +} + +/* + * For every destionation file lookup the newest file in chain and + * copy it. + * Additional pain is external directories. + */ +void +merge_non_data_file(parray *parent_chain, pgBackup *full_backup, + pgBackup *dest_backup, pgFile *dest_file, pgFile *tmp_file, + const char *full_database_dir, const char *to_external_prefix, + bool no_sync) +{ + int i; + char to_fullpath[MAXPGPATH]; + char to_fullpath_tmp[MAXPGPATH]; /* used for backup */ + char from_fullpath[MAXPGPATH]; + pgBackup *from_backup = NULL; + pgFile *from_file = NULL; + + /* We need to make full path to destination file */ + if (dest_file->external_dir_num) + { + char temp[MAXPGPATH]; + makeExternalDirPathByNum(temp, to_external_prefix, + dest_file->external_dir_num); + join_path_components(to_fullpath, temp, dest_file->rel_path); + } + else + join_path_components(to_fullpath, full_database_dir, dest_file->rel_path); + + snprintf(to_fullpath_tmp, MAXPGPATH, "%s_tmp", to_fullpath); + + /* + * Iterate over parent chain starting from direct parent of destination + * backup to oldest backup in chain, and look for the first + * full copy of destination file. + * Full copy is latest possible destination file with size equal(!) + * or greater than zero. + */ + for (i = 0; i < parray_num(parent_chain); i++) + { + pgFile **res_file = NULL; + from_backup = (pgBackup *) parray_get(parent_chain, i); + + /* lookup file in intermediate backup */ + res_file = parray_bsearch(from_backup->files, dest_file, pgFileCompareRelPathWithExternal); + from_file = (res_file) ? *res_file : NULL; + + /* + * It should not be possible not to find source file in intermediate + * backup, without encountering full copy first. + */ + if (!from_file) + { + elog(ERROR, "Failed to locate non-data file \"%s\" in backup %s", + dest_file->rel_path, backup_id_of(from_backup)); + continue; + } + + if (from_file->write_size > 0) + break; + } + + /* sanity */ + if (!from_backup) + elog(ERROR, "Failed to found a backup containing full copy of non-data file \"%s\"", + dest_file->rel_path); + + if (!from_file) + elog(ERROR, "Failed to locate a full copy of non-data file \"%s\"", dest_file->rel_path); + + /* set path to source file */ + if (from_file->external_dir_num) + { + char temp[MAXPGPATH]; + char external_prefix[MAXPGPATH]; + + join_path_components(external_prefix, from_backup->root_dir, EXTERNAL_DIR); + makeExternalDirPathByNum(temp, external_prefix, dest_file->external_dir_num); + + join_path_components(from_fullpath, temp, from_file->rel_path); + } + else + { + char backup_database_dir[MAXPGPATH]; + join_path_components(backup_database_dir, from_backup->root_dir, DATABASE_DIR); + join_path_components(from_fullpath, backup_database_dir, from_file->rel_path); + } + + /* Copy file to FULL backup directory into temp file */ + backup_non_data_file(tmp_file, NULL, from_fullpath, + to_fullpath_tmp, BACKUP_MODE_FULL, 0, false); + + /* sync temp file to disk */ + if (!no_sync && fio_sync(to_fullpath_tmp, FIO_BACKUP_HOST) != 0) + elog(ERROR, "Cannot sync merge temp file \"%s\": %s", + to_fullpath_tmp, strerror(errno)); + + /* Do atomic rename from second temp file to destination file */ + if (rename(to_fullpath_tmp, to_fullpath) == -1) + elog(ERROR, "Could not rename file \"%s\" to \"%s\": %s", + to_fullpath_tmp, to_fullpath, strerror(errno)); + +} + +/* + * If file format in incremental chain is compatible + * with current storage format. + * If not, then in-place merge is not possible. + * + * Consider the following examples: + * STORAGE_FORMAT_VERSION = 2.4.4 + * 2.3.3 \ + * 2.3.4 \ disable in-place merge, because + * 2.4.1 / current STORAGE_FORMAT_VERSION > 2.3.3 + * 2.4.3 / + * + * 2.4.4 \ enable in_place merge, because + * 2.4.5 / current STORAGE_FORMAT_VERSION == 2.4.4 + * + * 2.4.5 \ enable in_place merge, because + * 2.4.6 / current STORAGE_FORMAT_VERSION < 2.4.5 + * + */ +bool +is_forward_compatible(parray *parent_chain) +{ + int i; + pgBackup *oldest_ver_backup = NULL; + uint32 oldest_ver_in_chain = parse_program_version(PROGRAM_VERSION); + + for (i = 0; i < parray_num(parent_chain); i++) + { + pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); + uint32 current_version = parse_program_version(backup->program_version); + + if (!oldest_ver_backup) + oldest_ver_backup = backup; + + if (current_version < oldest_ver_in_chain) + { + oldest_ver_in_chain = current_version; + oldest_ver_backup = backup; + } + } + + if (oldest_ver_in_chain < parse_program_version(STORAGE_FORMAT_VERSION)) + { + elog(WARNING, "In-place merge is disabled because of storage format incompatibility. " + "Backup %s storage format version: %s, " + "current storage format version: %s", + backup_id_of(oldest_ver_backup), + oldest_ver_backup->program_version, + STORAGE_FORMAT_VERSION); + return false; + } + + return true; +} diff --git a/src/parsexlog.c b/src/parsexlog.c index 9008b5285..7df169fbf 100644 --- a/src/parsexlog.c +++ b/src/parsexlog.c @@ -29,7 +29,10 @@ * RmgrNames is an array of resource manager names, to make error messages * a bit nicer. */ -#if PG_VERSION_NUM >= 100000 +#if PG_VERSION_NUM >= 150000 +#define PG_RMGR(symname,name,redo,desc,identify,startup,cleanup,mask,decode) \ + name, +#elif PG_VERSION_NUM >= 100000 #define PG_RMGR(symname,name,redo,desc,identify,startup,cleanup,mask) \ name, #else @@ -138,6 +141,9 @@ typedef struct */ bool got_target; + /* Should we read record, located at endpoint position */ + bool inclusive_endpoint; + /* * Return value from the thread. * 0 means there is no error, 1 - there is an error. @@ -145,10 +151,16 @@ typedef struct int ret; } xlog_thread_arg; +static XLogRecord* WalReadRecord(XLogReaderState *xlogreader, XLogRecPtr startpoint, char **errormsg); +static XLogReaderState* WalReaderAllocate(uint32 wal_seg_size, XLogReaderData *reader_data); + static int SimpleXLogPageRead(XLogReaderState *xlogreader, XLogRecPtr targetPagePtr, - int reqLen, XLogRecPtr targetRecPtr, char *readBuf, - TimeLineID *pageTLI); + int reqLen, XLogRecPtr targetRecPtr, char *readBuf +#if PG_VERSION_NUM < 130000 + ,TimeLineID *pageTLI +#endif + ); static XLogReaderState *InitXLogPageRead(XLogReaderData *reader_data, const char *archivedir, TimeLineID tli, uint32 segment_size, @@ -162,7 +174,8 @@ static bool RunXLogThreads(const char *archivedir, XLogRecPtr startpoint, XLogRecPtr endpoint, bool consistent_read, xlog_record_function process_record, - XLogRecTarget *last_rec); + XLogRecTarget *last_rec, + bool inclusive_endpoint); //static XLogReaderState *InitXLogThreadRead(xlog_thread_arg *arg); static bool SwitchThreadToNextWal(XLogReaderState *xlogreader, xlog_thread_arg *arg); @@ -231,28 +244,123 @@ static XLogRecPtr wal_target_lsn = InvalidXLogRecPtr; * Pagemap extracting is processed using threads. Each thread reads single WAL * file. */ -void -extractPageMap(const char *archivedir, TimeLineID tli, uint32 wal_seg_size, - XLogRecPtr startpoint, XLogRecPtr endpoint) +bool +extractPageMap(const char *archivedir, uint32 wal_seg_size, + XLogRecPtr startpoint, TimeLineID start_tli, + XLogRecPtr endpoint, TimeLineID end_tli, + parray *tli_list) { - bool extract_isok = true; - time_t start_time, - end_time; - - elog(LOG, "Compiling pagemap"); - time(&start_time); - - extract_isok = RunXLogThreads(archivedir, 0, InvalidTransactionId, - InvalidXLogRecPtr, tli, wal_seg_size, - startpoint, endpoint, false, extractPageInfo, - NULL); - - time(&end_time); - if (extract_isok) - elog(LOG, "Pagemap compiled, time elapsed %.0f sec", - difftime(end_time, start_time)); + bool extract_isok = false; + + if (start_tli == end_tli) + /* easy case */ + extract_isok = RunXLogThreads(archivedir, 0, InvalidTransactionId, + InvalidXLogRecPtr, end_tli, wal_seg_size, + startpoint, endpoint, false, extractPageInfo, + NULL, true); else - elog(ERROR, "Pagemap compiling failed"); + { + /* We have to process WAL located on several different xlog intervals, + * located on different timelines. + * + * Consider this example: + * t3 C-----X + * / + * t1 -A----*-------> + * + * A - prev backup START_LSN + * B - switchpoint for t2, available as t2->switchpoint + * C - switch for t3, available as t3->switchpoint + * X - current backup START_LSN + * + * Intervals to be parsed: + * - [A,B) on t1 + * - [B,C) on t2 + * - [C,X] on t3 + */ + int i; + parray *interval_list = parray_new(); + timelineInfo *end_tlinfo = NULL; + timelineInfo *tmp_tlinfo = NULL; + XLogRecPtr prev_switchpoint = InvalidXLogRecPtr; + + /* We must find TLI information about final timeline (t3 in example) */ + for (i = 0; i < parray_num(tli_list); i++) + { + tmp_tlinfo = parray_get(tli_list, i); + + if (tmp_tlinfo->tli == end_tli) + { + end_tlinfo = tmp_tlinfo; + break; + } + } + + /* Iterate over timelines backward, + * starting with end_tli and ending with start_tli. + * For every timeline calculate LSN-interval that must be parsed. + */ + + tmp_tlinfo = end_tlinfo; + while (tmp_tlinfo) + { + lsnInterval *wal_interval = pgut_malloc(sizeof(lsnInterval)); + wal_interval->tli = tmp_tlinfo->tli; + + if (tmp_tlinfo->tli == end_tli) + { + wal_interval->begin_lsn = tmp_tlinfo->switchpoint; + wal_interval->end_lsn = endpoint; + } + else if (tmp_tlinfo->tli == start_tli) + { + wal_interval->begin_lsn = startpoint; + wal_interval->end_lsn = prev_switchpoint; + } + else + { + wal_interval->begin_lsn = tmp_tlinfo->switchpoint; + wal_interval->end_lsn = prev_switchpoint; + } + + parray_append(interval_list, wal_interval); + + if (tmp_tlinfo->tli == start_tli) + break; + + prev_switchpoint = tmp_tlinfo->switchpoint; + tmp_tlinfo = tmp_tlinfo->parent_link; + } + + for (i = parray_num(interval_list) - 1; i >= 0; i--) + { + bool inclusive_endpoint; + lsnInterval *tmp_interval = (lsnInterval *) parray_get(interval_list, i); + + /* In case of replica promotion, endpoints of intermediate + * timelines can be unreachable. + */ + inclusive_endpoint = false; + + /* ... but not the end timeline */ + if (tmp_interval->tli == end_tli) + inclusive_endpoint = true; + + extract_isok = RunXLogThreads(archivedir, 0, InvalidTransactionId, + InvalidXLogRecPtr, tmp_interval->tli, wal_seg_size, + tmp_interval->begin_lsn, tmp_interval->end_lsn, + false, extractPageInfo, NULL, inclusive_endpoint); + if (!extract_isok) + break; + + pg_free(tmp_interval); + } + pg_free(interval_list); + } + + return extract_isok; } /* @@ -272,7 +380,7 @@ validate_backup_wal_from_start_to_stop(pgBackup *backup, got_endpoint = RunXLogThreads(archivedir, 0, InvalidTransactionId, InvalidXLogRecPtr, tli, xlog_seg_size, backup->start_lsn, backup->stop_lsn, - false, NULL, NULL); + false, NULL, NULL, true); if (!got_endpoint) { @@ -280,11 +388,11 @@ validate_backup_wal_from_start_to_stop(pgBackup *backup, * If we don't have WAL between start_lsn and stop_lsn, * the backup is definitely corrupted. Update its status. */ - write_backup_status(backup, BACKUP_STATUS_CORRUPT); + write_backup_status(backup, BACKUP_STATUS_CORRUPT, true); elog(WARNING, "There are not enough WAL records to consistenly restore " "backup %s from START LSN: %X/%X to STOP LSN: %X/%X", - base36enc(backup->start_time), + backup_id_of(backup), (uint32) (backup->start_lsn >> 32), (uint32) (backup->start_lsn), (uint32) (backup->stop_lsn >> 32), @@ -302,25 +410,20 @@ validate_wal(pgBackup *backup, const char *archivedir, time_t target_time, TransactionId target_xid, XLogRecPtr target_lsn, TimeLineID tli, uint32 wal_seg_size) { - const char *backup_id; XLogRecTarget last_rec; char last_timestamp[100], target_timestamp[100]; bool all_wal = false; - char backup_xlog_path[MAXPGPATH]; - - /* We need free() this later */ - backup_id = base36enc(backup->start_time); if (!XRecOffIsValid(backup->start_lsn)) elog(ERROR, "Invalid start_lsn value %X/%X of backup %s", (uint32) (backup->start_lsn >> 32), (uint32) (backup->start_lsn), - backup_id); + backup_id_of(backup)); if (!XRecOffIsValid(backup->stop_lsn)) elog(ERROR, "Invalid stop_lsn value %X/%X of backup %s", (uint32) (backup->stop_lsn >> 32), (uint32) (backup->stop_lsn), - backup_id); + backup_id_of(backup)); /* * Check that the backup has all wal files needed @@ -328,8 +431,11 @@ validate_wal(pgBackup *backup, const char *archivedir, */ if (backup->stream) { - pgBackupGetPath2(backup, backup_xlog_path, lengthof(backup_xlog_path), - DATABASE_DIR, PG_XLOG_DIR); + char backup_database_dir[MAXPGPATH]; + char backup_xlog_path[MAXPGPATH]; + + join_path_components(backup_database_dir, backup->root_dir, DATABASE_DIR); + join_path_components(backup_xlog_path, backup_database_dir, PG_XLOG_DIR); validate_backup_wal_from_start_to_stop(backup, backup_xlog_path, tli, wal_seg_size); @@ -340,7 +446,7 @@ validate_wal(pgBackup *backup, const char *archivedir, if (backup->status == BACKUP_STATUS_CORRUPT) { - elog(WARNING, "Backup %s WAL segments are corrupted", backup_id); + elog(WARNING, "Backup %s WAL segments are corrupted", backup_id_of(backup)); return; } /* @@ -351,7 +457,7 @@ validate_wal(pgBackup *backup, const char *archivedir, !XRecOffIsValid(target_lsn)) { /* Recovery target is not given so exit */ - elog(INFO, "Backup %s WAL segments are valid", backup_id); + elog(INFO, "Backup %s WAL segments are valid", backup_id_of(backup)); return; } @@ -359,7 +465,7 @@ validate_wal(pgBackup *backup, const char *archivedir, * If recovery target is provided, ensure that archive files exist in * archive directory. */ - if (dir_is_empty(archivedir, FIO_BACKUP_HOST)) + if (dir_is_empty(archivedir, FIO_LOCAL_HOST)) elog(ERROR, "WAL archive is empty. You cannot restore backup to a recovery target without WAL archive."); /* @@ -373,7 +479,7 @@ validate_wal(pgBackup *backup, const char *archivedir, last_rec.rec_xid = backup->recovery_xid; last_rec.rec_lsn = backup->stop_lsn; - time2iso(last_timestamp, lengthof(last_timestamp), backup->recovery_time); + time2iso(last_timestamp, lengthof(last_timestamp), backup->recovery_time, false); if ((TransactionIdIsValid(target_xid) && target_xid == last_rec.rec_xid) || (target_time != 0 && backup->recovery_time >= target_time) @@ -383,10 +489,10 @@ validate_wal(pgBackup *backup, const char *archivedir, all_wal = all_wal || RunXLogThreads(archivedir, target_time, target_xid, target_lsn, tli, wal_seg_size, backup->stop_lsn, - InvalidXLogRecPtr, true, validateXLogRecord, &last_rec); + InvalidXLogRecPtr, true, validateXLogRecord, &last_rec, true); if (last_rec.rec_time > 0) time2iso(last_timestamp, lengthof(last_timestamp), - timestamptz_to_time_t(last_rec.rec_time)); + timestamptz_to_time_t(last_rec.rec_time), false); /* There are all needed WAL records */ if (all_wal) @@ -401,7 +507,7 @@ validate_wal(pgBackup *backup, const char *archivedir, (uint32) (last_rec.rec_lsn >> 32), (uint32) last_rec.rec_lsn); if (target_time > 0) - time2iso(target_timestamp, lengthof(target_timestamp), target_time); + time2iso(target_timestamp, lengthof(target_timestamp), target_time, false); if (TransactionIdIsValid(target_xid) && target_time != 0) elog(ERROR, "Not enough WAL records to time %s and xid " XID_FMT, target_timestamp, target_xid); @@ -425,7 +531,7 @@ validate_wal(pgBackup *backup, const char *archivedir, bool read_recovery_info(const char *archivedir, TimeLineID tli, uint32 wal_seg_size, XLogRecPtr start_lsn, XLogRecPtr stop_lsn, - time_t *recovery_time, TransactionId *recovery_xid) + time_t *recovery_time) { XLogRecPtr startpoint = stop_lsn; XLogReaderState *xlogreader; @@ -450,7 +556,13 @@ read_recovery_info(const char *archivedir, TimeLineID tli, uint32 wal_seg_size, TimestampTz last_time = 0; char *errormsg; - record = XLogReadRecord(xlogreader, startpoint, &errormsg); +#if PG_VERSION_NUM >= 130000 + if (XLogRecPtrIsInvalid(startpoint)) + startpoint = SizeOfXLogShortPHD; + XLogBeginRead(xlogreader, startpoint); +#endif + + record = WalReadRecord(xlogreader, startpoint, &errormsg); if (record == NULL) { XLogRecPtr errptr; @@ -472,7 +584,6 @@ read_recovery_info(const char *archivedir, TimeLineID tli, uint32 wal_seg_size, if (getRecordTimestamp(xlogreader, &last_time)) { *recovery_time = timestamptz_to_time_t(last_time); - *recovery_xid = XLogRecGetXid(xlogreader); /* Found timestamp in WAL record 'record' */ res = true; @@ -515,7 +626,13 @@ wal_contains_lsn(const char *archivedir, XLogRecPtr target_lsn, xlogreader->system_identifier = instance_config.system_identifier; - res = XLogReadRecord(xlogreader, target_lsn, &errormsg) != NULL; +#if PG_VERSION_NUM >= 130000 + if (XLogRecPtrIsInvalid(target_lsn)) + target_lsn = SizeOfXLogShortPHD; + XLogBeginRead(xlogreader, target_lsn); +#endif + + res = WalReadRecord(xlogreader, target_lsn, &errormsg) != NULL; /* Didn't find 'target_lsn' and there is no error, return false */ if (errormsg) @@ -529,16 +646,174 @@ wal_contains_lsn(const char *archivedir, XLogRecPtr target_lsn, } /* - * Get LSN of last or prior record within the WAL segment with number 'segno'. - * If 'start_lsn' - * is in the segment with number 'segno' then start from 'start_lsn', otherwise - * start from offset 0 within the segment. + * Get LSN of a first record within the WAL segment with number 'segno'. + */ +XLogRecPtr +get_first_record_lsn(const char *archivedir, XLogSegNo segno, + TimeLineID tli, uint32 wal_seg_size, int timeout) +{ + XLogReaderState *xlogreader; + XLogReaderData reader_data; + XLogRecPtr record = InvalidXLogRecPtr; + XLogRecPtr startpoint; + char wal_segment[MAXFNAMELEN]; + int attempts = 0; + + if (segno <= 1) + elog(ERROR, "Invalid WAL segment number " UINT64_FORMAT, segno); + + GetXLogFileName(wal_segment, tli, segno, instance_config.xlog_seg_size); + + xlogreader = InitXLogPageRead(&reader_data, archivedir, tli, wal_seg_size, + false, false, true); + if (xlogreader == NULL) + elog(ERROR, "Out of memory"); + xlogreader->system_identifier = instance_config.system_identifier; + + /* Set startpoint to 0 in segno */ + GetXLogRecPtr(segno, 0, wal_seg_size, startpoint); + +#if PG_VERSION_NUM >= 130000 + if (XLogRecPtrIsInvalid(startpoint)) + startpoint = SizeOfXLogShortPHD; + XLogBeginRead(xlogreader, startpoint); +#endif + + while (attempts <= timeout) + { + record = XLogFindNextRecord(xlogreader, startpoint); + + if (XLogRecPtrIsInvalid(record)) + record = InvalidXLogRecPtr; + else + { + elog(LOG, "First record in WAL segment \"%s\": %X/%X", wal_segment, + (uint32) (record >> 32), (uint32) (record)); + break; + } + + attempts++; + sleep(1); + } + + /* cleanup */ + CleanupXLogPageRead(xlogreader); + XLogReaderFree(xlogreader); + + return record; +} + + +/* + * Get LSN of the record next after target lsn. + */ +XLogRecPtr +get_next_record_lsn(const char *archivedir, XLogSegNo segno, + TimeLineID tli, uint32 wal_seg_size, int timeout, + XLogRecPtr target) +{ + XLogReaderState *xlogreader; + XLogReaderData reader_data; + XLogRecPtr startpoint, found; + XLogRecPtr res = InvalidXLogRecPtr; + char wal_segment[MAXFNAMELEN]; + int attempts = 0; + + if (segno <= 1) + elog(ERROR, "Invalid WAL segment number " UINT64_FORMAT, segno); + + GetXLogFileName(wal_segment, tli, segno, instance_config.xlog_seg_size); + + xlogreader = InitXLogPageRead(&reader_data, archivedir, tli, wal_seg_size, + false, false, true); + if (xlogreader == NULL) + elog(ERROR, "Out of memory"); + xlogreader->system_identifier = instance_config.system_identifier; + + /* Set startpoint to 0 in segno */ + GetXLogRecPtr(segno, 0, wal_seg_size, startpoint); + +#if PG_VERSION_NUM >= 130000 + if (XLogRecPtrIsInvalid(startpoint)) + startpoint = SizeOfXLogShortPHD; + XLogBeginRead(xlogreader, startpoint); +#endif + + found = XLogFindNextRecord(xlogreader, startpoint); + + if (XLogRecPtrIsInvalid(found)) + { + if (xlogreader->errormsg_buf[0] != '\0') + elog(WARNING, "Could not read WAL record at %X/%X: %s", + (uint32) (startpoint >> 32), (uint32) (startpoint), + xlogreader->errormsg_buf); + else + elog(WARNING, "Could not read WAL record at %X/%X", + (uint32) (startpoint >> 32), (uint32) (startpoint)); + PrintXLogCorruptionMsg(&reader_data, ERROR); + } + startpoint = found; + + while (attempts <= timeout) + { + XLogRecord *record; + char *errormsg; + + if (interrupted) + elog(ERROR, "Interrupted during WAL reading"); + + record = WalReadRecord(xlogreader, startpoint, &errormsg); + + if (record == NULL) + { + XLogRecPtr errptr; + + errptr = XLogRecPtrIsInvalid(startpoint) ? xlogreader->EndRecPtr : + startpoint; + + if (errormsg) + elog(WARNING, "Could not read WAL record at %X/%X: %s", + (uint32) (errptr >> 32), (uint32) (errptr), + errormsg); + else + elog(WARNING, "Could not read WAL record at %X/%X", + (uint32) (errptr >> 32), (uint32) (errptr)); + PrintXLogCorruptionMsg(&reader_data, ERROR); + } + + if (xlogreader->ReadRecPtr >= target) + { + elog(LOG, "Record %X/%X is next after target LSN %X/%X", + (uint32) (xlogreader->ReadRecPtr >> 32), (uint32) (xlogreader->ReadRecPtr), + (uint32) (target >> 32), (uint32) (target)); + res = xlogreader->ReadRecPtr; + break; + } + else + startpoint = InvalidXLogRecPtr; + } + + /* cleanup */ + CleanupXLogPageRead(xlogreader); + XLogReaderFree(xlogreader); + + return res; +} + + +/* + * Get LSN of a record prior to target_lsn. + * If 'start_lsn' is in the segment with number 'segno' then start from 'start_lsn', + * otherwise start from offset 0 within the segment. + * + * Returns LSN of a record which EndRecPtr is greater or equal to target_lsn. + * If 'seek_prev_segment' is true, then look for prior record in prior WAL segment. * - * Returns LSN which points to end+1 of the last WAL record if seek_prev_segment - * is true. Otherwise returns LSN of the record prior to stop_lsn. + * it's unclear that "last" in "last_wal_lsn" refers to the + * "closest to stop_lsn backward or forward, depending on seek_prev_segment setting". */ XLogRecPtr -get_last_wal_lsn(const char *archivedir, XLogRecPtr start_lsn, +get_prior_record_lsn(const char *archivedir, XLogRecPtr start_lsn, XLogRecPtr stop_lsn, TimeLineID tli, bool seek_prev_segment, uint32 wal_seg_size) { @@ -570,12 +845,26 @@ get_last_wal_lsn(const char *archivedir, XLogRecPtr start_lsn, */ GetXLogSegNo(start_lsn, start_segno, wal_seg_size); if (start_segno == segno) + { startpoint = start_lsn; +#if PG_VERSION_NUM >= 130000 + if (XLogRecPtrIsInvalid(startpoint)) + startpoint = SizeOfXLogShortPHD; + XLogBeginRead(xlogreader, startpoint); +#endif + } else { XLogRecPtr found; GetXLogRecPtr(segno, 0, wal_seg_size, startpoint); + +#if PG_VERSION_NUM >= 130000 + if (XLogRecPtrIsInvalid(startpoint)) + startpoint = SizeOfXLogShortPHD; + XLogBeginRead(xlogreader, startpoint); +#endif + found = XLogFindNextRecord(xlogreader, startpoint); if (XLogRecPtrIsInvalid(found)) @@ -596,12 +885,11 @@ get_last_wal_lsn(const char *archivedir, XLogRecPtr start_lsn, { XLogRecord *record; char *errormsg; - XLogSegNo next_segno = 0; if (interrupted) elog(ERROR, "Interrupted during WAL reading"); - record = XLogReadRecord(xlogreader, startpoint, &errormsg); + record = WalReadRecord(xlogreader, startpoint, &errormsg); if (record == NULL) { XLogRecPtr errptr; @@ -619,23 +907,18 @@ get_last_wal_lsn(const char *archivedir, XLogRecPtr start_lsn, PrintXLogCorruptionMsg(&reader_data, ERROR); } - /* continue reading at next record */ - startpoint = InvalidXLogRecPtr; - - GetXLogSegNo(xlogreader->EndRecPtr, next_segno, wal_seg_size); - if (next_segno > segno) - break; - - if (seek_prev_segment) + if (xlogreader->EndRecPtr >= stop_lsn) { - /* end+1 of last record read */ - res = xlogreader->EndRecPtr; - } - else + elog(LOG, "Record %X/%X has endpoint %X/%X which is equal or greater than requested LSN %X/%X", + (uint32) (xlogreader->ReadRecPtr >> 32), (uint32) (xlogreader->ReadRecPtr), + (uint32) (xlogreader->EndRecPtr >> 32), (uint32) (xlogreader->EndRecPtr), + (uint32) (stop_lsn >> 32), (uint32) (stop_lsn)); res = xlogreader->ReadRecPtr; - - if (xlogreader->EndRecPtr >= stop_lsn) break; + } + + /* continue reading at next record */ + startpoint = InvalidXLogRecPtr; } CleanupXLogPageRead(xlogreader); @@ -665,8 +948,11 @@ get_gz_error(gzFile gzf) /* XLogreader callback function, to read a WAL page */ static int SimpleXLogPageRead(XLogReaderState *xlogreader, XLogRecPtr targetPagePtr, - int reqLen, XLogRecPtr targetRecPtr, char *readBuf, - TimeLineID *pageTLI) + int reqLen, XLogRecPtr targetRecPtr, char *readBuf +#if PG_VERSION_NUM < 130000 + ,TimeLineID *pageTLI +#endif + ) { XLogReaderData *reader_data; uint32 targetPageOff; @@ -726,20 +1012,35 @@ SimpleXLogPageRead(XLogReaderState *xlogreader, XLogRecPtr targetPagePtr, if (!reader_data->xlogexists) { char xlogfname[MAXFNAMELEN]; + char partial_file[MAXPGPATH]; - GetXLogFileName(xlogfname, reader_data->tli, reader_data->xlogsegno, - wal_seg_size); - snprintf(reader_data->xlogpath, MAXPGPATH, "%s/%s", wal_archivedir, - xlogfname); + GetXLogFileName(xlogfname, reader_data->tli, reader_data->xlogsegno, wal_seg_size); - if (fileExists(reader_data->xlogpath, FIO_BACKUP_HOST)) + join_path_components(reader_data->xlogpath, wal_archivedir, xlogfname); + snprintf(reader_data->gz_xlogpath, MAXPGPATH, "%s.gz", reader_data->xlogpath); + + /* We fall back to using .partial segment in case if we are running + * multi-timeline incremental backup right after standby promotion. + * TODO: it should be explicitly enabled. + */ + snprintf(partial_file, MAXPGPATH, "%s.partial", reader_data->xlogpath); + + /* If segment do not exists, but the same + * segment with '.partial' suffix does, use it instead */ + if (!fileExists(reader_data->xlogpath, FIO_LOCAL_HOST) && + fileExists(partial_file, FIO_LOCAL_HOST)) + { + snprintf(reader_data->xlogpath, MAXPGPATH, "%s", partial_file); + } + + if (fileExists(reader_data->xlogpath, FIO_LOCAL_HOST)) { elog(LOG, "Thread [%d]: Opening WAL segment \"%s\"", reader_data->thread_num, reader_data->xlogpath); reader_data->xlogexists = true; reader_data->xlogfile = fio_open(reader_data->xlogpath, - O_RDONLY | PG_BINARY, FIO_BACKUP_HOST); + O_RDONLY | PG_BINARY, FIO_LOCAL_HOST); if (reader_data->xlogfile < 0) { @@ -751,29 +1052,23 @@ SimpleXLogPageRead(XLogReaderState *xlogreader, XLogRecPtr targetPagePtr, } #ifdef HAVE_LIBZ /* Try to open compressed WAL segment */ - else + else if (fileExists(reader_data->gz_xlogpath, FIO_LOCAL_HOST)) { - snprintf(reader_data->gz_xlogpath, sizeof(reader_data->gz_xlogpath), - "%s.gz", reader_data->xlogpath); - if (fileExists(reader_data->gz_xlogpath, FIO_BACKUP_HOST)) - { - elog(LOG, "Thread [%d]: Opening compressed WAL segment \"%s\"", - reader_data->thread_num, reader_data->gz_xlogpath); + elog(LOG, "Thread [%d]: Opening compressed WAL segment \"%s\"", + reader_data->thread_num, reader_data->gz_xlogpath); - reader_data->xlogexists = true; - reader_data->gz_xlogfile = fio_gzopen(reader_data->gz_xlogpath, - "rb", -1, FIO_BACKUP_HOST); - if (reader_data->gz_xlogfile == NULL) - { - elog(WARNING, "Thread [%d]: Could not open compressed WAL segment \"%s\": %s", - reader_data->thread_num, reader_data->gz_xlogpath, - strerror(errno)); - return -1; - } + reader_data->xlogexists = true; + reader_data->gz_xlogfile = fio_gzopen(reader_data->gz_xlogpath, + "rb", -1, FIO_LOCAL_HOST); + if (reader_data->gz_xlogfile == NULL) + { + elog(WARNING, "Thread [%d]: Could not open compressed WAL segment \"%s\": %s", + reader_data->thread_num, reader_data->gz_xlogpath, + strerror(errno)); + return -1; } } #endif - /* Exit without error if WAL segment doesn't exist */ if (!reader_data->xlogexists) return -1; @@ -791,7 +1086,9 @@ SimpleXLogPageRead(XLogReaderState *xlogreader, XLogRecPtr targetPagePtr, reader_data->prev_page_off == targetPageOff) { memcpy(readBuf, reader_data->page_buf, XLOG_BLCKSZ); +#if PG_VERSION_NUM < 130000 *pageTLI = reader_data->tli; +#endif return XLOG_BLCKSZ; } @@ -835,7 +1132,9 @@ SimpleXLogPageRead(XLogReaderState *xlogreader, XLogRecPtr targetPagePtr, memcpy(reader_data->page_buf, readBuf, XLOG_BLCKSZ); reader_data->prev_page_off = targetPageOff; +#if PG_VERSION_NUM < 130000 *pageTLI = reader_data->tli; +#endif return XLOG_BLCKSZ; } @@ -860,12 +1159,7 @@ InitXLogPageRead(XLogReaderData *reader_data, const char *archivedir, if (allocate_reader) { -#if PG_VERSION_NUM >= 110000 - xlogreader = XLogReaderAllocate(wal_seg_size, &SimpleXLogPageRead, - reader_data); -#else - xlogreader = XLogReaderAllocate(&SimpleXLogPageRead, reader_data); -#endif + xlogreader = WalReaderAllocate(wal_seg_size, reader_data); if (xlogreader == NULL) elog(ERROR, "Out of memory"); xlogreader->system_identifier = instance_config.system_identifier; @@ -896,7 +1190,7 @@ RunXLogThreads(const char *archivedir, time_t target_time, TransactionId target_xid, XLogRecPtr target_lsn, TimeLineID tli, uint32 segment_size, XLogRecPtr startpoint, XLogRecPtr endpoint, bool consistent_read, xlog_record_function process_record, - XLogRecTarget *last_rec) + XLogRecTarget *last_rec, bool inclusive_endpoint) { pthread_t *threads; xlog_thread_arg *thread_args; @@ -905,17 +1199,31 @@ RunXLogThreads(const char *archivedir, time_t target_time, XLogSegNo endSegNo = 0; bool result = true; - if (!XRecOffIsValid(startpoint)) + if (!XRecOffIsValid(startpoint) && !XRecOffIsNull(startpoint)) elog(ERROR, "Invalid startpoint value %X/%X", (uint32) (startpoint >> 32), (uint32) (startpoint)); + if (process_record) + elog(LOG, "Extracting pagemap from tli %i on range from %X/%X to %X/%X", + tli, + (uint32) (startpoint >> 32), (uint32) (startpoint), + (uint32) (endpoint >> 32), (uint32) (endpoint)); + if (!XLogRecPtrIsInvalid(endpoint)) { - if (!XRecOffIsValid(endpoint)) +// if (XRecOffIsNull(endpoint) && !inclusive_endpoint) + if (XRecOffIsNull(endpoint)) + { + GetXLogSegNo(endpoint, endSegNo, segment_size); + endSegNo--; + } + else if (!XRecOffIsValid(endpoint)) + { elog(ERROR, "Invalid endpoint value %X/%X", (uint32) (endpoint >> 32), (uint32) (endpoint)); - - GetXLogSegNo(endpoint, endSegNo, segment_size); + } + else + GetXLogSegNo(endpoint, endSegNo, segment_size); } /* Initialize static variables for workers */ @@ -950,6 +1258,7 @@ RunXLogThreads(const char *archivedir, time_t target_time, arg->startpoint = startpoint; arg->endpoint = endpoint; arg->endSegNo = endSegNo; + arg->inclusive_endpoint = inclusive_endpoint; arg->got_target = false; /* By default there is some error */ arg->ret = 1; @@ -980,6 +1289,11 @@ RunXLogThreads(const char *archivedir, time_t target_time, if (thread_args[i].ret == 1) result = false; } + thread_interrupted = false; + +// TODO: we must detect difference between actual error (failed to read WAL) and interrupt signal +// if (interrupted) +// elog(ERROR, "Interrupted during WAL parsing"); /* Release threads here, use thread_args only below */ pfree(threads); @@ -1050,16 +1364,18 @@ XLogThreadWorker(void *arg) uint32 prev_page_off = 0; bool need_read = true; -#if PG_VERSION_NUM >= 110000 - xlogreader = XLogReaderAllocate(wal_seg_size, &SimpleXLogPageRead, - reader_data); -#else - xlogreader = XLogReaderAllocate(&SimpleXLogPageRead, reader_data); -#endif + xlogreader = WalReaderAllocate(wal_seg_size, reader_data); + if (xlogreader == NULL) elog(ERROR, "Thread [%d]: out of memory", reader_data->thread_num); xlogreader->system_identifier = instance_config.system_identifier; +#if PG_VERSION_NUM >= 130000 + if (XLogRecPtrIsInvalid(thread_arg->startpoint)) + thread_arg->startpoint = SizeOfXLogShortPHD; + XLogBeginRead(xlogreader, thread_arg->startpoint); +#endif + found = XLogFindNextRecord(xlogreader, thread_arg->startpoint); /* @@ -1112,7 +1428,7 @@ XLogThreadWorker(void *arg) !SwitchThreadToNextWal(xlogreader, thread_arg)) break; - record = XLogReadRecord(xlogreader, thread_arg->startpoint, &errormsg); + record = WalReadRecord(xlogreader, thread_arg->startpoint, &errormsg); if (record == NULL) { @@ -1123,7 +1439,18 @@ XLogThreadWorker(void *arg) * Usually SimpleXLogPageRead() does it by itself. But here we need * to do it manually to support threads. */ +#if PG_VERSION_NUM >= 150000 + if (reader_data->need_switch && ( + errormsg == NULL || + /* + * Pg15 now informs if "contrecord" is missing. + * TODO: probably we should abort reading logs at this moment. + * But we continue as we did with bug present in Pg < 15. + */ + !XLogRecPtrIsInvalid(xlogreader->abortedRecPtr))) +#else if (reader_data->need_switch && errormsg == NULL) +#endif { if (SwitchThreadToNextWal(xlogreader, thread_arg)) continue; @@ -1165,6 +1492,18 @@ XLogThreadWorker(void *arg) reader_data->thread_num, (uint32) (errptr >> 32), (uint32) (errptr)); + /* In we failed to read record located at endpoint position, + * and endpoint is not inclusive, do not consider this as an error. + */ + if (!thread_arg->inclusive_endpoint && + errptr == thread_arg->endpoint) + { + elog(LOG, "Thread [%d]: Endpoint %X/%X is not inclusive, switch to the next timeline", + reader_data->thread_num, + (uint32) (thread_arg->endpoint >> 32), (uint32) (thread_arg->endpoint)); + break; + } + /* * If we don't have all WAL files from prev backup start_lsn to current * start_lsn, we won't be able to build page map and PAGE backup will @@ -1249,9 +1588,14 @@ SwitchThreadToNextWal(XLogReaderState *xlogreader, xlog_thread_arg *arg) reader_data = (XLogReaderData *) xlogreader->private_data; reader_data->need_switch = false; +start: /* Critical section */ pthread_lock(&wal_segment_mutex); Assert(segno_next); + + if (reader_data->xlogsegno > segno_next) + segno_next = reader_data->xlogsegno; + reader_data->xlogsegno = segno_next; segnum_read++; segno_next++; @@ -1265,6 +1609,7 @@ SwitchThreadToNextWal(XLogReaderState *xlogreader, xlog_thread_arg *arg) GetXLogRecPtr(reader_data->xlogsegno, 0, wal_seg_size, arg->startpoint); /* We need to close previously opened file if it wasn't closed earlier */ CleanupXLogPageRead(xlogreader); + xlogreader->currRecPtr = InvalidXLogRecPtr; /* Skip over the page header and contrecord if any */ found = XLogFindNextRecord(xlogreader, arg->startpoint); @@ -1274,6 +1619,8 @@ SwitchThreadToNextWal(XLogReaderState *xlogreader, xlog_thread_arg *arg) */ if (XLogRecPtrIsInvalid(found)) { + if (reader_data->need_switch) + goto start; /* * Check if we need to stop reading. We stop if other thread found a * target segment. @@ -1440,7 +1787,12 @@ extractPageInfo(XLogReaderState *record, XLogReaderData *reader_data, /* Is this a special record type that I recognize? */ - if (rmid == RM_DBASE_ID && rminfo == XLOG_DBASE_CREATE) + if (rmid == RM_DBASE_ID +#if PG_VERSION_NUM >= 150000 + && (rminfo == XLOG_DBASE_CREATE_WAL_LOG || rminfo == XLOG_DBASE_CREATE_FILE_COPY)) +#else + && rminfo == XLOG_DBASE_CREATE) +#endif { /* * New databases can be safely ignored. They would be completely @@ -1469,6 +1821,18 @@ extractPageInfo(XLogReaderState *record, XLogReaderData *reader_data, * source system. */ } + else if (rmid == RM_XACT_ID && + ((rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT || + (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_COMMIT_PREPARED || + (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_ABORT || + (rminfo & XLOG_XACT_OPMASK) == XLOG_XACT_ABORT_PREPARED)) + { + /* + * These records can include "dropped rels". We can safely ignore + * them, we will see that they are missing and copy them from the + * source. + */ + } else if (info & XLR_SPECIAL_REL_UPDATE) { /* @@ -1482,13 +1846,21 @@ extractPageInfo(XLogReaderState *record, XLogReaderData *reader_data, RmgrNames[rmid], info); } +#if PG_VERSION_NUM >= 150000 + for (block_id = 0; block_id <= record->record->max_block_id; block_id++) +#else for (block_id = 0; block_id <= record->max_block_id; block_id++) +#endif { RelFileNode rnode; ForkNumber forknum; BlockNumber blkno; +#if PG_VERSION_NUM >= 150000 + if (!XLogRecGetBlockTagExtended(record, block_id, &rnode, &forknum, &blkno, NULL)) +#else if (!XLogRecGetBlockTag(record, block_id, &rnode, &forknum, &blkno)) +#endif continue; /* We only care about the main fork; others are copied as is */ @@ -1556,3 +1928,53 @@ getRecordTimestamp(XLogReaderState *record, TimestampTz *recordXtime) return false; } +bool validate_wal_segment(TimeLineID tli, XLogSegNo segno, const char *prefetch_dir, uint32 wal_seg_size) +{ + XLogRecPtr startpoint; + XLogRecPtr endpoint; + + bool rc; + int tmp_num_threads = num_threads; + num_threads = 1; + + /* calculate startpoint and endpoint */ + GetXLogRecPtr(segno, 0, wal_seg_size, startpoint); + GetXLogRecPtr(segno+1, 0, wal_seg_size, endpoint); + + /* disable multi-threading */ + num_threads = 1; + + rc = RunXLogThreads(prefetch_dir, 0, InvalidTransactionId, + InvalidXLogRecPtr, tli, wal_seg_size, + startpoint, endpoint, false, NULL, NULL, true); + + num_threads = tmp_num_threads; + + return rc; +} + +static XLogRecord* WalReadRecord(XLogReaderState *xlogreader, XLogRecPtr startpoint, char **errormsg) +{ + +#if PG_VERSION_NUM >= 130000 + return XLogReadRecord(xlogreader, errormsg); +#else + return XLogReadRecord(xlogreader, startpoint, errormsg); +#endif + +} + +static XLogReaderState* WalReaderAllocate(uint32 wal_seg_size, XLogReaderData *reader_data) +{ + +#if PG_VERSION_NUM >= 130000 + return XLogReaderAllocate(wal_seg_size, NULL, + XL_ROUTINE(.page_read = &SimpleXLogPageRead), + reader_data); +#elif PG_VERSION_NUM >= 110000 + return XLogReaderAllocate(wal_seg_size, &SimpleXLogPageRead, + reader_data); +#else + return XLogReaderAllocate(&SimpleXLogPageRead, reader_data); +#endif +} diff --git a/src/pg_probackup.c b/src/pg_probackup.c index 915b4ca8c..fa67ddff5 100644 --- a/src/pg_probackup.c +++ b/src/pg_probackup.c @@ -2,13 +2,46 @@ * * pg_probackup.c: Backup/Recovery manager for PostgreSQL. * + * This is an entry point for the program. + * Parse command name and it's options, verify them and call a + * do_***() function that implements the command. + * + * Avoid using global variables in the code. + * Pass all needed information as funciton arguments: + * + + * + * TODO (see pg_probackup_state.h): + * + * Functions that work with a backup catalog accept catalogState, + * which currently only contains pathes to backup catalog subdirectories + * + function specific options. + * + * Functions that work with an instance accept instanceState argument, which + * includes catalogState, instance_name, + * info about pgdata associated with the instance (see pgState), + * various instance config options, and list of backups belonging to the instance. + * + function specific options. + * + * Functions that work with multiple backups in the catalog + * accept instanceState and info needed to determine the range of backups to handle. + * + function specific options. + * + * Functions that work with a single backup accept backupState argument, + * which includes link to the instanceState, backup_id and backup-specific info. + * + function specific options. + * + * Functions that work with a postgreSQL instance (i.e. checkdb) accept pgState, + * which includes info about pgdata directory and connection. + * * Portions Copyright (c) 2009-2013, NIPPON TELEGRAPH AND TELEPHONE CORPORATION - * Portions Copyright (c) 2015-2019, Postgres Professional + * Portions Copyright (c) 2015-2021, Postgres Professional * *------------------------------------------------------------------------- */ #include "pg_probackup.h" +#include "pg_probackup_state.h" #include "pg_getopt.h" #include "streamutil.h" @@ -27,159 +60,203 @@ const char *PROGRAM_FULL_PATH = NULL; const char *PROGRAM_URL = "https://github.com/postgrespro/pg_probackup"; const char *PROGRAM_EMAIL = "https://github.com/postgrespro/pg_probackup/issues"; -typedef enum ProbackupSubcmd -{ - NO_CMD = 0, - INIT_CMD, - ADD_INSTANCE_CMD, - DELETE_INSTANCE_CMD, - ARCHIVE_PUSH_CMD, - ARCHIVE_GET_CMD, - BACKUP_CMD, - RESTORE_CMD, - VALIDATE_CMD, - DELETE_CMD, - MERGE_CMD, - SHOW_CMD, - SET_CONFIG_CMD, - SHOW_CONFIG_CMD, - CHECKDB_CMD -} ProbackupSubcmd; - - +/* ================ catalogState =========== */ /* directory options */ -char *backup_path = NULL; -/* - * path or to the data files in the backup catalog - * $BACKUP_PATH/backups/instance_name - */ -char backup_instance_path[MAXPGPATH]; -/* - * path or to the wal files in the backup catalog - * $BACKUP_PATH/wal/instance_name - */ -char arclog_path[MAXPGPATH] = ""; +/* TODO make it local variable, pass as an argument to all commands that need it. */ +static char *backup_path = NULL; + +static CatalogState *catalogState = NULL; +/* ================ catalogState (END) =========== */ -/* colon separated external directories list ("/path1:/path2") */ -char *externaldir = NULL; /* common options */ -static char *backup_id_string = NULL; int num_threads = 1; bool stream_wal = false; +bool no_color = false; +bool show_color = true; +bool is_archive_cmd = false; +pid_t my_pid = 0; +__thread int my_thread_num = 1; bool progress = false; +bool no_sync = false; +time_t start_time = INVALID_BACKUP_ID; #if PG_VERSION_NUM >= 100000 char *replication_slot = NULL; -#endif bool temp_slot = false; +#endif +bool perm_slot = false; /* backup options */ -bool backup_logs = false; -bool smooth_checkpoint; -char *remote_agent; - +bool backup_logs = false; +bool smooth_checkpoint; +bool remote_agent = false; +static char *backup_note = NULL; +/* catchup options */ +static char *catchup_source_pgdata = NULL; +static char *catchup_destination_pgdata = NULL; /* restore options */ static char *target_time = NULL; static char *target_xid = NULL; static char *target_lsn = NULL; static char *target_inclusive = NULL; -static TimeLineID target_tli; +static char *target_tli_string; /* timeline number, "current" or "latest"*/ static char *target_stop; static bool target_immediate; static char *target_name = NULL; static char *target_action = NULL; +static char *primary_conninfo = NULL; + static pgRecoveryTarget *recovery_target_options = NULL; +static pgRestoreParams *restore_params = NULL; -bool restore_as_replica = false; +time_t current_time = 0; +static bool restore_as_replica = false; bool no_validate = false; +IncrRestoreMode incremental_mode = INCR_NONE; bool skip_block_validation = false; bool skip_external_dirs = false; +/* array for datnames, provided via db-include and db-exclude */ +static parray *datname_exclude_list = NULL; +static parray *datname_include_list = NULL; +/* arrays for --exclude-path's */ +static parray *exclude_absolute_paths_list = NULL; +static parray *exclude_relative_paths_list = NULL; +static char* gl_waldir_path = NULL; +static bool allow_partial_incremental = false; + /* checkdb options */ bool need_amcheck = false; bool heapallindexed = false; +bool checkunique = false; bool amcheck_parent = false; /* delete options */ bool delete_wal = false; bool delete_expired = false; bool merge_expired = false; -bool force_delete = false; +bool force = false; bool dry_run = false; - +static char *delete_status = NULL; /* compression options */ -bool compress_shortcut = false; +static bool compress_shortcut = false; + +/* ================ instanceState =========== */ +static char *instance_name; -/* other options */ -char *instance_name; +static InstanceState *instanceState = NULL; + +/* ================ instanceState (END) =========== */ /* archive push options */ +int batch_size = 1; static char *wal_file_path; static char *wal_file_name; -static bool file_overwrite = false; +static bool file_overwrite = false; +static bool no_ready_rename = false; +static char archive_push_xlog_dir[MAXPGPATH] = ""; + +/* archive get options */ +static char *prefetch_dir; +bool no_validate_wal = false; /* show options */ ShowFormat show_format = SHOW_PLAIN; +bool show_archive = false; +static bool show_base_units = false; + +/* set-backup options */ +int64 ttl = -1; +static char *expire_time_string = NULL; +static pgSetBackupParams *set_backup_params = NULL; -/* current settings */ +/* ================ backupState =========== */ +static char *backup_id_string = NULL; pgBackup current; -static ProbackupSubcmd backup_subcmd = NO_CMD; +/* ================ backupState (END) =========== */ static bool help_opt = false; +static void opt_incr_restore_mode(ConfigOption *opt, const char *arg); static void opt_backup_mode(ConfigOption *opt, const char *arg); static void opt_show_format(ConfigOption *opt, const char *arg); -static void compress_init(void); +static void compress_init(ProbackupSubcmd const subcmd); + +static void opt_datname_exclude_list(ConfigOption *opt, const char *arg); +static void opt_datname_include_list(ConfigOption *opt, const char *arg); +static void opt_exclude_path(ConfigOption *opt, const char *arg); /* * Short name should be non-printable ASCII character. + * Use values between 128 and 255. */ static ConfigOption cmd_options[] = { /* directory options */ - { 'b', 130, "help", &help_opt, SOURCE_CMD_STRICT }, + { 'b', 130, "help", &help_opt, SOURCE_CMD_STRICT }, { 's', 'B', "backup-path", &backup_path, SOURCE_CMD_STRICT }, /* common options */ { 'u', 'j', "threads", &num_threads, SOURCE_CMD_STRICT }, { 'b', 131, "stream", &stream_wal, SOURCE_CMD_STRICT }, { 'b', 132, "progress", &progress, SOURCE_CMD_STRICT }, { 's', 'i', "backup-id", &backup_id_string, SOURCE_CMD_STRICT }, + { 'b', 133, "no-sync", &no_sync, SOURCE_CMD_STRICT }, + { 'b', 134, "no-color", &no_color, SOURCE_CMD_STRICT }, /* backup options */ - { 'b', 133, "backup-pg-log", &backup_logs, SOURCE_CMD_STRICT }, + { 'b', 180, "backup-pg-log", &backup_logs, SOURCE_CMD_STRICT }, { 'f', 'b', "backup-mode", opt_backup_mode, SOURCE_CMD_STRICT }, { 'b', 'C', "smooth-checkpoint", &smooth_checkpoint, SOURCE_CMD_STRICT }, { 's', 'S', "slot", &replication_slot, SOURCE_CMD_STRICT }, - { 'b', 234, "temp-slot", &temp_slot, SOURCE_CMD_STRICT }, - { 'b', 134, "delete-wal", &delete_wal, SOURCE_CMD_STRICT }, - { 'b', 135, "delete-expired", &delete_expired, SOURCE_CMD_STRICT }, - { 'b', 235, "merge-expired", &merge_expired, SOURCE_CMD_STRICT }, - { 'b', 237, "dry-run", &dry_run, SOURCE_CMD_STRICT }, +#if PG_VERSION_NUM >= 100000 + { 'b', 181, "temp-slot", &temp_slot, SOURCE_CMD_STRICT }, +#endif + { 'b', 'P', "perm-slot", &perm_slot, SOURCE_CMD_STRICT }, + { 'b', 182, "delete-wal", &delete_wal, SOURCE_CMD_STRICT }, + { 'b', 183, "delete-expired", &delete_expired, SOURCE_CMD_STRICT }, + { 'b', 184, "merge-expired", &merge_expired, SOURCE_CMD_STRICT }, + { 'b', 185, "dry-run", &dry_run, SOURCE_CMD_STRICT }, + { 's', 238, "note", &backup_note, SOURCE_CMD_STRICT }, + { 'U', 241, "start-time", &start_time, SOURCE_CMD_STRICT }, + /* catchup options */ + { 's', 239, "source-pgdata", &catchup_source_pgdata, SOURCE_CMD_STRICT }, + { 's', 240, "destination-pgdata", &catchup_destination_pgdata, SOURCE_CMD_STRICT }, + { 'f', 'x', "exclude-path", opt_exclude_path, SOURCE_CMD_STRICT }, /* restore options */ { 's', 136, "recovery-target-time", &target_time, SOURCE_CMD_STRICT }, { 's', 137, "recovery-target-xid", &target_xid, SOURCE_CMD_STRICT }, { 's', 144, "recovery-target-lsn", &target_lsn, SOURCE_CMD_STRICT }, { 's', 138, "recovery-target-inclusive", &target_inclusive, SOURCE_CMD_STRICT }, - { 'u', 139, "recovery-target-timeline", &target_tli, SOURCE_CMD_STRICT }, + { 's', 139, "recovery-target-timeline", &target_tli_string, SOURCE_CMD_STRICT }, { 's', 157, "recovery-target", &target_stop, SOURCE_CMD_STRICT }, { 'f', 'T', "tablespace-mapping", opt_tablespace_map, SOURCE_CMD_STRICT }, { 'f', 155, "external-mapping", opt_externaldir_map, SOURCE_CMD_STRICT }, { 's', 141, "recovery-target-name", &target_name, SOURCE_CMD_STRICT }, { 's', 142, "recovery-target-action", &target_action, SOURCE_CMD_STRICT }, - { 'b', 'R', "restore-as-replica", &restore_as_replica, SOURCE_CMD_STRICT }, { 'b', 143, "no-validate", &no_validate, SOURCE_CMD_STRICT }, { 'b', 154, "skip-block-validation", &skip_block_validation, SOURCE_CMD_STRICT }, { 'b', 156, "skip-external-dirs", &skip_external_dirs, SOURCE_CMD_STRICT }, + { 'f', 158, "db-include", opt_datname_include_list, SOURCE_CMD_STRICT }, + { 'f', 159, "db-exclude", opt_datname_exclude_list, SOURCE_CMD_STRICT }, + { 'b', 'R', "restore-as-replica", &restore_as_replica, SOURCE_CMD_STRICT }, + { 's', 160, "primary-conninfo", &primary_conninfo, SOURCE_CMD_STRICT }, + { 's', 'S', "primary-slot-name",&replication_slot, SOURCE_CMD_STRICT }, + { 'f', 'I', "incremental-mode", opt_incr_restore_mode, SOURCE_CMD_STRICT }, + { 's', 'X', "waldir", &gl_waldir_path, SOURCE_CMD_STRICT }, + { 'b', 242, "destroy-all-other-dbs", &allow_partial_incremental, SOURCE_CMD_STRICT }, /* checkdb options */ { 'b', 195, "amcheck", &need_amcheck, SOURCE_CMD_STRICT }, { 'b', 196, "heapallindexed", &heapallindexed, SOURCE_CMD_STRICT }, + { 'b', 198, "checkunique", &checkunique, SOURCE_CMD_STRICT }, { 'b', 197, "parent", &amcheck_parent, SOURCE_CMD_STRICT }, /* delete options */ { 'b', 145, "wal", &delete_wal, SOURCE_CMD_STRICT }, { 'b', 146, "expired", &delete_expired, SOURCE_CMD_STRICT }, + { 's', 172, "status", &delete_status, SOURCE_CMD_STRICT }, + /* TODO not implemented yet */ - { 'b', 147, "force", &force_delete, SOURCE_CMD_STRICT }, + { 'b', 147, "force", &force, SOURCE_CMD_STRICT }, /* compression options */ { 'b', 148, "compress", &compress_shortcut, SOURCE_CMD_STRICT }, /* connection options */ @@ -191,61 +268,66 @@ static ConfigOption cmd_options[] = { 's', 150, "wal-file-path", &wal_file_path, SOURCE_CMD_STRICT }, { 's', 151, "wal-file-name", &wal_file_name, SOURCE_CMD_STRICT }, { 'b', 152, "overwrite", &file_overwrite, SOURCE_CMD_STRICT }, + { 'b', 153, "no-ready-rename", &no_ready_rename, SOURCE_CMD_STRICT }, + { 'i', 162, "batch-size", &batch_size, SOURCE_CMD_STRICT }, + /* archive-get options */ + { 's', 163, "prefetch-dir", &prefetch_dir, SOURCE_CMD_STRICT }, + { 'b', 164, "no-validate-wal", &no_validate_wal, SOURCE_CMD_STRICT }, /* show options */ - { 'f', 153, "format", opt_show_format, SOURCE_CMD_STRICT }, - - /* options for backward compatibility */ + { 'f', 165, "format", opt_show_format, SOURCE_CMD_STRICT }, + { 'b', 166, "archive", &show_archive, SOURCE_CMD_STRICT }, + /* show-config options */ + { 'b', 167, "no-scale-units", &show_base_units,SOURCE_CMD_STRICT }, + /* set-backup options */ + { 'I', 170, "ttl", &ttl, SOURCE_CMD_STRICT, SOURCE_DEFAULT, 0, OPTION_UNIT_S, option_get_value}, + { 's', 171, "expire-time", &expire_time_string, SOURCE_CMD_STRICT }, + + /* options for backward compatibility + * TODO: remove in 3.0.0 + */ { 's', 136, "time", &target_time, SOURCE_CMD_STRICT }, { 's', 137, "xid", &target_xid, SOURCE_CMD_STRICT }, { 's', 138, "inclusive", &target_inclusive, SOURCE_CMD_STRICT }, - { 'u', 139, "timeline", &target_tli, SOURCE_CMD_STRICT }, + { 's', 139, "timeline", &target_tli_string, SOURCE_CMD_STRICT }, { 's', 144, "lsn", &target_lsn, SOURCE_CMD_STRICT }, { 'b', 140, "immediate", &target_immediate, SOURCE_CMD_STRICT }, { 0 } }; -static void -setMyLocation(void) -{ - -#ifdef WIN32 - if (IsSshProtocol()) - elog(ERROR, "Currently remote operations on Windows are not supported"); -#endif - - MyLocation = IsSshProtocol() - ? (backup_subcmd == ARCHIVE_PUSH_CMD || backup_subcmd == ARCHIVE_GET_CMD) - ? FIO_DB_HOST - : (backup_subcmd == BACKUP_CMD || backup_subcmd == RESTORE_CMD || backup_subcmd == ADD_INSTANCE_CMD) - ? FIO_BACKUP_HOST - : FIO_LOCAL_HOST - : FIO_LOCAL_HOST; -} - /* * Entry point of pg_probackup command. */ int main(int argc, char *argv[]) { - char *command = NULL, - *command_name; - /* Check if backup_path is directory. */ - struct stat stat_buf; - int rc; + char *command = NULL; + ProbackupSubcmd backup_subcmd = NO_CMD; PROGRAM_NAME_FULL = argv[0]; + /* Check terminal presense and initialize ANSI escape codes for Windows */ + init_console(); + /* Initialize current backup */ pgBackupInit(¤t); /* Initialize current instance configuration */ - init_config(&instance_config); + //TODO get git of this global variable craziness + init_config(&instance_config, instance_name); PROGRAM_NAME = get_progname(argv[0]); + set_pglocale_pgservice(argv[0], PG_TEXTDOMAIN("pg_probackup")); PROGRAM_FULL_PATH = palloc0(MAXPGPATH); + // Setting C locale for numeric values in order to impose dot-based floating-point representation + memorize_environment_locale(); + setlocale(LC_NUMERIC, "C"); + + /* Get current time */ + current_time = time(NULL); + + my_pid = getpid(); //set_pglocale_pgservice(argv[0], "pgscripts"); #if PG_VERSION_NUM >= 110000 @@ -264,92 +346,67 @@ main(int argc, char *argv[]) /* Parse subcommands and non-subcommand options */ if (argc > 1) { - if (strcmp(argv[1], "archive-push") == 0) - backup_subcmd = ARCHIVE_PUSH_CMD; - else if (strcmp(argv[1], "archive-get") == 0) - backup_subcmd = ARCHIVE_GET_CMD; - else if (strcmp(argv[1], "add-instance") == 0) - backup_subcmd = ADD_INSTANCE_CMD; - else if (strcmp(argv[1], "del-instance") == 0) - backup_subcmd = DELETE_INSTANCE_CMD; - else if (strcmp(argv[1], "init") == 0) - backup_subcmd = INIT_CMD; - else if (strcmp(argv[1], "backup") == 0) - backup_subcmd = BACKUP_CMD; - else if (strcmp(argv[1], "restore") == 0) - backup_subcmd = RESTORE_CMD; - else if (strcmp(argv[1], "validate") == 0) - backup_subcmd = VALIDATE_CMD; - else if (strcmp(argv[1], "delete") == 0) - backup_subcmd = DELETE_CMD; - else if (strcmp(argv[1], "merge") == 0) - backup_subcmd = MERGE_CMD; - else if (strcmp(argv[1], "show") == 0) - backup_subcmd = SHOW_CMD; - else if (strcmp(argv[1], "set-config") == 0) - backup_subcmd = SET_CONFIG_CMD; - else if (strcmp(argv[1], "show-config") == 0) - backup_subcmd = SHOW_CONFIG_CMD; - else if (strcmp(argv[1], "checkdb") == 0) - backup_subcmd = CHECKDB_CMD; -#ifdef WIN32 - else if (strcmp(argv[1], "ssh") == 0) - launch_ssh(argv); -#endif - else if (strcmp(argv[1], "agent") == 0 && argc > 2) - { - remote_agent = argv[2]; - if (strcmp(remote_agent, PROGRAM_VERSION) != 0) - { - uint32 agent_version = parse_program_version(remote_agent); - elog(agent_version < AGENT_PROTOCOL_VERSION ? ERROR : WARNING, - "Agent version %s doesn't match master pg_probackup version %s", - remote_agent, PROGRAM_VERSION); - } - fio_communicate(STDIN_FILENO, STDOUT_FILENO); - return 0; - } - else if (strcmp(argv[1], "--help") == 0 || - strcmp(argv[1], "-?") == 0 || - strcmp(argv[1], "help") == 0) + backup_subcmd = parse_subcmd(argv[1]); + switch(backup_subcmd) { - if (argc > 2) - help_command(argv[2]); - else - help_pg_probackup(); - } - else if (strcmp(argv[1], "--version") == 0 - || strcmp(argv[1], "version") == 0 - || strcmp(argv[1], "-V") == 0) - { -#ifdef PGPRO_VERSION - fprintf(stderr, "%s %s (Postgres Pro %s %s)\n", - PROGRAM_NAME, PROGRAM_VERSION, - PGPRO_VERSION, PGPRO_EDITION); + case SSH_CMD: +#ifdef WIN32 + launch_ssh(argv); + break; #else - fprintf(stderr, "%s %s (PostgreSQL %s)\n", - PROGRAM_NAME, PROGRAM_VERSION, PG_VERSION); + elog(ERROR, "\"ssh\" command implemented only for Windows"); + break; #endif - exit(0); + case AGENT_CMD: + /* 'No forward compatibility' sanity: + * /old/binary -> ssh execute -> /newer/binary agent version_num + * If we are executed as an agent for older binary, then exit with error + */ + if (argc > 2) + elog(ERROR, "Version mismatch, pg_probackup binary with version '%s' " + "is launched as an agent for pg_probackup binary with version '%s'", + PROGRAM_VERSION, argv[2]); + remote_agent = true; + fio_communicate(STDIN_FILENO, STDOUT_FILENO); + return 0; + case HELP_CMD: + if (argc > 2) + { + /* 'pg_probackup help command' style */ + help_command(parse_subcmd(argv[2])); + exit(0); + } + else + { + help_pg_probackup(); + exit(0); + } + break; + case VERSION_CMD: + help_print_version(); + exit(0); + case NO_CMD: + elog(ERROR, "Unknown subcommand \"%s\"", argv[1]); + default: + /* Silence compiler warnings */ + break; } - else - elog(ERROR, "Unknown subcommand \"%s\"", argv[1]); } - - if (backup_subcmd == NO_CMD) - elog(ERROR, "No subcommand specified"); + else + elog(ERROR, "No subcommand specified. Please run with \"help\" argument to see possible subcommands."); /* * Make command string before getopt_long() will call. It permutes the * content of argv. */ /* TODO why do we do that only for some commands? */ - command_name = pstrdup(argv[1]); if (backup_subcmd == BACKUP_CMD || backup_subcmd == RESTORE_CMD || backup_subcmd == VALIDATE_CMD || backup_subcmd == DELETE_CMD || - backup_subcmd == MERGE_CMD) + backup_subcmd == MERGE_CMD || + backup_subcmd == SET_CONFIG_CMD || + backup_subcmd == SET_BACKUP_CMD) { int i, len = 0, @@ -380,12 +437,34 @@ main(int argc, char *argv[]) /* Parse command line only arguments */ config_get_opt(argc, argv, cmd_options, instance_options); + if (backup_subcmd == SET_CONFIG_CMD) + { + int i; + for (i = 0; i < argc; i++) + { + if (strncmp("--log-format-console", argv[i], strlen("--log-format-console")) == 0) + { + elog(ERROR, "Option 'log-format-console' set only from terminal\n"); + } + } + } + pgut_init(); + if (no_color) + show_color = false; + if (help_opt) - help_command(command_name); + { + /* 'pg_probackup command --help' style */ + help_command(backup_subcmd); + exit(0); + } + + /* set location based on cmdline options only */ + setMyLocation(backup_subcmd); - /* backup_path is required for all pg_probackup commands except help and checkdb */ + /* ===== catalogState ======*/ if (backup_path == NULL) { /* @@ -393,12 +472,8 @@ main(int argc, char *argv[]) * from environment variable */ backup_path = getenv("BACKUP_PATH"); - if (backup_path == NULL && backup_subcmd != CHECKDB_CMD) - elog(ERROR, "required parameter not specified: BACKUP_PATH (-B, --backup-path)"); } - setMyLocation(); - if (backup_path != NULL) { canonicalize_path(backup_path); @@ -407,27 +482,54 @@ main(int argc, char *argv[]) if (!is_absolute_path(backup_path)) elog(ERROR, "-B, --backup-path must be an absolute path"); - /* Ensure that backup_path is a path to a directory */ - rc = stat(backup_path, &stat_buf); - if (rc != -1 && !S_ISDIR(stat_buf.st_mode)) - elog(ERROR, "-B, --backup-path must be a path to directory"); + catalogState = pgut_new(CatalogState); + strncpy(catalogState->catalog_path, backup_path, MAXPGPATH); + join_path_components(catalogState->backup_subdir_path, + catalogState->catalog_path, BACKUPS_DIR); + join_path_components(catalogState->wal_subdir_path, + catalogState->catalog_path, WAL_SUBDIR); } - /* Ensure that backup_path is an absolute path */ - if (backup_path && !is_absolute_path(backup_path)) - elog(ERROR, "-B, --backup-path must be an absolute path"); + /* backup_path is required for all pg_probackup commands except help, version, checkdb and catchup */ + if (backup_path == NULL && + backup_subcmd != CHECKDB_CMD && + backup_subcmd != HELP_CMD && + backup_subcmd != VERSION_CMD && + backup_subcmd != CATCHUP_CMD) + elog(ERROR, + "No backup catalog path specified.\n" + "Please specify it either using environment variable BACKUP_PATH or\n" + "command line option --backup-path (-B)"); + /* ===== catalogState (END) ======*/ + + /* ===== instanceState ======*/ /* * Option --instance is required for all commands except - * init, show, checkdb and validate + * init, show, checkdb, validate and catchup */ if (instance_name == NULL) { if (backup_subcmd != INIT_CMD && backup_subcmd != SHOW_CMD && - backup_subcmd != VALIDATE_CMD && backup_subcmd != CHECKDB_CMD) - elog(ERROR, "required parameter not specified: --instance"); + backup_subcmd != VALIDATE_CMD && backup_subcmd != CHECKDB_CMD && backup_subcmd != CATCHUP_CMD) + elog(ERROR, "Required parameter not specified: --instance"); + } + else + { + instanceState = pgut_new(InstanceState); + instanceState->catalog_state = catalogState; + + strncpy(instanceState->instance_name, instance_name, MAXPGPATH); + join_path_components(instanceState->instance_backup_subdir_path, + catalogState->backup_subdir_path, instanceState->instance_name); + join_path_components(instanceState->instance_wal_subdir_path, + catalogState->wal_subdir_path, instanceState->instance_name); + join_path_components(instanceState->instance_config_path, + instanceState->instance_backup_subdir_path, BACKUP_CATALOG_CONF_FILE); + } + /* ===== instanceState (END) ======*/ /* * If --instance option was passed, construct paths for backup data and @@ -435,20 +537,33 @@ main(int argc, char *argv[]) */ if ((backup_path != NULL) && instance_name) { - sprintf(backup_instance_path, "%s/%s/%s", - backup_path, BACKUPS_DIR, instance_name); - sprintf(arclog_path, "%s/%s/%s", backup_path, "wal", instance_name); - /* * Ensure that requested backup instance exists. - * for all commands except init, which doesn't take this parameter - * and add-instance which creates new instance. + * for all commands except init, which doesn't take this parameter, + * add-instance, which creates new instance, + * and archive-get, which just do not require it at this point */ - if (backup_subcmd != INIT_CMD && backup_subcmd != ADD_INSTANCE_CMD) + if (backup_subcmd != INIT_CMD && backup_subcmd != ADD_INSTANCE_CMD && + backup_subcmd != ARCHIVE_GET_CMD) { - if (fio_access(backup_instance_path, F_OK, FIO_BACKUP_HOST) != 0) + struct stat st; + + if (fio_stat(instanceState->instance_backup_subdir_path, + &st, true, FIO_BACKUP_HOST) != 0) + { + elog(WARNING, "Failed to access directory \"%s\": %s", + instanceState->instance_backup_subdir_path, strerror(errno)); + + // TODO: redundant message, should we get rid of it? elog(ERROR, "Instance '%s' does not exist in this backup catalog", instance_name); + } + else + { + /* Ensure that backup_path is a path to a directory */ + if (!S_ISDIR(st.st_mode)) + elog(ERROR, "-B, --backup-path must be a path to directory"); + } } } @@ -459,24 +574,49 @@ main(int argc, char *argv[]) */ if (instance_name) { - char path[MAXPGPATH]; /* Read environment variables */ config_get_opt_env(instance_options); /* Read options from configuration file */ - if (backup_subcmd != ADD_INSTANCE_CMD) + if (backup_subcmd != ADD_INSTANCE_CMD && + backup_subcmd != ARCHIVE_GET_CMD) { - join_path_components(path, backup_instance_path, - BACKUP_CATALOG_CONF_FILE); - if (backup_subcmd == CHECKDB_CMD) - config_read_opt(path, instance_options, ERROR, true, true); + config_read_opt(instanceState->instance_config_path, instance_options, ERROR, true, true); else - config_read_opt(path, instance_options, ERROR, true, false); + config_read_opt(instanceState->instance_config_path, instance_options, ERROR, true, false); + + /* + * We can determine our location only after reading the configuration file, + * unless we are running arcive-push/archive-get - they are allowed to trust + * cmdline only. + */ + setMyLocation(backup_subcmd); } - setMyLocation(); + } + else if (backup_subcmd == CATCHUP_CMD) + { + config_get_opt_env(instance_options); + } + + /* + * Disable logging into file for archive-push and archive-get. + * Note, that we should NOT use fio_is_remote() here, + * because it will launch ssh connection and we do not + * want it, because it will kill archive-get prefetch + * performance. + * + * TODO: make logging into file possible via ssh + */ + if (fio_is_remote_simple(FIO_BACKUP_HOST) && + (backup_subcmd == ARCHIVE_GET_CMD || + backup_subcmd == ARCHIVE_PUSH_CMD)) + { + instance_config.logger.log_level_file = LOG_OFF; + is_archive_cmd = true; } + /* Just read environment variables */ if (backup_path == NULL && backup_subcmd == CHECKDB_CMD) config_get_opt_env(instance_options); @@ -486,7 +626,17 @@ main(int argc, char *argv[]) backup_path != NULL && instance_name == NULL && instance_config.pgdata == NULL) - elog(ERROR, "required parameter not specified: --instance"); + elog(ERROR, "Required parameter not specified: --instance"); + + /* Check checkdb command options consistency */ + if (backup_subcmd == CHECKDB_CMD && + !need_amcheck) + { + if (heapallindexed) + elog(ERROR, "--heapallindexed can only be used with --amcheck option"); + if (checkunique) + elog(ERROR, "--checkunique can only be used with --amcheck option"); + } /* Usually checkdb for file logging requires log_directory * to be specified explicitly, but if backup_dir and instance name are provided, @@ -500,6 +650,13 @@ main(int argc, char *argv[]) "You must specify --log-directory option when running checkdb with " "--log-level-file option enabled."); + if (backup_subcmd == CATCHUP_CMD && + instance_config.logger.log_level_file != LOG_OFF && + instance_config.logger.log_directory == NULL) + elog(ERROR, "Cannot save catchup logs to a file. " + "You must specify --log-directory option when running catchup with " + "--log-level-file option enabled."); + /* Initialize logger */ init_logger(backup_path, &instance_config.logger); @@ -528,16 +685,25 @@ main(int argc, char *argv[]) if (instance_config.pgdata != NULL) canonicalize_path(instance_config.pgdata); if (instance_config.pgdata != NULL && + (backup_subcmd != ARCHIVE_GET_CMD && backup_subcmd != CATCHUP_CMD) && !is_absolute_path(instance_config.pgdata)) elog(ERROR, "-D, --pgdata must be an absolute path"); #if PG_VERSION_NUM >= 110000 /* Check xlog-seg-size option */ if (instance_name && - backup_subcmd != INIT_CMD && backup_subcmd != SHOW_CMD && + backup_subcmd != INIT_CMD && backup_subcmd != ADD_INSTANCE_CMD && backup_subcmd != SET_CONFIG_CMD && !IsValidWalSegSize(instance_config.xlog_seg_size)) - elog(ERROR, "Invalid WAL segment size %u", instance_config.xlog_seg_size); + { + /* If we are working with instance of PG<11 using PG11 binary, + * then xlog_seg_size is equal to zero. Manually set it to 16MB. + */ + if (instance_config.xlog_seg_size == 0) + instance_config.xlog_seg_size = DEFAULT_XLOG_SEG_SIZE; + else + elog(ERROR, "Invalid WAL segment size %u", instance_config.xlog_seg_size); + } #endif /* Sanity check of --backup-id option */ @@ -547,9 +713,10 @@ main(int argc, char *argv[]) backup_subcmd != VALIDATE_CMD && backup_subcmd != DELETE_CMD && backup_subcmd != MERGE_CMD && + backup_subcmd != SET_BACKUP_CMD && backup_subcmd != SHOW_CMD) elog(ERROR, "Cannot use -i (--backup-id) option together with the \"%s\" command", - command_name); + get_subcmd_name(backup_subcmd)); current.backup_id = base36dec(backup_id_string); if (current.backup_id == 0) @@ -567,17 +734,6 @@ main(int argc, char *argv[]) if (instance_config.conn_opt.pguser != NULL) dbuser = pstrdup(instance_config.conn_opt.pguser); - /* setup exclusion list for file search */ - if (!backup_logs) - { - int i; - - for (i = 0; pgdata_exclude_dir[i]; i++); /* find first empty slot */ - - /* Set 'pg_log' in first empty slot */ - pgdata_exclude_dir[i] = PG_LOG_DIR; - } - if (backup_subcmd == VALIDATE_CMD || backup_subcmd == RESTORE_CMD) { /* @@ -586,77 +742,325 @@ main(int argc, char *argv[]) */ recovery_target_options = parseRecoveryTargetOptions(target_time, target_xid, - target_inclusive, target_tli, target_lsn, + target_inclusive, target_tli_string, target_lsn, (target_stop != NULL) ? target_stop : (target_immediate) ? "immediate" : NULL, - target_name, target_action, no_validate); + target_name, target_action); + + if (force && backup_subcmd != RESTORE_CMD) + elog(ERROR, "You cannot specify \"--force\" flag with the \"%s\" command", + get_subcmd_name(backup_subcmd)); + + if (force) + no_validate = true; + + /* keep all params in one structure */ + restore_params = pgut_new(pgRestoreParams); + restore_params->is_restore = (backup_subcmd == RESTORE_CMD); + restore_params->force = force; + restore_params->no_validate = no_validate; + restore_params->restore_as_replica = restore_as_replica; + restore_params->recovery_settings_mode = DEFAULT; + + restore_params->primary_slot_name = replication_slot; + restore_params->skip_block_validation = skip_block_validation; + restore_params->skip_external_dirs = skip_external_dirs; + restore_params->partial_db_list = NULL; + restore_params->partial_restore_type = NONE; + restore_params->primary_conninfo = primary_conninfo; + restore_params->incremental_mode = incremental_mode; + restore_params->allow_partial_incremental = allow_partial_incremental; + + /* handle partial restore parameters */ + if (datname_exclude_list && datname_include_list) + elog(ERROR, "You cannot specify '--db-include' and '--db-exclude' together"); + + if (datname_exclude_list) + { + restore_params->partial_restore_type = EXCLUDE; + restore_params->partial_db_list = datname_exclude_list; + } + else if (datname_include_list) + { + restore_params->partial_restore_type = INCLUDE; + restore_params->partial_db_list = datname_include_list; + } + + if (gl_waldir_path) + { + /* clean up xlog directory name, check it's absolute */ + canonicalize_path(gl_waldir_path); + if (!is_absolute_path(gl_waldir_path)) + { + elog(ERROR, "WAL directory location must be an absolute path"); + } + if (strlen(gl_waldir_path) > MAXPGPATH) + elog(ERROR, "Value specified to --waldir is too long"); + + } + restore_params->waldir = gl_waldir_path; + } + /* + * Parse set-backup options into set_backup_params structure. + */ + if (backup_subcmd == SET_BACKUP_CMD || backup_subcmd == BACKUP_CMD) + { + time_t expire_time = 0; + + if (expire_time_string && ttl >= 0) + elog(ERROR, "You cannot specify '--expire-time' and '--ttl' options together"); + + /* Parse string to seconds */ + if (expire_time_string) + { + if (!parse_time(expire_time_string, &expire_time, false)) + elog(ERROR, "Invalid value for '--expire-time' option: '%s'", + expire_time_string); + } + + if (expire_time > 0 || ttl >= 0 || backup_note) + { + set_backup_params = pgut_new(pgSetBackupParams); + set_backup_params->ttl = ttl; + set_backup_params->expire_time = expire_time; + set_backup_params->note = backup_note; + + if (backup_note && strlen(backup_note) > MAX_NOTE_SIZE) + elog(ERROR, "Backup note cannot exceed %u bytes", MAX_NOTE_SIZE); + } + } + + /* checking required options */ + if (backup_subcmd == CATCHUP_CMD) + { + if (catchup_source_pgdata == NULL) + elog(ERROR, "You must specify \"--source-pgdata\" option with the \"%s\" command", get_subcmd_name(backup_subcmd)); + if (catchup_destination_pgdata == NULL) + elog(ERROR, "You must specify \"--destination-pgdata\" option with the \"%s\" command", get_subcmd_name(backup_subcmd)); + if (current.backup_mode == BACKUP_MODE_INVALID) + elog(ERROR, "No backup mode specified.\n" + "Please specify it either using environment variable BACKUP_MODE or\n" + "command line option --backup-mode (-b)"); + if (current.backup_mode != BACKUP_MODE_FULL && current.backup_mode != BACKUP_MODE_DIFF_PTRACK && current.backup_mode != BACKUP_MODE_DIFF_DELTA) + elog(ERROR, "Only \"FULL\", \"PTRACK\" and \"DELTA\" modes are supported with the \"%s\" command", get_subcmd_name(backup_subcmd)); + if (!stream_wal) + elog(INFO, "--stream is required, forcing stream mode"); + current.stream = stream_wal = true; + if (instance_config.external_dir_str) + elog(ERROR, "External directories not supported fom \"%s\" command", get_subcmd_name(backup_subcmd)); + // TODO check instance_config.conn_opt + } + + /* sanity */ + if (backup_subcmd == VALIDATE_CMD && restore_params->no_validate) + elog(ERROR, "You cannot specify \"--no-validate\" option with the \"%s\" command", + get_subcmd_name(backup_subcmd)); + + if (backup_subcmd == ARCHIVE_PUSH_CMD) + { + /* Check archive-push parameters and construct archive_push_xlog_dir + * + * There are 4 cases: + * 1. no --wal-file-path specified -- use cwd, ./PG_XLOG_DIR for wal files + * (and ./PG_XLOG_DIR/archive_status for .done files inside do_archive_push()) + * in this case we can use batches and threads + * 2. --wal-file-path is specified and it is the same dir as stored in pg_probackup.conf (instance_config.pgdata) + * in this case we can use this path, as well as batches and thread + * 3. --wal-file-path is specified and it isn't same dir as stored in pg_probackup.conf but control file present with correct system_id + * in this case we can use this path, as well as batches and thread + * (replica for example, see test_archive_push_sanity) + * 4. --wal-file-path is specified and it is different from instance_config.pgdata and no control file found + * disable optimizations and work with user specified path + */ + bool check_system_id = true; + uint64 system_id; + char current_dir[MAXPGPATH]; + + if (wal_file_name == NULL) + elog(ERROR, "Required parameter is not specified: --wal-file-name %%f"); + + if (instance_config.pgdata == NULL) + elog(ERROR, "Cannot read pg_probackup.conf for this instance"); + + /* TODO may be remove in preference of checking inside compress_init()? */ + if (instance_config.compress_alg == PGLZ_COMPRESS) + elog(ERROR, "Cannot use pglz for WAL compression"); + + if (!getcwd(current_dir, sizeof(current_dir))) + elog(ERROR, "getcwd() error"); + + if (wal_file_path == NULL) + { + /* 1st case */ + system_id = get_system_identifier(current_dir, FIO_DB_HOST, false); + join_path_components(archive_push_xlog_dir, current_dir, XLOGDIR); + } + else + { + /* + * Usually we get something like + * wal_file_path = "pg_wal/0000000100000000000000A1" + * wal_file_name = "0000000100000000000000A1" + * instance_config.pgdata = "/pgdata/.../node/data" + * We need to strip wal_file_name from wal_file_path, add XLOGDIR to instance_config.pgdata + * and compare this directories. + * Note, that pg_wal can be symlink (see test_waldir_outside_pgdata_archiving) + */ + char *stripped_wal_file_path = pgut_str_strip_trailing_filename(wal_file_path, wal_file_name); + join_path_components(archive_push_xlog_dir, instance_config.pgdata, XLOGDIR); + if (fio_is_same_file(stripped_wal_file_path, archive_push_xlog_dir, true, FIO_DB_HOST)) + { + /* 2nd case */ + system_id = get_system_identifier(instance_config.pgdata, FIO_DB_HOST, false); + /* archive_push_xlog_dir already have right value */ + } + else + { + if (strlen(stripped_wal_file_path) < MAXPGPATH) + strncpy(archive_push_xlog_dir, stripped_wal_file_path, MAXPGPATH); + else + elog(ERROR, "Value specified to --wal_file_path is too long"); + + system_id = get_system_identifier(current_dir, FIO_DB_HOST, true); + /* 3rd case if control file present -- i.e. system_id != 0 */ + + if (system_id == 0) + { + /* 4th case */ + check_system_id = false; + + if (batch_size > 1 || num_threads > 1 || !no_ready_rename) + { + elog(WARNING, "Supplied --wal_file_path is outside pgdata, force safe values for options: --batch-size=1 -j 1 --no-ready-rename"); + batch_size = 1; + num_threads = 1; + no_ready_rename = true; + } + } + } + pfree(stripped_wal_file_path); + } + + if (check_system_id && system_id != instance_config.system_identifier) + elog(ERROR, "Refuse to push WAL segment %s into archive. Instance parameters mismatch." + "Instance '%s' should have SYSTEM_ID = " UINT64_FORMAT " instead of " UINT64_FORMAT, + wal_file_name, instanceState->instance_name, instance_config.system_identifier, system_id); + } + +#if PG_VERSION_NUM >= 100000 + if (temp_slot && perm_slot) + elog(ERROR, "You cannot specify \"--perm-slot\" option with the \"--temp-slot\" option"); + + /* if slot name was not provided for temp slot, use default slot name */ + if (!replication_slot && temp_slot) + replication_slot = DEFAULT_TEMP_SLOT_NAME; +#endif + if (!replication_slot && perm_slot) + replication_slot = DEFAULT_PERMANENT_SLOT_NAME; + if (num_threads < 1) num_threads = 1; - compress_init(); + if (batch_size < 1) + batch_size = 1; + + compress_init(backup_subcmd); /* do actual operation */ switch (backup_subcmd) { case ARCHIVE_PUSH_CMD: - return do_archive_push(wal_file_path, wal_file_name, file_overwrite); + do_archive_push(instanceState, &instance_config, archive_push_xlog_dir, wal_file_name, + batch_size, file_overwrite, no_sync, no_ready_rename); + break; case ARCHIVE_GET_CMD: - return do_archive_get(wal_file_path, wal_file_name); + do_archive_get(instanceState, &instance_config, prefetch_dir, + wal_file_path, wal_file_name, batch_size, !no_validate_wal); + break; case ADD_INSTANCE_CMD: - return do_add_instance(); + return do_add_instance(instanceState, &instance_config); case DELETE_INSTANCE_CMD: - return do_delete_instance(); + return do_delete_instance(instanceState); case INIT_CMD: - return do_init(); + return do_init(catalogState); case BACKUP_CMD: { - time_t start_time = time(NULL); - current.stream = stream_wal; + if (start_time != INVALID_BACKUP_ID) + elog(WARNING, "Please do not use the --start-time option to start backup. " + "This is a service option required to work with other extensions. " + "We do not guarantee future support for this flag."); + /* sanity */ if (current.backup_mode == BACKUP_MODE_INVALID) - elog(ERROR, "required parameter not specified: BACKUP_MODE " - "(-b, --backup-mode)"); + elog(ERROR, "No backup mode specified.\n" + "Please specify it either using environment variable BACKUP_MODE or\n" + "command line option --backup-mode (-b)"); - return do_backup(start_time, no_validate); + return do_backup(instanceState, set_backup_params, + no_validate, no_sync, backup_logs, start_time); } + case CATCHUP_CMD: + return do_catchup(catchup_source_pgdata, catchup_destination_pgdata, num_threads, !no_sync, + exclude_absolute_paths_list, exclude_relative_paths_list); case RESTORE_CMD: - return do_restore_or_validate(current.backup_id, - recovery_target_options, - true); + return do_restore_or_validate(instanceState, current.backup_id, + recovery_target_options, + restore_params, no_sync); case VALIDATE_CMD: - if (current.backup_id == 0 && target_time == 0 && target_xid == 0) - return do_validate_all(); + if (current.backup_id == 0 && target_time == 0 && target_xid == 0 && !target_lsn) + { + /* sanity */ + if (datname_exclude_list || datname_include_list) + elog(ERROR, "You must specify parameter (-i, --backup-id) for partial validation"); + + return do_validate_all(catalogState, instanceState); + } else - return do_restore_or_validate(current.backup_id, + /* PITR validation and, optionally, partial validation */ + return do_restore_or_validate(instanceState, current.backup_id, recovery_target_options, - false); + restore_params, + no_sync); case SHOW_CMD: - return do_show(current.backup_id); + return do_show(catalogState, instanceState, current.backup_id, show_archive); case DELETE_CMD: + if (delete_expired && backup_id_string) - elog(ERROR, "You cannot specify --delete-expired and --backup-id options together"); + elog(ERROR, "You cannot specify --delete-expired and (-i, --backup-id) options together"); if (merge_expired && backup_id_string) - elog(ERROR, "You cannot specify --merge-expired and --backup-id options together"); - if (!delete_expired && !merge_expired && !delete_wal && !backup_id_string) + elog(ERROR, "You cannot specify --merge-expired and (-i, --backup-id) options together"); + if (delete_status && backup_id_string) + elog(ERROR, "You cannot specify --status and (-i, --backup-id) options together"); + if (!delete_expired && !merge_expired && !delete_wal && delete_status == NULL && !backup_id_string) elog(ERROR, "You must specify at least one of the delete options: " - "--expired |--wal |--merge-expired |--delete-invalid |--backup_id"); + "--delete-expired |--delete-wal |--merge-expired |--status |(-i, --backup-id)"); if (!backup_id_string) - return do_retention(); + { + if (delete_status) + do_delete_status(instanceState, &instance_config, delete_status); + else + do_retention(instanceState, no_validate, no_sync); + } else - do_delete(current.backup_id); + do_delete(instanceState, current.backup_id); break; case MERGE_CMD: - do_merge(current.backup_id); + do_merge(instanceState, current.backup_id, no_validate, no_sync); break; case SHOW_CONFIG_CMD: - do_show_config(); + do_show_config(show_base_units); break; case SET_CONFIG_CMD: - do_set_config(false); + do_set_config(instanceState, false); + break; + case SET_BACKUP_CMD: + if (!backup_id_string) + elog(ERROR, "You must specify parameter (-i, --backup-id) for 'set-backup' command"); + do_set_backup(instanceState, current.backup_id, set_backup_params); break; case CHECKDB_CMD: do_checkdb(need_amcheck, @@ -665,11 +1069,43 @@ main(int argc, char *argv[]) case NO_CMD: /* Should not happen */ elog(ERROR, "Unknown subcommand"); + case SSH_CMD: + case AGENT_CMD: + /* Может перейти на использование какого-нибудь do_agent() для однобразия? */ + case HELP_CMD: + case VERSION_CMD: + /* Silence compiler warnings, these already handled earlier */ + break; } + free_environment_locale(); + return 0; } +static void +opt_incr_restore_mode(ConfigOption *opt, const char *arg) +{ + if (pg_strcasecmp(arg, "none") == 0) + { + incremental_mode = INCR_NONE; + return; + } + else if (pg_strcasecmp(arg, "checksum") == 0) + { + incremental_mode = INCR_CHECKSUM; + return; + } + else if (pg_strcasecmp(arg, "lsn") == 0) + { + incremental_mode = INCR_LSN; + return; + } + + /* Backup mode is invalid, so leave with an error */ + elog(ERROR, "Invalid value for '--incremental-mode' option: '%s'", arg); +} + static void opt_backup_mode(ConfigOption *opt, const char *arg) { @@ -704,17 +1140,18 @@ opt_show_format(ConfigOption *opt, const char *arg) * Initialize compress and sanity checks for compress. */ static void -compress_init(void) +compress_init(ProbackupSubcmd const subcmd) { /* Default algorithm is zlib */ if (compress_shortcut) instance_config.compress_alg = ZLIB_COMPRESS; - if (backup_subcmd != SET_CONFIG_CMD) + if (subcmd != SET_CONFIG_CMD) { if (instance_config.compress_level != COMPRESS_LEVEL_DEFAULT && instance_config.compress_alg == NOT_DEFINED_COMPRESS) - elog(ERROR, "Cannot specify compress-level option without compress-alg option"); + elog(ERROR, "Cannot specify compress-level option alone without " + "compress-algorithm option"); } if (instance_config.compress_level < 0 || instance_config.compress_level > 9) @@ -723,7 +1160,7 @@ compress_init(void) if (instance_config.compress_alg == ZLIB_COMPRESS && instance_config.compress_level == 0) elog(WARNING, "Compression level 0 will lead to data bloat!"); - if (backup_subcmd == BACKUP_CMD || backup_subcmd == ARCHIVE_PUSH_CMD) + if (subcmd == BACKUP_CMD || subcmd == ARCHIVE_PUSH_CMD) { #ifndef HAVE_LIBZ if (instance_config.compress_alg == ZLIB_COMPRESS) @@ -734,3 +1171,46 @@ compress_init(void) elog(ERROR, "Multithread backup does not support pglz compression"); } } + +static void +opt_parser_add_to_parray_helper(parray **list, const char *str) +{ + char *elem = NULL; + + if (*list == NULL) + *list = parray_new(); + + elem = pgut_malloc(strlen(str) + 1); + strcpy(elem, str); + + parray_append(*list, elem); +} + +/* Construct array of datnames, provided by user via db-exclude option */ +void +opt_datname_exclude_list(ConfigOption *opt, const char *arg) +{ + /* TODO add sanity for database name */ + opt_parser_add_to_parray_helper(&datname_exclude_list, arg); +} + +/* Construct array of datnames, provided by user via db-include option */ +void +opt_datname_include_list(ConfigOption *opt, const char *arg) +{ + if (strcmp(arg, "template0") == 0 || + strcmp(arg, "template1") == 0) + elog(ERROR, "Databases 'template0' and 'template1' cannot be used for partial restore or validation"); + + opt_parser_add_to_parray_helper(&datname_include_list, arg); +} + +/* Parse --exclude-path option */ +void +opt_exclude_path(ConfigOption *opt, const char *arg) +{ + if (is_absolute_path(arg)) + opt_parser_add_to_parray_helper(&exclude_absolute_paths_list, arg); + else + opt_parser_add_to_parray_helper(&exclude_relative_paths_list, arg); +} diff --git a/src/pg_probackup.h b/src/pg_probackup.h index e646d760d..ae99e0605 100644 --- a/src/pg_probackup.h +++ b/src/pg_probackup.h @@ -3,19 +3,25 @@ * pg_probackup.h: Backup/Recovery manager for PostgreSQL. * * Portions Copyright (c) 2009-2013, NIPPON TELEGRAPH AND TELEPHONE CORPORATION - * Portions Copyright (c) 2015-2018, Postgres Professional + * Portions Copyright (c) 2015-2022, Postgres Professional * *------------------------------------------------------------------------- */ #ifndef PG_PROBACKUP_H #define PG_PROBACKUP_H + #include "postgres_fe.h" #include "libpq-fe.h" #include "libpq-int.h" #include "access/xlog_internal.h" #include "utils/pg_crc.h" +#include "catalog/pg_control.h" + +#if PG_VERSION_NUM >= 120000 +#include "common/logging.h" +#endif #ifdef FRONTEND #undef FRONTEND @@ -33,6 +39,25 @@ #include "utils/file.h" #include "datapagemap.h" +#include "utils/thread.h" + +#include "pg_probackup_state.h" + + +#ifdef WIN32 +#define __thread __declspec(thread) +#else +#include +#endif + +#if PG_VERSION_NUM >= 150000 +// _() is explicitly undefined in libpq-int.h +// https://github.com/postgres/postgres/commit/28ec316787674dd74d00b296724a009b6edc2fb0 +#define _(s) gettext(s) +#endif + +/* Wrap the code that we're going to delete after refactoring in this define*/ +#define REFACTORE_ME /* pgut client variables and full path */ extern const char *PROGRAM_NAME; @@ -42,8 +67,9 @@ extern const char *PROGRAM_URL; extern const char *PROGRAM_EMAIL; /* Directory/File names */ -#define DATABASE_DIR "database" +#define DATABASE_DIR "database" #define BACKUPS_DIR "backups" +#define WAL_SUBDIR "wal" #if PG_VERSION_NUM >= 100000 #define PG_XLOG_DIR "pg_wal" #define PG_LOG_DIR "log" @@ -55,16 +81,28 @@ extern const char *PROGRAM_EMAIL; #define PG_GLOBAL_DIR "global" #define BACKUP_CONTROL_FILE "backup.control" #define BACKUP_CATALOG_CONF_FILE "pg_probackup.conf" -#define BACKUP_CATALOG_PID "backup.pid" +#define BACKUP_LOCK_FILE "backup.pid" +#define BACKUP_RO_LOCK_FILE "backup_ro.pid" #define DATABASE_FILE_LIST "backup_content.control" #define PG_BACKUP_LABEL_FILE "backup_label" -#define PG_BLACK_LIST "black_list" -#define PG_TABLESPACE_MAP_FILE "tablespace_map" +#define PG_TABLESPACE_MAP_FILE "tablespace_map" +#define RELMAPPER_FILENAME "pg_filenode.map" #define EXTERNAL_DIR "external_directories/externaldir" +#define DATABASE_MAP "database_map" +#define HEADER_MAP "page_header_map" +#define HEADER_MAP_TMP "page_header_map_tmp" +#define XLOG_CONTROL_BAK_FILE XLOG_CONTROL_FILE".pbk.bak" + +/* default replication slot names */ +#define DEFAULT_TEMP_SLOT_NAME "pg_probackup_slot"; +#define DEFAULT_PERMANENT_SLOT_NAME "pg_probackup_perm_slot"; /* Timeout defaults */ #define ARCHIVE_TIMEOUT_DEFAULT 300 #define REPLICA_TIMEOUT_DEFAULT 300 +#define LOCK_TIMEOUT 60 +#define LOCK_STALE_TIMEOUT 30 +#define LOG_FREQ 10 /* Directory/File permission */ #define DIR_PERMISSION (0700) @@ -73,6 +111,8 @@ extern const char *PROGRAM_EMAIL; /* 64-bit xid support for PGPRO_EE */ #ifndef PGPRO_EE #define XID_FMT "%u" +#elif !defined(XID_FMT) +#define XID_FMT UINT64_FORMAT #endif #ifndef STDIN_FILENO @@ -80,10 +120,97 @@ extern const char *PROGRAM_EMAIL; #define STDOUT_FILENO 1 #endif +/* stdio buffer size */ +#define STDIO_BUFSIZE 65536 + +#define ERRMSG_MAX_LEN 2048 +#define CHUNK_SIZE (128 * 1024) +#define LARGE_CHUNK_SIZE (4 * 1024 * 1024) +#define OUT_BUF_SIZE (512 * 1024) + +/* retry attempts */ +#define PAGE_READ_ATTEMPTS 300 + +/* max size of note, that can be added to backup */ +#define MAX_NOTE_SIZE 1024 + /* Check if an XLogRecPtr value is pointed to 0 offset */ #define XRecOffIsNull(xlrp) \ ((xlrp) % XLOG_BLCKSZ == 0) +/* log(2**64) / log(36) = 12.38 => max 13 char + '\0' */ +#define base36bufsize 14 + +/* Text Coloring macro */ +#define TC_LEN 11 +#define TC_RED "\033[0;31m" +#define TC_RED_BOLD "\033[1;31m" +#define TC_BLUE "\033[0;34m" +#define TC_BLUE_BOLD "\033[1;34m" +#define TC_GREEN "\033[0;32m" +#define TC_GREEN_BOLD "\033[1;32m" +#define TC_YELLOW "\033[0;33m" +#define TC_YELLOW_BOLD "\033[1;33m" +#define TC_MAGENTA "\033[0;35m" +#define TC_MAGENTA_BOLD "\033[1;35m" +#define TC_CYAN "\033[0;36m" +#define TC_CYAN_BOLD "\033[1;36m" +#define TC_RESET "\033[0m" + +typedef struct RedoParams +{ + TimeLineID tli; + XLogRecPtr lsn; + uint32 checksum_version; +} RedoParams; + +typedef struct PageState +{ + uint16 checksum; + XLogRecPtr lsn; +} PageState; + +typedef struct db_map_entry +{ + Oid dbOid; + char *datname; +} db_map_entry; + +/* State of pgdata in the context of its compatibility for incremental restore */ +typedef enum DestDirIncrCompatibility +{ + POSTMASTER_IS_RUNNING, + SYSTEM_ID_MISMATCH, + BACKUP_LABEL_EXISTS, + PARTIAL_INCREMENTAL_FORBIDDEN, + DEST_IS_NOT_OK, + DEST_OK +} DestDirIncrCompatibility; + +typedef enum IncrRestoreMode +{ + INCR_NONE, + INCR_CHECKSUM, + INCR_LSN +} IncrRestoreMode; + +typedef enum PartialRestoreType +{ + NONE, + INCLUDE, + EXCLUDE, +} PartialRestoreType; + +typedef enum RecoverySettingsMode +{ + DEFAULT, /* not set */ + DONTWRITE, /* explicitly forbid to update recovery settings */ + //TODO Should we always clean/preserve old recovery settings, + // or make it configurable? + PITR_REQUESTED, /* can be set based on other parameters + * if not explicitly forbidden */ +} RecoverySettingsMode; + typedef enum CompressAlg { NOT_DEFINED_COMPRESS = 0, @@ -92,6 +219,18 @@ typedef enum CompressAlg ZLIB_COMPRESS, } CompressAlg; +typedef enum ForkName +{ + none, + vm, + fsm, + cfm, + init, + ptrack, + cfs_bck, + cfm_bck +} ForkName; + #define INIT_FILE_CRC32(use_crc32c, crc) \ do { \ if (use_crc32c) \ @@ -114,44 +253,71 @@ do { \ FIN_TRADITIONAL_CRC32(crc); \ } while (0) +#define pg_off_t unsigned long long + /* Information about single file (or dir) in backup */ typedef struct pgFile { - char *name; /* file or directory name */ + char *name; /* file or directory name */ mode_t mode; /* protection (file type and permission) */ size_t size; /* size of the file */ + time_t mtime; /* file st_mtime attribute, can be used only + during backup */ size_t read_size; /* size of the portion read (if only some pages are backed up, it's different from size) */ int64 write_size; /* size of the backed-up file. BYTES_INVALID means that the file existed but was not backed up because not modified since last backup. */ + size_t uncompressed_size; /* size of the backed-up file before compression + * and adding block headers. + */ /* we need int64 here to store '-1' value */ pg_crc32 crc; /* CRC value of the file, regular file only */ - char *linked; /* path of the linked file */ + char *rel_path; /* relative path of the file */ + char *linked; /* path of the linked file */ bool is_datafile; /* true if the file is PostgreSQL data file */ - char *path; /* absolute path of the file */ - char *rel_path; /* relative path of the file */ Oid tblspcOid; /* tblspcOid extracted from path, if applicable */ Oid dbOid; /* dbOid extracted from path, if applicable */ Oid relOid; /* relOid extracted from path, if applicable */ - char *forkName; /* forkName extracted from path, if applicable */ + ForkName forkName; /* forkName extracted from path, if applicable */ int segno; /* Segment number for ptrack */ - int n_blocks; /* size of the file in blocks, readed during DELTA backup */ + int n_blocks; /* number of blocks in the data file in data directory */ bool is_cfs; /* Flag to distinguish files compressed by CFS*/ - bool is_database; - int external_dir_num; /* Number of external directory. 0 if not external */ - bool exists_in_prev; /* Mark files, both data and regular, that exists in previous backup */ - CompressAlg compress_alg; /* compression algorithm applied to the file */ - volatile pg_atomic_flag lock; /* lock for synchronization of parallel threads */ - datapagemap_t pagemap; /* bitmap of pages updated since previous backup */ - bool pagemap_isabsent; /* Used to mark files with unknown state of pagemap, - * i.e. datafiles without _ptrack */ + struct pgFile *cfs_chain; /* linked list of CFS segment's cfm, bck, cfm_bck related files */ + int external_dir_num; /* Number of external directory. 0 if not external */ + bool exists_in_prev; /* Mark files, both data and regular, that exists in previous backup */ + CompressAlg compress_alg; /* compression algorithm applied to the file */ + volatile pg_atomic_flag lock;/* lock for synchronization of parallel threads */ + datapagemap_t pagemap; /* bitmap of pages updated since previous backup + may take up to 16kB per file */ + bool pagemap_isabsent; /* Used to mark files with unknown state of pagemap, + * i.e. datafiles without _ptrack */ + /* Coordinates in header map */ + int n_headers; /* number of blocks in the data file in backup */ + pg_crc32 hdr_crc; /* CRC value of header file: name_hdr */ + pg_off_t hdr_off; /* offset in header map */ + int hdr_size; /* length of headers */ + bool excluded; /* excluded via --exclude-path option */ + bool skip_cfs_nested; /* mark to skip in processing treads as nested to cfs_chain */ + bool remove_from_list; /* tmp flag to clean up files list from temp and unlogged tables */ } pgFile; +typedef struct page_map_entry +{ + const char *path; /* file or directory name */ + char *pagemap; + size_t pagemapsize; +} page_map_entry; + /* Special values of datapagemap_t bitmapsize */ #define PageBitmapIsEmpty 0 /* Used to mark unchanged datafiles */ +/* Return codes for check_tablespace_mapping */ +#define NoTblspc 0 +#define EmptyTblspc 1 +#define NotEmptyTblspc 2 + /* Current state of backup */ typedef enum BackupStatus { @@ -160,6 +326,8 @@ typedef enum BackupStatus BACKUP_STATUS_ERROR, /* aborted because of unexpected error */ BACKUP_STATUS_RUNNING, /* running backup */ BACKUP_STATUS_MERGING, /* merging backups */ + BACKUP_STATUS_MERGED, /* backup has been successfully merged and now awaits + * the assignment of new start_time */ BACKUP_STATUS_DELETING, /* data files are being deleted */ BACKUP_STATUS_DELETED, /* data files have been deleted */ BACKUP_STATUS_DONE, /* completed but not validated yet */ @@ -188,9 +356,14 @@ typedef enum ShowFormat #define BYTES_INVALID (-1) /* file didn`t changed since previous backup, DELTA backup do not rely on it */ #define FILE_NOT_FOUND (-2) /* file disappeared during backup */ #define BLOCKNUM_INVALID (-1) -#define PROGRAM_VERSION "2.1.5" -#define AGENT_PROTOCOL_VERSION 20105 +#define PROGRAM_VERSION "2.5.15" +/* update when remote agent API or behaviour changes */ +#define AGENT_PROTOCOL_VERSION 20509 +#define AGENT_PROTOCOL_VERSION_STR "2.5.9" + +/* update only when changing storage format */ +#define STORAGE_FORMAT_VERSION "2.4.4" typedef struct ConnectionOptions { @@ -206,6 +379,14 @@ typedef struct ConnectionArgs PGcancel *cancel_conn; } ConnectionArgs; +/* Store values for --remote-* option for 'restore_command' constructor */ +typedef struct ArchiveOptions +{ + const char *host; + const char *port; + const char *user; +} ArchiveOptions; + /* * An instance configuration. It can be stored in a configuration file or passed * from command line. @@ -221,11 +402,14 @@ typedef struct InstanceConfig ConnectionOptions conn_opt; ConnectionOptions master_conn_opt; - uint32 replica_timeout; + uint32 replica_timeout; //Deprecated. Not used anywhere /* Wait timeout for WAL segment archiving */ uint32 archive_timeout; + /* cmdline to be used as restore_command */ + char *restore_command; + /* Logger parameters */ LoggerConfig logger; @@ -235,13 +419,18 @@ typedef struct InstanceConfig /* Retention options. 0 disables the option. */ uint32 retention_redundancy; uint32 retention_window; + uint32 wal_depth; CompressAlg compress_alg; int compress_level; + + /* Archive description */ + ArchiveOptions archive; } InstanceConfig; extern ConfigOption instance_options[]; extern InstanceConfig instance_config; +extern time_t current_time; typedef struct PGNodeInfo { @@ -249,12 +438,29 @@ typedef struct PGNodeInfo uint32 wal_block_size; uint32 checksum_version; bool is_superuser; + bool pgpro_support; int server_version; char server_version_str[100]; + int ptrack_version_num; + bool is_ptrack_enabled; + const char *ptrack_schema; /* used only for ptrack 2.x */ + } PGNodeInfo; +/* structure used for access to block header map */ +typedef struct HeaderMap +{ + char path[MAXPGPATH]; + char path_tmp[MAXPGPATH]; /* used only in merge */ + FILE *fp; /* used only for writing */ + char *buf; /* buffer */ + pg_off_t offset; /* current position in fp */ + pthread_mutex_t mutex; + +} HeaderMap; + typedef struct pgBackup pgBackup; /* Information about single backup stored in backup.conf */ @@ -262,19 +468,26 @@ struct pgBackup { BackupMode backup_mode; /* Mode - one of BACKUP_MODE_xxx above*/ time_t backup_id; /* Identifier of the backup. - * Currently it's the same as start_time */ + * By default it's the same as start_time + * but can be increased if same backup_id + * already exists. It can be also set by + * start_time parameter */ BackupStatus status; /* Status - one of BACKUP_STATUS_xxx above*/ TimeLineID tli; /* timeline of start and stop backup lsns */ XLogRecPtr start_lsn; /* backup's starting transaction log location */ XLogRecPtr stop_lsn; /* backup's finishing transaction log location */ - time_t start_time; /* since this moment backup has status - * BACKUP_STATUS_RUNNING */ + time_t start_time; /* UTC time of backup creation */ + time_t merge_dest_backup; /* start_time of incremental backup with + * which this backup is merging with. + * Only available for FULL backups + * with MERGING or MERGED statuses */ time_t merge_time; /* the moment when merge was started or 0 */ time_t end_time; /* the moment when backup was finished, or the moment * when we realized that backup is broken */ time_t recovery_time; /* Earliest moment for which you can restore * the state of the database cluster using * this backup */ + time_t expire_time; /* Backup expiration date */ TransactionId recovery_xid; /* Earliest xid for which you can restore * the state of the database cluster using * this backup */ @@ -285,8 +498,17 @@ struct pgBackup * BYTES_INVALID means nothing was backed up. */ int64 data_bytes; - /* Size of WAL files in archive needed to restore this backup */ + /* Size of WAL files needed to replay on top of this + * backup to reach the consistency. + */ int64 wal_bytes; + /* Size of data files before applying compression and block header, + * WAL files are not included. + */ + int64 uncompressed_bytes; + + /* Size of data files in PGDATA at the moment of backup. */ + int64 pgdata_bytes; CompressAlg compress_alg; int compress_level; @@ -309,6 +531,20 @@ struct pgBackup * in the format suitable for recovery.conf */ char *external_dir_str; /* List of external directories, * separated by ':' */ + char *root_dir; /* Full path for root backup directory: + backup_path/instance_name/backup_id */ + char *database_dir; /* Full path to directory with data files: + backup_path/instance_name/backup_id/database */ + parray *files; /* list of files belonging to this backup + * must be populated explicitly */ + char *note; + + pg_crc32 content_crc; + + /* map used for access to page headers */ + HeaderMap hdr_map; + + char backup_id_encoded[base36bufsize]; }; /* Recovery target for restore and validate subcommands */ @@ -329,11 +565,53 @@ typedef struct pgRecoveryTarget const char *target_stop; const char *target_name; const char *target_action; - bool no_validate; + const char *target_tli_string; /* timeline number, "current" or "latest" from recovery_target_timeline option*/ } pgRecoveryTarget; +/* Options needed for restore and validate commands */ +typedef struct pgRestoreParams +{ + bool force; + bool is_restore; + bool no_validate; + bool restore_as_replica; + //TODO maybe somehow add restore_as_replica as one of RecoverySettingsModes + RecoverySettingsMode recovery_settings_mode; + bool skip_external_dirs; + bool skip_block_validation; //Start using it + const char *restore_command; + const char *primary_slot_name; + const char *primary_conninfo; + + /* options for incremental restore */ + IncrRestoreMode incremental_mode; + XLogRecPtr shift_lsn; + + /* options for partial restore */ + PartialRestoreType partial_restore_type; + parray *partial_db_list; + bool allow_partial_incremental; + + char* waldir; +} pgRestoreParams; + +/* Options needed for set-backup command */ +typedef struct pgSetBackupParams +{ + int64 ttl; /* amount of time backup must be pinned + * -1 - do nothing + * 0 - disable pinning + */ + time_t expire_time; /* Point in time until backup + * must be pinned. + */ + char *note; +} pgSetBackupParams; + typedef struct { + PGNodeInfo *nodeInfo; + const char *from_root; const char *to_root; const char *external_prefix; @@ -343,8 +621,8 @@ typedef struct parray *external_dirs; XLogRecPtr prev_start_lsn; - ConnectionArgs conn_arg; int thread_num; + HeaderMap *hdr_map; /* * Return value from the thread. @@ -353,6 +631,64 @@ typedef struct int ret; } backup_files_arg; +typedef struct timelineInfo timelineInfo; + +/* struct to collect info about timelines in WAL archive */ +struct timelineInfo { + + TimeLineID tli; /* this timeline */ + TimeLineID parent_tli; /* parent timeline. 0 if none */ + timelineInfo *parent_link; /* link to parent timeline */ + XLogRecPtr switchpoint; /* if this timeline has a parent, then + * switchpoint contains switchpoint LSN, + * otherwise 0 */ + XLogSegNo begin_segno; /* first present segment in this timeline */ + XLogSegNo end_segno; /* last present segment in this timeline */ + size_t n_xlog_files; /* number of segments (only really existing) + * does not include lost segments */ + size_t size; /* space on disk taken by regular WAL files */ + parray *backups; /* array of pgBackup sturctures with info + * about backups belonging to this timeline */ + parray *xlog_filelist; /* array of ordinary WAL segments, '.partial' + * and '.backup' files belonging to this timeline */ + parray *lost_segments; /* array of intervals of lost segments */ + parray *keep_segments; /* array of intervals of segments used by WAL retention */ + pgBackup *closest_backup; /* link to valid backup, closest to timeline */ + pgBackup *oldest_backup; /* link to oldest backup on timeline */ + XLogRecPtr anchor_lsn; /* LSN belonging to the oldest segno to keep for 'wal-depth' */ + TimeLineID anchor_tli; /* timeline of anchor_lsn */ +}; + +typedef struct xlogInterval +{ + XLogSegNo begin_segno; + XLogSegNo end_segno; +} xlogInterval; + +typedef struct lsnInterval +{ + TimeLineID tli; + XLogRecPtr begin_lsn; + XLogRecPtr end_lsn; +} lsnInterval; + +typedef enum xlogFileType +{ + SEGMENT, + TEMP_SEGMENT, + PARTIAL_SEGMENT, + BACKUP_HISTORY_FILE +} xlogFileType; + +typedef struct xlogFile +{ + pgFile file; + XLogSegNo segno; + xlogFileType type; + bool keep; /* Used to prevent removal of WAL segments + * required by ARCHIVE backups. */ +} xlogFile; + /* * When copying datafiles to backup we validate and compress them block @@ -364,20 +700,26 @@ typedef struct BackupPageHeader int32 compressed_size; } BackupPageHeader; -/* Special value for compressed_size field */ -#define PageIsTruncated -2 -#define SkipCurrentPage -3 -#define PageIsCorrupted -4 /* used by checkdb */ +/* 4MB for 1GB file */ +typedef struct BackupPageHeader2 +{ + XLogRecPtr lsn; + int32 block; /* block number */ + int32 pos; /* position in backup file */ + uint16 checksum; +} BackupPageHeader2; +typedef struct StopBackupCallbackParams +{ + PGconn *conn; + int server_version; +} StopBackupCallbackParams; -/* - * return pointer that exceeds the length of prefix from character string. - * ex. str="/xxx/yyy/zzz", prefix="/xxx/yyy", return="zzz". - * - * Deprecated. Do not use this in new code. - */ -#define GetRelativePath(str, prefix) \ - ((strlen(str) <= strlen(prefix)) ? "" : str + strlen(prefix) + 1) +/* Special value for compressed_size field */ +#define PageIsOk 0 +#define SkipCurrentPage -1 +#define PageIsTruncated -2 +#define PageIsCorrupted -3 /* used by checkdb */ /* * Return timeline, xlog ID and record offset from an LSN of the type @@ -392,6 +734,9 @@ typedef struct BackupPageHeader strcmp((fname) + XLOG_FNAME_LEN, ".gz") == 0) #if PG_VERSION_NUM >= 110000 + +#define WalSegmentOffset(xlogptr, wal_segsz_bytes) \ + XLogSegmentOffset(xlogptr, wal_segsz_bytes) #define GetXLogSegNo(xlrp, logSegNo, wal_segsz_bytes) \ XLByteToSeg(xlrp, logSegNo, wal_segsz_bytes) #define GetXLogRecPtr(segno, offset, wal_segsz_bytes, dest) \ @@ -400,7 +745,19 @@ typedef struct BackupPageHeader XLogFileName(fname, tli, logSegNo, wal_segsz_bytes) #define IsInXLogSeg(xlrp, logSegNo, wal_segsz_bytes) \ XLByteInSeg(xlrp, logSegNo, wal_segsz_bytes) +#define GetXLogSegName(fname, logSegNo, wal_segsz_bytes) \ + snprintf(fname, 20, "%08X%08X", \ + (uint32) ((logSegNo) / XLogSegmentsPerXLogId(wal_segsz_bytes)), \ + (uint32) ((logSegNo) % XLogSegmentsPerXLogId(wal_segsz_bytes))) + +#define GetXLogSegNoFromScrath(logSegNo, log, seg, wal_segsz_bytes) \ + logSegNo = (uint64) log * XLogSegmentsPerXLogId(wal_segsz_bytes) + seg + +#define GetXLogFromFileName(fname, tli, logSegNo, wal_segsz_bytes) \ + XLogFromFileName(fname, tli, logSegNo, wal_segsz_bytes) #else +#define WalSegmentOffset(xlogptr, wal_segsz_bytes) \ + ((xlogptr) & ((XLogSegSize) - 1)) #define GetXLogSegNo(xlrp, logSegNo, wal_segsz_bytes) \ XLByteToSeg(xlrp, logSegNo) #define GetXLogRecPtr(segno, offset, wal_segsz_bytes, dest) \ @@ -409,57 +766,103 @@ typedef struct BackupPageHeader XLogFileName(fname, tli, logSegNo) #define IsInXLogSeg(xlrp, logSegNo, wal_segsz_bytes) \ XLByteInSeg(xlrp, logSegNo) +#define GetXLogSegName(fname, logSegNo, wal_segsz_bytes) \ + snprintf(fname, 20, "%08X%08X",\ + (uint32) ((logSegNo) / XLogSegmentsPerXLogId), \ + (uint32) ((logSegNo) % XLogSegmentsPerXLogId)) + +#define GetXLogSegNoFromScrath(logSegNo, log, seg, wal_segsz_bytes) \ + logSegNo = (uint64) log * XLogSegmentsPerXLogId + seg + +#define GetXLogFromFileName(fname, tli, logSegNo, wal_segsz_bytes) \ + XLogFromFileName(fname, tli, logSegNo) #endif -#define IsSshProtocol() (instance_config.remote.host && strcmp(instance_config.remote.proto, "ssh") == 0) +#define IsPartialCompressXLogFileName(fname) \ + (strlen(fname) == XLOG_FNAME_LEN + strlen(".gz.partial") && \ + strspn(fname, "0123456789ABCDEF") == XLOG_FNAME_LEN && \ + strcmp((fname) + XLOG_FNAME_LEN, ".gz.partial") == 0) + +#define IsTempXLogFileName(fname) \ + (strlen(fname) == XLOG_FNAME_LEN + strlen(".part") && \ + strspn(fname, "0123456789ABCDEF") == XLOG_FNAME_LEN && \ + strcmp((fname) + XLOG_FNAME_LEN, ".part") == 0) -/* directory options */ -extern char *backup_path; -extern char backup_instance_path[MAXPGPATH]; -extern char arclog_path[MAXPGPATH]; +#define IsTempPartialXLogFileName(fname) \ + (strlen(fname) == XLOG_FNAME_LEN + strlen(".partial.part") && \ + strspn(fname, "0123456789ABCDEF") == XLOG_FNAME_LEN && \ + strcmp((fname) + XLOG_FNAME_LEN, ".partial.part") == 0) + +#define IsTempCompressXLogFileName(fname) \ + (strlen(fname) == XLOG_FNAME_LEN + strlen(".gz.part") && \ + strspn(fname, "0123456789ABCDEF") == XLOG_FNAME_LEN && \ + strcmp((fname) + XLOG_FNAME_LEN, ".gz.part") == 0) + +#define IsSshProtocol() (instance_config.remote.host && strcmp(instance_config.remote.proto, "ssh") == 0) /* common options */ +extern pid_t my_pid; +extern __thread int my_thread_num; extern int num_threads; extern bool stream_wal; +extern bool show_color; extern bool progress; -#if PG_VERSION_NUM >= 100000 +extern bool is_archive_cmd; /* true for archive-{get,push} */ /* In pre-10 'replication_slot' is defined in receivelog.h */ extern char *replication_slot; -#endif +#if PG_VERSION_NUM >= 100000 extern bool temp_slot; +#endif +extern bool perm_slot; /* backup options */ extern bool smooth_checkpoint; /* remote probackup options */ -extern char* remote_agent; +extern bool remote_agent; -extern bool is_ptrack_support; extern bool exclusive_backup; -/* restore options */ -extern bool restore_as_replica; -extern bool skip_block_validation; -extern bool skip_external_dirs; - /* delete options */ extern bool delete_wal; extern bool delete_expired; extern bool merge_expired; -extern bool force_delete; extern bool dry_run; -/* compression options */ -extern bool compress_shortcut; +/* ===== instanceState ===== */ -/* other options */ -extern char *instance_name; +typedef struct InstanceState +{ + /* catalog, this instance belongs to */ + CatalogState *catalog_state; + + char instance_name[MAXPGPATH]; //previously global var instance_name + /* $BACKUP_PATH/backups/instance_name */ + char instance_backup_subdir_path[MAXPGPATH]; + + /* $BACKUP_PATH/backups/instance_name/BACKUP_CATALOG_CONF_FILE */ + char instance_config_path[MAXPGPATH]; + + /* $BACKUP_PATH/backups/instance_name */ + char instance_wal_subdir_path[MAXPGPATH]; // previously global var arclog_path + + /* TODO: Make it more specific */ + PGconn *conn; + + + //TODO split into some more meaningdul parts + InstanceConfig *config; +} InstanceState; + +/* ===== instanceState (END) ===== */ /* show options */ extern ShowFormat show_format; /* checkdb options */ extern bool heapallindexed; +extern bool checkunique; +extern bool skip_block_validation; /* current settings */ extern pgBackup current; @@ -467,12 +870,9 @@ extern pgBackup current; /* argv of the process */ extern char** commands_args; -/* in dir.c */ -/* exclude directory list for $PGDATA file listing */ -extern const char *pgdata_exclude_dir[]; - /* in backup.c */ -extern int do_backup(time_t start_time, bool no_validate); +extern int do_backup(InstanceState *instanceState, pgSetBackupParams *set_backup_params, + bool no_validate, bool no_sync, bool backup_logs, time_t start_time); extern void do_checkdb(bool need_amcheck, ConnectionOptions conn_opt, char *pgdata); extern BackupMode parse_backup_mode(const char *value); @@ -480,50 +880,83 @@ extern const char *deparse_backup_mode(BackupMode mode); extern void process_block_change(ForkNumber forknum, RelFileNode rnode, BlockNumber blkno); -extern char *pg_ptrack_get_block(ConnectionArgs *arguments, - Oid dbOid, Oid tblsOid, Oid relOid, - BlockNumber blknum, - size_t *result_size); +/* in catchup.c */ +extern int do_catchup(const char *source_pgdata, const char *dest_pgdata, int num_threads, bool sync_dest_files, + parray *exclude_absolute_paths_list, parray *exclude_relative_paths_list); + /* in restore.c */ -extern int do_restore_or_validate(time_t target_backup_id, +extern int do_restore_or_validate(InstanceState *instanceState, + time_t target_backup_id, pgRecoveryTarget *rt, - bool is_restore); -extern bool satisfy_timeline(const parray *timelines, const pgBackup *backup); + pgRestoreParams *params, + bool no_sync); +extern bool satisfy_timeline(const parray *timelines, TimeLineID tli, XLogRecPtr lsn); extern bool satisfy_recovery_target(const pgBackup *backup, const pgRecoveryTarget *rt); extern pgRecoveryTarget *parseRecoveryTargetOptions( const char *target_time, const char *target_xid, - const char *target_inclusive, TimeLineID target_tli, const char* target_lsn, + const char *target_inclusive, const char *target_tli_string, const char* target_lsn, const char *target_stop, const char *target_name, - const char *target_action, bool no_validate); + const char *target_action); + +extern parray *get_dbOid_exclude_list(pgBackup *backup, parray *datname_list, + PartialRestoreType partial_restore_type); + +extern const char* backup_id_of(pgBackup *backup); +extern void reset_backup_id(pgBackup *backup); + +extern parray *get_backup_filelist(pgBackup *backup, bool strict); +extern parray *read_timeline_history(const char *arclog_path, TimeLineID targetTLI, bool strict); +extern bool tliIsPartOfHistory(const parray *timelines, TimeLineID tli); +extern DestDirIncrCompatibility check_incremental_compatibility(const char *pgdata, uint64 system_identifier, + IncrRestoreMode incremental_mode, + parray *partial_db_list, + bool allow_partial_incremental); + +/* in remote.c */ +extern void check_remote_agent_compatibility(int agent_version, + char *compatibility_str, size_t compatibility_str_max_size); +extern size_t prepare_compatibility_str(char* compatibility_buf, size_t compatibility_buf_size); /* in merge.c */ -extern void do_merge(time_t backup_id); +extern void do_merge(InstanceState *instanceState, time_t backup_id, bool no_validate, bool no_sync); extern void merge_backups(pgBackup *backup, pgBackup *next_backup); +extern void merge_chain(InstanceState *instanceState, parray *parent_chain, + pgBackup *full_backup, pgBackup *dest_backup, + bool no_validate, bool no_sync); + +extern parray *read_database_map(pgBackup *backup); /* in init.c */ -extern int do_init(void); -extern int do_add_instance(void); +extern int do_init(CatalogState *catalogState); +extern int do_add_instance(InstanceState *instanceState, InstanceConfig *instance); /* in archive.c */ -extern int do_archive_push(char *wal_file_path, char *wal_file_name, - bool overwrite); -extern int do_archive_get(char *wal_file_path, char *wal_file_name); - +extern void do_archive_push(InstanceState *instanceState, InstanceConfig *instance, char *pg_xlog_dir, + char *wal_file_name, int batch_size, bool overwrite, + bool no_sync, bool no_ready_rename); +extern void do_archive_get(InstanceState *instanceState, InstanceConfig *instance, const char *prefetch_dir_arg, char *wal_file_path, + char *wal_file_name, int batch_size, bool validate_wal); /* in configure.c */ -extern void do_show_config(void); -extern void do_set_config(bool missing_ok); -extern void init_config(InstanceConfig *config); +extern void do_show_config(bool show_base_units); +extern void do_set_config(InstanceState *instanceState, bool missing_ok); +extern void init_config(InstanceConfig *config, const char *instance_name); +extern InstanceConfig *readInstanceConfigFile(InstanceState *instanceState); /* in show.c */ -extern int do_show(time_t requested_backup_id); +extern int do_show(CatalogState *catalogState, InstanceState *instanceState, + time_t requested_backup_id, bool show_archive); +extern void memorize_environment_locale(void); +extern void free_environment_locale(void); /* in delete.c */ -extern void do_delete(time_t backup_id); +extern void do_delete(InstanceState *instanceState, time_t backup_id); extern void delete_backup_files(pgBackup *backup); -extern int do_retention(void); -extern int do_delete_instance(void); +extern void do_retention(InstanceState *instanceState, bool no_validate, bool no_sync); +extern int do_delete_instance(InstanceState *instanceState); +extern void do_delete_status(InstanceState *instanceState, + InstanceConfig *instance_config, const char *status); /* in fetch.c */ extern char *slurpFile(const char *datadir, @@ -534,37 +967,65 @@ extern char *slurpFile(const char *datadir, extern char *fetchFile(PGconn *conn, const char *filename, size_t *filesize); /* in help.c */ +extern void help_print_version(void); extern void help_pg_probackup(void); -extern void help_command(char *command); +extern void help_command(ProbackupSubcmd const subcmd); /* in validate.c */ -extern void pgBackupValidate(pgBackup* backup); -extern int do_validate_all(void); +extern void pgBackupValidate(pgBackup* backup, pgRestoreParams *params); +extern int do_validate_all(CatalogState *catalogState, InstanceState *instanceState); +extern int validate_one_page(Page page, BlockNumber absolute_blkno, + XLogRecPtr stop_lsn, PageState *page_st, + uint32 checksum_version); +extern bool validate_tablespace_map(pgBackup *backup, bool no_validate); + +extern parray* get_history_streaming(ConnectionOptions *conn_opt, TimeLineID tli, parray *backup_list); + +/* return codes for validate_one_page */ +/* TODO: use enum */ +#define PAGE_IS_VALID (-1) +#define PAGE_IS_NOT_FOUND (-2) +#define PAGE_IS_ZEROED (-3) +#define PAGE_HEADER_IS_INVALID (-4) +#define PAGE_CHECKSUM_MISMATCH (-5) +#define PAGE_LSN_FROM_FUTURE (-6) /* in catalog.c */ -extern pgBackup *read_backup(time_t timestamp); -extern void write_backup(pgBackup *backup); -extern void write_backup_status(pgBackup *backup, BackupStatus status); +extern pgBackup *read_backup(const char *root_dir); +extern void write_backup(pgBackup *backup, bool strict); +extern void write_backup_status(pgBackup *backup, BackupStatus status, + bool strict); extern void write_backup_data_bytes(pgBackup *backup); -extern bool lock_backup(pgBackup *backup); +extern bool lock_backup(pgBackup *backup, bool strict, bool exclusive); + +extern const char *pgBackupGetBackupMode(pgBackup *backup, bool show_color); +extern void pgBackupGetBackupModeColor(pgBackup *backup, char *mode); -extern const char *pgBackupGetBackupMode(pgBackup *backup); +extern parray *catalog_get_instance_list(CatalogState *catalogState); -extern parray *catalog_get_backup_list(time_t requested_backup_id); +extern parray *catalog_get_backup_list(InstanceState *instanceState, time_t requested_backup_id); extern void catalog_lock_backup_list(parray *backup_list, int from_idx, - int to_idx); + int to_idx, bool strict, bool exclusive); extern pgBackup *catalog_get_last_data_backup(parray *backup_list, TimeLineID tli, time_t current_start_time); -extern void pgBackupWriteControl(FILE *out, pgBackup *backup); +extern pgBackup *get_multi_timeline_parent(parray *backup_list, parray *tli_list, + TimeLineID current_tli, time_t current_start_time, + InstanceConfig *instance); +extern timelineInfo *timelineInfoNew(TimeLineID tli); +extern void timelineInfoFree(void *tliInfo); +extern parray *catalog_get_timelines(InstanceState *instanceState, InstanceConfig *instance); +extern void do_set_backup(InstanceState *instanceState, time_t backup_id, + pgSetBackupParams *set_backup_params); +extern void pin_backup(pgBackup *target_backup, + pgSetBackupParams *set_backup_params); +extern void add_note(pgBackup *target_backup, char *note); +extern void pgBackupWriteControl(FILE *out, pgBackup *backup, bool utc); extern void write_backup_filelist(pgBackup *backup, parray *files, - const char *root, parray *external_list); + const char *root, parray *external_list, bool sync); -extern void pgBackupGetPath(const pgBackup *backup, char *path, size_t len, - const char *subdir); -extern void pgBackupGetPath2(const pgBackup *backup, char *path, size_t len, - const char *subdir1, const char *subdir2); -extern int pgBackupCreateDir(pgBackup *backup); + +extern void pgBackupInitDir(pgBackup *backup, const char *backup_instance_path); extern void pgNodeInit(PGNodeInfo *node); extern void pgBackupInit(pgBackup *backup); extern void pgBackupFree(void *backup); @@ -574,10 +1035,14 @@ extern int pgBackupCompareIdEqual(const void *l, const void *r); extern pgBackup* find_parent_full_backup(pgBackup *current_backup); extern int scan_parent_chain(pgBackup *current_backup, pgBackup **result_backup); +/* return codes for scan_parent_chain */ +#define ChainIsBroken 0 +#define ChainIsInvalid 1 +#define ChainIsOk 2 + extern bool is_parent(time_t parent_backup_time, pgBackup *child_backup, bool inclusive); extern bool is_prolific(parray *backup_list, pgBackup *target_backup); -extern bool in_backup_list(parray *backup_list, pgBackup *target_backup); -extern int get_backup_index_number(parray *backup_list, pgBackup *backup); +extern void append_children(parray *backup_list, pgBackup *target_backup, parray *append_list); extern bool launch_agent(void); extern void launch_ssh(char* argv[]); extern void wait_ssh(void); @@ -589,26 +1054,36 @@ extern CompressAlg parse_compress_alg(const char *arg); extern const char* deparse_compress_alg(int alg); /* in dir.c */ +extern bool get_control_value_int64(const char *str, const char *name, int64 *value_int64, bool is_mandatory); +extern bool get_control_value_str(const char *str, const char *name, + char *value_str, size_t value_str_size, bool is_mandatory); extern void dir_list_file(parray *files, const char *root, bool exclude, - bool follow_symlink, bool add_root, int external_dir_num, fio_location location); + bool follow_symlink, bool add_root, bool backup_logs, + bool skip_hidden, int external_dir_num, fio_location location); +extern const char *get_tablespace_mapping(const char *dir); extern void create_data_directories(parray *dest_files, const char *data_dir, const char *backup_dir, bool extract_tablespaces, - fio_location location); + bool incremental, + fio_location location, + const char *waldir_path); -extern void read_tablespace_map(parray *files, const char *backup_dir); +extern void read_tablespace_map(parray *links, const char *backup_dir); extern void opt_tablespace_map(ConfigOption *opt, const char *arg); extern void opt_externaldir_map(ConfigOption *opt, const char *arg); -extern void check_tablespace_mapping(pgBackup *backup); -extern void check_external_dir_mapping(pgBackup *backup); +extern int check_tablespace_mapping(pgBackup *backup, bool incremental, bool force, bool pgdata_is_empty, bool no_validate); +extern void check_external_dir_mapping(pgBackup *backup, bool incremental); extern char *get_external_remap(char *current_dir); +extern void print_database_map(FILE *out, parray *database_list); +extern void write_database_map(pgBackup *backup, parray *database_list, + parray *backup_file_list); +extern void db_map_entry_free(void *map); + extern void print_file_list(FILE *out, const parray *files, const char *root, const char *external_prefix, parray *external_list); -extern parray *dir_read_file_list(const char *root, const char *external_prefix, - const char *file_txt, fio_location location); extern parray *make_external_directory_list(const char *colon_separated_dirs, bool remap); extern void free_dir_list(parray *list); @@ -616,7 +1091,7 @@ extern void makeExternalDirPathByNum(char *ret_path, const char *pattern_path, const int dir_num); extern bool backup_contains_external(const char *dir, parray *dirs_list); -extern int dir_create_dir(const char *path, mode_t mode); +extern int dir_create_dir(const char *path, mode_t mode, bool strict); extern bool dir_is_empty(const char *path, fio_location location); extern bool fileExists(const char *path, fio_location location); @@ -625,86 +1100,272 @@ extern size_t pgFileSize(const char *path); extern pgFile *pgFileNew(const char *path, const char *rel_path, bool follow_symlink, int external_dir_num, fio_location location); -extern pgFile *pgFileInit(const char *path, const char *rel_path); -extern void pgFileDelete(pgFile *file); +extern pgFile *pgFileInit(const char *rel_path); +extern void pgFileDelete(mode_t mode, const char *full_path); +extern void fio_pgFileDelete(pgFile *file, const char *full_path); + extern void pgFileFree(void *file); -extern pg_crc32 pgFileGetCRC(const char *file_path, bool use_crc32c, - bool raise_on_deleted, size_t *bytes_read, fio_location location); + +extern pg_crc32 pgFileGetCRC(const char *file_path, bool use_crc32c, bool missing_ok); +extern pg_crc32 pgFileGetCRCTruncated(const char *file_path, bool use_crc32c, bool missing_ok); +extern pg_crc32 pgFileGetCRCgz(const char *file_path, bool use_crc32c, bool missing_ok); + +extern int pgFileMapComparePath(const void *f1, const void *f2); extern int pgFileCompareName(const void *f1, const void *f2); -extern int pgFileComparePath(const void *f1, const void *f2); -extern int pgFileComparePathWithExternal(const void *f1, const void *f2); +extern int pgFileCompareNameWithString(const void *f1, const void *f2); +extern int pgFileCompareRelPathWithString(const void *f1, const void *f2); extern int pgFileCompareRelPathWithExternal(const void *f1, const void *f2); -extern int pgFileComparePathDesc(const void *f1, const void *f2); -extern int pgFileComparePathWithExternalDesc(const void *f1, const void *f2); +extern int pgFileCompareRelPathWithExternalDesc(const void *f1, const void *f2); extern int pgFileCompareLinked(const void *f1, const void *f2); extern int pgFileCompareSize(const void *f1, const void *f2); +extern int pgFileCompareSizeDesc(const void *f1, const void *f2); +extern int pgCompareString(const void *str1, const void *str2); +extern int pgPrefixCompareString(const void *str1, const void *str2); +extern int pgCompareOid(const void *f1, const void *f2); +extern void pfilearray_clear_locks(parray *file_list); +extern bool set_forkname(pgFile *file); /* in data.c */ -extern bool check_data_file(ConnectionArgs* arguments, pgFile* file, uint32 checksum_version); -extern bool backup_data_file(backup_files_arg* arguments, - const char *to_path, pgFile *file, - XLogRecPtr prev_backup_start_lsn, - BackupMode backup_mode, - CompressAlg calg, int clevel, - bool missing_ok); -extern void restore_data_file(const char *to_path, - pgFile *file, bool allow_truncate, - bool write_header, - uint32 backup_version); -extern bool copy_file(fio_location from_location, const char *to_root, - fio_location to_location, pgFile *file, bool missing_ok); - -extern bool check_file_pages(pgFile *file, XLogRecPtr stop_lsn, - uint32 checksum_version, uint32 backup_version); +extern bool check_data_file(ConnectionArgs *arguments, pgFile *file, + const char *from_fullpath, uint32 checksum_version); + + +extern void catchup_data_file(pgFile *file, const char *from_fullpath, const char *to_fullpath, + XLogRecPtr sync_lsn, BackupMode backup_mode, + uint32 checksum_version, size_t prev_size); +extern void backup_data_file(pgFile *file, const char *from_fullpath, const char *to_fullpath, + XLogRecPtr prev_backup_start_lsn, BackupMode backup_mode, + CompressAlg calg, int clevel, uint32 checksum_version, + HeaderMap *hdr_map, bool missing_ok); +extern void backup_non_data_file(pgFile *file, pgFile *prev_file, + const char *from_fullpath, const char *to_fullpath, + BackupMode backup_mode, time_t parent_backup_time, + bool missing_ok); +extern void backup_non_data_file_internal(const char *from_fullpath, + fio_location from_location, + const char *to_fullpath, pgFile *file, + bool missing_ok); + +extern size_t restore_data_file(parray *parent_chain, pgFile *dest_file, FILE *out, + const char *to_fullpath, bool use_bitmap, PageState *checksum_map, + XLogRecPtr shift_lsn, datapagemap_t *lsn_map, bool use_headers); +extern size_t restore_data_file_internal(FILE *in, FILE *out, pgFile *file, uint32 backup_version, + const char *from_fullpath, const char *to_fullpath, int nblocks, + datapagemap_t *map, PageState *checksum_map, int checksum_version, + datapagemap_t *lsn_map, BackupPageHeader2 *headers); +extern size_t restore_non_data_file(parray *parent_chain, pgBackup *dest_backup, + pgFile *dest_file, FILE *out, const char *to_fullpath, + bool already_exists); +extern void restore_non_data_file_internal(FILE *in, FILE *out, pgFile *file, + const char *from_fullpath, const char *to_fullpath); +extern bool create_empty_file(fio_location from_location, const char *to_root, + fio_location to_location, pgFile *file); + +extern PageState *get_checksum_map(const char *fullpath, uint32 checksum_version, + int n_blocks, XLogRecPtr dest_stop_lsn, BlockNumber segmentno); +extern datapagemap_t *get_lsn_map(const char *fullpath, uint32 checksum_version, + int n_blocks, XLogRecPtr shift_lsn, BlockNumber segmentno); +extern bool validate_file_pages(pgFile *file, const char *fullpath, XLogRecPtr stop_lsn, + uint32 checksum_version, uint32 backup_version, HeaderMap *hdr_map); + +extern BackupPageHeader2* get_data_file_headers(HeaderMap *hdr_map, pgFile *file, uint32 backup_version, bool strict); +extern void write_page_headers(BackupPageHeader2 *headers, pgFile *file, HeaderMap *hdr_map, bool is_merge); +extern void init_header_map(pgBackup *backup); +extern void cleanup_header_map(HeaderMap *hdr_map); /* parsexlog.c */ -extern void extractPageMap(const char *archivedir, - TimeLineID tli, uint32 seg_size, - XLogRecPtr startpoint, XLogRecPtr endpoint); +extern bool extractPageMap(const char *archivedir, uint32 wal_seg_size, + XLogRecPtr startpoint, TimeLineID start_tli, + XLogRecPtr endpoint, TimeLineID end_tli, + parray *tli_list); extern void validate_wal(pgBackup *backup, const char *archivedir, time_t target_time, TransactionId target_xid, XLogRecPtr target_lsn, TimeLineID tli, uint32 seg_size); +extern bool validate_wal_segment(TimeLineID tli, XLogSegNo segno, + const char *prefetch_dir, uint32 wal_seg_size); extern bool read_recovery_info(const char *archivedir, TimeLineID tli, uint32 seg_size, XLogRecPtr start_lsn, XLogRecPtr stop_lsn, - time_t *recovery_time, - TransactionId *recovery_xid); + time_t *recovery_time); extern bool wal_contains_lsn(const char *archivedir, XLogRecPtr target_lsn, TimeLineID target_tli, uint32 seg_size); -extern XLogRecPtr get_last_wal_lsn(const char *archivedir, XLogRecPtr start_lsn, +extern XLogRecPtr get_prior_record_lsn(const char *archivedir, XLogRecPtr start_lsn, XLogRecPtr stop_lsn, TimeLineID tli, bool seek_prev_segment, uint32 seg_size); +extern XLogRecPtr get_first_record_lsn(const char *archivedir, XLogRecPtr start_lsn, + TimeLineID tli, uint32 wal_seg_size, int timeout); +extern XLogRecPtr get_next_record_lsn(const char *archivedir, XLogSegNo segno, TimeLineID tli, + uint32 wal_seg_size, int timeout, XLogRecPtr target); + /* in util.c */ -extern TimeLineID get_current_timeline(bool safe); +extern TimeLineID get_current_timeline(PGconn *conn); +extern TimeLineID get_current_timeline_from_control(const char *pgdata_path, fio_location location, bool safe); extern XLogRecPtr get_checkpoint_location(PGconn *conn); -extern uint64 get_system_identifier(const char *pgdata_path); +extern uint64 get_system_identifier(const char *pgdata_path, fio_location location, bool safe); extern uint64 get_remote_system_identifier(PGconn *conn); extern uint32 get_data_checksum_version(bool safe); extern pg_crc32c get_pgcontrol_checksum(const char *pgdata_path); -extern uint32 get_xlog_seg_size(char *pgdata_path); +extern uint32 get_xlog_seg_size(const char *pgdata_path); +extern void get_redo(const char *pgdata_path, fio_location pgdata_location, RedoParams *redo); extern void set_min_recovery_point(pgFile *file, const char *backup_path, XLogRecPtr stop_backup_lsn); -extern void copy_pgcontrol_file(const char *from_root, fio_location location, const char *to_root, fio_location to_location, - pgFile *file); +extern void get_control_file_or_back_file(const char *pgdata_path, fio_location location, + ControlFileData *control); +extern void copy_pgcontrol_file(const char *from_fullpath, fio_location from_location, + const char *to_fullpath, fio_location to_location, pgFile *file); -extern void time2iso(char *buf, size_t len, time_t time); +extern void time2iso(char *buf, size_t len, time_t time, bool utc); extern const char *status2str(BackupStatus status); -extern const char *base36enc(long unsigned int value); -extern char *base36enc_dup(long unsigned int value); +const char *status2str_color(BackupStatus status); +extern BackupStatus str2status(const char *status); +extern const char *base36enc_to(long unsigned int value, char buf[ARG_SIZE_HINT base36bufsize]); +/* Abuse C99 Compound Literal's lifetime */ +#define base36enc(value) (base36enc_to((value), (char[base36bufsize]){0})) extern long unsigned int base36dec(const char *text); extern uint32 parse_server_version(const char *server_version_str); extern uint32 parse_program_version(const char *program_version); extern bool parse_page(Page page, XLogRecPtr *lsn); -int32 do_compress(void* dst, size_t dst_size, void const* src, size_t src_size, - CompressAlg alg, int level, const char **errormsg); +extern int32 do_compress(void* dst, size_t dst_size, void const* src, size_t src_size, + CompressAlg alg, int level, const char **errormsg); +extern int32 do_decompress(void* dst, size_t dst_size, void const* src, size_t src_size, + CompressAlg alg, const char **errormsg); extern void pretty_size(int64 size, char *buf, size_t len); - +extern void pretty_time_interval(double time, char *buf, size_t len); extern PGconn *pgdata_basic_setup(ConnectionOptions conn_opt, PGNodeInfo *nodeInfo); -extern void check_system_identifiers(PGconn *conn, char *pgdata); +extern void check_system_identifiers(PGconn *conn, const char *pgdata); extern void parse_filelist_filenames(parray *files, const char *root); +/* in ptrack.c */ +extern void make_pagemap_from_ptrack_2(parray* files, PGconn* backup_conn, + const char *ptrack_schema, + int ptrack_version_num, + XLogRecPtr lsn); +extern void get_ptrack_version(PGconn *backup_conn, PGNodeInfo *nodeInfo); +extern bool pg_is_ptrack_enabled(PGconn *backup_conn, int ptrack_version_num); + +extern XLogRecPtr get_last_ptrack_lsn(PGconn *backup_conn, PGNodeInfo *nodeInfo); +extern parray * pg_ptrack_get_pagemapset(PGconn *backup_conn, const char *ptrack_schema, + int ptrack_version_num, XLogRecPtr lsn); + +/* open local file to writing */ +extern FILE* open_local_file_rw(const char *to_fullpath, char **out_buf, uint32 buf_size); + +extern int send_pages(const char *to_fullpath, const char *from_fullpath, + pgFile *file, XLogRecPtr prev_backup_start_lsn, CompressAlg calg, int clevel, + uint32 checksum_version, bool use_pagemap, BackupPageHeader2 **headers, + BackupMode backup_mode); +extern int copy_pages(const char *to_fullpath, const char *from_fullpath, + pgFile *file, XLogRecPtr prev_backup_start_lsn, + uint32 checksum_version, bool use_pagemap, + BackupMode backup_mode); + +/* FIO */ +extern void setMyLocation(ProbackupSubcmd const subcmd); +extern void fio_delete(mode_t mode, const char *fullpath, fio_location location); +extern int fio_send_pages(const char *to_fullpath, const char *from_fullpath, pgFile *file, + XLogRecPtr horizonLsn, int calg, int clevel, uint32 checksum_version, + bool use_pagemap, BlockNumber *err_blknum, char **errormsg, + BackupPageHeader2 **headers); +extern int fio_copy_pages(const char *to_fullpath, const char *from_fullpath, pgFile *file, + XLogRecPtr horizonLsn, int calg, int clevel, uint32 checksum_version, + bool use_pagemap, BlockNumber *err_blknum, char **errormsg); +/* return codes for fio_send_pages */ +extern int fio_send_file_gz(const char *from_fullpath, FILE* out, char **errormsg); +extern int fio_send_file(const char *from_fullpath, FILE* out, bool cut_zero_tail, + pgFile *file, char **errormsg); +extern int fio_send_file_local(const char *from_fullpath, FILE* out, bool cut_zero_tail, + pgFile *file, char **errormsg); + +extern void fio_list_dir(parray *files, const char *root, bool exclude, bool follow_symlink, + bool add_root, bool backup_logs, bool skip_hidden, int external_dir_num); + +extern bool pgut_rmtree(const char *path, bool rmtopdir, bool strict); + +extern void pgut_setenv(const char *key, const char *val); +extern void pgut_unsetenv(const char *key); + +extern PageState *fio_get_checksum_map(const char *fullpath, uint32 checksum_version, int n_blocks, + XLogRecPtr dest_stop_lsn, BlockNumber segmentno, fio_location location); + +extern datapagemap_t *fio_get_lsn_map(const char *fullpath, uint32 checksum_version, + int n_blocks, XLogRecPtr horizonLsn, BlockNumber segmentno, + fio_location location); +extern pid_t fio_check_postmaster(const char *pgdata, fio_location location); + +extern int32 fio_decompress(void* dst, void const* src, size_t size, int compress_alg, char **errormsg); + +/* return codes for fio_send_pages() and fio_send_file() */ +#define SEND_OK (0) +#define FILE_MISSING (-1) +#define OPEN_FAILED (-2) +#define READ_FAILED (-3) +#define WRITE_FAILED (-4) +#define ZLIB_ERROR (-5) +#define REMOTE_ERROR (-6) +#define PAGE_CORRUPTION (-8) + +/* Check if specified location is local for current node */ +extern bool fio_is_remote(fio_location location); +extern bool fio_is_remote_simple(fio_location location); + +extern void get_header_errormsg(Page page, char **errormsg); +extern void get_checksum_errormsg(Page page, char **errormsg, + BlockNumber absolute_blkno); + +extern bool +datapagemap_is_set(datapagemap_t *map, BlockNumber blkno); + +extern void +datapagemap_print_debug(datapagemap_t *map); + +/* in stream.c */ +extern XLogRecPtr stop_backup_lsn; +extern void start_WAL_streaming(PGconn *backup_conn, char *stream_dst_path, + ConnectionOptions *conn_opt, + XLogRecPtr startpos, TimeLineID starttli, + bool is_backup); +extern int wait_WAL_streaming_end(parray *backup_files_list); +extern parray* parse_tli_history_buffer(char *history, TimeLineID tli); + +/* external variables and functions, implemented in backup.c */ +typedef struct PGStopBackupResult +{ + /* + * We will use values of snapshot_xid and invocation_time if there are + * no transactions between start_lsn and stop_lsn. + */ + TransactionId snapshot_xid; + time_t invocation_time; + /* + * Fields that store pg_catalog.pg_stop_backup() result + */ + XLogRecPtr lsn; + size_t backup_label_content_len; + char *backup_label_content; + size_t tablespace_map_content_len; + char *tablespace_map_content; +} PGStopBackupResult; + +extern bool backup_in_progress; +extern parray *backup_files_list; + +extern void pg_start_backup(const char *label, bool smooth, pgBackup *backup, + PGNodeInfo *nodeInfo, PGconn *conn); +extern void pg_silent_client_messages(PGconn *conn); +extern void pg_create_restore_point(PGconn *conn, time_t backup_start_time); +extern void pg_stop_backup_send(PGconn *conn, int server_version, bool is_started_on_replica, bool is_exclusive, char **query_text); +extern void pg_stop_backup_consume(PGconn *conn, int server_version, + bool is_exclusive, uint32 timeout, const char *query_text, + PGStopBackupResult *result); +extern void pg_stop_backup_write_file_helper(const char *path, const char *filename, const char *error_msg_filename, + const void *data, size_t len, parray *file_list); +extern XLogRecPtr wait_wal_lsn(const char *wal_segment_dir, XLogRecPtr lsn, bool is_start_lsn, TimeLineID tli, + bool in_prev_segment, bool segment_only, + int timeout_elevel, bool in_stream_dir); +extern void wait_wal_and_calculate_stop_lsn(const char *xlog_path, XLogRecPtr stop_lsn, pgBackup *backup); +extern int64 calculate_datasize_of_filelist(parray *filelist); #endif /* PG_PROBACKUP_H */ diff --git a/src/pg_probackup_state.h b/src/pg_probackup_state.h new file mode 100644 index 000000000..56d852537 --- /dev/null +++ b/src/pg_probackup_state.h @@ -0,0 +1,31 @@ +/*------------------------------------------------------------------------- + * + * pg_probackup_state.h: Definitions of internal pg_probackup states + * + * Portions Copyright (c) 2021, Postgres Professional + * + *------------------------------------------------------------------------- + */ +#ifndef PG_PROBACKUP_STATE_H +#define PG_PROBACKUP_STATE_H + +/* ====== CatalogState ======= */ + +typedef struct CatalogState +{ + /* $BACKUP_PATH */ + char catalog_path[MAXPGPATH]; //previously global var backup_path + /* $BACKUP_PATH/backups */ + char backup_subdir_path[MAXPGPATH]; + /* $BACKUP_PATH/wal */ + char wal_subdir_path[MAXPGPATH]; // previously global var arclog_path +} CatalogState; + +/* ====== CatalogState (END) ======= */ + + +/* ===== instanceState ===== */ + +/* ===== instanceState (END) ===== */ + +#endif /* PG_PROBACKUP_STATE_H */ diff --git a/src/ptrack.c b/src/ptrack.c new file mode 100644 index 000000000..d27629e45 --- /dev/null +++ b/src/ptrack.c @@ -0,0 +1,309 @@ +/*------------------------------------------------------------------------- + * + * ptrack.c: support functions for ptrack backups + * + * Copyright (c) 2021 Postgres Professional + * + *------------------------------------------------------------------------- + */ + +#include "pg_probackup.h" + +#if PG_VERSION_NUM < 110000 +#include "catalog/catalog.h" +#endif +#include "catalog/pg_tablespace.h" + +/* + * Macro needed to parse ptrack. + * NOTE Keep those values synchronized with definitions in ptrack.h + */ +#define PTRACK_BITS_PER_HEAPBLOCK 1 +#define HEAPBLOCKS_PER_BYTE (BITS_PER_BYTE / PTRACK_BITS_PER_HEAPBLOCK) + +/* + * Parse a string like "2.1" into int + * result: int by formula major_number * 100 + minor_number + * or -1 if string cannot be parsed + */ +static int +ptrack_parse_version_string(const char *version_str) +{ + int ma, mi; + int sscanf_readed_count; + if (sscanf(version_str, "%u.%2u%n", &ma, &mi, &sscanf_readed_count) != 2) + return -1; + if (sscanf_readed_count != strlen(version_str)) + return -1; + return ma * 100 + mi; +} + +/* Check if the instance supports compatible version of ptrack, + * fill-in version number if it does. + * Also for ptrack 2.x save schema namespace. + */ +void +get_ptrack_version(PGconn *backup_conn, PGNodeInfo *nodeInfo) +{ + PGresult *res_db; + char *ptrack_version_str; + int ptrack_version_num; + + res_db = pgut_execute(backup_conn, + "SELECT extnamespace::regnamespace, extversion " + "FROM pg_catalog.pg_extension WHERE extname = 'ptrack'::name", + 0, NULL); + + if (PQntuples(res_db) > 0) + { + /* ptrack 2.x is supported, save schema name and version */ + nodeInfo->ptrack_schema = pgut_strdup(PQgetvalue(res_db, 0, 0)); + + if (nodeInfo->ptrack_schema == NULL) + elog(ERROR, "Failed to obtain schema name of ptrack extension"); + + ptrack_version_str = PQgetvalue(res_db, 0, 1); + } + else + { + /* ptrack 1.x is supported, save version */ + PQclear(res_db); + res_db = pgut_execute(backup_conn, + "SELECT proname FROM pg_catalog.pg_proc WHERE proname='ptrack_version'::name", + 0, NULL); + + if (PQntuples(res_db) == 0) + { + /* ptrack is not supported */ + PQclear(res_db); + return; + } + + /* + * it's ok not to have permission to call this old function in PGPRO-11 version (ok_error = true) + * see deprication notice https://postgrespro.com/docs/postgrespro/11/release-pro-11-9-1 + */ + res_db = pgut_execute_extended(backup_conn, + "SELECT pg_catalog.ptrack_version()", + 0, NULL, true, true); + if (PQntuples(res_db) == 0) + { + PQclear(res_db); + elog(WARNING, "Can't call pg_catalog.ptrack_version(), it is assumed that there is no ptrack extension installed."); + return; + } + ptrack_version_str = PQgetvalue(res_db, 0, 0); + } + + ptrack_version_num = ptrack_parse_version_string(ptrack_version_str); + if (ptrack_version_num == -1) + /* leave default nodeInfo->ptrack_version_num = 0 from pgNodeInit() */ + elog(WARNING, "Cannot parse ptrack version string \"%s\"", + ptrack_version_str); + else + nodeInfo->ptrack_version_num = ptrack_version_num; + + /* ptrack 1.X is buggy, so fall back to DELTA backup strategy for safety */ + if (nodeInfo->ptrack_version_num < 200) + { + if (current.backup_mode == BACKUP_MODE_DIFF_PTRACK) + { + elog(WARNING, "Update your ptrack to the version 2.1 or upper. Current version is %s. " + "Fall back to DELTA backup.", + ptrack_version_str); + current.backup_mode = BACKUP_MODE_DIFF_DELTA; + } + } + + PQclear(res_db); +} + +/* + * Check if ptrack is enabled in target instance + */ +bool +pg_is_ptrack_enabled(PGconn *backup_conn, int ptrack_version_num) +{ + PGresult *res_db; + bool result = false; + + if (ptrack_version_num > 200) + { + res_db = pgut_execute(backup_conn, "SHOW ptrack.map_size", 0, NULL); + result = strcmp(PQgetvalue(res_db, 0, 0), "0") != 0 && + strcmp(PQgetvalue(res_db, 0, 0), "-1") != 0; + PQclear(res_db); + } + else if (ptrack_version_num == 200) + { + res_db = pgut_execute(backup_conn, "SHOW ptrack_map_size", 0, NULL); + result = strcmp(PQgetvalue(res_db, 0, 0), "0") != 0; + PQclear(res_db); + } + else + { + result = false; + } + + return result; +} + +/* + * Get lsn of the moment when ptrack was enabled the last time. + */ +XLogRecPtr +get_last_ptrack_lsn(PGconn *backup_conn, PGNodeInfo *nodeInfo) + +{ + PGresult *res; + uint32 lsn_hi; + uint32 lsn_lo; + XLogRecPtr lsn; + + char query[128]; + + if (nodeInfo->ptrack_version_num == 200) + sprintf(query, "SELECT %s.pg_ptrack_control_lsn()", nodeInfo->ptrack_schema); + else + sprintf(query, "SELECT %s.ptrack_init_lsn()", nodeInfo->ptrack_schema); + + res = pgut_execute(backup_conn, query, 0, NULL); + + /* Extract timeline and LSN from results of pg_start_backup() */ + XLogDataFromLSN(PQgetvalue(res, 0, 0), &lsn_hi, &lsn_lo); + /* Calculate LSN */ + lsn = ((uint64) lsn_hi) << 32 | lsn_lo; + + PQclear(res); + return lsn; +} + +/* ---------------------------- + * Ptrack 2.* support functions + * ---------------------------- + */ + +/* + * Fetch a list of changed files with their ptrack maps. + */ +parray * +pg_ptrack_get_pagemapset(PGconn *backup_conn, const char *ptrack_schema, + int ptrack_version_num, XLogRecPtr lsn) +{ + PGresult *res; + char lsn_buf[17 + 1]; + char *params[1]; + parray *pagemapset = NULL; + int i; + char query[512]; + + snprintf(lsn_buf, sizeof lsn_buf, "%X/%X", (uint32) (lsn >> 32), (uint32) lsn); + params[0] = pstrdup(lsn_buf); + + if (!ptrack_schema) + elog(ERROR, "Schema name of ptrack extension is missing"); + + if (ptrack_version_num == 200) + sprintf(query, "SELECT path, pagemap FROM %s.pg_ptrack_get_pagemapset($1) ORDER BY 1", + ptrack_schema); + else + sprintf(query, "SELECT path, pagemap FROM %s.ptrack_get_pagemapset($1) ORDER BY 1", + ptrack_schema); + + res = pgut_execute(backup_conn, query, 1, (const char **) params); + pfree(params[0]); + + if (PQnfields(res) != 2) + elog(ERROR, "Cannot get ptrack pagemapset"); + + /* sanity ? */ + + /* Construct database map */ + for (i = 0; i < PQntuples(res); i++) + { + page_map_entry *pm_entry = (page_map_entry *) pgut_malloc(sizeof(page_map_entry)); + + /* get path */ + pm_entry->path = pgut_strdup(PQgetvalue(res, i, 0)); + + /* get bytea */ + pm_entry->pagemap = (char *) PQunescapeBytea((unsigned char *) PQgetvalue(res, i, 1), + &pm_entry->pagemapsize); + + if (pagemapset == NULL) + pagemapset = parray_new(); + + parray_append(pagemapset, pm_entry); + } + + PQclear(res); + + return pagemapset; +} + +/* + * Given a list of files in the instance to backup, build a pagemap for each + * data file that has ptrack. Result is saved in the pagemap field of pgFile. + * + * We fetch a list of changed files with their ptrack maps. After that files + * are merged with their bitmaps. File without bitmap is treated as unchanged. + */ +void +make_pagemap_from_ptrack_2(parray *files, + PGconn *backup_conn, + const char *ptrack_schema, + int ptrack_version_num, + XLogRecPtr lsn) +{ + parray *filemaps; + int file_i = 0; + page_map_entry *dummy_map = NULL; + + /* Receive all available ptrack bitmaps at once */ + filemaps = pg_ptrack_get_pagemapset(backup_conn, ptrack_schema, + ptrack_version_num, lsn); + + if (filemaps != NULL) + parray_qsort(filemaps, pgFileMapComparePath); + else + return; + + dummy_map = (page_map_entry *) pgut_malloc(sizeof(page_map_entry)); + + /* Iterate over files and look for corresponding pagemap if any */ + for (file_i = 0; file_i < parray_num(files); file_i++) + { + pgFile *file = (pgFile *) parray_get(files, file_i); + page_map_entry **res_map = NULL; + page_map_entry *map = NULL; + + /* + * For now nondata files are not entitled to have pagemap + * TODO It's possible to use ptrack for incremental backup of + * relation forks. Not implemented yet. + */ + if (!file->is_datafile || file->is_cfs) + continue; + + /* Consider only files from PGDATA (this check is probably redundant) */ + if (file->external_dir_num != 0) + continue; + + if (filemaps) + { + dummy_map->path = file->rel_path; + res_map = parray_bsearch(filemaps, dummy_map, pgFileMapComparePath); + map = (res_map) ? *res_map : NULL; + } + + /* Found map */ + if (map) + { + elog(VERBOSE, "Using ptrack pagemap for file \"%s\"", file->rel_path); + file->pagemap.bitmapsize = map->pagemapsize; + file->pagemap.bitmap = map->pagemap; + } + } + + free(dummy_map); +} diff --git a/src/restore.c b/src/restore.c index c3c279b78..f9310dcee 100644 --- a/src/restore.c +++ b/src/restore.c @@ -3,7 +3,7 @@ * restore.c: restore DB cluster and archived WAL. * * Portions Copyright (c) 2009-2013, NIPPON TELEGRAPH AND TELEPHONE CORPORATION - * Portions Copyright (c) 2015-2019, Postgres Professional + * Portions Copyright (c) 2015-2022, Postgres Professional * *------------------------------------------------------------------------- */ @@ -19,12 +19,18 @@ typedef struct { - parray *files; - pgBackup *backup; - parray *external_dirs; - char *external_prefix; - parray *dest_external_dirs; + parray *pgdata_files; parray *dest_files; + pgBackup *dest_backup; + parray *dest_external_dirs; + parray *parent_chain; + parray *dbOid_exclude_list; + bool skip_external_dirs; + const char *to_root; + size_t restored_bytes; + bool use_bitmap; + IncrRestoreMode incremental_mode; + XLogRecPtr shift_lsn; /* used only in LSN incremental_mode */ /* * Return value from the thread. @@ -33,50 +39,196 @@ typedef struct int ret; } restore_files_arg; -static void restore_backup(pgBackup *backup, parray *dest_external_dirs, - parray *dest_files); -static void create_recovery_conf(time_t backup_id, +static bool control_downloaded = false; +static ControlFileData instance_control; + +static void +print_recovery_settings(InstanceState *instanceState, FILE *fp, pgBackup *backup, + pgRestoreParams *params, pgRecoveryTarget *rt); +static void +print_standby_settings_common(FILE *fp, pgBackup *backup, pgRestoreParams *params); + +#if PG_VERSION_NUM >= 120000 +static void +update_recovery_options(InstanceState *instanceState, pgBackup *backup, + pgRestoreParams *params, pgRecoveryTarget *rt); +#else +static void +update_recovery_options_before_v12(InstanceState *instanceState, pgBackup *backup, + pgRestoreParams *params, pgRecoveryTarget *rt); +#endif + +static void create_recovery_conf(InstanceState *instanceState, time_t backup_id, pgRecoveryTarget *rt, - pgBackup *backup); -static parray *read_timeline_history(TimeLineID targetTLI); + pgBackup *backup, + pgRestoreParams *params); static void *restore_files(void *arg); +static void set_orphan_status(parray *backups, pgBackup *parent_backup); + +static void restore_chain(pgBackup *dest_backup, parray *parent_chain, + parray *dbOid_exclude_list, pgRestoreParams *params, + const char *pgdata_path, bool no_sync, bool cleanup_pgdata, + bool backup_has_tblspc); + +/* + * Iterate over backup list to find all ancestors of the broken parent_backup + * and update their status to BACKUP_STATUS_ORPHAN + */ +static void +set_orphan_status(parray *backups, pgBackup *parent_backup) +{ + /* chain is intact, but at least one parent is invalid */ + int j; + + for (j = 0; j < parray_num(backups); j++) + { + + pgBackup *backup = (pgBackup *) parray_get(backups, j); + + if (is_parent(parent_backup->start_time, backup, false)) + { + if (backup->status == BACKUP_STATUS_OK || + backup->status == BACKUP_STATUS_DONE) + { + write_backup_status(backup, BACKUP_STATUS_ORPHAN, true); + + elog(WARNING, + "Backup %s is orphaned because his parent %s has status: %s", + backup_id_of(backup), + backup_id_of(parent_backup), + status2str(parent_backup->status)); + } + else + { + elog(WARNING, "Backup %s has parent %s with status: %s", + backup_id_of(backup), + backup_id_of(parent_backup), + status2str(parent_backup->status)); + } + } + } +} /* * Entry point of pg_probackup RESTORE and VALIDATE subcommands. */ int -do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, - bool is_restore) +do_restore_or_validate(InstanceState *instanceState, time_t target_backup_id, pgRecoveryTarget *rt, + pgRestoreParams *params, bool no_sync) { int i = 0; int j = 0; - parray *backups; + parray *backups = NULL; pgBackup *tmp_backup = NULL; pgBackup *current_backup = NULL; pgBackup *dest_backup = NULL; pgBackup *base_full_backup = NULL; pgBackup *corrupted_backup = NULL; - char *action = is_restore ? "Restore":"Validate"; + char *action = params->is_restore ? "Restore":"Validate"; parray *parent_chain = NULL; + parray *dbOid_exclude_list = NULL; + bool pgdata_is_empty = true; + bool cleanup_pgdata = false; + bool backup_has_tblspc = true; /* backup contain tablespace */ + XLogRecPtr shift_lsn = InvalidXLogRecPtr; - if (is_restore) + if (instanceState == NULL) + elog(ERROR, "Required parameter not specified: --instance"); + + if (params->is_restore) { if (instance_config.pgdata == NULL) - elog(ERROR, - "required parameter not specified: PGDATA (-D, --pgdata)"); + elog(ERROR, "No postgres data directory specified.\n" + "Please specify it either using environment variable PGDATA or\n" + "command line option --pgdata (-D)"); + /* Check if restore destination empty */ if (!dir_is_empty(instance_config.pgdata, FIO_DB_HOST)) - elog(ERROR, "restore destination is not empty: \"%s\"", - instance_config.pgdata); - } + { + /* if destination directory is empty, then incremental restore may be disabled */ + pgdata_is_empty = false; + + /* Check that remote system is NOT running and systemd id is the same as ours */ + if (params->incremental_mode != INCR_NONE) + { + DestDirIncrCompatibility rc; + const char *message = NULL; + bool ok_to_go = true; + + elog(INFO, "Running incremental restore into nonempty directory: \"%s\"", + instance_config.pgdata); + + rc = check_incremental_compatibility(instance_config.pgdata, + instance_config.system_identifier, + params->incremental_mode, + params->partial_db_list, + params->allow_partial_incremental); + if (rc == POSTMASTER_IS_RUNNING) + { + /* Even with force flag it is unwise to run + * incremental restore over running instance + */ + message = "Postmaster is running."; + ok_to_go = false; + } + else if (rc == SYSTEM_ID_MISMATCH) + { + /* + * In force mode it is possible to ignore system id mismatch + * by just wiping clean the destination directory. + */ + if (params->incremental_mode != INCR_NONE && params->force) + cleanup_pgdata = true; + else + { + message = "System ID mismatch."; + ok_to_go = false; + } + } + else if (rc == BACKUP_LABEL_EXISTS) + { + /* + * A big no-no for lsn-based incremental restore + * If there is backup label in PGDATA, then this cluster was probably + * restored from backup, but not started yet. Which means that values + * in pg_control are not synchronized with PGDATA and so we cannot use + * incremental restore in LSN mode, because it is relying on pg_control + * to calculate switchpoint. + */ + if (params->incremental_mode == INCR_LSN) + { + message = "Backup label exists. Cannot use incremental restore in LSN mode."; + ok_to_go = false; + } + } + else if (rc == DEST_IS_NOT_OK) + { + /* + * Something else is wrong. For example, postmaster.pid is mangled, + * so we cannot be sure that postmaster is running or not. + * It is better to just error out. + */ + message = "We cannot be sure about the database state."; + ok_to_go = false; + } else if (rc == PARTIAL_INCREMENTAL_FORBIDDEN) + { + message = "Partial incremental restore into non-empty PGDATA is forbidden."; + ok_to_go = false; + } - if (instance_name == NULL) - elog(ERROR, "required parameter not specified: --instance"); + if (!ok_to_go) + elog(ERROR, "Incremental restore is not allowed: %s", message); + } + else + elog(ERROR, "Restore destination is not empty: \"%s\"", + instance_config.pgdata); + } + } elog(LOG, "%s begin.", action); /* Get list of all backups sorted in order of descending start time */ - backups = catalog_get_backup_list(INVALID_BACKUP_ID); + backups = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); /* Find backup range we should restore or validate. */ while ((i < parray_num(backups)) && !dest_backup) @@ -91,15 +243,21 @@ do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, /* * [PGPRO-1164] If BACKUP_ID is not provided for restore command, * we must find the first valid(!) backup. + + * If target_backup_id is not provided, we can be sure that + * PITR for restore or validate is requested. + * So we can assume that user is more interested in recovery to specific point + * in time and NOT interested in revalidation of invalid backups. + * So based on that assumptions we should choose only OK and DONE backups + * as candidates for validate and restore. */ - if (is_restore && - target_backup_id == INVALID_BACKUP_ID && + if (target_backup_id == INVALID_BACKUP_ID && (current_backup->status != BACKUP_STATUS_OK && current_backup->status != BACKUP_STATUS_DONE)) { elog(WARNING, "Skipping backup %s, because it has non-valid status: %s", - base36enc(current_backup->start_time), status2str(current_backup->status)); + backup_id_of(current_backup), status2str(current_backup->status)); continue; } @@ -127,26 +285,30 @@ do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, if ((current_backup->status == BACKUP_STATUS_ORPHAN || current_backup->status == BACKUP_STATUS_CORRUPT || current_backup->status == BACKUP_STATUS_RUNNING) - && !rt->no_validate) + && (!params->no_validate || params->force)) elog(WARNING, "Backup %s has status: %s", - base36enc(current_backup->start_time), status2str(current_backup->status)); + backup_id_of(current_backup), status2str(current_backup->status)); else elog(ERROR, "Backup %s has status: %s", - base36enc(current_backup->start_time), status2str(current_backup->status)); + backup_id_of(current_backup), status2str(current_backup->status)); } if (rt->target_tli) { parray *timelines; - elog(LOG, "target timeline ID = %u", rt->target_tli); + // elog(LOG, "target timeline ID = %u", rt->target_tli); /* Read timeline history files from archives */ - timelines = read_timeline_history(rt->target_tli); + timelines = read_timeline_history(instanceState->instance_wal_subdir_path, + rt->target_tli, true); + + if (!timelines) + elog(ERROR, "Failed to get history file for target timeline %i", rt->target_tli); - if (!satisfy_timeline(timelines, current_backup)) + if (!satisfy_timeline(timelines, current_backup->tli, current_backup->stop_lsn)) { if (target_backup_id != INVALID_BACKUP_ID) - elog(ERROR, "target backup %s does not satisfy target timeline", + elog(ERROR, "Target backup %s does not satisfy target timeline", base36enc(target_backup_id)); else /* Try to find another backup that satisfies target timeline */ @@ -195,16 +357,16 @@ do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, result = scan_parent_chain(dest_backup, &tmp_backup); - if (result == 0) + if (result == ChainIsBroken) { /* chain is broken, determine missing backup ID * and orphinize all his descendants */ - char *missing_backup_id; - time_t missing_backup_start_time; + const char *missing_backup_id; + time_t missing_backup_start_time; missing_backup_start_time = tmp_backup->parent_backup; - missing_backup_id = base36enc_dup(tmp_backup->parent_backup); + missing_backup_id = base36enc(tmp_backup->parent_backup); for (j = 0; j < parray_num(backups); j++) { @@ -218,72 +380,34 @@ do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, if (backup->status == BACKUP_STATUS_OK || backup->status == BACKUP_STATUS_DONE) { - write_backup_status(backup, BACKUP_STATUS_ORPHAN); + write_backup_status(backup, BACKUP_STATUS_ORPHAN, true); elog(WARNING, "Backup %s is orphaned because his parent %s is missing", - base36enc(backup->start_time), missing_backup_id); + backup_id_of(backup), missing_backup_id); } else { elog(WARNING, "Backup %s has missing parent %s", - base36enc(backup->start_time), missing_backup_id); + backup_id_of(backup), missing_backup_id); } } } - pg_free(missing_backup_id); /* No point in doing futher */ - elog(ERROR, "%s of backup %s failed.", action, base36enc(dest_backup->start_time)); + elog(ERROR, "%s of backup %s failed.", action, backup_id_of(dest_backup)); } - else if (result == 1) + else if (result == ChainIsInvalid) { /* chain is intact, but at least one parent is invalid */ - char *parent_backup_id; - - /* parent_backup_id contain human-readable backup ID of oldest invalid backup */ - parent_backup_id = base36enc_dup(tmp_backup->start_time); - - for (j = 0; j < parray_num(backups); j++) - { - - pgBackup *backup = (pgBackup *) parray_get(backups, j); - - if (is_parent(tmp_backup->start_time, backup, false)) - { - if (backup->status == BACKUP_STATUS_OK || - backup->status == BACKUP_STATUS_DONE) - { - write_backup_status(backup, BACKUP_STATUS_ORPHAN); - - elog(WARNING, - "Backup %s is orphaned because his parent %s has status: %s", - base36enc(backup->start_time), - parent_backup_id, - status2str(tmp_backup->status)); - } - else - { - elog(WARNING, "Backup %s has parent %s with status: %s", - base36enc(backup->start_time), parent_backup_id, - status2str(tmp_backup->status)); - } - } - } - pg_free(parent_backup_id); + set_orphan_status(backups, tmp_backup); tmp_backup = find_parent_full_backup(dest_backup); /* sanity */ if (!tmp_backup) elog(ERROR, "Parent full backup for the given backup %s was not found", - base36enc(dest_backup->start_time)); + backup_id_of(dest_backup)); } - /* - * We have found full backup by link, - * now we need to walk the list to find its index. - * - * TODO I think we should rewrite it someday to use double linked list - * and avoid relying on sort order anymore. - */ + /* We have found full backup */ base_full_backup = tmp_backup; } @@ -294,13 +418,29 @@ do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, * Ensure that directories provided in tablespace mapping are valid * i.e. empty or not exist. */ - if (is_restore) + if (params->is_restore) { - check_tablespace_mapping(dest_backup); + int rc = check_tablespace_mapping(dest_backup, + params->incremental_mode != INCR_NONE, params->force, + pgdata_is_empty, params->no_validate); + + /* backup contain no tablespaces */ + if (rc == NoTblspc) + backup_has_tblspc = false; + + if (params->incremental_mode != INCR_NONE && !cleanup_pgdata && pgdata_is_empty && (rc != NotEmptyTblspc)) + { + elog(INFO, "Destination directory and tablespace directories are empty, " + "disable incremental restore"); + params->incremental_mode = INCR_NONE; + } /* no point in checking external directories if their restore is not requested */ - if (!skip_external_dirs) - check_external_dir_mapping(dest_backup); + //TODO: + // - make check_external_dir_mapping more like check_tablespace_mapping + // - honor force flag in case of incremental restore just like check_tablespace_mapping + if (!params->skip_external_dirs) + check_external_dir_mapping(dest_backup, params->incremental_mode != INCR_NONE); } /* At this point we are sure that parent chain is whole @@ -314,19 +454,133 @@ do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, */ tmp_backup = dest_backup; - while(tmp_backup->parent_backup_link) + while (tmp_backup) { parray_append(parent_chain, tmp_backup); tmp_backup = tmp_backup->parent_backup_link; } - parray_append(parent_chain, base_full_backup); + /* + * Determine the shift-LSN + * Consider the example A: + * + * + * /----D----------F-> + * -A--B---C---*-------X-----> + * + * [A,F] - incremental chain + * X - the state of pgdata + * F - destination backup + * * - switch point + * + * When running incremental restore in 'lsn' mode, we get a bitmap of pages, + * whose LSN is less than shift-LSN (backup C stop_lsn). + * So when restoring file, we can skip restore of pages coming from + * A, B and C. + * Pages from D and F cannot be skipped due to incremental restore. + * + * Consider the example B: + * + * + * /----------X----> + * ----*---A---B---C--> + * + * [A,C] - incremental chain + * X - the state of pgdata + * C - destination backup + * * - switch point + * + * Incremental restore in shift mode IS NOT POSSIBLE in this case. + * We must be able to differentiate the scenario A and scenario B. + * + */ + if (params->is_restore && params->incremental_mode == INCR_LSN) + { + RedoParams redo; + parray *timelines = NULL; + get_redo(instance_config.pgdata, FIO_DB_HOST, &redo); + + if (redo.checksum_version == 0) + elog(ERROR, "Incremental restore in 'lsn' mode require " + "data_checksums to be enabled in destination data directory"); + if (!control_downloaded) + get_control_file_or_back_file(instance_config.pgdata, FIO_DB_HOST, + &instance_control); + + timelines = read_timeline_history(instanceState->instance_wal_subdir_path, + redo.tli, false); + + if (!timelines) + elog(WARNING, "Failed to get history for redo timeline %i, " + "multi-timeline incremental restore in 'lsn' mode is impossible", redo.tli); + + tmp_backup = dest_backup; + + while (tmp_backup) + { + /* Candidate, whose stop_lsn if less than shift LSN, is found */ + if (tmp_backup->stop_lsn < redo.lsn) + { + /* if candidate timeline is the same as redo TLI, + * then we are good to go. + */ + if (redo.tli == tmp_backup->tli) + { + elog(INFO, "Backup %s is chosen as shiftpoint, its Stop LSN will be used as shift LSN", + backup_id_of(tmp_backup)); + + shift_lsn = tmp_backup->stop_lsn; + break; + } + + if (!timelines) + { + elog(WARNING, "Redo timeline %i differs from target timeline %i, " + "in this case, to safely run incremental restore in 'lsn' mode, " + "the history file for timeline %i is mandatory", + redo.tli, tmp_backup->tli, redo.tli); + break; + } + + /* check whether the candidate tli is a part of redo TLI history */ + if (tliIsPartOfHistory(timelines, tmp_backup->tli)) + { + shift_lsn = tmp_backup->stop_lsn; + break; + } + else + elog(INFO, "Backup %s cannot be a shiftpoint, " + "because its tli %i is not in history of redo timeline %i", + backup_id_of(tmp_backup), tmp_backup->tli, redo.tli); + } + + tmp_backup = tmp_backup->parent_backup_link; + } + + if (XLogRecPtrIsInvalid(shift_lsn)) + elog(ERROR, "Cannot perform incremental restore of backup chain %s in 'lsn' mode, " + "because destination directory redo point %X/%X on tli %i is out of reach", + backup_id_of(dest_backup), + (uint32) (redo.lsn >> 32), (uint32) redo.lsn, redo.tli); + else + elog(INFO, "Destination directory redo point %X/%X on tli %i is " + "within reach of backup %s with Stop LSN %X/%X on tli %i", + (uint32) (redo.lsn >> 32), (uint32) redo.lsn, redo.tli, + backup_id_of(tmp_backup), + (uint32) (tmp_backup->stop_lsn >> 32), (uint32) tmp_backup->stop_lsn, + tmp_backup->tli); + + elog(INFO, "shift LSN: %X/%X", + (uint32) (shift_lsn >> 32), (uint32) shift_lsn); + + } + params->shift_lsn = shift_lsn; /* for validation or restore with enabled validation */ - if (!is_restore || !rt->no_validate) + if (!params->is_restore || !params->no_validate) { if (dest_backup->backup_mode != BACKUP_MODE_FULL) - elog(INFO, "Validating parents for backup %s", base36enc(dest_backup->start_time)); + elog(INFO, "Validating parents for backup %s", backup_id_of(dest_backup)); /* * Validate backups from base_full_backup to dest_backup. @@ -335,21 +589,16 @@ do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, { tmp_backup = (pgBackup *) parray_get(parent_chain, i); - /* Do not interrupt, validate the next backup */ - if (!lock_backup(tmp_backup)) + /* lock every backup in chain in read-only mode */ + if (!lock_backup(tmp_backup, true, false)) { - if (is_restore) - elog(ERROR, "Cannot lock backup %s directory", - base36enc(tmp_backup->start_time)); - else - { - elog(WARNING, "Cannot lock backup %s directory, skip validation", - base36enc(tmp_backup->start_time)); - continue; - } + elog(ERROR, "Cannot lock backup %s directory", + backup_id_of(tmp_backup)); } - pgBackupValidate(tmp_backup); + /* validate datafiles only */ + pgBackupValidate(tmp_backup, params); + /* After pgBackupValidate() only following backup * states are possible: ERROR, RUNNING, CORRUPT and OK. * Validate WAL only for OK, because there is no point @@ -366,6 +615,7 @@ do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, } /* There is no point in wal validation of corrupted backups */ + // TODO: there should be a way for a user to request only(!) WAL validation if (!corrupted_backup) { /* @@ -373,36 +623,13 @@ do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, * We pass base_full_backup timeline as last argument to this function, * because it's needed to form the name of xlog file. */ - validate_wal(dest_backup, arclog_path, rt->target_time, + validate_wal(dest_backup, instanceState->instance_wal_subdir_path, rt->target_time, rt->target_xid, rt->target_lsn, - base_full_backup->tli, instance_config.xlog_seg_size); + dest_backup->tli, instance_config.xlog_seg_size); } /* Orphanize every OK descendant of corrupted backup */ else - { - char *corrupted_backup_id; - corrupted_backup_id = base36enc_dup(corrupted_backup->start_time); - - for (j = 0; j < parray_num(backups); j++) - { - pgBackup *backup = (pgBackup *) parray_get(backups, j); - - if (is_parent(corrupted_backup->start_time, backup, false)) - { - if (backup->status == BACKUP_STATUS_OK || - backup->status == BACKUP_STATUS_DONE) - { - write_backup_status(backup, BACKUP_STATUS_ORPHAN); - - elog(WARNING, "Backup %s is orphaned because his parent %s has status: %s", - base36enc(backup->start_time), - corrupted_backup_id, - status2str(corrupted_backup->status)); - } - } - } - free(corrupted_backup_id); - } + set_orphan_status(backups, corrupted_backup); } /* @@ -412,222 +639,418 @@ do_restore_or_validate(time_t target_backup_id, pgRecoveryTarget *rt, if (dest_backup->status == BACKUP_STATUS_OK || dest_backup->status == BACKUP_STATUS_DONE) { - if (rt->no_validate) - elog(WARNING, "Backup %s is used without validation.", base36enc(dest_backup->start_time)); + if (params->no_validate) + elog(WARNING, "Backup %s is used without validation.", backup_id_of(dest_backup)); else - elog(INFO, "Backup %s is valid.", base36enc(dest_backup->start_time)); + elog(INFO, "Backup %s is valid.", backup_id_of(dest_backup)); } else if (dest_backup->status == BACKUP_STATUS_CORRUPT) - elog(ERROR, "Backup %s is corrupt.", base36enc(dest_backup->start_time)); + { + if (params->force) + elog(WARNING, "Backup %s is corrupt.", backup_id_of(dest_backup)); + else + elog(ERROR, "Backup %s is corrupt.", backup_id_of(dest_backup)); + } else if (dest_backup->status == BACKUP_STATUS_ORPHAN) - elog(ERROR, "Backup %s is orphan.", base36enc(dest_backup->start_time)); + { + if (params->force) + elog(WARNING, "Backup %s is orphan.", backup_id_of(dest_backup)); + else + elog(ERROR, "Backup %s is orphan.", backup_id_of(dest_backup)); + } else elog(ERROR, "Backup %s has status: %s", - base36enc(dest_backup->start_time), status2str(dest_backup->status)); + backup_id_of(dest_backup), status2str(dest_backup->status)); /* We ensured that all backups are valid, now restore if required - * TODO: before restore - lock entire parent chain */ - if (is_restore) + if (params->is_restore) { - parray *dest_external_dirs = NULL; - parray *dest_files; - char control_file[MAXPGPATH], - dest_backup_path[MAXPGPATH]; - int i; - - /* - * Preparations for actual restoring. - */ - pgBackupGetPath(dest_backup, control_file, lengthof(control_file), - DATABASE_FILE_LIST); - dest_files = dir_read_file_list(NULL, NULL, control_file, - FIO_BACKUP_HOST); - parray_qsort(dest_files, pgFileCompareRelPathWithExternal); - - /* - * Restore dest_backup internal directories. - */ - pgBackupGetPath(dest_backup, dest_backup_path, - lengthof(dest_backup_path), NULL); - create_data_directories(dest_files, instance_config.pgdata, dest_backup_path, true, - FIO_DB_HOST); - - /* - * Restore dest_backup external directories. - */ - if (dest_backup->external_dir_str && !skip_external_dirs) - { - dest_external_dirs = make_external_directory_list( - dest_backup->external_dir_str, - true); - if (parray_num(dest_external_dirs) > 0) - elog(LOG, "Restore external directories"); - - for (i = 0; i < parray_num(dest_external_dirs); i++) - fio_mkdir(parray_get(dest_external_dirs, i), - DIR_PERMISSION, FIO_DB_HOST); - } - /* - * At least restore backups files starting from the parent backup. + * Get a list of dbOids to skip if user requested the partial restore. + * It is important that we do this after(!) validation so + * database_map can be trusted. + * NOTE: database_map could be missing for legal reasons, e.g. missing + * permissions on pg_database during `backup` and, as long as user + * do not request partial restore, it`s OK. + * + * If partial restore is requested and database map doesn't exist, + * throw an error. */ - for (i = parray_num(parent_chain) - 1; i >= 0; i--) - { - pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); - - if (rt->lsn_string && - parse_server_version(backup->server_version) < 100000) - elog(ERROR, "Backup %s was created for version %s which doesn't support recovery_target_lsn", - base36enc(dest_backup->start_time), + if (params->partial_db_list) + dbOid_exclude_list = get_dbOid_exclude_list(dest_backup, params->partial_db_list, + params->partial_restore_type); + + if (rt->lsn_string && + parse_server_version(dest_backup->server_version) < 100000) + elog(ERROR, "Backup %s was created for version %s which doesn't support recovery_target_lsn", + backup_id_of(dest_backup), dest_backup->server_version); - /* - * Backup was locked during validation if no-validate wasn't - * specified. - */ - if (rt->no_validate && !lock_backup(backup)) - elog(ERROR, "Cannot lock backup directory"); - - restore_backup(backup, dest_external_dirs, dest_files); - } - - if (dest_external_dirs != NULL) - free_dir_list(dest_external_dirs); + if (instance_config.remote.host) + elog(INFO, "Restoring the database from backup %s on %s", backup_id_of(dest_backup), instance_config.remote.host); + else + elog(INFO, "Restoring the database from backup %s", backup_id_of(dest_backup)); - parray_walk(dest_files, pgFileFree); - parray_free(dest_files); + restore_chain(dest_backup, parent_chain, dbOid_exclude_list, params, + instance_config.pgdata, no_sync, cleanup_pgdata, backup_has_tblspc); + //TODO rename and update comment /* Create recovery.conf with given recovery target parameters */ - create_recovery_conf(target_backup_id, rt, dest_backup); + create_recovery_conf(instanceState, target_backup_id, rt, dest_backup, params); } + /* ssh connection to longer needed */ + fio_disconnect(); + + elog(INFO, "%s of backup %s completed.", + action, backup_id_of(dest_backup)); + /* cleanup */ parray_walk(backups, pgBackupFree); parray_free(backups); parray_free(parent_chain); - elog(INFO, "%s of backup %s completed.", - action, base36enc(dest_backup->start_time)); return 0; } /* - * Restore one backup. + * Restore backup chain. + * Flag 'cleanup_pgdata' demands the removing of already existing content in PGDATA. */ void -restore_backup(pgBackup *backup, parray *dest_external_dirs, parray *dest_files) +restore_chain(pgBackup *dest_backup, parray *parent_chain, + parray *dbOid_exclude_list, pgRestoreParams *params, + const char *pgdata_path, bool no_sync, bool cleanup_pgdata, + bool backup_has_tblspc) { - char timestamp[100]; - char database_path[MAXPGPATH]; - char external_prefix[MAXPGPATH]; - char list_path[MAXPGPATH]; - parray *files; - parray *external_dirs = NULL; int i; + parray *pgdata_files = NULL; + parray *dest_files = NULL; + parray *external_dirs = NULL; + pgFile *dest_pg_control_file = NULL; + char dest_pg_control_fullpath[MAXPGPATH]; + char dest_pg_control_bak_fullpath[MAXPGPATH]; + /* arrays with meta info for multi threaded backup */ pthread_t *threads; restore_files_arg *threads_args; bool restore_isok = true; + bool use_bitmap = true; + + /* fancy reporting */ + char pretty_dest_bytes[20]; + char pretty_total_bytes[20]; + size_t dest_bytes = 0; + size_t total_bytes = 0; + char pretty_time[20]; + time_t start_time, end_time; + + /* Preparations for actual restoring */ + dest_files = get_backup_filelist(dest_backup, true); + + /* Lock backup chain and make sanity checks */ + for (i = parray_num(parent_chain) - 1; i >= 0; i--) + { + pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); + + if (!lock_backup(backup, true, false)) + elog(ERROR, "Cannot lock backup %s", backup_id_of(backup)); + + if (backup->status != BACKUP_STATUS_OK && + backup->status != BACKUP_STATUS_DONE) + { + if (params->force) + elog(WARNING, "Backup %s is not valid, restore is forced", + backup_id_of(backup)); + else + elog(ERROR, "Backup %s cannot be restored because it is not valid", + backup_id_of(backup)); + } + + /* confirm block size compatibility */ + if (backup->block_size != BLCKSZ) + elog(ERROR, + "BLCKSZ(%d) is not compatible(%d expected)", + backup->block_size, BLCKSZ); + + if (backup->wal_block_size != XLOG_BLCKSZ) + elog(ERROR, + "XLOG_BLCKSZ(%d) is not compatible(%d expected)", + backup->wal_block_size, XLOG_BLCKSZ); + + /* populate backup filelist */ + if (backup->start_time != dest_backup->start_time) + backup->files = get_backup_filelist(backup, true); + else + backup->files = dest_files; - if (backup->status != BACKUP_STATUS_OK && - backup->status != BACKUP_STATUS_DONE) - elog(ERROR, "Backup %s cannot be restored because it is not valid", - base36enc(backup->start_time)); + /* + * this sorting is important, because we rely on it to find + * destination file in intermediate backups file lists + * using bsearch. + */ + parray_qsort(backup->files, pgFileCompareRelPathWithExternal); + } + + /* If dest backup version is older than 2.4.0, then bitmap optimization + * is impossible to use, because bitmap restore rely on pgFile.n_blocks, + * which is not always available in old backups. + */ + if (parse_program_version(dest_backup->program_version) < 20400) + { + use_bitmap = false; - /* confirm block size compatibility */ - if (backup->block_size != BLCKSZ) - elog(ERROR, - "BLCKSZ(%d) is not compatible(%d expected)", - backup->block_size, BLCKSZ); - if (backup->wal_block_size != XLOG_BLCKSZ) - elog(ERROR, - "XLOG_BLCKSZ(%d) is not compatible(%d expected)", - backup->wal_block_size, XLOG_BLCKSZ); + if (params->incremental_mode != INCR_NONE) + elog(ERROR, "Incremental restore is not possible for backups older than 2.3.0 version"); + } - time2iso(timestamp, lengthof(timestamp), backup->start_time); - elog(LOG, "Restoring database from backup %s", timestamp); + /* There is no point in bitmap restore, when restoring a single FULL backup, + * unless we are running incremental-lsn restore, then bitmap is mandatory. + */ + if (use_bitmap && parray_num(parent_chain) == 1) + { + if (params->incremental_mode == INCR_NONE) + use_bitmap = false; + else + use_bitmap = true; + } - if (backup->external_dir_str) - external_dirs = make_external_directory_list(backup->external_dir_str, - true); + /* + * Restore dest_backup internal directories. + */ + create_data_directories(dest_files, instance_config.pgdata, + dest_backup->root_dir, backup_has_tblspc, + params->incremental_mode != INCR_NONE, + FIO_DB_HOST, params->waldir); /* - * Get list of files which need to be restored. + * Restore dest_backup external directories. */ - pgBackupGetPath(backup, database_path, lengthof(database_path), DATABASE_DIR); - pgBackupGetPath(backup, external_prefix, lengthof(external_prefix), - EXTERNAL_DIR); - pgBackupGetPath(backup, list_path, lengthof(list_path), DATABASE_FILE_LIST); - files = dir_read_file_list(database_path, external_prefix, list_path, - FIO_BACKUP_HOST); + if (dest_backup->external_dir_str && !params->skip_external_dirs) + { + external_dirs = make_external_directory_list(dest_backup->external_dir_str, true); + + if (!external_dirs) + elog(ERROR, "Failed to get a list of external directories"); + + if (parray_num(external_dirs) > 0) + elog(LOG, "Restore external directories"); - /* Restore directories in do_backup_instance way */ - parray_qsort(files, pgFileComparePath); + for (i = 0; i < parray_num(external_dirs); i++) + fio_mkdir(parray_get(external_dirs, i), + DIR_PERMISSION, FIO_DB_HOST); + } /* - * Make external directories before restore - * and setup threads at the same time + * Setup directory structure for external directories */ - for (i = 0; i < parray_num(files); i++) + for (i = 0; i < parray_num(dest_files); i++) { - pgFile *file = (pgFile *) parray_get(files, i); + pgFile *file = (pgFile *) parray_get(dest_files, i); - /* - * If the entry was an external directory, create it in the backup. - */ - if (!skip_external_dirs && - file->external_dir_num && S_ISDIR(file->mode) && - /* Do not create unnecessary external directories */ - parray_bsearch(dest_files, file, pgFileCompareRelPathWithExternal)) + if (S_ISDIR(file->mode)) + total_bytes += 4096; + + if (!params->skip_external_dirs && + file->external_dir_num && S_ISDIR(file->mode)) { char *external_path; + char dirpath[MAXPGPATH]; - if (!external_dirs || - parray_num(external_dirs) < file->external_dir_num - 1) + if (parray_num(external_dirs) < file->external_dir_num - 1) elog(ERROR, "Inconsistent external directory backup metadata"); - external_path = parray_get(external_dirs, - file->external_dir_num - 1); - if (backup_contains_external(external_path, dest_external_dirs)) + external_path = parray_get(external_dirs, file->external_dir_num - 1); + join_path_components(dirpath, external_path, file->rel_path); + + elog(LOG, "Create external directory \"%s\"", dirpath); + fio_mkdir(dirpath, file->mode, FIO_DB_HOST); + } + } + + /* setup threads */ + pfilearray_clear_locks(dest_files); + + /* Get list of files in destination directory and remove redundant files */ + if (params->incremental_mode != INCR_NONE || cleanup_pgdata) + { + pgdata_files = parray_new(); + + elog(INFO, "Extracting the content of destination directory for incremental restore"); + + time(&start_time); + fio_list_dir(pgdata_files, pgdata_path, false, true, false, false, true, 0); + + /* + * TODO: + * 1. Currently we are cleaning the tablespaces in check_tablespace_mapping and PGDATA here. + * It would be great to do all this work in one place. + * + * 2. In case of tablespace remapping we do not cleanup the old tablespace path, + * it is just left as it is. + * Lookup tests.incr_restore.IncrRestoreTest.test_incr_restore_with_tablespace_5 + */ + + /* get external dirs content */ + if (external_dirs) + { + for (i = 0; i < parray_num(external_dirs); i++) + { + char *external_path = parray_get(external_dirs, i); + parray *external_files = parray_new(); + + fio_list_dir(external_files, external_path, + false, true, false, false, true, i+1); + + parray_concat(pgdata_files, external_files); + parray_free(external_files); + } + } + + parray_qsort(pgdata_files, pgFileCompareRelPathWithExternalDesc); + + time(&end_time); + pretty_time_interval(difftime(end_time, start_time), + pretty_time, lengthof(pretty_time)); + + elog(INFO, "Destination directory content extracted, time elapsed: %s", + pretty_time); + + elog(INFO, "Removing redundant files in destination directory"); + time(&start_time); + for (i = 0; i < parray_num(pgdata_files); i++) + { + bool redundant = true; + pgFile *file = (pgFile *) parray_get(pgdata_files, i); + + if (parray_bsearch(dest_backup->files, file, pgFileCompareRelPathWithExternal)) + redundant = false; + + /* pg_filenode.map are always restored, because it's crc cannot be trusted */ + if (file->external_dir_num == 0 && + pg_strcasecmp(file->name, RELMAPPER_FILENAME) == 0) + redundant = true; + + /* global/pg_control.pbk.bak are always keeped, because it's needed for restart failed incremental restore */ + if (file->external_dir_num == 0 && + pg_strcasecmp(file->rel_path, XLOG_CONTROL_BAK_FILE) == 0) + redundant = false; + + /* do not delete the useful internal directories */ + if (S_ISDIR(file->mode) && !redundant) + continue; + + /* if file does not exists in destination list, then we can safely unlink it */ + if (cleanup_pgdata || redundant) { - char container_dir[MAXPGPATH]; - char dirpath[MAXPGPATH]; - char *dir_name; - - makeExternalDirPathByNum(container_dir, external_prefix, - file->external_dir_num); - dir_name = GetRelativePath(file->path, container_dir); - elog(VERBOSE, "Create directory \"%s\"", dir_name); - join_path_components(dirpath, external_path, dir_name); - fio_mkdir(dirpath, DIR_PERMISSION, FIO_DB_HOST); + char fullpath[MAXPGPATH]; + + join_path_components(fullpath, pgdata_path, file->rel_path); + + fio_delete(file->mode, fullpath, FIO_DB_HOST); + elog(LOG, "Deleted file \"%s\"", fullpath); + + /* shrink pgdata list */ + pgFileFree(file); + parray_remove(pgdata_files, i); + i--; } } - /* setup threads */ - pg_atomic_clear_flag(&file->lock); + if (cleanup_pgdata) + { + /* Destination PGDATA and tablespaces were cleaned up, so it's the regular restore from this point */ + params->incremental_mode = INCR_NONE; + parray_free(pgdata_files); + pgdata_files = NULL; + } + + time(&end_time); + pretty_time_interval(difftime(end_time, start_time), + pretty_time, lengthof(pretty_time)); + + /* At this point PDATA do not contain files, that do not exists in dest backup file list */ + elog(INFO, "Redundant files are removed, time elapsed: %s", pretty_time); } + + /* + * Close ssh connection belonging to the main thread + * to avoid the possibility of been killed for idleness + */ + fio_disconnect(); + threads = (pthread_t *) palloc(sizeof(pthread_t) * num_threads); threads_args = (restore_files_arg *) palloc(sizeof(restore_files_arg) * num_threads); + if (dest_backup->stream) + dest_bytes = dest_backup->pgdata_bytes + dest_backup->wal_bytes; + else + dest_bytes = dest_backup->pgdata_bytes; - /* Restore files into target directory */ + pretty_size(dest_bytes, pretty_dest_bytes, lengthof(pretty_dest_bytes)); + /* + * [Issue #313] + * find pg_control file (in already sorted earlier dest_files, see parray_qsort(backup->files...)) + * and exclude it from list for future special processing + */ + { + int control_file_elem_index; + pgFile search_key; + MemSet(&search_key, 0, sizeof(pgFile)); + /* pgFileCompareRelPathWithExternal uses only .rel_path and .external_dir_num for comparision */ + search_key.rel_path = XLOG_CONTROL_FILE; + search_key.external_dir_num = 0; + control_file_elem_index = parray_bsearch_index(dest_files, &search_key, pgFileCompareRelPathWithExternal); + + if (control_file_elem_index < 0) + elog(ERROR, "File \"%s\" not found in backup %s", XLOG_CONTROL_FILE, base36enc(dest_backup->start_time)); + dest_pg_control_file = (pgFile *) parray_get(dest_files, control_file_elem_index); + parray_remove(dest_files, control_file_elem_index); + + join_path_components(dest_pg_control_fullpath, pgdata_path, XLOG_CONTROL_FILE); + join_path_components(dest_pg_control_bak_fullpath, pgdata_path, XLOG_CONTROL_BAK_FILE); + /* + * rename (if it exist) dest control file before restoring + * if it doesn't exist, that mean, that we already restoring in a previously failed + * pgdata, where XLOG_CONTROL_BAK_FILE exist + */ + if (params->incremental_mode != INCR_NONE) + { + if (fio_access(dest_pg_control_fullpath,F_OK,FIO_DB_HOST) == 0){ + if (fio_rename(dest_pg_control_fullpath, dest_pg_control_bak_fullpath, FIO_DB_HOST) < 0) + elog(WARNING, "Cannot rename file \"%s\" to \"%s\": %s", + dest_pg_control_fullpath, dest_pg_control_bak_fullpath, strerror(errno)); + } + } + } + + elog(INFO, "Start restoring backup files. PGDATA size: %s", pretty_dest_bytes); + time(&start_time); thread_interrupted = false; + + /* Restore files into target directory */ for (i = 0; i < num_threads; i++) { restore_files_arg *arg = &(threads_args[i]); - arg->files = files; - arg->backup = backup; - arg->external_dirs = external_dirs; - arg->external_prefix = external_prefix; - arg->dest_external_dirs = dest_external_dirs; arg->dest_files = dest_files; + arg->pgdata_files = pgdata_files; + arg->dest_backup = dest_backup; + arg->dest_external_dirs = external_dirs; + arg->parent_chain = parent_chain; + arg->dbOid_exclude_list = dbOid_exclude_list; + arg->skip_external_dirs = params->skip_external_dirs; + arg->to_root = pgdata_path; + arg->use_bitmap = use_bitmap; + arg->incremental_mode = params->incremental_mode; + arg->shift_lsn = params->shift_lsn; + threads_args[i].restored_bytes = 0; /* By default there are some error */ threads_args[i].ret = 1; /* Useless message TODO: rewrite */ - elog(LOG, "Start thread for num:%zu", parray_num(files)); + elog(LOG, "Start thread %i", i + 1); pthread_create(&threads[i], NULL, restore_files, arg); } @@ -638,21 +1061,129 @@ restore_backup(pgBackup *backup, parray *dest_external_dirs, parray *dest_files) pthread_join(threads[i], NULL); if (threads_args[i].ret == 1) restore_isok = false; + + total_bytes += threads_args[i].restored_bytes; } - if (!restore_isok) - elog(ERROR, "Data files restoring failed"); - pfree(threads); - pfree(threads_args); + /* [Issue #313] copy pg_control at very end */ + if (restore_isok) + { + FILE *out = NULL; + elog(progress ? INFO : LOG, "Progress: Restore file \"%s\"", + dest_pg_control_file->rel_path); + + out = fio_fopen(dest_pg_control_fullpath, PG_BINARY_R "+", FIO_DB_HOST); + + total_bytes += restore_non_data_file(parent_chain, + dest_backup, + dest_pg_control_file, + out, + dest_pg_control_fullpath, false); + fio_fclose(out); + /* Now backup control file can be deleted */ + if (params->incremental_mode != INCR_NONE) + { + pgFile *dst_control; + dst_control = pgFileNew(dest_pg_control_bak_fullpath, XLOG_CONTROL_BAK_FILE, + true,0, FIO_BACKUP_HOST); + fio_delete(dst_control->mode, dest_pg_control_bak_fullpath, FIO_LOCAL_HOST); + pgFileFree(dst_control); + } + } + + time(&end_time); + pretty_time_interval(difftime(end_time, start_time), + pretty_time, lengthof(pretty_time)); + pretty_size(total_bytes, pretty_total_bytes, lengthof(pretty_total_bytes)); + + if (restore_isok) + { + elog(INFO, "Backup files are restored. Transfered bytes: %s, time elapsed: %s", + pretty_total_bytes, pretty_time); + + elog(INFO, "Restore incremental ratio (less is better): %.f%% (%s/%s)", + ((float) total_bytes / dest_bytes) * 100, + pretty_total_bytes, pretty_dest_bytes); + } + else + elog(ERROR, "Backup files restoring failed. Transfered bytes: %s, time elapsed: %s", + pretty_total_bytes, pretty_time); + + /* Close page header maps */ + for (i = parray_num(parent_chain) - 1; i >= 0; i--) + { + pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); + cleanup_header_map(&(backup->hdr_map)); + } + + if (no_sync) + elog(WARNING, "Restored files are not synced to disk"); + else + { + elog(INFO, "Syncing restored files to disk"); + time(&start_time); + + for (i = 0; i < parray_num(dest_files); i++) + { + char to_fullpath[MAXPGPATH]; + pgFile *dest_file = (pgFile *) parray_get(dest_files, i); + + if (S_ISDIR(dest_file->mode)) + continue; + + /* skip external files if ordered to do so */ + if (dest_file->external_dir_num > 0 && + params->skip_external_dirs) + continue; + + /* construct fullpath */ + if (dest_file->external_dir_num == 0) + { + if (strcmp(PG_TABLESPACE_MAP_FILE, dest_file->rel_path) == 0) + continue; + if (strcmp(DATABASE_MAP, dest_file->rel_path) == 0) + continue; + join_path_components(to_fullpath, pgdata_path, dest_file->rel_path); + } + else + { + char *external_path = parray_get(external_dirs, dest_file->external_dir_num - 1); + join_path_components(to_fullpath, external_path, dest_file->rel_path); + } + + /* TODO: write test for case: file to be synced is missing */ + if (fio_sync(to_fullpath, FIO_DB_HOST) != 0) + elog(ERROR, "Failed to sync file \"%s\": %s", to_fullpath, strerror(errno)); + } + + time(&end_time); + pretty_time_interval(difftime(end_time, start_time), + pretty_time, lengthof(pretty_time)); + elog(INFO, "Restored backup files are synced, time elapsed: %s", pretty_time); + } /* cleanup */ - parray_walk(files, pgFileFree); - parray_free(files); + pfree(threads); + pfree(threads_args); if (external_dirs != NULL) free_dir_list(external_dirs); - elog(LOG, "Restore %s backup completed", base36enc(backup->start_time)); + if (pgdata_files) + { + parray_walk(pgdata_files, pgFileFree); + parray_free(pgdata_files); + } + + if(dest_pg_control_file) pgFileFree(dest_pg_control_file); + + for (i = parray_num(parent_chain) - 1; i >= 0; i--) + { + pgBackup *backup = (pgBackup *) parray_get(parent_chain, i); + + parray_walk(backup->files, pgFileFree); + parray_free(backup->files); + } } /* @@ -661,201 +1192,597 @@ restore_backup(pgBackup *backup, parray *dest_external_dirs, parray *dest_files) static void * restore_files(void *arg) { - int i; - restore_files_arg *arguments = (restore_files_arg *)arg; + int i; + uint64 n_files; + char to_fullpath[MAXPGPATH]; + FILE *out = NULL; + char *out_buf = pgut_malloc(STDIO_BUFSIZE); - for (i = 0; i < parray_num(arguments->files); i++) + restore_files_arg *arguments = (restore_files_arg *) arg; + + n_files = (unsigned long) parray_num(arguments->dest_files); + + for (i = 0; i < parray_num(arguments->dest_files); i++) { - char from_root[MAXPGPATH]; - pgFile *file = (pgFile *) parray_get(arguments->files, i); + bool already_exists = false; + PageState *checksum_map = NULL; /* it should take ~1.5MB at most */ + datapagemap_t *lsn_map = NULL; /* it should take 16kB at most */ + char *errmsg = NULL; /* remote agent error message */ + pgFile *dest_file = (pgFile *) parray_get(arguments->dest_files, i); - if (!pg_atomic_test_set_flag(&file->lock)) + /* Directories were created before */ + if (S_ISDIR(dest_file->mode)) continue; - pgBackupGetPath(arguments->backup, from_root, - lengthof(from_root), DATABASE_DIR); + if (!pg_atomic_test_set_flag(&dest_file->lock)) + continue; /* check for interrupt */ if (interrupted || thread_interrupted) - elog(ERROR, "Interrupted during restore database"); + elog(ERROR, "Interrupted during restore"); - if (progress) - elog(INFO, "Progress: (%d/%lu). Process file %s ", - i + 1, (unsigned long) parray_num(arguments->files), - file->rel_path); + elog(progress ? INFO : LOG, "Progress: (%d/%lu). Restore file \"%s\"", + i + 1, n_files, dest_file->rel_path); - /* - * For PAGE and PTRACK backups skip datafiles which haven't changed - * since previous backup and thus were not backed up. - * We cannot do the same when restoring DELTA backup because we need information - * about every datafile to correctly truncate them. - */ - if (file->write_size == BYTES_INVALID) + /* Only files from pgdata can be skipped by partial restore */ + if (arguments->dbOid_exclude_list && dest_file->external_dir_num == 0) { - /* data file, only PAGE and PTRACK can skip */ - if (((file->is_datafile && !file->is_cfs) && - (arguments->backup->backup_mode == BACKUP_MODE_DIFF_PAGE || - arguments->backup->backup_mode == BACKUP_MODE_DIFF_PTRACK)) || - /* non-data file can be skipped regardless of backup type */ - !(file->is_datafile && !file->is_cfs)) + /* Check if the file belongs to the database we exclude */ + if (parray_bsearch(arguments->dbOid_exclude_list, + &dest_file->dbOid, pgCompareOid)) { - elog(VERBOSE, "The file didn`t change. Skip restore: \"%s\"", file->path); + /* + * We cannot simply skip the file, because it may lead to + * failure during WAL redo; hence, create empty file. + */ + create_empty_file(FIO_BACKUP_HOST, + arguments->to_root, FIO_DB_HOST, dest_file); + + elog(LOG, "Skip file due to partial restore: \"%s\"", + dest_file->rel_path); continue; } } - /* Directories were created before */ - if (S_ISDIR(file->mode)) + /* Do not restore tablespace_map file */ + if ((dest_file->external_dir_num == 0) && + strcmp(PG_TABLESPACE_MAP_FILE, dest_file->rel_path) == 0) + { + elog(LOG, "Skip tablespace_map"); continue; + } - /* Do not restore tablespace_map file */ - if (path_is_prefix_of_path(PG_TABLESPACE_MAP_FILE, file->rel_path)) + /* Do not restore database_map file */ + if ((dest_file->external_dir_num == 0) && + strcmp(DATABASE_MAP, dest_file->rel_path) == 0) { - elog(VERBOSE, "Skip tablespace_map"); + elog(LOG, "Skip database_map"); continue; } /* Do no restore external directory file if a user doesn't want */ - if (skip_external_dirs && file->external_dir_num > 0) + if (arguments->skip_external_dirs && dest_file->external_dir_num > 0) continue; - /* Skip unnecessary file */ - if (parray_bsearch(arguments->dest_files, file, - pgFileCompareRelPathWithExternal) == NULL) - continue; + /* set fullpath of destination file */ + if (dest_file->external_dir_num == 0) + join_path_components(to_fullpath, arguments->to_root, dest_file->rel_path); + else + { + char *external_path = parray_get(arguments->dest_external_dirs, + dest_file->external_dir_num - 1); + join_path_components(to_fullpath, external_path, dest_file->rel_path); + } + + if (arguments->incremental_mode != INCR_NONE && + parray_bsearch(arguments->pgdata_files, dest_file, pgFileCompareRelPathWithExternalDesc)) + { + already_exists = true; + } /* - * restore the file. - * We treat datafiles separately, cause they were backed up block by - * block and have BackupPageHeader meta information, so we cannot just - * copy the file from backup. + * Handle incremental restore case for data files. + * If file is already exists in pgdata, then + * we scan it block by block and get + * array of checksums for every page. */ - elog(VERBOSE, "Restoring file %s, is_datafile %i, is_cfs %i", - file->path, file->is_datafile?1:0, file->is_cfs?1:0); - if (file->is_datafile && !file->is_cfs) + if (already_exists && + dest_file->is_datafile && !dest_file->is_cfs && + dest_file->n_blocks > 0) { - char to_path[MAXPGPATH]; - - join_path_components(to_path, instance_config.pgdata, - file->rel_path); - restore_data_file(to_path, file, - arguments->backup->backup_mode == BACKUP_MODE_DIFF_DELTA, - false, - parse_program_version(arguments->backup->program_version)); + if (arguments->incremental_mode == INCR_LSN) + { + lsn_map = fio_get_lsn_map(to_fullpath, arguments->dest_backup->checksum_version, + dest_file->n_blocks, arguments->shift_lsn, + dest_file->segno * RELSEG_SIZE, FIO_DB_HOST); + } + else if (arguments->incremental_mode == INCR_CHECKSUM) + { + checksum_map = fio_get_checksum_map(to_fullpath, arguments->dest_backup->checksum_version, + dest_file->n_blocks, arguments->dest_backup->stop_lsn, + dest_file->segno * RELSEG_SIZE, FIO_DB_HOST); + } } - else if (file->external_dir_num) + + /* + * Open dest file and truncate it to zero, if destination + * file already exists and dest file size is zero, or + * if file do not exist + */ + if ((already_exists && dest_file->write_size == 0) || !already_exists) + out = fio_fopen(to_fullpath, PG_BINARY_W, FIO_DB_HOST); + /* + * If file already exists and dest size is not zero, + * then open it for reading and writing. + */ + else + out = fio_fopen(to_fullpath, PG_BINARY_R "+", FIO_DB_HOST); + + if (out == NULL) + elog(ERROR, "Cannot open restore target file \"%s\": %s", + to_fullpath, strerror(errno)); + + /* update file permission */ + if (fio_chmod(to_fullpath, dest_file->mode, FIO_DB_HOST) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", to_fullpath, + strerror(errno)); + + if (!dest_file->is_datafile || dest_file->is_cfs) + elog(LOG, "Restoring non-data file: \"%s\"", to_fullpath); + else + elog(LOG, "Restoring data file: \"%s\"", to_fullpath); + + // If destination file is 0 sized, then just close it and go for the next + if (dest_file->write_size == 0) + goto done; + + /* Restore destination file */ + if (dest_file->is_datafile && !dest_file->is_cfs) { - char *external_path = parray_get(arguments->external_dirs, - file->external_dir_num - 1); - if (backup_contains_external(external_path, - arguments->dest_external_dirs)) - copy_file(FIO_BACKUP_HOST, - external_path, FIO_DB_HOST, file, false); + /* enable stdio buffering for local destination data file */ + if (!fio_is_remote_file(out)) + setvbuf(out, out_buf, _IOFBF, STDIO_BUFSIZE); + /* Destination file is data file */ + arguments->restored_bytes += restore_data_file(arguments->parent_chain, + dest_file, out, to_fullpath, + arguments->use_bitmap, checksum_map, + arguments->shift_lsn, lsn_map, true); } - else if (strcmp(file->name, "pg_control") == 0) - copy_pgcontrol_file(from_root, FIO_BACKUP_HOST, - instance_config.pgdata, FIO_DB_HOST, - file); else - copy_file(FIO_BACKUP_HOST, - instance_config.pgdata, FIO_DB_HOST, - file, false); - - /* print size of restored file */ - if (file->write_size != BYTES_INVALID) - elog(VERBOSE, "Restored file %s : " INT64_FORMAT " bytes", - file->path, file->write_size); + { + /* disable stdio buffering for local destination non-data file */ + if (!fio_is_remote_file(out)) + setvbuf(out, NULL, _IONBF, BUFSIZ); + /* Destination file is non-data file */ + arguments->restored_bytes += restore_non_data_file(arguments->parent_chain, + arguments->dest_backup, dest_file, out, to_fullpath, + already_exists); + } + +done: + /* Writing is asynchronous in case of restore in remote mode, so check the agent status */ + if (fio_check_error_file(out, &errmsg)) + elog(ERROR, "Cannot write to the remote file \"%s\": %s", to_fullpath, errmsg); + + /* close file */ + if (fio_fclose(out) != 0) + elog(ERROR, "Cannot close file \"%s\": %s", to_fullpath, + strerror(errno)); + + /* free pagemap used for restore optimization */ + pg_free(dest_file->pagemap.bitmap); + + if (lsn_map) + pg_free(lsn_map->bitmap); + + pg_free(lsn_map); + pg_free(checksum_map); } + free(out_buf); + + /* ssh connection to longer needed */ + fio_disconnect(); + /* Data files restoring is successful */ arguments->ret = 0; return NULL; } -/* Create recovery.conf with given recovery target parameters */ +/* + * Create recovery.conf (postgresql.auto.conf in case of PG12) + * with given recovery target parameters + */ static void -create_recovery_conf(time_t backup_id, +create_recovery_conf(InstanceState *instanceState, time_t backup_id, pgRecoveryTarget *rt, - pgBackup *backup) + pgBackup *backup, + pgRestoreParams *params) { - char path[MAXPGPATH]; - FILE *fp; - bool need_restore_conf; bool target_latest; + bool target_immediate; + bool restore_command_provided = false; - target_latest = rt->target_stop != NULL && - strcmp(rt->target_stop, "latest") == 0; - need_restore_conf = !backup->stream || - (rt->time_string || rt->xid_string || rt->lsn_string) || target_latest; + if (instance_config.restore_command && + (pg_strcasecmp(instance_config.restore_command, "none") != 0)) + { + restore_command_provided = true; + } - /* No need to generate recovery.conf at all. */ - if (!(need_restore_conf || restore_as_replica)) - return; + /* restore-target='latest' support */ + target_latest = (rt->target_tli_string != NULL && + strcmp(rt->target_tli_string, "latest") == 0) || + (rt->target_stop != NULL && + strcmp(rt->target_stop, "latest") == 0); + + target_immediate = rt->target_stop != NULL && + strcmp(rt->target_stop, "immediate") == 0; + + /* + * Note that setting restore_command alone interpreted + * as PITR with target - "until all available WAL is replayed". + * We do this because of the following case: + * The user is restoring STREAM backup as replica but + * also relies on WAL archive to catch-up with master. + * If restore_command is provided, then it should be + * added to recovery config. + * In this scenario, "would be" replica will replay + * all WAL segments available in WAL archive, after that + * it will try to connect to master via repprotocol. + * + * The risk is obvious, what if masters current state is + * in "the past" relatively to latest state in the archive? + * We will get a replica that is "in the future" to the master. + * We accept this risk because its probability is low. + */ + if (!backup->stream || rt->time_string || + rt->xid_string || rt->lsn_string || rt->target_name || + target_immediate || target_latest || restore_command_provided) + params->recovery_settings_mode = PITR_REQUESTED; + /* + * The recovery-target-timeline option can be 'latest' for streaming backups. + * This operation requires a WAL archive for PITR. + */ + if (rt->target_tli && backup->stream && params->recovery_settings_mode != PITR_REQUESTED) + elog(WARNING, "The '--recovery-target-timeline' option applied for STREAM backup. " + "The timeline number will be ignored."); elog(LOG, "----------------------------------------"); - elog(LOG, "creating recovery.conf"); - snprintf(path, lengthof(path), "%s/recovery.conf", instance_config.pgdata); +#if PG_VERSION_NUM >= 120000 + update_recovery_options(instanceState, backup, params, rt); +#else + update_recovery_options_before_v12(instanceState, backup, params, rt); +#endif +} + + +/* TODO get rid of using global variables: instance_config */ +static void +print_recovery_settings(InstanceState *instanceState, FILE *fp, pgBackup *backup, + pgRestoreParams *params, pgRecoveryTarget *rt) +{ + char restore_command_guc[16384]; + fio_fprintf(fp, "## recovery settings\n"); + + /* If restore_command is provided, use it. Otherwise construct it from scratch. */ + if (instance_config.restore_command && + (pg_strcasecmp(instance_config.restore_command, "none") != 0)) + sprintf(restore_command_guc, "%s", instance_config.restore_command); + else + { + /* default cmdline, ok for local restore */ + sprintf(restore_command_guc, "\"%s\" archive-get -B \"%s\" --instance \"%s\" " + "--wal-file-path=%%p --wal-file-name=%%f", + PROGRAM_FULL_PATH ? PROGRAM_FULL_PATH : PROGRAM_NAME, + /* TODO What is going on here? Why do we use catalog path as wal-file-path? */ + instanceState->catalog_state->catalog_path, instanceState->instance_name); + + /* append --remote-* parameters provided via --archive-* settings */ + if (instance_config.archive.host) + { + strcat(restore_command_guc, " --remote-host="); + strcat(restore_command_guc, instance_config.archive.host); + } + + if (instance_config.archive.port) + { + strcat(restore_command_guc, " --remote-port="); + strcat(restore_command_guc, instance_config.archive.port); + } + + if (instance_config.archive.user) + { + strcat(restore_command_guc, " --remote-user="); + strcat(restore_command_guc, instance_config.archive.user); + } + } + + /* + * We've already checked that only one of the four following mutually + * exclusive options is specified, so the order of calls is insignificant. + */ + if (rt->target_name) + fio_fprintf(fp, "recovery_target_name = '%s'\n", rt->target_name); + + if (rt->time_string) + fio_fprintf(fp, "recovery_target_time = '%s'\n", rt->time_string); + + if (rt->xid_string) + fio_fprintf(fp, "recovery_target_xid = '%s'\n", rt->xid_string); + + if (rt->lsn_string) + fio_fprintf(fp, "recovery_target_lsn = '%s'\n", rt->lsn_string); + + if (rt->target_stop && (strcmp(rt->target_stop, "immediate") == 0)) + fio_fprintf(fp, "recovery_target = '%s'\n", rt->target_stop); + + if (rt->inclusive_specified) + fio_fprintf(fp, "recovery_target_inclusive = '%s'\n", + rt->target_inclusive ? "true" : "false"); + + if (rt->target_tli) + fio_fprintf(fp, "recovery_target_timeline = '%u'\n", rt->target_tli); + else + { + if (rt->target_tli_string) + fio_fprintf(fp, "recovery_target_timeline = '%s'\n", rt->target_tli_string); + else if (rt->target_stop && (strcmp(rt->target_stop, "latest") == 0)) + fio_fprintf(fp, "recovery_target_timeline = 'latest'\n"); +#if PG_VERSION_NUM >= 120000 + else + { + /* + * In PG12 default recovery target timeline was changed to 'latest', which + * is extremely risky. Explicitly preserve old behavior of recovering to current + * timneline for PG12. + */ + fio_fprintf(fp, "recovery_target_timeline = 'current'\n"); + } +#endif + } + + if (rt->target_action) + fio_fprintf(fp, "recovery_target_action = '%s'\n", rt->target_action); + else + /* default recovery_target_action is 'pause' */ + fio_fprintf(fp, "recovery_target_action = '%s'\n", "pause"); + + elog(LOG, "Setting restore_command to '%s'", restore_command_guc); + fio_fprintf(fp, "restore_command = '%s'\n", restore_command_guc); +} + +static void +print_standby_settings_common(FILE *fp, pgBackup *backup, pgRestoreParams *params) +{ + fio_fprintf(fp, "\n## standby settings\n"); + if (params->primary_conninfo) + fio_fprintf(fp, "primary_conninfo = '%s'\n", params->primary_conninfo); + else if (backup->primary_conninfo) + fio_fprintf(fp, "primary_conninfo = '%s'\n", backup->primary_conninfo); + + if (params->primary_slot_name != NULL) + fio_fprintf(fp, "primary_slot_name = '%s'\n", params->primary_slot_name); +} + +#if PG_VERSION_NUM < 120000 +static void +update_recovery_options_before_v12(InstanceState *instanceState, pgBackup *backup, + pgRestoreParams *params, pgRecoveryTarget *rt) +{ + FILE *fp; + char path[MAXPGPATH]; + + /* + * If PITR is not requested and instance is not restored as replica, + * then recovery.conf should not be created. + */ + if (params->recovery_settings_mode != PITR_REQUESTED && + !params->restore_as_replica) + { + return; + } + + elog(LOG, "update recovery settings in recovery.conf"); + join_path_components(path, instance_config.pgdata, "recovery.conf"); + fp = fio_fopen(path, "w", FIO_DB_HOST); if (fp == NULL) - elog(ERROR, "cannot open recovery.conf \"%s\": %s", path, + elog(ERROR, "Cannot open file \"%s\": %s", path, strerror(errno)); + if (fio_chmod(path, FILE_PERMISSION, FIO_DB_HOST) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", path, strerror(errno)); + fio_fprintf(fp, "# recovery.conf generated by pg_probackup %s\n", PROGRAM_VERSION); - if (need_restore_conf) - { + if (params->recovery_settings_mode == PITR_REQUESTED) + print_recovery_settings(instanceState, fp, backup, params, rt); - fio_fprintf(fp, "restore_command = '%s archive-get -B %s --instance %s " - "--wal-file-path %%p --wal-file-name %%f'\n", - PROGRAM_FULL_PATH ? PROGRAM_FULL_PATH : PROGRAM_NAME, - backup_path, instance_name); + if (params->restore_as_replica) + { + print_standby_settings_common(fp, backup, params); + fio_fprintf(fp, "standby_mode = 'on'\n"); + } - /* - * We've already checked that only one of the four following mutually - * exclusive options is specified, so the order of calls is insignificant. - */ - if (rt->target_name) - fio_fprintf(fp, "recovery_target_name = '%s'\n", rt->target_name); + if (fio_fflush(fp) != 0 || + fio_fclose(fp)) + elog(ERROR, "Cannot write file \"%s\": %s", path, + strerror(errno)); +} +#endif - if (rt->time_string) - fio_fprintf(fp, "recovery_target_time = '%s'\n", rt->time_string); +/* + * Read postgresql.auto.conf, clean old recovery options, + * to avoid unexpected intersections. + * Write recovery options for this backup. + */ +#if PG_VERSION_NUM >= 120000 +static void +update_recovery_options(InstanceState *instanceState, pgBackup *backup, + pgRestoreParams *params, pgRecoveryTarget *rt) - if (rt->xid_string) - fio_fprintf(fp, "recovery_target_xid = '%s'\n", rt->xid_string); +{ + char postgres_auto_path[MAXPGPATH]; + char postgres_auto_path_tmp[MAXPGPATH]; + char path[MAXPGPATH]; + FILE *fp = NULL; + FILE *fp_tmp = NULL; + struct stat st; + char current_time_str[100]; + /* postgresql.auto.conf parsing */ + char line[16384] = "\0"; + char *buf = NULL; + int buf_len = 0; + int buf_len_max = 16384; - if (rt->lsn_string) - fio_fprintf(fp, "recovery_target_lsn = '%s'\n", rt->lsn_string); + elog(LOG, "update recovery settings in postgresql.auto.conf"); - if (rt->target_stop && !target_latest) - fio_fprintf(fp, "recovery_target = '%s'\n", rt->target_stop); + time2iso(current_time_str, lengthof(current_time_str), current_time, false); - if (rt->inclusive_specified) - fio_fprintf(fp, "recovery_target_inclusive = '%s'\n", - rt->target_inclusive ? "true" : "false"); + join_path_components(postgres_auto_path, instance_config.pgdata, "postgresql.auto.conf"); - if (rt->target_tli) - fio_fprintf(fp, "recovery_target_timeline = '%u'\n", rt->target_tli); + if (fio_stat(postgres_auto_path, &st, false, FIO_DB_HOST) < 0) + { + /* file not found is not an error case */ + if (errno != ENOENT) + elog(ERROR, "Cannot stat file \"%s\": %s", postgres_auto_path, + strerror(errno)); + st.st_size = 0; + } - if (rt->target_action) - fio_fprintf(fp, "recovery_target_action = '%s'\n", rt->target_action); + /* Kludge for 0-sized postgresql.auto.conf file. TODO: make something more intelligent */ + if (st.st_size > 0) + { + fp = fio_open_stream(postgres_auto_path, FIO_DB_HOST); + if (fp == NULL) + elog(ERROR, "Cannot open \"%s\": %s", postgres_auto_path, strerror(errno)); } - if (restore_as_replica) + sprintf(postgres_auto_path_tmp, "%s.tmp", postgres_auto_path); + fp_tmp = fio_fopen(postgres_auto_path_tmp, "w", FIO_DB_HOST); + if (fp_tmp == NULL) + elog(ERROR, "Cannot open \"%s\": %s", postgres_auto_path_tmp, strerror(errno)); + + while (fp && fgets(line, lengthof(line), fp)) { - fio_fprintf(fp, "standby_mode = 'on'\n"); + /* ignore "include 'probackup_recovery.conf'" directive */ + if (strstr(line, "include") && + strstr(line, "probackup_recovery.conf")) + { + continue; + } - if (backup->primary_conninfo) - fio_fprintf(fp, "primary_conninfo = '%s'\n", backup->primary_conninfo); + /* ignore already existing recovery options */ + if (strstr(line, "restore_command") || + strstr(line, "recovery_target")) + { + continue; + } + + if (!buf) + buf = pgut_malloc(buf_len_max); + + /* avoid buffer overflow */ + if ((buf_len + strlen(line)) >= buf_len_max) + { + buf_len_max += (buf_len + strlen(line)) *2; + buf = pgut_realloc(buf, buf_len_max); + } + + buf_len += snprintf(buf+buf_len, sizeof(line), "%s", line); } - if (fio_fflush(fp) != 0 || - fio_fclose(fp)) - elog(ERROR, "cannot write recovery.conf \"%s\": %s", path, - strerror(errno)); + /* close input postgresql.auto.conf */ + if (fp) + fio_close_stream(fp); + + /* Write data to postgresql.auto.conf.tmp */ + if (buf_len > 0 && + (fio_fwrite(fp_tmp, buf, buf_len) != buf_len)) + elog(ERROR, "Cannot write to \"%s\": %s", + postgres_auto_path_tmp, strerror(errno)); + + if (fio_fflush(fp_tmp) != 0 || + fio_fclose(fp_tmp)) + elog(ERROR, "Cannot write file \"%s\": %s", postgres_auto_path_tmp, + strerror(errno)); + pg_free(buf); + + if (fio_rename(postgres_auto_path_tmp, postgres_auto_path, FIO_DB_HOST) < 0) + elog(ERROR, "Cannot rename file \"%s\" to \"%s\": %s", + postgres_auto_path_tmp, postgres_auto_path, strerror(errno)); + + if (fio_chmod(postgres_auto_path, FILE_PERMISSION, FIO_DB_HOST) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", postgres_auto_path, strerror(errno)); + + if (params) + { + fp = fio_fopen(postgres_auto_path, "a", FIO_DB_HOST); + if (fp == NULL) + elog(ERROR, "Cannot open file \"%s\": %s", postgres_auto_path, + strerror(errno)); + + fio_fprintf(fp, "\n# recovery settings added by pg_probackup restore of backup %s at '%s'\n", + backup_id_of(backup), current_time_str); + + if (params->recovery_settings_mode == PITR_REQUESTED) + print_recovery_settings(instanceState, fp, backup, params, rt); + + if (params->restore_as_replica) + print_standby_settings_common(fp, backup, params); + + if (fio_fflush(fp) != 0 || + fio_fclose(fp)) + elog(ERROR, "Cannot write file \"%s\": %s", postgres_auto_path, + strerror(errno)); + + /* + * Create "recovery.signal" to mark this recovery as PITR for PostgreSQL. + * In older versions presense of recovery.conf alone was enough. + * To keep behaviour consistent with older versions, + * we are forced to create "recovery.signal" + * even when only restore_command is provided. + * Presense of "recovery.signal" by itself determine only + * one thing: do PostgreSQL must switch to a new timeline + * after successfull recovery or not? + */ + if (params->recovery_settings_mode == PITR_REQUESTED) + { + elog(LOG, "creating recovery.signal file"); + join_path_components(path, instance_config.pgdata, "recovery.signal"); + + fp = fio_fopen(path, PG_BINARY_W, FIO_DB_HOST); + if (fp == NULL) + elog(ERROR, "Cannot open file \"%s\": %s", path, + strerror(errno)); + + if (fio_fflush(fp) != 0 || + fio_fclose(fp)) + elog(ERROR, "Cannot write file \"%s\": %s", path, + strerror(errno)); + } + + if (params->restore_as_replica) + { + elog(LOG, "creating standby.signal file"); + join_path_components(path, instance_config.pgdata, "standby.signal"); + + fp = fio_fopen(path, PG_BINARY_W, FIO_DB_HOST); + if (fp == NULL) + elog(ERROR, "Cannot open file \"%s\": %s", path, + strerror(errno)); + + if (fio_fflush(fp) != 0 || + fio_fclose(fp)) + elog(ERROR, "Cannot write file \"%s\": %s", path, + strerror(errno)); + } + } } +#endif /* * Try to read a timeline's history file. @@ -867,7 +1794,7 @@ create_recovery_conf(time_t backup_id, * based on readTimeLineHistory() in timeline.c */ parray * -read_timeline_history(TimeLineID targetTLI) +read_timeline_history(const char *arclog_path, TimeLineID targetTLI, bool strict) { parray *result; char path[MAXPGPATH]; @@ -887,12 +1814,15 @@ read_timeline_history(TimeLineID targetTLI) if (fd == NULL) { if (errno != ENOENT) - elog(ERROR, "could not open file \"%s\": %s", path, + elog(ERROR, "Could not open file \"%s\": %s", path, strerror(errno)); /* There is no history file for target timeline */ - elog(ERROR, "recovery target timeline %u does not exist", - targetTLI); + if (strict) + elog(ERROR, "Recovery target timeline %u does not exist", + targetTLI); + else + return NULL; } } @@ -923,12 +1853,12 @@ read_timeline_history(TimeLineID targetTLI) { /* expect a numeric timeline ID as first field of line */ elog(ERROR, - "syntax error in history file: %s. Expected a numeric timeline ID.", + "Syntax error in history file: %s. Expected a numeric timeline ID.", fline); } if (nfields != 3) elog(ERROR, - "syntax error in history file: %s. Expected a transaction log switchpoint location.", + "Syntax error in history file: %s. Expected a transaction log switchpoint location.", fline); if (last_timeline && tli <= last_timeline->tli) @@ -946,12 +1876,23 @@ read_timeline_history(TimeLineID targetTLI) /* we ignore the remainder of each line */ } + if (fd && (ferror(fd))) + elog(ERROR, "Failed to read from file: \"%s\"", path); + if (fd) fclose(fd); if (last_timeline && targetTLI <= last_timeline->tli) elog(ERROR, "Timeline IDs must be less than child timeline's ID."); + /* History file is empty or corrupted */ + if (parray_num(result) == 0 && targetTLI != 1) + { + elog(WARNING, "History file is corrupted or missing: \"%s\"", path); + pg_free(result); + return NULL; + } + /* append target timeline */ entry = pgut_new(TimeLineHistoryEntry); entry->tli = targetTLI; @@ -980,20 +1921,46 @@ satisfy_recovery_target(const pgBackup *backup, const pgRecoveryTarget *rt) /* TODO description */ bool -satisfy_timeline(const parray *timelines, const pgBackup *backup) +satisfy_timeline(const parray *timelines, TimeLineID tli, XLogRecPtr lsn) { int i; + elog(VERBOSE, "satisfy_timeline() checking: tli = %X, lsn = %X/%X", + tli, (uint32) (lsn >> 32), (uint32) lsn); for (i = 0; i < parray_num(timelines); i++) { TimeLineHistoryEntry *timeline; timeline = (TimeLineHistoryEntry *) parray_get(timelines, i); - if (backup->tli == timeline->tli && + elog(VERBOSE, "satisfy_timeline() check %i entry: timeline->tli = %X, timeline->end = %X/%X", + i, timeline->tli, (uint32) (timeline->end >> 32), (uint32) timeline->end); + if (tli == timeline->tli && (XLogRecPtrIsInvalid(timeline->end) || - backup->stop_lsn < timeline->end)) + lsn <= timeline->end)) + return true; + } + return false; +} + +/* timelines represents a history of one particular timeline, + * we must determine whether a target tli is part of that history. + * + * /--------* + * ---------*--------------> + */ +bool +tliIsPartOfHistory(const parray *timelines, TimeLineID tli) +{ + int i; + + for (i = 0; i < parray_num(timelines); i++) + { + TimeLineHistoryEntry *timeline = (TimeLineHistoryEntry *) parray_get(timelines, i); + + if (tli == timeline->tli) return true; } + return false; } /* @@ -1004,12 +1971,11 @@ pgRecoveryTarget * parseRecoveryTargetOptions(const char *target_time, const char *target_xid, const char *target_inclusive, - TimeLineID target_tli, + const char *target_tli_string, const char *target_lsn, const char *target_stop, const char *target_name, - const char *target_action, - bool no_validate) + const char *target_action) { bool dummy_bool; /* @@ -1033,7 +1999,7 @@ parseRecoveryTargetOptions(const char *target_time, if (parse_time(target_time, &dummy_time, false)) rt->target_time = dummy_time; else - elog(ERROR, "Invalid value for --recovery-target-time option %s", + elog(ERROR, "Invalid value for '--recovery-target-time' option '%s'", target_time); } @@ -1051,7 +2017,7 @@ parseRecoveryTargetOptions(const char *target_time, #endif rt->target_xid = dummy_xid; else - elog(ERROR, "Invalid value for --recovery-target-xid option %s", + elog(ERROR, "Invalid value for '--recovery-target-xid' option '%s'", target_xid); } @@ -1064,7 +2030,7 @@ parseRecoveryTargetOptions(const char *target_time, if (parse_lsn(target_lsn, &dummy_lsn)) rt->target_lsn = dummy_lsn; else - elog(ERROR, "Invalid value of --recovery-target-lsn option %s", + elog(ERROR, "Invalid value of '--recovery-target-lsn' option '%s'", target_lsn); } @@ -1074,24 +2040,35 @@ parseRecoveryTargetOptions(const char *target_time, if (parse_bool(target_inclusive, &dummy_bool)) rt->target_inclusive = dummy_bool; else - elog(ERROR, "Invalid value for --recovery-target-inclusive option %s", + elog(ERROR, "Invalid value for '--recovery-target-inclusive' option '%s'", target_inclusive); } - rt->target_tli = target_tli; + rt->target_tli_string = target_tli_string; + rt->target_tli = 0; + /* target_tli can contains timeline number, "current" or "latest" */ + if(target_tli_string && strcmp(target_tli_string, "current") != 0 && strcmp(target_tli_string, "latest") != 0) + { + errno = 0; + rt->target_tli = strtoul(target_tli_string, NULL, 10); + if (errno == EINVAL || errno == ERANGE || !rt->target_tli) + { + elog(ERROR, "Invalid value for '--recovery-target-timeline' option '%s'", + target_tli_string); + } + } + if (target_stop) { if ((strcmp(target_stop, "immediate") != 0) && (strcmp(target_stop, "latest") != 0)) - elog(ERROR, "Invalid value for --recovery-target option %s", + elog(ERROR, "Invalid value for '--recovery-target' option '%s'", target_stop); recovery_target_specified++; rt->target_stop = target_stop; } - rt->no_validate = no_validate; - if (target_name) { recovery_target_specified++; @@ -1103,20 +2080,17 @@ parseRecoveryTargetOptions(const char *target_time, if ((strcmp(target_action, "pause") != 0) && (strcmp(target_action, "promote") != 0) && (strcmp(target_action, "shutdown") != 0)) - elog(ERROR, "Invalid value for --recovery-target-action option %s", + elog(ERROR, "Invalid value for '--recovery-target-action' option '%s'", target_action); rt->target_action = target_action; } - else - { - /* Default recovery target action is pause */ - rt->target_action = "pause"; - } /* More than one mutually exclusive option was defined. */ if (recovery_target_specified > 1) - elog(ERROR, "At most one of --recovery-target, --recovery-target-name, --recovery-target-time, --recovery-target-xid, or --recovery-target-lsn can be specified"); + elog(ERROR, "At most one of '--recovery-target', '--recovery-target-name', " + "'--recovery-target-time', '--recovery-target-xid' or " + "'--recovery-target-lsn' options can be specified"); /* * If none of the options is defined, '--recovery-target-inclusive' option @@ -1124,7 +2098,266 @@ parseRecoveryTargetOptions(const char *target_time, */ if (!(rt->xid_string || rt->time_string || rt->lsn_string) && rt->target_inclusive) - elog(ERROR, "--recovery-target-inclusive option applies when either --recovery-target-time, --recovery-target-xid or --recovery-target-lsn is specified"); + elog(ERROR, "The '--recovery-target-inclusive' option can be applied only when " + "either of '--recovery-target-time', '--recovery-target-xid' or " + "'--recovery-target-lsn' options is specified"); + + /* If none of the options is defined, '--recovery-target-action' is meaningless */ + if (rt->target_action && recovery_target_specified == 0) + elog(ERROR, "The '--recovery-target-action' option can be applied only when " + "either of '--recovery-target', '--recovery-target-time', '--recovery-target-xid', " + "'--recovery-target-lsn' or '--recovery-target-name' options is specified"); + + /* TODO: sanity for recovery-target-timeline */ return rt; } + +/* + * Return array of dbOids of databases that should not be restored + * Regardless of what option user used, db-include or db-exclude, + * we always convert it into exclude_list. + */ +parray * +get_dbOid_exclude_list(pgBackup *backup, parray *datname_list, + PartialRestoreType partial_restore_type) +{ + int i; + int j; +// pg_crc32 crc; + parray *database_map = NULL; + parray *dbOid_exclude_list = NULL; + pgFile *database_map_file = NULL; + char path[MAXPGPATH]; + char database_map_path[MAXPGPATH]; + parray *files = NULL; + + files = get_backup_filelist(backup, true); + + /* look for 'database_map' file in backup_content.control */ + for (i = 0; i < parray_num(files); i++) + { + pgFile *file = (pgFile *) parray_get(files, i); + + if ((file->external_dir_num == 0) && + strcmp(DATABASE_MAP, file->name) == 0) + { + database_map_file = file; + break; + } + } + + if (!database_map_file) + elog(ERROR, "Backup %s doesn't contain a database_map, partial restore is impossible.", + backup_id_of(backup)); + + join_path_components(path, backup->root_dir, DATABASE_DIR); + join_path_components(database_map_path, path, DATABASE_MAP); + + /* check database_map CRC */ +// crc = pgFileGetCRC(database_map_path, true, true, NULL, FIO_LOCAL_HOST); +// +// if (crc != database_map_file->crc) +// elog(ERROR, "Invalid CRC of backup file \"%s\" : %X. Expected %X", +// database_map_file->path, crc, database_map_file->crc); + + /* get database_map from file */ + database_map = read_database_map(backup); + + /* partial restore requested but database_map is missing */ + if (!database_map) + elog(ERROR, "Backup %s has empty or mangled database_map, partial restore is impossible.", + backup_id_of(backup)); + + /* + * So we have a list of datnames and a database_map for it. + * We must construct a list of dbOids to exclude. + */ + if (partial_restore_type == INCLUDE) + { + /* For 'include', keep dbOid of every datname NOT specified by user */ + for (i = 0; i < parray_num(datname_list); i++) + { + bool found_match = false; + char *datname = (char *) parray_get(datname_list, i); + + for (j = 0; j < parray_num(database_map); j++) + { + db_map_entry *db_entry = (db_map_entry *) parray_get(database_map, j); + + /* got a match */ + if (strcmp(db_entry->datname, datname) == 0) + { + found_match = true; + /* for db-include we must exclude db_entry from database_map */ + parray_remove(database_map, j); + j--; + } + } + /* If specified datname is not found in database_map, error out */ + if (!found_match) + elog(ERROR, "Failed to find a database '%s' in database_map of backup %s", + datname, backup_id_of(backup)); + } + + /* At this moment only databases to exclude are left in the map */ + for (j = 0; j < parray_num(database_map); j++) + { + db_map_entry *db_entry = (db_map_entry *) parray_get(database_map, j); + + if (!dbOid_exclude_list) + dbOid_exclude_list = parray_new(); + parray_append(dbOid_exclude_list, &db_entry->dbOid); + } + } + else if (partial_restore_type == EXCLUDE) + { + /* For exclude, job is easier - find dbOid for every specified datname */ + for (i = 0; i < parray_num(datname_list); i++) + { + bool found_match = false; + char *datname = (char *) parray_get(datname_list, i); + + for (j = 0; j < parray_num(database_map); j++) + { + db_map_entry *db_entry = (db_map_entry *) parray_get(database_map, j); + + /* got a match */ + if (strcmp(db_entry->datname, datname) == 0) + { + found_match = true; + /* for db-exclude we must add dbOid to exclude list */ + if (!dbOid_exclude_list) + dbOid_exclude_list = parray_new(); + parray_append(dbOid_exclude_list, &db_entry->dbOid); + } + } + /* If specified datname is not found in database_map, error out */ + if (!found_match) + elog(ERROR, "Failed to find a database '%s' in database_map of backup %s", + datname, backup_id_of(backup)); + } + } + + /* extra sanity: ensure that list is not empty */ + if (!dbOid_exclude_list || parray_num(dbOid_exclude_list) < 1) + elog(ERROR, "Failed to find a match in database_map of backup %s for partial restore", + backup_id_of(backup)); + + /* clean backup filelist */ + if (files) + { + parray_walk(files, pgFileFree); + parray_free(files); + } + + /* sort dbOid array in ASC order */ + parray_qsort(dbOid_exclude_list, pgCompareOid); + + return dbOid_exclude_list; +} + +/* Check that instance is suitable for incremental restore + * Depending on type of incremental restore requirements are differs. + * + * TODO: add PG_CONTROL_IS_MISSING + */ +DestDirIncrCompatibility +check_incremental_compatibility(const char *pgdata, uint64 system_identifier, + IncrRestoreMode incremental_mode, + parray *partial_db_list, + bool allow_partial_incremental) +{ + uint64 system_id_pgdata; + bool system_id_match = false; + bool success = true; + bool postmaster_is_up = false; + bool backup_label_exists = false; + pid_t pid; + char backup_label[MAXPGPATH]; + + /* check postmaster pid */ + pid = fio_check_postmaster(pgdata, FIO_DB_HOST); + + if (pid == 1) /* postmaster.pid is mangled */ + { + char pid_file[MAXPGPATH]; + + join_path_components(pid_file, pgdata, "postmaster.pid"); + elog(WARNING, "Pid file \"%s\" is mangled, cannot determine whether postmaster is running or not", + pid_file); + success = false; + } + else if (pid > 1) /* postmaster is up */ + { + elog(WARNING, "Postmaster with pid %u is running in destination directory \"%s\"", + pid, pgdata); + success = false; + postmaster_is_up = true; + } + + /* check that PG_VERSION is the same */ + + /* slurp pg_control and check that system ID is the same + * check that instance is not running + * if lsn_based, check that there is no backup_label files is around AND + * get redo point lsn from destination pg_control. + + * It is really important to be sure that pg_control is in cohesion with + * data files content, because based on pg_control information we will + * choose a backup suitable for lsn based incremental restore. + */ + elog(LOG, "Trying to read pg_control file in destination directory"); + + get_control_file_or_back_file(pgdata, FIO_DB_HOST, &instance_control); + control_downloaded = true; + + system_id_pgdata = instance_control.system_identifier; + + if (system_id_pgdata == instance_config.system_identifier) + system_id_match = true; + else + elog(WARNING, "Backup catalog was initialized for system id %lu, " + "but destination directory system id is %lu", + system_identifier, system_id_pgdata); + + /* + * TODO: maybe there should be some other signs, pointing to pg_control + * desynchronization with cluster state. + */ + if (incremental_mode == INCR_LSN) + { + join_path_components(backup_label, pgdata, "backup_label"); + if (fio_access(backup_label, F_OK, FIO_DB_HOST) == 0) + { + elog(WARNING, "Destination directory contains \"backup_control\" file. " + "This does NOT mean that you should delete this file and retry, only that " + "incremental restore in 'lsn' mode may produce incorrect result, when applied " + "to cluster with pg_control not synchronized with cluster state." + "Consider to use incremental restore in 'checksum' mode"); + success = false; + backup_label_exists = true; + } + } + + if (postmaster_is_up) + return POSTMASTER_IS_RUNNING; + + /* PG_CONTROL MISSING */ + + /* PG_CONTROL unreadable */ + + if (!system_id_match) + return SYSTEM_ID_MISMATCH; + + if (backup_label_exists) + return BACKUP_LABEL_EXISTS; + + if (partial_db_list && !allow_partial_incremental) + return PARTIAL_INCREMENTAL_FORBIDDEN; + /* some other error condition */ + if (!success) + return DEST_IS_NOT_OK; + + return DEST_OK; +} diff --git a/src/show.c b/src/show.c index 4ba95ebc8..810262df6 100644 --- a/src/show.c +++ b/src/show.c @@ -3,7 +3,7 @@ * show.c: show backup information. * * Portions Copyright (c) 2009-2011, NIPPON TELEGRAPH AND TELEPHONE CORPORATION - * Portions Copyright (c) 2015-2019, Postgres Professional + * Portions Copyright (c) 2015-2022, Postgres Professional * *------------------------------------------------------------------------- */ @@ -12,10 +12,14 @@ #include #include +#include #include #include "utils/json.h" +#define half_rounded(x) (((x) + ((x) < 0 ? 0 : 1)) / 2) + +/* struct to align fields printed in plain format */ typedef struct ShowBackendRow { const char *instance; @@ -27,197 +31,264 @@ typedef struct ShowBackendRow char tli[20]; char duration[20]; char data_bytes[20]; + char wal_bytes[20]; + char zratio[20]; char start_lsn[20]; char stop_lsn[20]; const char *status; } ShowBackendRow; +/* struct to align fields printed in plain format */ +typedef struct ShowArchiveRow +{ + char tli[20]; + char parent_tli[20]; + char switchpoint[20]; + char min_segno[MAXFNAMELEN]; + char max_segno[MAXFNAMELEN]; + char n_segments[20]; + char size[20]; + char zratio[20]; + const char *status; + char n_backups[20]; +} ShowArchiveRow; static void show_instance_start(void); static void show_instance_end(void); -static void show_instance(time_t requested_backup_id, bool show_name); -static int show_backup(time_t requested_backup_id); +static void show_instance(InstanceState *instanceState, time_t requested_backup_id, bool show_name); +static void print_backup_json_object(PQExpBuffer buf, pgBackup *backup); +static int show_backup(InstanceState *instanceState, time_t requested_backup_id); + +static void show_instance_plain(const char *instance_name, parray *backup_list, bool show_name); +static void show_instance_json(const char *instance_name, parray *backup_list); -static void show_instance_plain(parray *backup_list, bool show_name); -static void show_instance_json(parray *backup_list); +static void show_instance_archive(InstanceState *instanceState, InstanceConfig *instance); +static void show_archive_plain(const char *instance_name, uint32 xlog_seg_size, + parray *timelines_list, bool show_name); +static void show_archive_json(const char *instance_name, uint32 xlog_seg_size, + parray *tli_list); +static bool backup_has_tablespace_map(pgBackup *backup); static PQExpBufferData show_buf; static bool first_instance = true; static int32 json_level = 0; +static const char* lc_env_locale; +typedef enum { + LOCALE_C, // Used for formatting output to unify the dot-based floating point representation + LOCALE_ENV // Default environment locale +} output_numeric_locale; + +#ifdef HAVE_USELOCALE +static locale_t env_locale, c_locale; +#endif +void memorize_environment_locale() { + lc_env_locale = (const char *)getenv("LC_NUMERIC"); + lc_env_locale = lc_env_locale != NULL ? lc_env_locale : "C"; +#ifdef HAVE_USELOCALE + env_locale = newlocale(LC_NUMERIC_MASK, lc_env_locale, (locale_t)0); + c_locale = newlocale(LC_NUMERIC_MASK, "C", (locale_t)0); +#else +#ifdef HAVE__CONFIGTHREADLOCALE + _configthreadlocale(_ENABLE_PER_THREAD_LOCALE); +#endif +#endif +} + +void free_environment_locale() { +#ifdef HAVE_USELOCALE + freelocale(env_locale); + freelocale(c_locale); +#endif +} + +static void set_output_numeric_locale(output_numeric_locale loc) { +#ifdef HAVE_USELOCALE + uselocale(loc == LOCALE_C ? c_locale : env_locale); +#else + setlocale(LC_NUMERIC, loc == LOCALE_C ? "C" : lc_env_locale); +#endif +} + +/* + * Entry point of pg_probackup SHOW subcommand. + */ int -do_show(time_t requested_backup_id) +do_show(CatalogState *catalogState, InstanceState *instanceState, + time_t requested_backup_id, bool show_archive) { - if (instance_name == NULL && + int i; + + if (instanceState == NULL && requested_backup_id != INVALID_BACKUP_ID) - elog(ERROR, "You must specify --instance to use --backup_id option"); + elog(ERROR, "You must specify --instance to use (-i, --backup-id) option"); - if (instance_name == NULL) - { - /* Show list of instances */ - char path[MAXPGPATH]; - DIR *dir; - struct dirent *dent; + if (show_archive && + requested_backup_id != INVALID_BACKUP_ID) + elog(ERROR, "You cannot specify --archive and (-i, --backup-id) options together"); - /* open directory and list contents */ - join_path_components(path, backup_path, BACKUPS_DIR); - dir = opendir(path); - if (dir == NULL) - elog(ERROR, "Cannot open directory \"%s\": %s", - path, strerror(errno)); + /* + * if instance is not specified, + * show information about all instances in this backup catalog + */ + if (instanceState == NULL) + { + parray *instances = catalog_get_instance_list(catalogState); show_instance_start(); - - while (errno = 0, (dent = readdir(dir)) != NULL) + for (i = 0; i < parray_num(instances); i++) { - char child[MAXPGPATH]; - struct stat st; - - /* skip entries point current dir or parent dir */ - if (strcmp(dent->d_name, ".") == 0 || - strcmp(dent->d_name, "..") == 0) - continue; - - join_path_components(child, path, dent->d_name); - - if (lstat(child, &st) == -1) - elog(ERROR, "Cannot stat file \"%s\": %s", - child, strerror(errno)); + instanceState = parray_get(instances, i); - if (!S_ISDIR(st.st_mode)) - continue; + if (interrupted) + elog(ERROR, "Interrupted during show"); - instance_name = dent->d_name; - sprintf(backup_instance_path, "%s/%s/%s", backup_path, BACKUPS_DIR, instance_name); - - show_instance(INVALID_BACKUP_ID, true); + if (show_archive) + show_instance_archive(instanceState, instanceState->config); + else + show_instance(instanceState, INVALID_BACKUP_ID, true); } - - if (errno) - elog(ERROR, "Cannot read directory \"%s\": %s", - path, strerror(errno)); - - if (closedir(dir)) - elog(ERROR, "Cannot close directory \"%s\": %s", - path, strerror(errno)); - show_instance_end(); return 0; } - else if (requested_backup_id == INVALID_BACKUP_ID || - show_format == SHOW_JSON) + /* always use */ + else if (show_format == SHOW_JSON || + requested_backup_id == INVALID_BACKUP_ID) { show_instance_start(); - show_instance(requested_backup_id, false); + + if (show_archive) + { + InstanceConfig *instance = readInstanceConfigFile(instanceState); + show_instance_archive(instanceState, instance); + } + else + show_instance(instanceState, requested_backup_id, false); + show_instance_end(); return 0; } else - return show_backup(requested_backup_id); + { + if (show_archive) + { + InstanceConfig *instance = readInstanceConfigFile(instanceState); + show_instance_archive(instanceState, instance); + } + else + show_backup(instanceState, requested_backup_id); + + return 0; + } } void pretty_size(int64 size, char *buf, size_t len) { - int exp = 0; + int64 limit = 10 * 1024; + int64 limit2 = limit * 2 - 1; /* minus means the size is invalid */ - if (size < 0) - { - strncpy(buf, "----", len); - return; - } +// if (size < 0) +// { +// strncpy(buf, "----", len); +// return; +// } - /* determine postfix */ - while (size > 9999) + if (size <= 0) { - ++exp; - size /= 1000; + strncpy(buf, "0", len); + return; } - switch (exp) + if (size < limit) + snprintf(buf, len, "%dB", (int) size); + else { - case 0: - snprintf(buf, len, "%dB", (int) size); - break; - case 1: - snprintf(buf, len, "%dkB", (int) size); - break; - case 2: - snprintf(buf, len, "%dMB", (int) size); - break; - case 3: - snprintf(buf, len, "%dGB", (int) size); - break; - case 4: - snprintf(buf, len, "%dTB", (int) size); - break; - case 5: - snprintf(buf, len, "%dPB", (int) size); - break; - default: - strncpy(buf, "***", len); - break; + size >>= 9; + if (size < limit2) + snprintf(buf, len, "%dkB", (int) half_rounded(size)); + else + { + size >>= 10; + if (size < limit2) + snprintf(buf, len, "%dMB", (int) half_rounded(size)); + else + { + size >>= 10; + if (size < limit2) + snprintf(buf, len, "%dGB", (int) half_rounded(size)); + else + { + size >>= 10; + snprintf(buf, len, "%dTB", (int) half_rounded(size)); + } + } + } } } -static TimeLineID -get_parent_tli(TimeLineID child_tli) +void +pretty_time_interval(double time, char *buf, size_t len) { - TimeLineID result = 0; - char path[MAXPGPATH]; - char fline[MAXPGPATH]; - FILE *fd; + int num_seconds = 0; + int milliseconds = 0; + int seconds = 0; + int minutes = 0; + int hours = 0; + int days = 0; - /* Timeline 1 does not have a history file and parent timeline */ - if (child_tli == 1) - return 0; + num_seconds = (int) time; - /* Search history file in archives */ - snprintf(path, lengthof(path), "%s/%08X.history", arclog_path, - child_tli); - fd = fopen(path, "rt"); - if (fd == NULL) + if (time <= 0) { - if (errno != ENOENT) - elog(ERROR, "could not open file \"%s\": %s", path, - strerror(errno)); - - /* Did not find history file, do not raise the error */ - return 0; + strncpy(buf, "0", len); + return; } - /* - * Parse the file... - */ - while (fgets(fline, sizeof(fline), fd) != NULL) + days = num_seconds / (24 * 3600); + num_seconds %= (24 * 3600); + + hours = num_seconds / 3600; + num_seconds %= 3600; + + minutes = num_seconds / 60; + num_seconds %= 60; + + seconds = num_seconds; + milliseconds = (int)((time - (int) time) * 1000.0); + + if (days > 0) { - /* skip leading whitespace and check for # comment */ - char *ptr; - char *endptr; + snprintf(buf, len, "%dd:%dh", days, hours); + return; + } - for (ptr = fline; *ptr; ptr++) - { - if (!IsSpace(*ptr)) - break; - } - if (*ptr == '\0' || *ptr == '#') - continue; + if (hours > 0) + { + snprintf(buf, len, "%dh:%dm", hours, minutes); + return; + } - /* expect a numeric timeline ID as first field of line */ - result = (TimeLineID) strtoul(ptr, &endptr, 0); - if (endptr == ptr) - elog(ERROR, - "syntax error(timeline ID) in history file: %s", - fline); + if (minutes > 0) + { + snprintf(buf, len, "%dm:%ds", minutes, seconds); + return; } - fclose(fd); + if (seconds > 0) + { + if (milliseconds > 0) + snprintf(buf, len, "%ds:%dms", seconds, milliseconds); + else + snprintf(buf, len, "%ds", seconds); + return; + } - /* TLI of the last line is parent TLI */ - return result; + snprintf(buf, len, "%dms", milliseconds); + return; } /* @@ -255,16 +326,16 @@ show_instance_end(void) * Show brief meta information about all backups in the backup instance. */ static void -show_instance(time_t requested_backup_id, bool show_name) +show_instance(InstanceState *instanceState, time_t requested_backup_id, bool show_name) { parray *backup_list; - backup_list = catalog_get_backup_list(requested_backup_id); + backup_list = catalog_get_backup_list(instanceState, requested_backup_id); if (show_format == SHOW_PLAIN) - show_instance_plain(backup_list, show_name); + show_instance_plain(instanceState->instance_name, backup_list, show_name); else if (show_format == SHOW_JSON) - show_instance_json(backup_list); + show_instance_json(instanceState->instance_name, backup_list); else elog(ERROR, "Invalid show format %d", (int) show_format); @@ -273,59 +344,264 @@ show_instance(time_t requested_backup_id, bool show_name) parray_free(backup_list); } +/* helper routine to print backup info as json object */ +static void +print_backup_json_object(PQExpBuffer buf, pgBackup *backup) +{ + TimeLineID parent_tli = 0; + char timestamp[100] = "----"; + char lsn[20]; + + json_add(buf, JT_BEGIN_OBJECT, &json_level); + + json_add_value(buf, "id", backup_id_of(backup), json_level, + true); + + if (backup->parent_backup != 0) + json_add_value(buf, "parent-backup-id", + base36enc(backup->parent_backup), json_level, true); + + json_add_value(buf, "backup-mode", pgBackupGetBackupMode(backup, false), + json_level, true); + + json_add_value(buf, "wal", backup->stream ? "STREAM": "ARCHIVE", + json_level, true); + + json_add_value(buf, "compress-alg", + deparse_compress_alg(backup->compress_alg), json_level, + true); + + json_add_key(buf, "compress-level", json_level); + appendPQExpBuffer(buf, "%d", backup->compress_level); + + json_add_value(buf, "from-replica", + backup->from_replica ? "true" : "false", json_level, + true); + + json_add_key(buf, "block-size", json_level); + appendPQExpBuffer(buf, "%u", backup->block_size); + + json_add_key(buf, "xlog-block-size", json_level); + appendPQExpBuffer(buf, "%u", backup->wal_block_size); + + json_add_key(buf, "checksum-version", json_level); + appendPQExpBuffer(buf, "%u", backup->checksum_version); + + json_add_value(buf, "program-version", backup->program_version, + json_level, true); + json_add_value(buf, "server-version", backup->server_version, + json_level, true); + + json_add_key(buf, "current-tli", json_level); + appendPQExpBuffer(buf, "%d", backup->tli); + + json_add_key(buf, "parent-tli", json_level); + + /* Only incremental backup can have Parent TLI */ + if (backup->parent_backup_link) + parent_tli = backup->parent_backup_link->tli; + + appendPQExpBuffer(buf, "%u", parent_tli); + + snprintf(lsn, lengthof(lsn), "%X/%X", + (uint32) (backup->start_lsn >> 32), (uint32) backup->start_lsn); + json_add_value(buf, "start-lsn", lsn, json_level, true); + + snprintf(lsn, lengthof(lsn), "%X/%X", + (uint32) (backup->stop_lsn >> 32), (uint32) backup->stop_lsn); + json_add_value(buf, "stop-lsn", lsn, json_level, true); + + time2iso(timestamp, lengthof(timestamp), backup->start_time, false); + json_add_value(buf, "start-time", timestamp, json_level, true); + + if (backup->end_time) + { + time2iso(timestamp, lengthof(timestamp), backup->end_time, false); + json_add_value(buf, "end-time", timestamp, json_level, true); + } + + json_add_key(buf, "recovery-xid", json_level); + appendPQExpBuffer(buf, XID_FMT, backup->recovery_xid); + + if (backup->recovery_time > 0) + { + time2iso(timestamp, lengthof(timestamp), backup->recovery_time, false); + json_add_value(buf, "recovery-time", timestamp, json_level, true); + } + + if (backup->expire_time > 0) + { + time2iso(timestamp, lengthof(timestamp), backup->expire_time, false); + json_add_value(buf, "expire-time", timestamp, json_level, true); + } + + if (backup->data_bytes != BYTES_INVALID) + { + json_add_key(buf, "data-bytes", json_level); + appendPQExpBuffer(buf, INT64_FORMAT, backup->data_bytes); + } + + if (backup->wal_bytes != BYTES_INVALID) + { + json_add_key(buf, "wal-bytes", json_level); + appendPQExpBuffer(buf, INT64_FORMAT, backup->wal_bytes); + } + + if (backup->uncompressed_bytes >= 0) + { + json_add_key(buf, "uncompressed-bytes", json_level); + appendPQExpBuffer(buf, INT64_FORMAT, backup->uncompressed_bytes); + } + + if (backup->pgdata_bytes >= 0) + { + json_add_key(buf, "pgdata-bytes", json_level); + appendPQExpBuffer(buf, INT64_FORMAT, backup->pgdata_bytes); + } + + if (backup->primary_conninfo) + json_add_value(buf, "primary_conninfo", backup->primary_conninfo, + json_level, true); + + if (backup->external_dir_str) + json_add_value(buf, "external-dirs", backup->external_dir_str, + json_level, true); + + json_add_value(buf, "status", status2str(backup->status), json_level, + true); + + if (backup->note) + json_add_value(buf, "note", backup->note, + json_level, true); + + if (backup->content_crc != 0) + { + json_add_key(buf, "content-crc", json_level); + appendPQExpBuffer(buf, "%u", backup->content_crc); + } + + /* print tablespaces list */ + if (backup_has_tablespace_map(backup)) + { + parray *links = parray_new(); + + json_add_key(buf, "tablespace_map", json_level); + json_add(buf, JT_BEGIN_ARRAY, &json_level); + + read_tablespace_map(links, backup->root_dir); + parray_qsort(links, pgFileCompareLinked); + + for (size_t i = 0; i < parray_num(links); i++){ + pgFile *link = (pgFile *) parray_get(links, i); + if (i) + appendPQExpBufferChar(buf, ','); + json_add(buf, JT_BEGIN_OBJECT, &json_level); + json_add_value(buf, "oid", link->name, json_level, true); + json_add_value(buf, "path", link->linked, json_level, true); + json_add(buf, JT_END_OBJECT, &json_level); + } + /* End of tablespaces */ + json_add(buf, JT_END_ARRAY, &json_level); + parray_walk(links, pgFileFree); + parray_free(links); + } + + json_add(buf, JT_END_OBJECT, &json_level); +} + /* * Show detailed meta information about specified backup. */ static int -show_backup(time_t requested_backup_id) +show_backup(InstanceState *instanceState, time_t requested_backup_id) { - pgBackup *backup; + int i; + pgBackup *backup = NULL; + parray *backups; + + //TODO pass requested_backup_id to the function + backups = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); + + /* Find requested backup */ + for (i = 0; i < parray_num(backups); i++) + { + pgBackup *tmp_backup = (pgBackup *) parray_get(backups, i); + + /* found target */ + if (tmp_backup->start_time == requested_backup_id) + { + backup = tmp_backup; + break; + } + } - backup = read_backup(requested_backup_id); if (backup == NULL) { + // TODO for 3.0: we should ERROR out here. elog(INFO, "Requested backup \"%s\" is not found.", /* We do not need free base36enc's result, we exit anyway */ base36enc(requested_backup_id)); + parray_walk(backups, pgBackupFree); + parray_free(backups); /* This is not error */ return 0; } if (show_format == SHOW_PLAIN) - pgBackupWriteControl(stdout, backup); + { + pgBackupWriteControl(stdout, backup, false); + + /* print tablespaces list */ + if (backup_has_tablespace_map(backup)) + { + parray *links = parray_new(); + + fio_fprintf(stdout, "\ntablespace_map = '"); + + read_tablespace_map(links, backup->root_dir); + parray_qsort(links, pgFileCompareLinked); + + for (size_t i = 0; i < parray_num(links); i++){ + pgFile *link = (pgFile *) parray_get(links, i); + fio_fprintf(stdout, "%s %s%s", link->name, link->linked, (i < parray_num(links) - 1) ? "; " : "'\n"); + } + parray_walk(links, pgFileFree); + parray_free(links); + } + } else elog(ERROR, "Invalid show format %d", (int) show_format); /* cleanup */ - pgBackupFree(backup); + parray_walk(backups, pgBackupFree); + parray_free(backups); return 0; } -/* - * Plain output. - */ - /* * Show instance backups in plain format. */ static void -show_instance_plain(parray *backup_list, bool show_name) +show_instance_plain(const char *instance_name, parray *backup_list, bool show_name) { -#define SHOW_FIELDS_COUNT 12 +#define SHOW_FIELDS_COUNT 14 int i; const char *names[SHOW_FIELDS_COUNT] = { "Instance", "Version", "ID", "Recovery Time", - "Mode", "WAL", "Current/Parent TLI", "Time", "Data", - "Start LSN", "Stop LSN", "Status" }; + "Mode", "WAL Mode", "TLI", "Time", "Data", "WAL", + "Zratio", "Start LSN", "Stop LSN", "Status" }; const char *field_formats[SHOW_FIELDS_COUNT] = { " %-*s ", " %-*s ", " %-*s ", " %-*s ", - " %-*s ", " %-*s ", " %-*s ", " %*s ", " %*s ", - " %*s ", " %*s ", " %-*s "}; + " %-*s ", " %-*s ", " %-*s ", " %*s ", " %*s ", " %*s ", + " %*s ", " %-*s ", " %-*s ", " %-*s "}; uint32 widths[SHOW_FIELDS_COUNT]; uint32 widths_sum = 0; ShowBackendRow *rows; - time_t current_time = time(NULL); + TimeLineID parent_tli = 0; + + // Since we've been printing a table, set LC_NUMERIC to its default environment value + set_output_numeric_locale(LOCALE_ENV); for (i = 0; i < SHOW_FIELDS_COUNT; i++) widths[i] = strlen(names[i]); @@ -341,6 +617,7 @@ show_instance_plain(parray *backup_list, bool show_name) pgBackup *backup = parray_get(backup_list, i); ShowBackendRow *row = &rows[i]; int cur = 0; + float zratio = 1; /* Instance */ row->instance = instance_name; @@ -355,48 +632,51 @@ show_instance_plain(parray *backup_list, bool show_name) /* ID */ snprintf(row->backup_id, lengthof(row->backup_id), "%s", - base36enc(backup->start_time)); + backup_id_of(backup)); widths[cur] = Max(widths[cur], strlen(row->backup_id)); cur++; /* Recovery Time */ if (backup->recovery_time != (time_t) 0) time2iso(row->recovery_time, lengthof(row->recovery_time), - backup->recovery_time); + backup->recovery_time, false); else - StrNCpy(row->recovery_time, "----", sizeof(row->recovery_time)); + strlcpy(row->recovery_time, "----", sizeof(row->recovery_time)); widths[cur] = Max(widths[cur], strlen(row->recovery_time)); cur++; /* Mode */ - row->mode = pgBackupGetBackupMode(backup); - widths[cur] = Max(widths[cur], strlen(row->mode)); + row->mode = pgBackupGetBackupMode(backup, show_color); + widths[cur] = Max(widths[cur], strlen(row->mode) - (show_color ? TC_LEN : 0)); cur++; - /* WAL */ + /* WAL mode*/ row->wal_mode = backup->stream ? "STREAM": "ARCHIVE"; widths[cur] = Max(widths[cur], strlen(row->wal_mode)); cur++; /* Current/Parent TLI */ - snprintf(row->tli, lengthof(row->tli), "%u / %u", + if (backup->parent_backup_link != NULL) + parent_tli = backup->parent_backup_link->tli; + + snprintf(row->tli, lengthof(row->tli), "%u/%u", backup->tli, - backup->backup_mode == BACKUP_MODE_FULL ? 0 : get_parent_tli(backup->tli)); + backup->backup_mode == BACKUP_MODE_FULL ? 0 : parent_tli); widths[cur] = Max(widths[cur], strlen(row->tli)); cur++; /* Time */ if (backup->status == BACKUP_STATUS_RUNNING) - snprintf(row->duration, lengthof(row->duration), "%.*lfs", 0, - difftime(current_time, backup->start_time)); + pretty_time_interval(difftime(current_time, backup->start_time), + row->duration, lengthof(row->duration)); else if (backup->merge_time != (time_t) 0) - snprintf(row->duration, lengthof(row->duration), "%.*lfs", 0, - difftime(backup->end_time, backup->merge_time)); + pretty_time_interval(difftime(backup->end_time, backup->merge_time), + row->duration, lengthof(row->duration)); else if (backup->end_time != (time_t) 0) - snprintf(row->duration, lengthof(row->duration), "%.*lfs", 0, - difftime(backup->end_time, backup->start_time)); + pretty_time_interval(difftime(backup->end_time, backup->start_time), + row->duration, lengthof(row->duration)); else - StrNCpy(row->duration, "----", sizeof(row->duration)); + strlcpy(row->duration, "----", sizeof(row->duration)); widths[cur] = Max(widths[cur], strlen(row->duration)); cur++; @@ -406,6 +686,25 @@ show_instance_plain(parray *backup_list, bool show_name) widths[cur] = Max(widths[cur], strlen(row->data_bytes)); cur++; + /* WAL */ + pretty_size(backup->wal_bytes, row->wal_bytes, + lengthof(row->wal_bytes)); + widths[cur] = Max(widths[cur], strlen(row->wal_bytes)); + cur++; + + /* Zratio (compression ratio) */ + if (backup->uncompressed_bytes != BYTES_INVALID && + (backup->uncompressed_bytes > 0 && backup->data_bytes > 0)) + { + zratio = (float)backup->uncompressed_bytes / (backup->data_bytes); + snprintf(row->zratio, lengthof(row->zratio), "%.2f", zratio); + } + else + snprintf(row->zratio, lengthof(row->zratio), "%.2f", zratio); + + widths[cur] = Max(widths[cur], strlen(row->zratio)); + cur++; + /* Start LSN */ snprintf(row->start_lsn, lengthof(row->start_lsn), "%X/%X", (uint32) (backup->start_lsn >> 32), @@ -421,8 +720,9 @@ show_instance_plain(parray *backup_list, bool show_name) cur++; /* Status */ - row->status = status2str(backup->status); - widths[cur] = Max(widths[cur], strlen(row->status)); + row->status = show_color ? status2str_color(backup->status) : status2str(backup->status); + widths[cur] = Max(widths[cur], strlen(row->status) - (show_color ? TC_LEN : 0)); + } for (i = 0; i < SHOW_FIELDS_COUNT; i++) @@ -472,7 +772,7 @@ show_instance_plain(parray *backup_list, bool show_name) row->recovery_time); cur++; - appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur] + (show_color ? TC_LEN : 0), row->mode); cur++; @@ -492,6 +792,14 @@ show_instance_plain(parray *backup_list, bool show_name) row->data_bytes); cur++; + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->wal_bytes); + cur++; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->zratio); + cur++; + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], row->start_lsn); cur++; @@ -500,7 +808,7 @@ show_instance_plain(parray *backup_list, bool show_name) row->stop_lsn); cur++; - appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur] + (show_color ? TC_LEN : 0), row->status); cur++; @@ -508,17 +816,15 @@ show_instance_plain(parray *backup_list, bool show_name) } pfree(rows); + // Restore the C locale + set_output_numeric_locale(LOCALE_C); } -/* - * Json output. - */ - /* * Show instance backups in json format. */ static void -show_instance_json(parray *backup_list) +show_instance_json(const char *instance_name, parray *backup_list) { int i; PQExpBuffer buf = &show_buf; @@ -540,118 +846,374 @@ show_instance_json(parray *backup_list) for (i = 0; i < parray_num(backup_list); i++) { pgBackup *backup = parray_get(backup_list, i); - TimeLineID parent_tli; - char timestamp[100] = "----"; - char lsn[20]; if (i != 0) appendPQExpBufferChar(buf, ','); - json_add(buf, JT_BEGIN_OBJECT, &json_level); + print_backup_json_object(buf, backup); + } - json_add_value(buf, "id", base36enc(backup->start_time), json_level, - true); + /* End of backups */ + json_add(buf, JT_END_ARRAY, &json_level); - if (backup->parent_backup != 0) - json_add_value(buf, "parent-backup-id", - base36enc(backup->parent_backup), json_level, true); + /* End of instance object */ + json_add(buf, JT_END_OBJECT, &json_level); - json_add_value(buf, "backup-mode", pgBackupGetBackupMode(backup), - json_level, true); + first_instance = false; +} - json_add_value(buf, "wal", backup->stream ? "STREAM": "ARCHIVE", - json_level, true); +/* + * show information about WAL archive of the instance + */ +static void +show_instance_archive(InstanceState *instanceState, InstanceConfig *instance) +{ + parray *timelineinfos; - json_add_value(buf, "compress-alg", - deparse_compress_alg(backup->compress_alg), json_level, - true); + timelineinfos = catalog_get_timelines(instanceState, instance); - json_add_key(buf, "compress-level", json_level); - appendPQExpBuffer(buf, "%d", backup->compress_level); + if (show_format == SHOW_PLAIN) + show_archive_plain(instanceState->instance_name, instance->xlog_seg_size, timelineinfos, true); + else if (show_format == SHOW_JSON) + show_archive_json(instanceState->instance_name, instance->xlog_seg_size, timelineinfos); + else + elog(ERROR, "Invalid show format %d", (int) show_format); +} - json_add_value(buf, "from-replica", - backup->from_replica ? "true" : "false", json_level, - true); +static void +show_archive_plain(const char *instance_name, uint32 xlog_seg_size, + parray *tli_list, bool show_name) +{ + char segno_tmp[MAXFNAMELEN]; + parray *actual_tli_list = parray_new(); +#define SHOW_ARCHIVE_FIELDS_COUNT 10 + int i; + const char *names[SHOW_ARCHIVE_FIELDS_COUNT] = + { "TLI", "Parent TLI", "Switchpoint", + "Min Segno", "Max Segno", "N segments", "Size", "Zratio", "N backups", "Status"}; + const char *field_formats[SHOW_ARCHIVE_FIELDS_COUNT] = + { " %-*s ", " %-*s ", " %-*s ", " %-*s ", + " %-*s ", " %-*s ", " %-*s ", " %-*s ", " %-*s ", " %-*s "}; + uint32 widths[SHOW_ARCHIVE_FIELDS_COUNT]; + uint32 widths_sum = 0; + ShowArchiveRow *rows; + + // Since we've been printing a table, set LC_NUMERIC to its default environment value + set_output_numeric_locale(LOCALE_ENV); + + for (i = 0; i < SHOW_ARCHIVE_FIELDS_COUNT; i++) + widths[i] = strlen(names[i]); + + /* Ignore empty timelines */ + for (i = 0; i < parray_num(tli_list); i++) + { + timelineInfo *tlinfo = (timelineInfo *) parray_get(tli_list, i); + + if (tlinfo->n_xlog_files > 0) + parray_append(actual_tli_list, tlinfo); + } + + rows = (ShowArchiveRow *) palloc0(parray_num(actual_tli_list) * + sizeof(ShowArchiveRow)); - json_add_key(buf, "block-size", json_level); - appendPQExpBuffer(buf, "%u", backup->block_size); + /* + * Fill row values and calculate maximum width of each field. + */ + for (i = 0; i < parray_num(actual_tli_list); i++) + { + timelineInfo *tlinfo = (timelineInfo *) parray_get(actual_tli_list, i); + ShowArchiveRow *row = &rows[i]; + int cur = 0; + float zratio = 0; - json_add_key(buf, "xlog-block-size", json_level); - appendPQExpBuffer(buf, "%u", backup->wal_block_size); + /* TLI */ + snprintf(row->tli, lengthof(row->tli), "%u", + tlinfo->tli); + widths[cur] = Max(widths[cur], strlen(row->tli)); + cur++; - json_add_key(buf, "checksum-version", json_level); - appendPQExpBuffer(buf, "%u", backup->checksum_version); + /* Parent TLI */ + snprintf(row->parent_tli, lengthof(row->parent_tli), "%u", + tlinfo->parent_tli); + widths[cur] = Max(widths[cur], strlen(row->parent_tli)); + cur++; - json_add_value(buf, "program-version", backup->program_version, - json_level, true); - json_add_value(buf, "server-version", backup->server_version, - json_level, true); + /* Switchpoint LSN */ + snprintf(row->switchpoint, lengthof(row->switchpoint), "%X/%X", + (uint32) (tlinfo->switchpoint >> 32), + (uint32) tlinfo->switchpoint); + widths[cur] = Max(widths[cur], strlen(row->switchpoint)); + cur++; - json_add_key(buf, "current-tli", json_level); - appendPQExpBuffer(buf, "%d", backup->tli); + /* Min Segno */ + GetXLogFileName(segno_tmp, tlinfo->tli, tlinfo->begin_segno, xlog_seg_size); + snprintf(row->min_segno, lengthof(row->min_segno), "%s",segno_tmp); + + widths[cur] = Max(widths[cur], strlen(row->min_segno)); + cur++; + + /* Max Segno */ + GetXLogFileName(segno_tmp, tlinfo->tli, tlinfo->end_segno, xlog_seg_size); + snprintf(row->max_segno, lengthof(row->max_segno), "%s", segno_tmp); + + widths[cur] = Max(widths[cur], strlen(row->max_segno)); + cur++; + + /* N files */ + snprintf(row->n_segments, lengthof(row->n_segments), "%lu", + tlinfo->n_xlog_files); + widths[cur] = Max(widths[cur], strlen(row->n_segments)); + cur++; + + /* Size */ + pretty_size(tlinfo->size, row->size, + lengthof(row->size)); + widths[cur] = Max(widths[cur], strlen(row->size)); + cur++; + + /* Zratio (compression ratio) */ + if (tlinfo->size != 0) + zratio = ((float)xlog_seg_size*tlinfo->n_xlog_files) / tlinfo->size; + + snprintf(row->zratio, lengthof(row->n_segments), "%.2f", zratio); + widths[cur] = Max(widths[cur], strlen(row->zratio)); + cur++; + + /* N backups */ + snprintf(row->n_backups, lengthof(row->n_backups), "%lu", + tlinfo->backups?parray_num(tlinfo->backups):0); + widths[cur] = Max(widths[cur], strlen(row->n_backups)); + cur++; + + /* Status */ + if (tlinfo->lost_segments == NULL) + row->status = "OK"; + else + row->status = "DEGRADED"; + widths[cur] = Max(widths[cur], strlen(row->status)); + cur++; + } + + for (i = 0; i < SHOW_ARCHIVE_FIELDS_COUNT; i++) + widths_sum += widths[i] + 2 /* two space */; + + if (show_name) + appendPQExpBuffer(&show_buf, "\nARCHIVE INSTANCE '%s'\n", instance_name); + + /* + * Print header. + */ + for (i = 0; i < widths_sum; i++) + appendPQExpBufferChar(&show_buf, '='); + appendPQExpBufferChar(&show_buf, '\n'); + + for (i = 0; i < SHOW_ARCHIVE_FIELDS_COUNT; i++) + { + appendPQExpBuffer(&show_buf, field_formats[i], widths[i], names[i]); + } + appendPQExpBufferChar(&show_buf, '\n'); + + for (i = 0; i < widths_sum; i++) + appendPQExpBufferChar(&show_buf, '='); + appendPQExpBufferChar(&show_buf, '\n'); + + /* + * Print values. + */ + for (i = parray_num(actual_tli_list) - 1; i >= 0; i--) + { + ShowArchiveRow *row = &rows[i]; + int cur = 0; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->tli); + cur++; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->parent_tli); + cur++; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->switchpoint); + cur++; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->min_segno); + cur++; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->max_segno); + cur++; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->n_segments); + cur++; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->size); + cur++; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->zratio); + cur++; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->n_backups); + cur++; + + appendPQExpBuffer(&show_buf, field_formats[cur], widths[cur], + row->status); + cur++; + appendPQExpBufferChar(&show_buf, '\n'); + } + + pfree(rows); + // Restore the C locale + set_output_numeric_locale(LOCALE_C); + //TODO: free timelines +} + +static void +show_archive_json(const char *instance_name, uint32 xlog_seg_size, + parray *tli_list) +{ + int i,j; + PQExpBuffer buf = &show_buf; + parray *actual_tli_list = parray_new(); + char segno_tmp[MAXFNAMELEN]; + + if (!first_instance) + appendPQExpBufferChar(buf, ','); + + /* Begin of instance object */ + json_add(buf, JT_BEGIN_OBJECT, &json_level); + + json_add_value(buf, "instance", instance_name, json_level, true); + json_add_key(buf, "timelines", json_level); + + /* Ignore empty timelines */ + + for (i = 0; i < parray_num(tli_list); i++) + { + timelineInfo *tlinfo = (timelineInfo *) parray_get(tli_list, i); + + if (tlinfo->n_xlog_files > 0) + parray_append(actual_tli_list, tlinfo); + } + + /* + * List timelines. + */ + json_add(buf, JT_BEGIN_ARRAY, &json_level); + + for (i = parray_num(actual_tli_list) - 1; i >= 0; i--) + { + timelineInfo *tlinfo = (timelineInfo *) parray_get(actual_tli_list, i); + char tmp_buf[MAXFNAMELEN]; + float zratio = 0; + + if (i != (parray_num(actual_tli_list) - 1)) + appendPQExpBufferChar(buf, ','); + + json_add(buf, JT_BEGIN_OBJECT, &json_level); + + json_add_key(buf, "tli", json_level); + appendPQExpBuffer(buf, "%u", tlinfo->tli); json_add_key(buf, "parent-tli", json_level); + appendPQExpBuffer(buf, "%u", tlinfo->parent_tli); + + snprintf(tmp_buf, lengthof(tmp_buf), "%X/%X", + (uint32) (tlinfo->switchpoint >> 32), (uint32) tlinfo->switchpoint); + json_add_value(buf, "switchpoint", tmp_buf, json_level, true); - /* Only incremental backup can have Parent TLI */ - if (backup->backup_mode == BACKUP_MODE_FULL) - parent_tli = 0; + GetXLogFileName(segno_tmp, tlinfo->tli, tlinfo->begin_segno, xlog_seg_size); + snprintf(tmp_buf, lengthof(tmp_buf), "%s", segno_tmp); + json_add_value(buf, "min-segno", tmp_buf, json_level, true); + + GetXLogFileName(segno_tmp, tlinfo->tli, tlinfo->end_segno, xlog_seg_size); + snprintf(tmp_buf, lengthof(tmp_buf), "%s", segno_tmp); + json_add_value(buf, "max-segno", tmp_buf, json_level, true); + + json_add_key(buf, "n-segments", json_level); + appendPQExpBuffer(buf, "%lu", tlinfo->n_xlog_files); + + json_add_key(buf, "size", json_level); + appendPQExpBuffer(buf, "%lu", tlinfo->size); + + json_add_key(buf, "zratio", json_level); + + if (tlinfo->size != 0) + zratio = ((float) xlog_seg_size * tlinfo->n_xlog_files) / tlinfo->size; + appendPQExpBuffer(buf, "%.2f", zratio); + + if (tlinfo->closest_backup != NULL) + snprintf(tmp_buf, lengthof(tmp_buf), "%s", + backup_id_of(tlinfo->closest_backup)); else - parent_tli = get_parent_tli(backup->tli); - appendPQExpBuffer(buf, "%u", parent_tli); + snprintf(tmp_buf, lengthof(tmp_buf), "%s", ""); - snprintf(lsn, lengthof(lsn), "%X/%X", - (uint32) (backup->start_lsn >> 32), (uint32) backup->start_lsn); - json_add_value(buf, "start-lsn", lsn, json_level, true); + json_add_value(buf, "closest-backup-id", tmp_buf, json_level, true); - snprintf(lsn, lengthof(lsn), "%X/%X", - (uint32) (backup->stop_lsn >> 32), (uint32) backup->stop_lsn); - json_add_value(buf, "stop-lsn", lsn, json_level, true); + if (tlinfo->lost_segments == NULL) + json_add_value(buf, "status", "OK", json_level, true); + else + json_add_value(buf, "status", "DEGRADED", json_level, true); - time2iso(timestamp, lengthof(timestamp), backup->start_time); - json_add_value(buf, "start-time", timestamp, json_level, true); + json_add_key(buf, "lost-segments", json_level); - if (backup->end_time) + if (tlinfo->lost_segments != NULL) { - time2iso(timestamp, lengthof(timestamp), backup->end_time); - json_add_value(buf, "end-time", timestamp, json_level, true); - } + json_add(buf, JT_BEGIN_ARRAY, &json_level); - json_add_key(buf, "recovery-xid", json_level); - appendPQExpBuffer(buf, XID_FMT, backup->recovery_xid); + for (j = 0; j < parray_num(tlinfo->lost_segments); j++) + { + xlogInterval *lost_segments = (xlogInterval *) parray_get(tlinfo->lost_segments, j); - if (backup->recovery_time > 0) - { - time2iso(timestamp, lengthof(timestamp), backup->recovery_time); - json_add_value(buf, "recovery-time", timestamp, json_level, true); - } + if (j != 0) + appendPQExpBufferChar(buf, ','); - if (backup->data_bytes != BYTES_INVALID) - { - json_add_key(buf, "data-bytes", json_level); - appendPQExpBuffer(buf, INT64_FORMAT, backup->data_bytes); + json_add(buf, JT_BEGIN_OBJECT, &json_level); + + GetXLogFileName(segno_tmp, tlinfo->tli, lost_segments->begin_segno, xlog_seg_size); + snprintf(tmp_buf, lengthof(tmp_buf), "%s", segno_tmp); + json_add_value(buf, "begin-segno", tmp_buf, json_level, true); + + GetXLogFileName(segno_tmp, tlinfo->tli, lost_segments->end_segno, xlog_seg_size); + snprintf(tmp_buf, lengthof(tmp_buf), "%s", segno_tmp); + json_add_value(buf, "end-segno", tmp_buf, json_level, true); + json_add(buf, JT_END_OBJECT, &json_level); + } + json_add(buf, JT_END_ARRAY, &json_level); } + else + appendPQExpBuffer(buf, "[]"); - if (backup->wal_bytes != BYTES_INVALID) + json_add_key(buf, "backups", json_level); + + if (tlinfo->backups != NULL) { - json_add_key(buf, "wal-bytes", json_level); - appendPQExpBuffer(buf, INT64_FORMAT, backup->wal_bytes); - } + json_add(buf, JT_BEGIN_ARRAY, &json_level); + for (j = 0; j < parray_num(tlinfo->backups); j++) + { + pgBackup *backup = parray_get(tlinfo->backups, j); - if (backup->primary_conninfo) - json_add_value(buf, "primary_conninfo", backup->primary_conninfo, - json_level, true); + if (j != 0) + appendPQExpBufferChar(buf, ','); - if (backup->external_dir_str) - json_add_value(buf, "external-dirs", backup->external_dir_str, - json_level, true); + print_backup_json_object(buf, backup); + } - json_add_value(buf, "status", status2str(backup->status), json_level, - true); + json_add(buf, JT_END_ARRAY, &json_level); + } + else + appendPQExpBuffer(buf, "[]"); + /* End of timeline */ json_add(buf, JT_END_OBJECT, &json_level); } - /* End of backups */ + /* End of timelines object */ json_add(buf, JT_END_ARRAY, &json_level); /* End of instance object */ @@ -659,3 +1221,10 @@ show_instance_json(parray *backup_list) first_instance = false; } + +static bool backup_has_tablespace_map(pgBackup *backup) +{ + char map_path[MAXPGPATH]; + join_path_components(map_path, backup->database_dir, PG_TABLESPACE_MAP_FILE); + return fileExists(map_path, FIO_BACKUP_HOST); +} diff --git a/src/stream.c b/src/stream.c new file mode 100644 index 000000000..77453e997 --- /dev/null +++ b/src/stream.c @@ -0,0 +1,779 @@ +/*------------------------------------------------------------------------- + * + * stream.c: pg_probackup specific code for WAL streaming + * + * Portions Copyright (c) 2015-2020, Postgres Professional + * + *------------------------------------------------------------------------- + */ + +#include "pg_probackup.h" +#include "receivelog.h" +#include "streamutil.h" +#include "access/timeline.h" + +#include +#include + +/* + * global variable needed by ReceiveXlogStream() + * + * standby_message_timeout controls how often we send a message + * back to the primary letting it know our progress, in milliseconds. + * + * in pg_probackup we use a default setting = 10 sec + */ +static int standby_message_timeout = 10 * 1000; + +/* stop_backup_lsn is set by pg_stop_backup() to stop streaming */ +XLogRecPtr stop_backup_lsn = InvalidXLogRecPtr; +static XLogRecPtr stop_stream_lsn = InvalidXLogRecPtr; + +/* + * How long we should wait for streaming end in seconds. + * Retrieved as checkpoint_timeout + checkpoint_timeout * 0.1 + */ +static uint32 stream_stop_timeout = 0; +/* Time in which we started to wait for streaming end */ +static time_t stream_stop_begin = 0; + +/* + * We need to wait end of WAL streaming before execute pg_stop_backup(). + */ +typedef struct +{ + char basedir[MAXPGPATH]; + PGconn *conn; + + /* + * Return value from the thread. + * 0 means there is no error, 1 - there is an error. + */ + int ret; + + XLogRecPtr startpos; + TimeLineID starttli; +} StreamThreadArg; + +static pthread_t stream_thread; +static StreamThreadArg stream_thread_arg = {"", NULL, 1}; + +static parray *xlog_files_list = NULL; +static bool do_crc = true; + +static void IdentifySystem(StreamThreadArg *stream_thread_arg); +static int checkpoint_timeout(PGconn *backup_conn); +static void *StreamLog(void *arg); +static bool stop_streaming(XLogRecPtr xlogpos, uint32 timeline, + bool segment_finished); +static void add_walsegment_to_filelist(parray *filelist, uint32 timeline, + XLogRecPtr xlogpos, char *basedir, + uint32 xlog_seg_size); +static void add_history_file_to_filelist(parray *filelist, uint32 timeline, + char *basedir); + +/* + * Run IDENTIFY_SYSTEM through a given connection and + * check system identifier and timeline are matching + */ +static void +IdentifySystem(StreamThreadArg *stream_thread_arg) +{ + PGresult *res; + + uint64 stream_conn_sysidentifier = 0; + char *stream_conn_sysidentifier_str; + TimeLineID stream_conn_tli = 0; + + if (!CheckServerVersionForStreaming(stream_thread_arg->conn)) + { + PQfinish(stream_thread_arg->conn); + /* + * Error message already written in CheckServerVersionForStreaming(). + * There's no hope of recovering from a version mismatch, so don't + * retry. + */ + elog(ERROR, "Cannot continue backup because stream connect has failed."); + } + + /* + * Identify server, obtain server system identifier and timeline + */ + res = pgut_execute(stream_thread_arg->conn, "IDENTIFY_SYSTEM", 0, NULL); + + if (PQresultStatus(res) != PGRES_TUPLES_OK) + { + elog(WARNING,"Could not send replication command \"%s\": %s", + "IDENTIFY_SYSTEM", PQerrorMessage(stream_thread_arg->conn)); + PQfinish(stream_thread_arg->conn); + elog(ERROR, "Cannot continue backup because stream connect has failed."); + } + + stream_conn_sysidentifier_str = PQgetvalue(res, 0, 0); + stream_conn_tli = atoll(PQgetvalue(res, 0, 1)); + + /* Additional sanity, primary for PG 9.5, + * where system id can be obtained only via "IDENTIFY SYSTEM" + */ + if (!parse_uint64(stream_conn_sysidentifier_str, &stream_conn_sysidentifier, 0)) + elog(ERROR, "%s is not system_identifier", stream_conn_sysidentifier_str); + + if (stream_conn_sysidentifier != instance_config.system_identifier) + elog(ERROR, "System identifier mismatch. Connected PostgreSQL instance has system id: " + "" UINT64_FORMAT ". Expected: " UINT64_FORMAT ".", + stream_conn_sysidentifier, instance_config.system_identifier); + + if (stream_conn_tli != current.tli) + elog(ERROR, "Timeline identifier mismatch. " + "Connected PostgreSQL instance has timeline id: %X. Expected: %X.", + stream_conn_tli, current.tli); + + PQclear(res); +} + +/* + * Retrieve checkpoint_timeout GUC value in seconds. + */ +static int +checkpoint_timeout(PGconn *backup_conn) +{ + PGresult *res; + const char *val; + const char *hintmsg; + int val_int; + + res = pgut_execute(backup_conn, "show checkpoint_timeout", 0, NULL); + val = PQgetvalue(res, 0, 0); + + if (!parse_int(val, &val_int, OPTION_UNIT_S, &hintmsg)) + { + PQclear(res); + if (hintmsg) + elog(ERROR, "Invalid value of checkout_timeout %s: %s", val, + hintmsg); + else + elog(ERROR, "Invalid value of checkout_timeout %s", val); + } + + PQclear(res); + + return val_int; +} + +/* + * CreateReplicationSlot_compat() -- wrapper for CreateReplicationSlot() used in StreamLog() + * src/bin/pg_basebackup/streamutil.c + * CreateReplicationSlot() has different signatures on different PG versions: + * PG 15 + * bool + * CreateReplicationSlot(PGconn *conn, const char *slot_name, const char *plugin, + * bool is_temporary, bool is_physical, bool reserve_wal, + * bool slot_exists_ok, bool two_phase) + * PG 11-14 + * bool + * CreateReplicationSlot(PGconn *conn, const char *slot_name, const char *plugin, + * bool is_temporary, bool is_physical, bool reserve_wal, + * bool slot_exists_ok) + * PG 9.5-10 + * CreateReplicationSlot(PGconn *conn, const char *slot_name, const char *plugin, + * bool is_physical, bool slot_exists_ok) + * NOTE: PG 9.6 and 10 support reserve_wal in + * pg_catalog.pg_create_physical_replication_slot(slot_name name [, immediately_reserve boolean]) + * and + * CREATE_REPLICATION_SLOT slot_name { PHYSICAL [ RESERVE_WAL ] | LOGICAL output_plugin } + * replication protocol command, but CreateReplicationSlot() C function doesn't + */ +static bool +CreateReplicationSlot_compat(PGconn *conn, const char *slot_name, const char *plugin, + bool is_temporary, bool is_physical, + bool slot_exists_ok) +{ +#if PG_VERSION_NUM >= 150000 + return CreateReplicationSlot(conn, slot_name, plugin, is_temporary, is_physical, + /* reserve_wal = */ true, slot_exists_ok, /* two_phase = */ false); +#elif PG_VERSION_NUM >= 110000 + return CreateReplicationSlot(conn, slot_name, plugin, is_temporary, is_physical, + /* reserve_wal = */ true, slot_exists_ok); +#elif PG_VERSION_NUM >= 100000 + /* + * PG-10 doesn't support creating temp_slot by calling CreateReplicationSlot(), but + * it will be created by setting StreamCtl.temp_slot later in StreamLog() + */ + if (!is_temporary) + return CreateReplicationSlot(conn, slot_name, plugin, /*is_temporary,*/ is_physical, /*reserve_wal,*/ slot_exists_ok); + else + return true; +#else + /* these parameters not supported in PG < 10 */ + Assert(!is_temporary); + return CreateReplicationSlot(conn, slot_name, plugin, /*is_temporary,*/ is_physical, /*reserve_wal,*/ slot_exists_ok); +#endif +} + +/* + * Start the log streaming + */ +static void * +StreamLog(void *arg) +{ + StreamThreadArg *stream_arg = (StreamThreadArg *) arg; + + /* + * Always start streaming at the beginning of a segment + */ + stream_arg->startpos -= stream_arg->startpos % instance_config.xlog_seg_size; + + xlog_files_list = parray_new(); + + /* Initialize timeout */ + stream_stop_begin = 0; + + /* Create repslot */ +#if PG_VERSION_NUM >= 100000 + if (temp_slot || perm_slot) + if (!CreateReplicationSlot_compat(stream_arg->conn, replication_slot, NULL, temp_slot, true, false)) +#else + if (perm_slot) + if (!CreateReplicationSlot_compat(stream_arg->conn, replication_slot, NULL, false, true, false)) +#endif + { + interrupted = true; + elog(ERROR, "Couldn't create physical replication slot %s", replication_slot); + } + + /* + * Start the replication + */ + if (replication_slot) + elog(LOG, "started streaming WAL at %X/%X (timeline %u) using%s slot %s", + (uint32) (stream_arg->startpos >> 32), (uint32) stream_arg->startpos, + stream_arg->starttli, +#if PG_VERSION_NUM >= 100000 + temp_slot ? " temporary" : "", +#else + "", +#endif + replication_slot); + else + elog(LOG, "started streaming WAL at %X/%X (timeline %u)", + (uint32) (stream_arg->startpos >> 32), (uint32) stream_arg->startpos, + stream_arg->starttli); + +#if PG_VERSION_NUM >= 90600 + { + StreamCtl ctl; + + MemSet(&ctl, 0, sizeof(ctl)); + + ctl.startpos = stream_arg->startpos; + ctl.timeline = stream_arg->starttli; + ctl.sysidentifier = NULL; + ctl.stream_stop = stop_streaming; + ctl.standby_message_timeout = standby_message_timeout; + ctl.partial_suffix = NULL; + ctl.synchronous = false; + ctl.mark_done = false; + +#if PG_VERSION_NUM >= 100000 +#if PG_VERSION_NUM >= 150000 + ctl.walmethod = CreateWalDirectoryMethod( + stream_arg->basedir, + PG_COMPRESSION_NONE, + 0, + false); +#else /* PG_VERSION_NUM >= 100000 && PG_VERSION_NUM < 150000 */ + ctl.walmethod = CreateWalDirectoryMethod( + stream_arg->basedir, +// (instance_config.compress_alg == NONE_COMPRESS) ? 0 : instance_config.compress_level, + 0, + false); +#endif /* PG_VERSION_NUM >= 150000 */ + ctl.replication_slot = replication_slot; + ctl.stop_socket = PGINVALID_SOCKET; + ctl.do_sync = false; /* We sync all files at the end of backup */ +// ctl.mark_done /* for future use in s3 */ +#if PG_VERSION_NUM >= 100000 && PG_VERSION_NUM < 110000 + /* StreamCtl.temp_slot used only for PG-10, in PG>10, temp_slots are created by calling CreateReplicationSlot() */ + ctl.temp_slot = temp_slot; +#endif /* PG_VERSION_NUM >= 100000 && PG_VERSION_NUM < 110000 */ +#else /* PG_VERSION_NUM < 100000 */ + ctl.basedir = (char *) stream_arg->basedir; +#endif /* PG_VERSION_NUM >= 100000 */ + + if (ReceiveXlogStream(stream_arg->conn, &ctl) == false) + { + interrupted = true; + elog(ERROR, "Problem in receivexlog"); + } + +#if PG_VERSION_NUM >= 100000 +#if PG_VERSION_NUM >= 160000 + if (!ctl.walmethod->ops->finish(ctl.walmethod)) +#else + if (!ctl.walmethod->finish()) +#endif + { + interrupted = true; + elog(ERROR, "Could not finish writing WAL files: %s", + strerror(errno)); + } +#endif /* PG_VERSION_NUM >= 100000 */ + } +#else /* PG_VERSION_NUM < 90600 */ + /* PG-9.5 */ + if (ReceiveXlogStream(stream_arg->conn, stream_arg->startpos, stream_arg->starttli, + NULL, (char *) stream_arg->basedir, stop_streaming, + standby_message_timeout, NULL, false, false) == false) + { + interrupted = true; + elog(ERROR, "Problem in receivexlog"); + } +#endif /* PG_VERSION_NUM >= 90600 */ + + /* be paranoid and sort xlog_files_list, + * so if stop_lsn segno is already in the list, + * then list must be sorted to detect duplicates. + */ + parray_qsort(xlog_files_list, pgFileCompareRelPathWithExternal); + + /* Add the last segment to the list */ + add_walsegment_to_filelist(xlog_files_list, stream_arg->starttli, + stop_stream_lsn, (char *) stream_arg->basedir, + instance_config.xlog_seg_size); + + /* append history file to walsegment filelist */ + add_history_file_to_filelist(xlog_files_list, stream_arg->starttli, (char *) stream_arg->basedir); + + /* + * TODO: remove redundant WAL segments + * walk pg_wal and remove files with segno greater that of stop_lsn`s segno +1 + */ + + elog(LOG, "finished streaming WAL at %X/%X (timeline %u)", + (uint32) (stop_stream_lsn >> 32), (uint32) stop_stream_lsn, stream_arg->starttli); + stream_arg->ret = 0; + + PQfinish(stream_arg->conn); + stream_arg->conn = NULL; + + return NULL; +} + +/* + * for ReceiveXlogStream + * + * The stream_stop callback will be called every time data + * is received, and whenever a segment is completed. If it returns + * true, the streaming will stop and the function + * return. As long as it returns false, streaming will continue + * indefinitely. + * + * Stop WAL streaming if current 'xlogpos' exceeds 'stop_backup_lsn', which is + * set by pg_stop_backup(). + * + */ +static bool +stop_streaming(XLogRecPtr xlogpos, uint32 timeline, bool segment_finished) +{ + static uint32 prevtimeline = 0; + static XLogRecPtr prevpos = InvalidXLogRecPtr; + + /* check for interrupt */ + if (interrupted || thread_interrupted) + elog(ERROR, "Interrupted during WAL streaming"); + + /* we assume that we get called once at the end of each segment */ + if (segment_finished) + { + elog(VERBOSE, _("finished segment at %X/%X (timeline %u)"), + (uint32) (xlogpos >> 32), (uint32) xlogpos, timeline); + + add_walsegment_to_filelist(xlog_files_list, timeline, xlogpos, + (char*) stream_thread_arg.basedir, + instance_config.xlog_seg_size); + } + + /* + * Note that we report the previous, not current, position here. After a + * timeline switch, xlogpos points to the beginning of the segment because + * that's where we always begin streaming. Reporting the end of previous + * timeline isn't totally accurate, because the next timeline can begin + * slightly before the end of the WAL that we received on the previous + * timeline, but it's close enough for reporting purposes. + */ + if (prevtimeline != 0 && prevtimeline != timeline) + elog(LOG, _("switched to timeline %u at %X/%X\n"), + timeline, (uint32) (prevpos >> 32), (uint32) prevpos); + + if (!XLogRecPtrIsInvalid(stop_backup_lsn)) + { + if (xlogpos >= stop_backup_lsn) + { + stop_stream_lsn = xlogpos; + return true; + } + + /* pg_stop_backup() was executed, wait for the completion of stream */ + if (stream_stop_begin == 0) + { + elog(INFO, "Wait for LSN %X/%X to be streamed", + (uint32) (stop_backup_lsn >> 32), (uint32) stop_backup_lsn); + + stream_stop_begin = time(NULL); + } + + if (time(NULL) - stream_stop_begin > stream_stop_timeout) + elog(ERROR, "Target LSN %X/%X could not be streamed in %d seconds", + (uint32) (stop_backup_lsn >> 32), (uint32) stop_backup_lsn, + stream_stop_timeout); + } + + prevtimeline = timeline; + prevpos = xlogpos; + + return false; +} + + +/* --- External API --- */ + +/* + * Maybe add a StreamOptions struct ? + * Backup conn only needed to calculate stream_stop_timeout. Think about refactoring it. + */ +parray* +get_history_streaming(ConnectionOptions *conn_opt, TimeLineID tli, parray *backup_list) +{ + PGresult *res; + PGconn *conn; + char *history; + char query[128]; + parray *result = NULL; + parray *tli_list = NULL; + timelineInfo *tlinfo = NULL; + int i,j; + + snprintf(query, sizeof(query), "TIMELINE_HISTORY %u", tli); + + /* + * Connect in replication mode to the server. + */ + conn = pgut_connect_replication(conn_opt->pghost, + conn_opt->pgport, + conn_opt->pgdatabase, + conn_opt->pguser, + false); + + if (!conn) + return NULL; + + res = PQexec(conn, query); + PQfinish(conn); + + if (PQresultStatus(res) != PGRES_TUPLES_OK) + { + elog(WARNING, "Could not send replication command \"%s\": %s", + query, PQresultErrorMessage(res)); + PQclear(res); + return NULL; + } + + /* + * The response to TIMELINE_HISTORY is a single row result set + * with two fields: filename and content + */ + + if (PQnfields(res) != 2 || PQntuples(res) != 1) + { + elog(WARNING, "Unexpected response to TIMELINE_HISTORY command: " + "got %d rows and %d fields, expected %d rows and %d fields", + PQntuples(res), PQnfields(res), 1, 2); + PQclear(res); + return NULL; + } + + history = pgut_strdup(PQgetvalue(res, 0, 1)); + result = parse_tli_history_buffer(history, tli); + + /* some cleanup */ + pg_free(history); + PQclear(res); + + if (result) + tlinfo = timelineInfoNew(tli); + else + return NULL; + + /* transform TimeLineHistoryEntry into timelineInfo */ + for (i = parray_num(result) -1; i >= 0; i--) + { + TimeLineHistoryEntry *tln = (TimeLineHistoryEntry *) parray_get(result, i); + + tlinfo->parent_tli = tln->tli; + tlinfo->switchpoint = tln->end; + + if (!tli_list) + tli_list = parray_new(); + + parray_append(tli_list, tlinfo); + + /* Next tli */ + tlinfo = timelineInfoNew(tln->tli); + + /* oldest tli */ + if (i == 0) + { + tlinfo->tli = tln->tli; + tlinfo->parent_tli = 0; + tlinfo->switchpoint = 0; + parray_append(tli_list, tlinfo); + } + } + + /* link parent to child */ + for (i = 0; i < parray_num(tli_list); i++) + { + tlinfo = (timelineInfo *) parray_get(tli_list, i); + + for (j = 0; j < parray_num(tli_list); j++) + { + timelineInfo *tlinfo_parent = (timelineInfo *) parray_get(tli_list, j); + + if (tlinfo->parent_tli == tlinfo_parent->tli) + { + tlinfo->parent_link = tlinfo_parent; + break; + } + } + } + + /* add backups to each timeline info */ + for (i = 0; i < parray_num(tli_list); i++) + { + tlinfo = parray_get(tli_list, i); + for (j = 0; j < parray_num(backup_list); j++) + { + pgBackup *backup = parray_get(backup_list, j); + if (tlinfo->tli == backup->tli) + { + if (tlinfo->backups == NULL) + tlinfo->backups = parray_new(); + parray_append(tlinfo->backups, backup); + } + } + } + + /* cleanup */ + parray_walk(result, pg_free); + pg_free(result); + + return tli_list; +} + +parray* +parse_tli_history_buffer(char *history, TimeLineID tli) +{ + char *curLine = history; + TimeLineHistoryEntry *entry; + TimeLineHistoryEntry *last_timeline = NULL; + parray *result = NULL; + + /* Parse timeline history buffer string by string */ + while (curLine) + { + char tempStr[1024]; + char *nextLine = strchr(curLine, '\n'); + int curLineLen = nextLine ? (nextLine-curLine) : strlen(curLine); + + memcpy(tempStr, curLine, curLineLen); + tempStr[curLineLen] = '\0'; // NUL-terminate! + curLine = nextLine ? (nextLine+1) : NULL; + + if (curLineLen > 0) + { + char *ptr; + TimeLineID tli; + uint32 switchpoint_hi; + uint32 switchpoint_lo; + int nfields; + + for (ptr = tempStr; *ptr; ptr++) + { + if (!isspace((unsigned char) *ptr)) + break; + } + if (*ptr == '\0' || *ptr == '#') + continue; + + nfields = sscanf(tempStr, "%u\t%X/%X", &tli, &switchpoint_hi, &switchpoint_lo); + + if (nfields < 1) + { + /* expect a numeric timeline ID as first field of line */ + elog(ERROR, "Syntax error in timeline history: \"%s\". Expected a numeric timeline ID.", tempStr); + } + if (nfields != 3) + elog(ERROR, "Syntax error in timeline history: \"%s\". Expected a transaction log switchpoint location.", tempStr); + + if (last_timeline && tli <= last_timeline->tli) + elog(ERROR, "Timeline IDs must be in increasing sequence: \"%s\"", tempStr); + + entry = pgut_new(TimeLineHistoryEntry); + entry->tli = tli; + entry->end = ((uint64) switchpoint_hi << 32) | switchpoint_lo; + + last_timeline = entry; + /* Build list with newest item first */ + if (!result) + result = parray_new(); + parray_append(result, entry); + elog(VERBOSE, "parse_tli_history_buffer() found entry: tli = %X, end = %X/%X", + tli, switchpoint_hi, switchpoint_lo); + + /* we ignore the remainder of each line */ + } + } + + return result; +} + +/* + * Maybe add a StreamOptions struct ? + * Backup conn only needed to calculate stream_stop_timeout. Think about refactoring it. + */ +void +start_WAL_streaming(PGconn *backup_conn, char *stream_dst_path, ConnectionOptions *conn_opt, + XLogRecPtr startpos, TimeLineID starttli, bool is_backup) +{ + /* calculate crc only when running backup, catchup has no need for it */ + do_crc = is_backup; + /* How long we should wait for streaming end after pg_stop_backup */ + stream_stop_timeout = checkpoint_timeout(backup_conn); + //TODO Add a comment about this calculation + stream_stop_timeout = stream_stop_timeout + stream_stop_timeout * 0.1; + + strlcpy(stream_thread_arg.basedir, stream_dst_path, sizeof(stream_thread_arg.basedir)); + + /* + * Connect in replication mode to the server. + */ + stream_thread_arg.conn = pgut_connect_replication(conn_opt->pghost, + conn_opt->pgport, + conn_opt->pgdatabase, + conn_opt->pguser, + true); + /* sanity check*/ + IdentifySystem(&stream_thread_arg); + + /* Set error exit code as default */ + stream_thread_arg.ret = 1; + /* we must use startpos as start_lsn from start_backup */ + stream_thread_arg.startpos = startpos; + stream_thread_arg.starttli = starttli; + + thread_interrupted = false; + pthread_create(&stream_thread, NULL, StreamLog, &stream_thread_arg); +} + +/* + * Wait for the completion of stream + * append list of streamed xlog files + * into backup_files_list (if it is not NULL) + */ +int +wait_WAL_streaming_end(parray *backup_files_list) +{ + pthread_join(stream_thread, NULL); + + if(backup_files_list != NULL) + parray_concat(backup_files_list, xlog_files_list); + parray_free(xlog_files_list); + return stream_thread_arg.ret; +} + +/* Append streamed WAL segment to filelist */ +void +add_walsegment_to_filelist(parray *filelist, uint32 timeline, XLogRecPtr xlogpos, char *basedir, uint32 xlog_seg_size) +{ + XLogSegNo xlog_segno; + char wal_segment_name[MAXFNAMELEN]; + char wal_segment_relpath[MAXPGPATH]; + char wal_segment_fullpath[MAXPGPATH]; + pgFile *file = NULL; + pgFile **existing_file = NULL; + + GetXLogSegNo(xlogpos, xlog_segno, xlog_seg_size); + + /* + * When xlogpos points to the zero offset (0/3000000), + * it means that previous segment was just successfully streamed. + * When xlogpos points to the positive offset, + * then current segment is successfully streamed. + */ + if (WalSegmentOffset(xlogpos, xlog_seg_size) == 0) + xlog_segno--; + + GetXLogFileName(wal_segment_name, timeline, xlog_segno, xlog_seg_size); + + join_path_components(wal_segment_fullpath, basedir, wal_segment_name); + join_path_components(wal_segment_relpath, PG_XLOG_DIR, wal_segment_name); + + file = pgFileNew(wal_segment_fullpath, wal_segment_relpath, false, 0, FIO_BACKUP_HOST); + + /* + * Check if file is already in the list + * stop_lsn segment can be added to this list twice, so + * try not to add duplicates + */ + + existing_file = (pgFile **) parray_bsearch(filelist, file, pgFileCompareRelPathWithExternal); + + if (existing_file) + { + if (do_crc) + (*existing_file)->crc = pgFileGetCRC(wal_segment_fullpath, true, false); + (*existing_file)->write_size = xlog_seg_size; + (*existing_file)->uncompressed_size = xlog_seg_size; + + return; + } + + if (do_crc) + file->crc = pgFileGetCRC(wal_segment_fullpath, true, false); + + /* Should we recheck it using stat? */ + file->write_size = xlog_seg_size; + file->uncompressed_size = xlog_seg_size; + + /* append file to filelist */ + parray_append(filelist, file); +} + +/* Append history file to filelist */ +void +add_history_file_to_filelist(parray *filelist, uint32 timeline, char *basedir) +{ + char filename[MAXFNAMELEN]; + char fullpath[MAXPGPATH]; + char relpath[MAXPGPATH]; + pgFile *file = NULL; + + /* Timeline 1 does not have a history file */ + if (timeline == 1) + return; + + snprintf(filename, lengthof(filename), "%08X.history", timeline); + join_path_components(fullpath, basedir, filename); + join_path_components(relpath, PG_XLOG_DIR, filename); + + file = pgFileNew(fullpath, relpath, false, 0, FIO_BACKUP_HOST); + + /* calculate crc */ + if (do_crc) + file->crc = pgFileGetCRC(fullpath, true, false); + file->write_size = file->size; + file->uncompressed_size = file->size; + + /* append file to filelist */ + parray_append(filelist, file); +} diff --git a/src/util.c b/src/util.c index 93790283d..3c0a33453 100644 --- a/src/util.c +++ b/src/util.c @@ -3,54 +3,51 @@ * util.c: log messages to log file or stderr, and misc code. * * Portions Copyright (c) 2009-2011, NIPPON TELEGRAPH AND TELEPHONE CORPORATION - * Portions Copyright (c) 2015-2019, Postgres Professional + * Portions Copyright (c) 2015-2021, Postgres Professional * *------------------------------------------------------------------------- */ #include "pg_probackup.h" -#include "catalog/pg_control.h" - #include #include #include -const char * -base36enc(long unsigned int value) +static const char *statusName[] = { - const char base36[36] = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ"; - /* log(2**64) / log(36) = 12.38 => max 13 char + '\0' */ - static char buffer[14]; - unsigned int offset = sizeof(buffer); + "UNKNOWN", + "OK", + "ERROR", + "RUNNING", + "MERGING", + "MERGED", + "DELETING", + "DELETED", + "DONE", + "ORPHAN", + "CORRUPT" +}; - buffer[--offset] = '\0'; - do { - buffer[--offset] = base36[value % 36]; - } while (value /= 36); - - return &buffer[offset]; -} - -/* - * Same as base36enc(), but the result must be released by the user. - */ -char * -base36enc_dup(long unsigned int value) +const char * +base36enc_to(long unsigned int value, char buf[ARG_SIZE_HINT base36bufsize]) { const char base36[36] = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ"; - /* log(2**64) / log(36) = 12.38 => max 13 char + '\0' */ - char buffer[14]; - unsigned int offset = sizeof(buffer); + char buffer[base36bufsize]; + char *p; - buffer[--offset] = '\0'; + p = &buffer[sizeof(buffer)-1]; + *p = '\0'; do { - buffer[--offset] = base36[value % 36]; + *(--p) = base36[value % 36]; } while (value /= 36); - return strdup(&buffer[offset]); + /* I know, it doesn't look safe */ + strncpy(buf, p, base36bufsize); + + return buf; } long unsigned int @@ -77,7 +74,7 @@ checkControlFile(ControlFileData *ControlFile) if ((ControlFile->pg_control_version % 65536 == 0 || ControlFile->pg_control_version % 65536 > 10000) && ControlFile->pg_control_version / 65536 != 0) - elog(ERROR, "possible byte ordering mismatch\n" + elog(ERROR, "Possible byte ordering mismatch\n" "The byte ordering used to store the pg_control file might not match the one\n" "used by this program. In that case the results below would be incorrect, and\n" "the PostgreSQL installation would be incompatible with this data directory."); @@ -96,7 +93,7 @@ digestControlFile(ControlFileData *ControlFile, char *src, size_t size) #endif if (size != ControlFileSize) - elog(ERROR, "unexpected control file size %d, expected %d", + elog(ERROR, "Unexpected control file size %d, expected %d", (int) size, ControlFileSize); memcpy(ControlFile, src, sizeof(ControlFileData)); @@ -109,7 +106,7 @@ digestControlFile(ControlFileData *ControlFile, char *src, size_t size) * Write ControlFile to pg_control */ static void -writeControlFile(ControlFileData *ControlFile, char *path, fio_location location) +writeControlFile(ControlFileData *ControlFile, const char *path, fio_location location) { int fd; char *buffer = NULL; @@ -121,7 +118,7 @@ writeControlFile(ControlFileData *ControlFile, char *path, fio_location location #endif /* copy controlFileSize */ - buffer = pg_malloc(ControlFileSize); + buffer = pg_malloc0(ControlFileSize); memcpy(buffer, ControlFile, sizeof(ControlFileData)); /* Write pg_control */ @@ -135,7 +132,7 @@ writeControlFile(ControlFileData *ControlFile, char *path, fio_location location elog(ERROR, "Failed to overwrite file: %s", path); if (fio_flush(fd) != 0) - elog(ERROR, "Failed to fsync file: %s", path); + elog(ERROR, "Failed to sync file: %s", path); fio_close(fd); pg_free(buffer); @@ -146,15 +143,44 @@ writeControlFile(ControlFileData *ControlFile, char *path, fio_location location * used by a node. */ TimeLineID -get_current_timeline(bool safe) +get_current_timeline(PGconn *conn) +{ + + PGresult *res; + TimeLineID tli = 0; + char *val; + + res = pgut_execute_extended(conn, + "SELECT timeline_id FROM pg_catalog.pg_control_checkpoint()", 0, NULL, true, true); + + if (PQresultStatus(res) == PGRES_TUPLES_OK) + val = PQgetvalue(res, 0, 0); + else + return get_current_timeline_from_control(instance_config.pgdata, FIO_DB_HOST, false); + + if (!parse_uint32(val, &tli, 0)) + { + PQclear(res); + elog(WARNING, "Invalid value of timeline_id %s", val); + + /* TODO 3.0 remove it and just error out */ + return get_current_timeline_from_control(instance_config.pgdata, FIO_DB_HOST, false); + } + + return tli; +} + +/* Get timeline from pg_control file */ +TimeLineID +get_current_timeline_from_control(const char *pgdata_path, fio_location location, bool safe) { ControlFileData ControlFile; char *buffer; size_t size; /* First fetch file... */ - buffer = slurpFile(instance_config.pgdata, XLOG_CONTROL_FILE, &size, - safe, FIO_DB_HOST); + buffer = slurpFile(pgdata_path, XLOG_CONTROL_FILE, &size, + safe, location); if (safe && buffer == NULL) return 0; @@ -164,6 +190,26 @@ get_current_timeline(bool safe) return ControlFile.checkPointCopy.ThisTimeLineID; } +void +get_control_file_or_back_file(const char *pgdata_path, fio_location location, ControlFileData *control) +{ + char *buffer; + size_t size; + + /* First fetch file... */ + buffer = slurpFile(pgdata_path, XLOG_CONTROL_FILE, &size, true, location); + + if (!buffer || size == 0){ + /* Error read XLOG_CONTROL_FILE or file is truncated, trying read backup */ + buffer = slurpFile(pgdata_path, XLOG_CONTROL_BAK_FILE, &size, true, location); + if (!buffer) + elog(ERROR, "Could not read %s and %s files\n", XLOG_CONTROL_FILE, XLOG_CONTROL_BAK_FILE); /* Maybe it should be PANIC? */ + } + digestControlFile(control, buffer, size); + pg_free(buffer); +} + + /* * Get last check point record ptr from pg_tonrol. */ @@ -205,15 +251,15 @@ get_checkpoint_location(PGconn *conn) } uint64 -get_system_identifier(const char *pgdata_path) +get_system_identifier(const char *pgdata_path, fio_location location, bool safe) { ControlFileData ControlFile; char *buffer; size_t size; /* First fetch file... */ - buffer = slurpFile(pgdata_path, XLOG_CONTROL_FILE, &size, false, FIO_DB_HOST); - if (buffer == NULL) + buffer = slurpFile(pgdata_path, XLOG_CONTROL_FILE, &size, safe, location); + if (safe && buffer == NULL) return 0; digestControlFile(&ControlFile, buffer, size); pg_free(buffer); @@ -255,7 +301,7 @@ get_remote_system_identifier(PGconn *conn) } uint32 -get_xlog_seg_size(char *pgdata_path) +get_xlog_seg_size(const char *pgdata_path) { #if PG_VERSION_NUM >= 110000 ControlFileData ControlFile; @@ -307,6 +353,39 @@ get_pgcontrol_checksum(const char *pgdata_path) return ControlFile.crc; } +void +get_redo(const char *pgdata_path, fio_location pgdata_location, RedoParams *redo) +{ + ControlFileData ControlFile; + char *buffer; + size_t size; + + /* First fetch file... */ + buffer = slurpFile(pgdata_path, XLOG_CONTROL_FILE, &size, false, pgdata_location); + + digestControlFile(&ControlFile, buffer, size); + pg_free(buffer); + + redo->lsn = ControlFile.checkPointCopy.redo; + redo->tli = ControlFile.checkPointCopy.ThisTimeLineID; + + if (ControlFile.minRecoveryPoint > 0 && + ControlFile.minRecoveryPoint < redo->lsn) + { + redo->lsn = ControlFile.minRecoveryPoint; + redo->tli = ControlFile.minRecoveryPointTLI; + } + + if (ControlFile.backupStartPoint > 0 && + ControlFile.backupStartPoint < redo->lsn) + { + redo->lsn = ControlFile.backupStartPoint; + redo->tli = ControlFile.checkPointCopy.ThisTimeLineID; + } + + redo->checksum_version = ControlFile.data_checksum_version; +} + /* * Rewrite minRecoveryPoint of pg_control in backup directory. minRecoveryPoint * 'as-is' is not to be trusted. @@ -341,7 +420,7 @@ set_min_recovery_point(pgFile *file, const char *backup_path, FIN_CRC32C(ControlFile.crc); /* overwrite pg_control */ - snprintf(fullpath, sizeof(fullpath), "%s/%s", backup_path, XLOG_CONTROL_FILE); + join_path_components(fullpath, backup_path, XLOG_CONTROL_FILE); writeControlFile(&ControlFile, fullpath, FIO_LOCAL_HOST); /* Update pg_control checksum in backup_list */ @@ -354,24 +433,23 @@ set_min_recovery_point(pgFile *file, const char *backup_path, * Copy pg_control file to backup. We do not apply compression to this file. */ void -copy_pgcontrol_file(const char *from_root, fio_location from_location, - const char *to_root, fio_location to_location, pgFile *file) +copy_pgcontrol_file(const char *from_fullpath, fio_location from_location, + const char *to_fullpath, fio_location to_location, pgFile *file) { ControlFileData ControlFile; char *buffer; size_t size; - char to_path[MAXPGPATH]; - buffer = slurpFile(from_root, XLOG_CONTROL_FILE, &size, false, from_location); + buffer = slurpFile(from_fullpath, "", &size, false, from_location); digestControlFile(&ControlFile, buffer, size); file->crc = ControlFile.crc; file->read_size = size; file->write_size = size; + file->uncompressed_size = size; - join_path_components(to_path, to_root, file->rel_path); - writeControlFile(&ControlFile, to_path, to_location); + writeControlFile(&ControlFile, to_fullpath, to_location); pg_free(buffer); } @@ -434,21 +512,92 @@ parse_program_version(const char *program_version) const char * status2str(BackupStatus status) { - static const char *statusName[] = - { - "UNKNOWN", - "OK", - "ERROR", - "RUNNING", - "MERGING", - "DELETING", - "DELETED", - "DONE", - "ORPHAN", - "CORRUPT" - }; if (status < BACKUP_STATUS_INVALID || BACKUP_STATUS_CORRUPT < status) return "UNKNOWN"; return statusName[status]; } + +const char * +status2str_color(BackupStatus status) +{ + char *status_str = pgut_malloc(20); + + /* UNKNOWN */ + if (status == BACKUP_STATUS_INVALID) + snprintf(status_str, 20, "%s%s%s", TC_YELLOW_BOLD, "UNKNOWN", TC_RESET); + /* CORRUPT, ERROR and ORPHAN */ + else if (status == BACKUP_STATUS_CORRUPT || status == BACKUP_STATUS_ERROR || + status == BACKUP_STATUS_ORPHAN) + snprintf(status_str, 20, "%s%s%s", TC_RED_BOLD, statusName[status], TC_RESET); + /* MERGING, MERGED, DELETING and DELETED */ + else if (status == BACKUP_STATUS_MERGING || status == BACKUP_STATUS_MERGED || + status == BACKUP_STATUS_DELETING || status == BACKUP_STATUS_DELETED) + snprintf(status_str, 20, "%s%s%s", TC_YELLOW_BOLD, statusName[status], TC_RESET); + /* OK and DONE */ + else + snprintf(status_str, 20, "%s%s%s", TC_GREEN_BOLD, statusName[status], TC_RESET); + + return status_str; +} + +BackupStatus +str2status(const char *status) +{ + BackupStatus i; + + for (i = BACKUP_STATUS_INVALID; i <= BACKUP_STATUS_CORRUPT; i++) + { + if (pg_strcasecmp(status, statusName[i]) == 0) return i; + } + + return BACKUP_STATUS_INVALID; +} + +bool +datapagemap_is_set(datapagemap_t *map, BlockNumber blkno) +{ + int offset; + int bitno; + + offset = blkno / 8; + bitno = blkno % 8; + + return (map->bitmapsize <= offset) ? false : (map->bitmap[offset] & (1 << bitno)) != 0; +} + +/* + * A debugging aid. Prints out the contents of the page map. + */ +void +datapagemap_print_debug(datapagemap_t *map) +{ + datapagemap_iterator_t *iter; + BlockNumber blocknum; + + iter = datapagemap_iterate(map); + while (datapagemap_next(iter, &blocknum)) + elog(VERBOSE, " block %u", blocknum); + + pg_free(iter); +} + +const char* +backup_id_of(pgBackup *backup) +{ + /* Change this Assert when backup_id will not be bound to start_time */ + Assert(backup->backup_id == backup->start_time || backup->start_time == 0); + + if (backup->backup_id_encoded[0] == '\x00') + { + base36enc_to(backup->backup_id, backup->backup_id_encoded); + } + return backup->backup_id_encoded; +} + +void +reset_backup_id(pgBackup *backup) +{ + backup->backup_id = INVALID_BACKUP_ID; + memset(backup->backup_id_encoded, 0, sizeof(backup->backup_id_encoded)); +} diff --git a/src/utils/configuration.c b/src/utils/configuration.c index cf5436f29..f049aa1be 100644 --- a/src/utils/configuration.c +++ b/src/utils/configuration.c @@ -18,7 +18,11 @@ #include "getopt_long.h" +#ifndef WIN32 +#include +#endif #include +#include #define MAXPG_LSNCOMPONENT 8 @@ -87,6 +91,63 @@ static const unit_conversion time_unit_conversion_table[] = {""} /* end of table marker */ }; +/* Order is important, keep it in sync with utils/configuration.h:enum ProbackupSubcmd declaration */ +static char const * const subcmd_names[] = +{ + "NO_CMD", + "init", + "add-instance", + "del-instance", + "archive-push", + "archive-get", + "backup", + "restore", + "validate", + "delete", + "merge", + "show", + "set-config", + "set-backup", + "show-config", + "checkdb", + "ssh", + "agent", + "help", + "version", + "catchup", +}; + +ProbackupSubcmd +parse_subcmd(char const * const subcmd_str) +{ + struct { + ProbackupSubcmd id; + char *name; + } + static const subcmd_additional_names[] = { + { HELP_CMD, "--help" }, + { HELP_CMD, "-?" }, + { VERSION_CMD, "--version" }, + { VERSION_CMD, "-V" }, + }; + + int i; + for(i = (int)NO_CMD + 1; i < sizeof(subcmd_names) / sizeof(subcmd_names[0]); ++i) + if(strcmp(subcmd_str, subcmd_names[i]) == 0) + return (ProbackupSubcmd)i; + for(i = 0; i < sizeof(subcmd_additional_names) / sizeof(subcmd_additional_names[0]); ++i) + if(strcmp(subcmd_str, subcmd_additional_names[i].name) == 0) + return subcmd_additional_names[i].id; + return NO_CMD; +} + +char const * +get_subcmd_name(ProbackupSubcmd const subcmd) +{ + Assert((int)subcmd < sizeof(subcmd_names) / sizeof(subcmd_names[0])); + return subcmd_names[(int)subcmd]; +} + /* * Reading functions. */ @@ -250,6 +311,14 @@ assign_option(ConfigOption *opt, const char *optarg, OptionSource src) case 's': if (orig_source != SOURCE_DEFAULT) free(*(char **) opt->var); + + /* 'none' and 'off' are always disable the string parameter */ + //if (optarg && (pg_strcasecmp(optarg, "none") == 0)) + //{ + // *(char **) opt->var = "none"; + // return; + //} + *(char **) opt->var = pgut_strdup(optarg); if (strcmp(optarg,"") != 0) return; @@ -426,6 +495,8 @@ get_username(void) /* * Process options passed from command line. + * TODO: currectly argument parsing treat missing argument for options + * as invalid option */ int config_get_opt(int argc, char **argv, ConfigOption cmd_options[], @@ -450,17 +521,22 @@ config_get_opt(int argc, char **argv, ConfigOption cmd_options[], optstring = longopts_to_optstring(longopts, cmd_len + len); + opterr = 0; /* Assign named options */ while ((c = getopt_long(argc, argv, optstring, longopts, &optindex)) != -1) { ConfigOption *opt; + if (c == '?') + { + elog(ERROR, "Option '%s' requires an argument. Try \"%s --help\" for more information.", + argv[optind-1], PROGRAM_NAME); + } opt = option_find(c, cmd_options); if (opt == NULL) opt = option_find(c, options); if (opt - && !remote_agent && opt->allowed < SOURCE_CMD && opt->allowed != SOURCE_CMD_STRICT) elog(ERROR, "Option %s cannot be specified in command line", opt->lname); @@ -468,6 +544,9 @@ config_get_opt(int argc, char **argv, ConfigOption cmd_options[], assign_option(opt, optarg, SOURCE_CMD); } + pgut_free(optstring); + pgut_free(longopts); + return optind; } @@ -480,9 +559,9 @@ config_read_opt(const char *path, ConfigOption options[], int elevel, bool strict, bool missing_ok) { FILE *fp; - char buf[1024]; + char buf[4096]; char key[1024]; - char value[1024]; + char value[2048]; int parsed_options = 0; if (!options) @@ -523,6 +602,9 @@ config_read_opt(const char *path, ConfigOption options[], int elevel, } } + if (ferror(fp)) + elog(ERROR, "Failed to read from file: \"%s\"", path); + fio_close_stream(fp); return parsed_options; @@ -596,6 +678,8 @@ config_set_opt(ConfigOption options[], void *var, OptionSource source) /* * Return value of the function in the string representation. Result is * allocated string. + * We can set GET_VAL_IN_BASE_UNITS flag in opt->flags + * before call option_get_value() to get option value in default units */ char * option_get_value(ConfigOption *opt) @@ -610,20 +694,33 @@ option_get_value(ConfigOption *opt) */ if (opt->flags & OPTION_UNIT) { - if (opt->type == 'i') - convert_from_base_unit(*((int32 *) opt->var), - opt->flags & OPTION_UNIT, &value, &unit); - else if (opt->type == 'i') - convert_from_base_unit(*((int64 *) opt->var), - opt->flags & OPTION_UNIT, &value, &unit); - else if (opt->type == 'u') - convert_from_base_unit_u(*((uint32 *) opt->var), - opt->flags & OPTION_UNIT, &value_u, &unit); - else if (opt->type == 'U') - convert_from_base_unit_u(*((uint64 *) opt->var), - opt->flags & OPTION_UNIT, &value_u, &unit); + if (opt->flags & GET_VAL_IN_BASE_UNITS){ + if (opt->type == 'i') + value = *((int32 *) opt->var); + else if (opt->type == 'I') + value = *((int64 *) opt->var); + else if (opt->type == 'u') + value_u = *((uint32 *) opt->var); + else if (opt->type == 'U') + value_u = *((uint64 *) opt->var); + unit = ""; + } + else + { + if (opt->type == 'i') + convert_from_base_unit(*((int32 *) opt->var), + opt->flags & OPTION_UNIT, &value, &unit); + else if (opt->type == 'I') + convert_from_base_unit(*((int64 *) opt->var), + opt->flags & OPTION_UNIT, &value, &unit); + else if (opt->type == 'u') + convert_from_base_unit_u(*((uint32 *) opt->var), + opt->flags & OPTION_UNIT, &value_u, &unit); + else if (opt->type == 'U') + convert_from_base_unit_u(*((uint64 *) opt->var), + opt->flags & OPTION_UNIT, &value_u, &unit); + } } - /* Get string representation itself */ switch (opt->type) { @@ -653,6 +750,10 @@ option_get_value(ConfigOption *opt) case 's': if (*((char **) opt->var) == NULL) return NULL; + /* 'none' and 'off' are always disable the string parameter */ + //if ((pg_strcasecmp(*((char **) opt->var), "none") == 0) || + // (pg_strcasecmp(*((char **) opt->var), "off") == 0)) + // return NULL; return pstrdup(*((char **) opt->var)); case 't': { @@ -662,7 +763,7 @@ option_get_value(ConfigOption *opt) if (t > 0) { timestamp = palloc(100); - time2iso(timestamp, 100, t); + time2iso(timestamp, 100, t, false); } else timestamp = palloc0(1 /* just null termination */); @@ -1079,6 +1180,8 @@ parse_uint64(const char *value, uint64 *result, int flags) * * If utc_default is true, then if timezone offset isn't specified tz will be * +00:00. + * + * TODO: '0' converted into '2000-01-01 00:00:00'. Example: set-backup --expire-time=0 */ bool parse_time(const char *value, time_t *result, bool utc_default) @@ -1092,8 +1195,11 @@ parse_time(const char *value, time_t *result, bool utc_default) struct tm tm; char junk[2]; + char *local_tz = getenv("TZ"); + /* tmp = replace( value, !isalnum, ' ' ) */ - tmp = pgut_malloc(strlen(value) + + 1); + tmp = pgut_malloc(strlen(value) + 1); + if(!tmp) return false; len = 0; fields_num = 1; @@ -1121,7 +1227,10 @@ parse_time(const char *value, time_t *result, bool utc_default) errno = 0; hr = strtol(value + 1, &cp, 10); if ((value + 1) == cp || errno == ERANGE) + { + pfree(tmp); return false; + } /* explicit delimiter? */ if (*cp == ':') @@ -1129,13 +1238,19 @@ parse_time(const char *value, time_t *result, bool utc_default) errno = 0; min = strtol(cp + 1, &cp, 10); if (errno == ERANGE) + { + pfree(tmp); return false; + } if (*cp == ':') { errno = 0; sec = strtol(cp + 1, &cp, 10); if (errno == ERANGE) + { + pfree(tmp); return false; + } } } /* otherwise, might have run things together... */ @@ -1150,11 +1265,20 @@ parse_time(const char *value, time_t *result, bool utc_default) /* Range-check the values; see notes in datatype/timestamp.h */ if (hr < 0 || hr > MAX_TZDISP_HOUR) + { + pfree(tmp); return false; + } if (min < 0 || min >= MINS_PER_HOUR) + { + pfree(tmp); return false; + } if (sec < 0 || sec >= SECS_PER_MINUTE) + { + pfree(tmp); return false; + } tz = (hr * MINS_PER_HOUR + min) * SECS_PER_MINUTE + sec; if (*value == '-') @@ -1167,7 +1291,12 @@ parse_time(const char *value, time_t *result, bool utc_default) } /* wrong format */ else if (!IsSpace(*value)) + { + pfree(tmp); return false; + } + else + value++; } tmp[len] = '\0'; @@ -1182,9 +1311,9 @@ parse_time(const char *value, time_t *result, bool utc_default) i = sscanf(tmp, "%04d %02d %02d %02d %02d %02d%1s", &tm.tm_year, &tm.tm_mon, &tm.tm_mday, &tm.tm_hour, &tm.tm_min, &tm.tm_sec, junk); - free(tmp); + pfree(tmp); - if (i < 1 || 6 < i) + if (i < 3 || i > 6) return false; /* adjust year */ @@ -1200,24 +1329,33 @@ parse_time(const char *value, time_t *result, bool utc_default) /* determine whether Daylight Saving Time is in effect */ tm.tm_isdst = -1; + /* + * If tz is not set, + * treat it as UTC if requested, otherwise as local timezone + */ + if (tz_set || utc_default) + { + /* set timezone to UTC */ + pgut_setenv("TZ", "UTC"); + tzset(); + } + + /* convert time to utc unix time */ *result = mktime(&tm); + /* return old timezone back if any */ + if (local_tz) + pgut_setenv("TZ", local_tz); + else + pgut_unsetenv("TZ"); + + tzset(); + /* adjust time zone */ if (tz_set || utc_default) { - time_t ltime = time(NULL); - struct tm *ptm = gmtime(<ime); - time_t gmt = mktime(ptm); - time_t offset; - /* UTC time */ *result -= tz; - - /* Get local time */ - ptm = localtime(<ime); - offset = ltime - gmt + (ptm->tm_isdst ? 3600 : 0); - - *result += offset; } return true; @@ -1322,16 +1460,16 @@ parse_lsn(const char *value, XLogRecPtr *result) len1 = strspn(value, "0123456789abcdefABCDEF"); if (len1 < 1 || len1 > MAXPG_LSNCOMPONENT || value[len1] != '/') - elog(ERROR, "invalid LSN \"%s\"", value); + elog(ERROR, "Invalid LSN \"%s\"", value); len2 = strspn(value + len1 + 1, "0123456789abcdefABCDEF"); if (len2 < 1 || len2 > MAXPG_LSNCOMPONENT || value[len1 + 1 + len2] != '\0') - elog(ERROR, "invalid LSN \"%s\"", value); + elog(ERROR, "Invalid LSN \"%s\"", value); if (sscanf(value, "%X/%X", &xlogid, &xrecoff) == 2) *result = (XLogRecPtr) ((uint64) xlogid << 32) | xrecoff; else { - elog(ERROR, "invalid LSN \"%s\"", value); + elog(ERROR, "Invalid LSN \"%s\"", value); return false; } @@ -1441,14 +1579,26 @@ convert_from_base_unit_u(uint64 base_value, int base_unit, * Convert time_t value to ISO-8601 format string. Always set timezone offset. */ void -time2iso(char *buf, size_t len, time_t time) +time2iso(char *buf, size_t len, time_t time, bool utc) { - struct tm *ptm = gmtime(&time); - time_t gmt = mktime(ptm); + struct tm *ptm = NULL; + time_t gmt; time_t offset; char *ptr = buf; + /* set timezone to UTC if requested */ + if (utc) + { + ptm = gmtime(&time); + strftime(ptr, len, "%Y-%m-%d %H:%M:%S+00", ptm); + return; + } + + ptm = gmtime(&time); + gmt = mktime(ptm); ptm = localtime(&time); + + /* adjust timezone offset */ offset = time - gmt + (ptm->tm_isdst ? 3600 : 0); strftime(ptr, len, "%Y-%m-%d %H:%M:%S", ptm); diff --git a/src/utils/configuration.h b/src/utils/configuration.h index 46b5d6c1b..59da29bd5 100644 --- a/src/utils/configuration.h +++ b/src/utils/configuration.h @@ -16,6 +16,32 @@ #define INFINITE_STR "INFINITE" +/* Order is important, keep it in sync with configuration.c:subcmd_names[] and help.c:help_command() */ +typedef enum ProbackupSubcmd +{ + NO_CMD = 0, + INIT_CMD, + ADD_INSTANCE_CMD, + DELETE_INSTANCE_CMD, + ARCHIVE_PUSH_CMD, + ARCHIVE_GET_CMD, + BACKUP_CMD, + RESTORE_CMD, + VALIDATE_CMD, + DELETE_CMD, + MERGE_CMD, + SHOW_CMD, + SET_CONFIG_CMD, + SET_BACKUP_CMD, + SHOW_CONFIG_CMD, + CHECKDB_CMD, + SSH_CMD, + AGENT_CMD, + HELP_CMD, + VERSION_CMD, + CATCHUP_CMD, +} ProbackupSubcmd; + typedef enum OptionSource { SOURCE_DEFAULT, @@ -35,14 +61,14 @@ typedef char *(*option_get_fn) (ConfigOption *opt); /* * type: - * b: bool (true) - * B: bool (false) + * b: bool (true) + * B: bool (false) * f: option_fn - * i: 32bit signed integer - * u: 32bit unsigned integer - * I: 64bit signed integer - * U: 64bit unsigned integer - * s: string + * i: 32bit signed integer + * u: 32bit unsigned integer + * I: 64bit signed integer + * U: 64bit unsigned integer + * s: string * t: time_t */ struct ConfigOption @@ -74,7 +100,10 @@ struct ConfigOption #define OPTION_UNIT_TIME 0xF0000 /* mask for time-related units */ #define OPTION_UNIT (OPTION_UNIT_MEMORY | OPTION_UNIT_TIME) +#define GET_VAL_IN_BASE_UNITS 0x80000000 /* bitflag to get memory and time values in default units*/ +extern ProbackupSubcmd parse_subcmd(char const * const subcmd_str); +extern char const *get_subcmd_name(ProbackupSubcmd const subcmd); extern int config_get_opt(int argc, char **argv, ConfigOption cmd_options[], ConfigOption options[]); extern int config_read_opt(const char *path, ConfigOption options[], int elevel, @@ -96,7 +125,7 @@ extern bool parse_int(const char *value, int *result, int flags, const char **hintmsg); extern bool parse_lsn(const char *value, XLogRecPtr *result); -extern void time2iso(char *buf, size_t len, time_t time); +extern void time2iso(char *buf, size_t len, time_t time, bool utc); extern void convert_from_base_unit(int64 base_value, int base_unit, int64 *value, const char **unit); diff --git a/src/utils/file.c b/src/utils/file.c index ba0575da3..fa08939f5 100644 --- a/src/utils/file.c +++ b/src/utils/file.c @@ -1,51 +1,146 @@ #include #include -#include - -#ifdef WIN32 -#define __thread __declspec(thread) -#else -#include -#endif #include "pg_probackup.h" +/* sys/stat.h must be included after pg_probackup.h (see problems with compilation for windows described in PGPRO-5750) */ +#include + #include "file.h" #include "storage/checksum.h" #define PRINTF_BUF_SIZE 1024 #define FILE_PERMISSIONS 0600 -#define PAGE_READ_ATTEMPTS 100 static __thread unsigned long fio_fdset = 0; static __thread void* fio_stdin_buffer; static __thread int fio_stdout = 0; static __thread int fio_stdin = 0; +static __thread int fio_stderr = 0; +static char *async_errormsg = NULL; + +#define PAGE_ZEROSEARCH_COARSE_GRANULARITY 4096 +#define PAGE_ZEROSEARCH_FINE_GRANULARITY 64 +static const char zerobuf[PAGE_ZEROSEARCH_COARSE_GRANULARITY] = {0}; fio_location MyLocation; typedef struct { BlockNumber nblocks; - BlockNumber segBlockNum; + BlockNumber segmentno; XLogRecPtr horizonLsn; uint32 checksumVersion; int calg; int clevel; + int bitmapsize; + int path_len; } fio_send_request; +typedef struct +{ + char path[MAXPGPATH]; + bool exclude; + bool follow_symlink; + bool add_root; + bool backup_logs; + bool exclusive_backup; + bool skip_hidden; + int external_dir_num; +} fio_list_dir_request; + +typedef struct +{ + mode_t mode; + size_t size; + time_t mtime; + bool is_datafile; + Oid tblspcOid; + Oid dbOid; + Oid relOid; + ForkName forkName; + int segno; + int external_dir_num; + int linked_len; +} fio_pgFile; + +typedef struct +{ + BlockNumber n_blocks; + BlockNumber segmentno; + XLogRecPtr stop_lsn; + uint32 checksumVersion; +} fio_checksum_map_request; + +typedef struct +{ + BlockNumber n_blocks; + BlockNumber segmentno; + XLogRecPtr shift_lsn; + uint32 checksumVersion; +} fio_lsn_map_request; + /* Convert FIO pseudo handle to index in file descriptor array */ #define fio_fileno(f) (((size_t)f - 1) | FIO_PIPE_MARKER) +#if defined(WIN32) +#undef open(a, b, c) +#undef fopen(a, b) +#endif + +void +setMyLocation(ProbackupSubcmd const subcmd) +{ + +#ifdef WIN32 + if (IsSshProtocol()) + elog(ERROR, "Currently remote operations on Windows are not supported"); +#endif + + MyLocation = IsSshProtocol() + ? (subcmd == ARCHIVE_PUSH_CMD || subcmd == ARCHIVE_GET_CMD) + ? FIO_DB_HOST + : (subcmd == BACKUP_CMD || subcmd == RESTORE_CMD || subcmd == ADD_INSTANCE_CMD || subcmd == CATCHUP_CMD) + ? FIO_BACKUP_HOST + : FIO_LOCAL_HOST + : FIO_LOCAL_HOST; +} + /* Use specified file descriptors as stdin/stdout for FIO functions */ -void fio_redirect(int in, int out) +void +fio_redirect(int in, int out, int err) { fio_stdin = in; fio_stdout = out; + fio_stderr = err; +} + +void +fio_error(int rc, int size, char const* file, int line) +{ + if (remote_agent) + { + fprintf(stderr, "%s:%d: processed %d bytes instead of %d: %s\n", file, line, rc, size, rc >= 0 ? "end of data" : strerror(errno)); + exit(EXIT_FAILURE); + } + else + { + char buf[PRINTF_BUF_SIZE+1]; +// Assert(false); + int err_size = read(fio_stderr, buf, PRINTF_BUF_SIZE); + if (err_size > 0) + { + buf[err_size] = '\0'; + elog(ERROR, "Agent error: %s", buf); + } + else + elog(ERROR, "Communication error: %s", rc >= 0 ? "end of data" : strerror(errno)); + } } /* Check if file descriptor is local or remote (created by FIO) */ -static bool fio_is_remote_fd(int fd) +static bool +fio_is_remote_fd(int fd) { return (fd & FIO_PIPE_MARKER) != 0; } @@ -86,14 +181,18 @@ fio_safestat(const char *path, struct stat *buf) #define stat(x, y) fio_safestat(x, y) -static ssize_t pread(int fd, void* buf, size_t size, off_t off) +/* TODO: use real pread on Linux */ +static ssize_t +pread(int fd, void* buf, size_t size, off_t off) { off_t rc = lseek(fd, off, SEEK_SET); if (rc != off) return -1; return read(fd, buf, size); } -static int remove_file_or_dir(char const* path) + +static int +remove_file_or_dir(char const* path) { int rc = remove(path); #ifdef WIN32 @@ -107,7 +206,8 @@ static int remove_file_or_dir(char const* path) #endif /* Check if specified location is local for current node */ -static bool fio_is_remote(fio_location location) +bool +fio_is_remote(fio_location location) { bool is_remote = MyLocation != FIO_LOCAL_HOST && location != FIO_LOCAL_HOST @@ -117,37 +217,54 @@ static bool fio_is_remote(fio_location location) return is_remote; } +/* Check if specified location is local for current node */ +bool +fio_is_remote_simple(fio_location location) +{ + bool is_remote = MyLocation != FIO_LOCAL_HOST + && location != FIO_LOCAL_HOST + && location != MyLocation; + return is_remote; +} + /* Try to read specified amount of bytes unless error or EOF are encountered */ -static ssize_t fio_read_all(int fd, void* buf, size_t size) +static ssize_t +fio_read_all(int fd, void* buf, size_t size) { size_t offs = 0; while (offs < size) { ssize_t rc = read(fd, (char*)buf + offs, size - offs); - if (rc < 0) { - if (errno == EINTR) { + if (rc < 0) + { + if (errno == EINTR) continue; - } + elog(ERROR, "fio_read_all error, fd %i: %s", fd, strerror(errno)); return rc; - } else if (rc == 0) { - break; } + else if (rc == 0) + break; + offs += rc; } return offs; } /* Try to write specified amount of bytes unless error is encountered */ -static ssize_t fio_write_all(int fd, void const* buf, size_t size) +static ssize_t +fio_write_all(int fd, void const* buf, size_t size) { size_t offs = 0; while (offs < size) { ssize_t rc = write(fd, (char*)buf + offs, size - offs); - if (rc <= 0) { - if (errno == EINTR) { + if (rc <= 0) + { + if (errno == EINTR) continue; - } + + elog(ERROR, "fio_write_all error, fd %i: %s", fd, strerror(errno)); + return rc; } offs += rc; @@ -155,8 +272,28 @@ static ssize_t fio_write_all(int fd, void const* buf, size_t size) return offs; } +/* Get version of remote agent */ +void +fio_get_agent_version(int* protocol, char* payload_buf, size_t payload_buf_size) +{ + fio_header hdr; + hdr.cop = FIO_AGENT_VERSION; + hdr.size = 0; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + if (hdr.size > payload_buf_size) + { + elog(ERROR, "Corrupted remote compatibility protocol: insufficient payload_buf_size=%zu", payload_buf_size); + } + + *protocol = hdr.arg; + IO_CHECK(fio_read_all(fio_stdin, payload_buf, hdr.size), hdr.size); +} + /* Open input stream. Remote file is fetched to the in-memory buffer and then accessed through Linux fmemopen */ -FILE* fio_open_stream(char const* path, fio_location location) +FILE* +fio_open_stream(char const* path, fio_location location) { FILE* f; if (fio_is_remote(location)) @@ -177,7 +314,7 @@ FILE* fio_open_stream(char const* path, fio_location location) IO_CHECK(fio_read_all(fio_stdin, fio_stdin_buffer, hdr.size), hdr.size); #ifdef WIN32 f = tmpfile(); - IO_CHECK(fwrite(f, 1, hdr.size, fio_stdin_buffer), hdr.size); + IO_CHECK(fwrite(fio_stdin_buffer, 1, hdr.size, f), hdr.size); SYS_CHECK(fseek(f, 0, SEEK_SET)); #else f = fmemopen(fio_stdin_buffer, hdr.size, "r"); @@ -196,7 +333,8 @@ FILE* fio_open_stream(char const* path, fio_location location) } /* Close input stream */ -int fio_close_stream(FILE* f) +int +fio_close_stream(FILE* f) { if (fio_stdin_buffer) { @@ -207,7 +345,8 @@ int fio_close_stream(FILE* f) } /* Open directory */ -DIR* fio_opendir(char const* path, fio_location location) +DIR* +fio_opendir(char const* path, fio_location location) { DIR* dir; if (fio_is_remote(location)) @@ -219,7 +358,8 @@ DIR* fio_opendir(char const* path, fio_location location) mask = fio_fdset; for (i = 0; (mask & 1) != 0; i++, mask >>= 1); if (i == FIO_FDMAX) { - return NULL; + elog(ERROR, "Descriptor pool for remote files is exhausted, " + "probably too many remote directories are opened"); } hdr.cop = FIO_OPENDIR; hdr.handle = i; @@ -234,6 +374,7 @@ DIR* fio_opendir(char const* path, fio_location location) if (hdr.arg != 0) { errno = hdr.arg; + fio_fdset &= ~(1 << hdr.handle); return NULL; } dir = (DIR*)(size_t)(i + 1); @@ -246,7 +387,8 @@ DIR* fio_opendir(char const* path, fio_location location) } /* Get next directory entry */ -struct dirent* fio_readdir(DIR *dir) +struct dirent* +fio_readdir(DIR *dir) { if (fio_is_remote_file((FILE*)dir)) { @@ -274,7 +416,8 @@ struct dirent* fio_readdir(DIR *dir) } /* Close directory */ -int fio_closedir(DIR *dir) +int +fio_closedir(DIR *dir) { if (fio_is_remote_file((FILE*)dir)) { @@ -294,7 +437,8 @@ int fio_closedir(DIR *dir) } /* Open file */ -int fio_open(char const* path, int mode, fio_location location) +int +fio_open(char const* path, int mode, fio_location location) { int fd; if (fio_is_remote(location)) @@ -305,23 +449,29 @@ int fio_open(char const* path, int mode, fio_location location) mask = fio_fdset; for (i = 0; (mask & 1) != 0; i++, mask >>= 1); - if (i == FIO_FDMAX) { - return -1; - } + if (i == FIO_FDMAX) + elog(ERROR, "Descriptor pool for remote files is exhausted, " + "probably too many remote files are opened"); + hdr.cop = FIO_OPEN; hdr.handle = i; hdr.size = strlen(path) + 1; - hdr.arg = mode & ~O_EXCL; + hdr.arg = mode; +// hdr.arg = mode & ~O_EXCL; +// elog(INFO, "PATH: %s MODE: %i, %i", path, mode, O_EXCL); +// elog(INFO, "MODE: %i", hdr.arg); fio_fdset |= 1 << i; IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); IO_CHECK(fio_write_all(fio_stdout, path, hdr.size), hdr.size); + /* check results */ IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); if (hdr.arg != 0) { errno = hdr.arg; + fio_fdset &= ~(1 << hdr.handle); return -1; } fd = i | FIO_PIPE_MARKER; @@ -340,16 +490,25 @@ fio_disconnect(void) { if (fio_stdin) { + fio_header hdr; + hdr.cop = FIO_DISCONNECT; + hdr.size = 0; + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + Assert(hdr.cop == FIO_DISCONNECTED); SYS_CHECK(close(fio_stdin)); SYS_CHECK(close(fio_stdout)); + SYS_CHECK(close(fio_stderr)); fio_stdin = 0; fio_stdout = 0; + fio_stderr = 0; wait_ssh(); } } /* Open stdio file */ -FILE* fio_fopen(char const* path, char const* mode, fio_location location) +FILE* +fio_fopen(char const* path, char const* mode, fio_location location) { FILE *f = NULL; @@ -394,7 +553,8 @@ FILE* fio_fopen(char const* path, char const* mode, fio_location location) } /* Format output to file stream */ -int fio_fprintf(FILE* f, char const* format, ...) +int +fio_fprintf(FILE* f, char const* format, ...) { int rc; va_list args; @@ -419,28 +579,26 @@ int fio_fprintf(FILE* f, char const* format, ...) return rc; } -/* Flush stream data (does nothing for remote file) */ -int fio_fflush(FILE* f) +/* Flush stream data (does nothing for remote file) */ +int +fio_fflush(FILE* f) { int rc = 0; if (!fio_is_remote_file(f)) - { rc = fflush(f); - if (rc == 0) { - rc = fsync(fileno(f)); - } - } return rc; } /* Sync file to the disk (does nothing for remote file) */ -int fio_flush(int fd) +int +fio_flush(int fd) { return fio_is_remote_fd(fd) ? 0 : fsync(fd); } /* Close output stream */ -int fio_fclose(FILE* f) +int +fio_fclose(FILE* f) { return fio_is_remote_file(f) ? fio_close(fio_fileno(f)) @@ -448,7 +606,8 @@ int fio_fclose(FILE* f) } /* Close file */ -int fio_close(int fd) +int +fio_close(int fd) { if (fio_is_remote_fd(fd)) { @@ -461,6 +620,15 @@ int fio_close(int fd) IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + /* Wait for response */ + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (hdr.arg != 0) + { + errno = hdr.arg; + return -1; + } + return 0; } else @@ -469,16 +637,36 @@ int fio_close(int fd) } } +/* Close remote file implementation */ +static void +fio_close_impl(int fd, int out) +{ + fio_header hdr; + + hdr.cop = FIO_CLOSE; + hdr.arg = 0; + + if (close(fd) != 0) + hdr.arg = errno; + + /* send header */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); +} + /* Truncate stdio file */ -int fio_ftruncate(FILE* f, off_t size) +int +fio_ftruncate(FILE* f, off_t size) { return fio_is_remote_file(f) ? fio_truncate(fio_fileno(f), size) : ftruncate(fileno(f), size); } -/* Truncate file */ -int fio_truncate(int fd, off_t size) +/* Truncate file + * TODO: make it synchronous + */ +int +fio_truncate(int fd, off_t size) { if (fio_is_remote_fd(fd)) { @@ -503,7 +691,8 @@ int fio_truncate(int fd, off_t size) /* * Read file from specified location. */ -int fio_pread(FILE* f, void* buf, off_t offs) +int +fio_pread(FILE* f, void* buf, off_t offs) { if (fio_is_remote_file(f)) { @@ -522,22 +711,35 @@ int fio_pread(FILE* f, void* buf, off_t offs) if (hdr.size != 0) IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + /* TODO: error handling */ + return hdr.arg; } else - return pread(fileno(f), buf, BLCKSZ, offs); + { + /* For local file, opened by fopen, we should use stdio functions */ + int rc = fseek(f, offs, SEEK_SET); + + if (rc < 0) + return rc; + + return fread(buf, 1, BLCKSZ, f); + } } /* Set position in stdio file */ -int fio_fseek(FILE* f, off_t offs) +int +fio_fseek(FILE* f, off_t offs) { return fio_is_remote_file(f) - ? fio_seek(fio_fileno(f), offs) + ? fio_seek(fio_fileno(f), offs) : fseek(f, offs, SEEK_SET); } /* Set position in file */ -int fio_seek(int fd, off_t offs) +/* TODO: make it synchronous or check async error */ +int +fio_seek(int fd, off_t offs) { if (fio_is_remote_fd(fd)) { @@ -558,16 +760,66 @@ int fio_seek(int fd, off_t offs) } } +/* seek is asynchronous */ +static void +fio_seek_impl(int fd, off_t offs) +{ + int rc; + + /* Quick exit for tainted agent */ + if (async_errormsg) + return; + + rc = lseek(fd, offs, SEEK_SET); + + if (rc < 0) + { + async_errormsg = pgut_malloc(ERRMSG_MAX_LEN); + snprintf(async_errormsg, ERRMSG_MAX_LEN, "%s", strerror(errno)); + } +} + /* Write data to stdio file */ -size_t fio_fwrite(FILE* f, void const* buf, size_t size) +size_t +fio_fwrite(FILE* f, void const* buf, size_t size) { - return fio_is_remote_file(f) - ? fio_write(fio_fileno(f), buf, size) - : fwrite(buf, 1, size, f); + if (fio_is_remote_file(f)) + return fio_write(fio_fileno(f), buf, size); + else + return fwrite(buf, 1, size, f); } -/* Write data to the file */ -ssize_t fio_write(int fd, void const* buf, size_t size) +/* + * Write buffer to descriptor by calling write(), + * If size of written data is less than buffer size, + * then try to write what is left. + * We do this to get honest errno if there are some problems + * with filesystem, since writing less than buffer size + * is not considered an error. + */ +static ssize_t +durable_write(int fd, const char* buf, size_t size) +{ + off_t current_pos = 0; + size_t bytes_left = size; + + while (bytes_left > 0) + { + int rc = write(fd, buf + current_pos, bytes_left); + + if (rc <= 0) + return rc; + + bytes_left -= rc; + current_pos += rc; + } + + return size; +} + +/* Write data to the file synchronously */ +ssize_t +fio_write(int fd, void const* buf, size_t size) { if (fio_is_remote_fd(fd)) { @@ -580,129 +832,413 @@ ssize_t fio_write(int fd, void const* buf, size_t size) IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); IO_CHECK(fio_write_all(fio_stdout, buf, size), size); + /* check results */ + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + /* set errno */ + if (hdr.arg > 0) + { + errno = hdr.arg; + return -1; + } + return size; } else { - return write(fd, buf, size); + return durable_write(fd, buf, size); } } -/* Read data from stdio file */ -ssize_t fio_fread(FILE* f, void* buf, size_t size) +static void +fio_write_impl(int fd, void const* buf, size_t size, int out) { - size_t rc; - if (fio_is_remote_file(f)) - return fio_read(fio_fileno(f), buf, size); - rc = fread(buf, 1, size, f); - return rc == 0 && !feof(f) ? -1 : rc; + int rc; + fio_header hdr; + + rc = durable_write(fd, buf, size); + + hdr.arg = 0; + hdr.size = 0; + + if (rc < 0) + hdr.arg = errno; + + /* send header */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + + return; } -/* Read data from file */ -ssize_t fio_read(int fd, void* buf, size_t size) +size_t +fio_fwrite_async(FILE* f, void const* buf, size_t size) +{ + return fio_is_remote_file(f) + ? fio_write_async(fio_fileno(f), buf, size) + : fwrite(buf, 1, size, f); +} + +/* Write data to the file */ +/* TODO: support async report error */ +ssize_t +fio_write_async(int fd, void const* buf, size_t size) { + if (size == 0) + return 0; + if (fio_is_remote_fd(fd)) { fio_header hdr; - hdr.cop = FIO_READ; + hdr.cop = FIO_WRITE_ASYNC; hdr.handle = fd & ~FIO_PIPE_MARKER; - hdr.size = 0; - hdr.arg = size; + hdr.size = size; IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); - - IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); - Assert(hdr.cop == FIO_SEND); - IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); - - return hdr.size; + IO_CHECK(fio_write_all(fio_stdout, buf, size), size); } else + return durable_write(fd, buf, size); + + return size; +} + +static void +fio_write_async_impl(int fd, void const* buf, size_t size, int out) +{ + /* Quick exit if agent is tainted */ + if (async_errormsg) + return; + + if (durable_write(fd, buf, size) <= 0) { - return read(fd, buf, size); + async_errormsg = pgut_malloc(ERRMSG_MAX_LEN); + snprintf(async_errormsg, ERRMSG_MAX_LEN, "%s", strerror(errno)); } } -/* Get information about stdio file */ -int fio_ffstat(FILE* f, struct stat* st) +int32 +fio_decompress(void* dst, void const* src, size_t size, int compress_alg, char **errormsg) { - return fio_is_remote_file(f) - ? fio_fstat(fio_fileno(f), st) - : fio_fstat(fileno(f), st); + const char *internal_errormsg = NULL; + int32 uncompressed_size = do_decompress(dst, BLCKSZ, + src, + size, + compress_alg, &internal_errormsg); + + if (uncompressed_size < 0 && internal_errormsg != NULL) + { + *errormsg = pgut_malloc(ERRMSG_MAX_LEN); + snprintf(*errormsg, ERRMSG_MAX_LEN, "An error occured during decompressing block: %s", internal_errormsg); + return -1; + } + + if (uncompressed_size != BLCKSZ) + { + *errormsg = pgut_malloc(ERRMSG_MAX_LEN); + snprintf(*errormsg, ERRMSG_MAX_LEN, "Page uncompressed to %d bytes != BLCKSZ", uncompressed_size); + return -1; + } + return uncompressed_size; } -/* Get information about file descriptor */ -int fio_fstat(int fd, struct stat* st) +/* Write data to the file */ +ssize_t +fio_fwrite_async_compressed(FILE* f, void const* buf, size_t size, int compress_alg) { - if (fio_is_remote_fd(fd)) + if (fio_is_remote_file(f)) { fio_header hdr; - hdr.cop = FIO_FSTAT; - hdr.handle = fd & ~FIO_PIPE_MARKER; - hdr.size = 0; + hdr.cop = FIO_WRITE_COMPRESSED_ASYNC; + hdr.handle = fio_fileno(f) & ~FIO_PIPE_MARKER; + hdr.size = size; + hdr.arg = compress_alg; IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, buf, size), size); - IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); - Assert(hdr.cop == FIO_FSTAT); - IO_CHECK(fio_read_all(fio_stdin, st, sizeof(*st)), sizeof(*st)); - - if (hdr.arg != 0) - { - errno = hdr.arg; - return -1; - } - return 0; + return size; } else { - return fstat(fd, st); + char *errormsg = NULL; + char decompressed_buf[BLCKSZ]; + int32 decompressed_size = fio_decompress(decompressed_buf, buf, size, compress_alg, &errormsg); + + if (decompressed_size < 0) + elog(ERROR, "%s", errormsg); + + return fwrite(decompressed_buf, 1, decompressed_size, f); } } -/* Get information about file */ -int fio_stat(char const* path, struct stat* st, bool follow_symlink, fio_location location) +static void +fio_write_compressed_impl(int fd, void const* buf, size_t size, int compress_alg) { - if (fio_is_remote(location)) - { - fio_header hdr; - size_t path_len = strlen(path) + 1; + int32 decompressed_size; + char decompressed_buf[BLCKSZ]; - hdr.cop = FIO_STAT; - hdr.handle = -1; - hdr.arg = follow_symlink; - hdr.size = path_len; + /* If the previous command already have failed, + * then there is no point in bashing a head against the wall + */ + if (async_errormsg) + return; - IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); - IO_CHECK(fio_write_all(fio_stdout, path, path_len), path_len); + /* decompress chunk */ + decompressed_size = fio_decompress(decompressed_buf, buf, size, compress_alg, &async_errormsg); - IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); - Assert(hdr.cop == FIO_STAT); - IO_CHECK(fio_read_all(fio_stdin, st, sizeof(*st)), sizeof(*st)); + if (decompressed_size < 0) + return; - if (hdr.arg != 0) - { - errno = hdr.arg; - return -1; - } - return 0; - } - else + if (durable_write(fd, decompressed_buf, decompressed_size) <= 0) { - return follow_symlink ? stat(path, st) : lstat(path, st); + async_errormsg = pgut_malloc(ERRMSG_MAX_LEN); + snprintf(async_errormsg, ERRMSG_MAX_LEN, "%s", strerror(errno)); } } -/* Check presence of the file */ -int fio_access(char const* path, int mode, fio_location location) +/* check if remote agent encountered any error during execution of async operations */ +int +fio_check_error_file(FILE* f, char **errmsg) { - if (fio_is_remote(location)) + if (fio_is_remote_file(f)) { fio_header hdr; - size_t path_len = strlen(path) + 1; - hdr.cop = FIO_ACCESS; + + hdr.cop = FIO_GET_ASYNC_ERROR; + hdr.size = 0; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + + /* check results */ + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (hdr.size > 0) + { + *errmsg = pgut_malloc(ERRMSG_MAX_LEN); + IO_CHECK(fio_read_all(fio_stdin, *errmsg, hdr.size), hdr.size); + return 1; + } + } + + return 0; +} + +/* check if remote agent encountered any error during execution of async operations */ +int +fio_check_error_fd(int fd, char **errmsg) +{ + if (fio_is_remote_fd(fd)) + { + fio_header hdr; + + hdr.cop = FIO_GET_ASYNC_ERROR; + hdr.size = 0; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + + /* check results */ + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (hdr.size > 0) + { + *errmsg = pgut_malloc(ERRMSG_MAX_LEN); + IO_CHECK(fio_read_all(fio_stdin, *errmsg, hdr.size), hdr.size); + return 1; + } + } + return 0; +} + +static void +fio_get_async_error_impl(int out) +{ + fio_header hdr; + hdr.cop = FIO_GET_ASYNC_ERROR; + + /* send error message */ + if (async_errormsg) + { + hdr.size = strlen(async_errormsg) + 1; + + /* send header */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + + /* send message itself */ + IO_CHECK(fio_write_all(out, async_errormsg, hdr.size), hdr.size); + + //TODO: should we reset the tainted state ? +// pg_free(async_errormsg); +// async_errormsg = NULL; + } + else + { + hdr.size = 0; + /* send header */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + } +} + +/* Read data from stdio file */ +ssize_t +fio_fread(FILE* f, void* buf, size_t size) +{ + size_t rc; + if (fio_is_remote_file(f)) + return fio_read(fio_fileno(f), buf, size); + rc = fread(buf, 1, size, f); + return rc == 0 && !feof(f) ? -1 : rc; +} + +/* Read data from file */ +ssize_t +fio_read(int fd, void* buf, size_t size) +{ + if (fio_is_remote_fd(fd)) + { + fio_header hdr; + + hdr.cop = FIO_READ; + hdr.handle = fd & ~FIO_PIPE_MARKER; + hdr.size = 0; + hdr.arg = size; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + Assert(hdr.cop == FIO_SEND); + IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + + return hdr.size; + } + else + { + return read(fd, buf, size); + } +} + +/* Get information about file */ +int +fio_stat(char const* path, struct stat* st, bool follow_symlink, fio_location location) +{ + if (fio_is_remote(location)) + { + fio_header hdr; + size_t path_len = strlen(path) + 1; + + hdr.cop = FIO_STAT; + hdr.handle = -1; + hdr.arg = follow_symlink; + hdr.size = path_len; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, path, path_len), path_len); + + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + Assert(hdr.cop == FIO_STAT); + IO_CHECK(fio_read_all(fio_stdin, st, sizeof(*st)), sizeof(*st)); + + if (hdr.arg != 0) + { + errno = hdr.arg; + return -1; + } + return 0; + } + else + { + return follow_symlink ? stat(path, st) : lstat(path, st); + } +} + +/* + * Compare, that filename1 and filename2 is the same file + * in windows compare only filenames + */ +bool +fio_is_same_file(char const* filename1, char const* filename2, bool follow_symlink, fio_location location) +{ + char *abs_name1 = make_absolute_path(filename1); + char *abs_name2 = make_absolute_path(filename2); + bool result = strcmp(abs_name1, abs_name2) == 0; + +#ifndef WIN32 + if (!result) + { + struct stat stat1, stat2; + + if (fio_stat(filename1, &stat1, follow_symlink, location) < 0) + { + if (errno == ENOENT) + return false; + elog(ERROR, "Can't stat file \"%s\": %s", filename1, strerror(errno)); + } + + if (fio_stat(filename2, &stat2, follow_symlink, location) < 0) + { + if (errno == ENOENT) + return false; + elog(ERROR, "Can't stat file \"%s\": %s", filename2, strerror(errno)); + } + + result = (stat1.st_ino == stat2.st_ino && stat1.st_dev == stat2.st_dev); + } +#endif + free(abs_name2); + free(abs_name1); + return result; +} + +/* + * Read value of a symbolic link + * this is a wrapper about readlink() syscall + * side effects: string truncation occur (and it + * can be checked by caller by comparing + * returned value >= valsiz) + */ +ssize_t +fio_readlink(const char *path, char *value, size_t valsiz, fio_location location) +{ + if (!fio_is_remote(location)) + { + /* readlink don't place trailing \0 */ + ssize_t len = readlink(path, value, valsiz); + value[len < valsiz ? len : valsiz] = '\0'; + return len; + } + else + { + fio_header hdr; + size_t path_len = strlen(path) + 1; + + hdr.cop = FIO_READLINK; + hdr.handle = -1; + Assert(valsiz <= UINT_MAX); /* max value of fio_header.arg */ + hdr.arg = valsiz; + hdr.size = path_len; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, path, path_len), path_len); + + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + Assert(hdr.cop == FIO_READLINK); + Assert(hdr.size <= valsiz); + IO_CHECK(fio_read_all(fio_stdin, value, hdr.size), hdr.size); + value[hdr.size < valsiz ? hdr.size : valsiz] = '\0'; + return hdr.size; + } +} + +/* Check presence of the file */ +int +fio_access(char const* path, int mode, fio_location location) +{ + if (fio_is_remote(location)) + { + fio_header hdr; + size_t path_len = strlen(path) + 1; + hdr.cop = FIO_ACCESS; hdr.handle = -1; hdr.size = path_len; hdr.arg = mode; @@ -727,7 +1263,8 @@ int fio_access(char const* path, int mode, fio_location location) } /* Create symbolic link */ -int fio_symlink(char const* target, char const* link_path, fio_location location) +int +fio_symlink(char const* target, char const* link_path, bool overwrite, fio_location location) { if (fio_is_remote(location)) { @@ -737,6 +1274,7 @@ int fio_symlink(char const* target, char const* link_path, fio_location location hdr.cop = FIO_SYMLINK; hdr.handle = -1; hdr.size = target_len + link_path_len; + hdr.arg = overwrite ? 1 : 0; IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); IO_CHECK(fio_write_all(fio_stdout, target, target_len), target_len); @@ -746,12 +1284,30 @@ int fio_symlink(char const* target, char const* link_path, fio_location location } else { + if (overwrite) + remove_file_or_dir(link_path); + return symlink(target, link_path); } } +static void +fio_symlink_impl(int out, char *buf, bool overwrite) +{ + char *linked_path = buf; + char *link_path = buf + strlen(buf) + 1; + + if (overwrite) + remove_file_or_dir(link_path); + + if (symlink(linked_path, link_path)) + elog(ERROR, "Could not create symbolic link \"%s\": %s", + link_path, strerror(errno)); +} + /* Rename file */ -int fio_rename(char const* old_path, char const* new_path, fio_location location) +int +fio_rename(char const* old_path, char const* new_path, fio_location location) { if (fio_is_remote(location)) { @@ -766,6 +1322,8 @@ int fio_rename(char const* old_path, char const* new_path, fio_location location IO_CHECK(fio_write_all(fio_stdout, old_path, old_path_len), old_path_len); IO_CHECK(fio_write_all(fio_stdout, new_path, new_path_len), new_path_len); + //TODO: wait for confirmation. + return 0; } else @@ -774,8 +1332,114 @@ int fio_rename(char const* old_path, char const* new_path, fio_location location } } +/* Sync file to disk */ +int +fio_sync(char const* path, fio_location location) +{ + if (fio_is_remote(location)) + { + fio_header hdr; + size_t path_len = strlen(path) + 1; + hdr.cop = FIO_SYNC; + hdr.handle = -1; + hdr.size = path_len; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, path, path_len), path_len); + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (hdr.arg != 0) + { + errno = hdr.arg; + return -1; + } + + return 0; + } + else + { + int fd; + + fd = open(path, O_WRONLY | PG_BINARY, FILE_PERMISSIONS); + if (fd < 0) + return -1; + + if (fsync(fd) < 0) + { + close(fd); + return -1; + } + close(fd); + + return 0; + } +} + +enum { + GET_CRC32_DECOMPRESS = 1, + GET_CRC32_MISSING_OK = 2, + GET_CRC32_TRUNCATED = 4 +}; + +/* Get crc32 of file */ +static pg_crc32 +fio_get_crc32_ex(const char *file_path, fio_location location, + bool decompress, bool missing_ok, bool truncated) +{ + if (decompress && truncated) + elog(ERROR, "Could not calculate CRC for compressed truncated file"); + + if (fio_is_remote(location)) + { + fio_header hdr; + size_t path_len = strlen(file_path) + 1; + pg_crc32 crc = 0; + hdr.cop = FIO_GET_CRC32; + hdr.handle = -1; + hdr.size = path_len; + hdr.arg = 0; + + if (decompress) + hdr.arg = GET_CRC32_DECOMPRESS; + if (missing_ok) + hdr.arg |= GET_CRC32_MISSING_OK; + if (truncated) + hdr.arg |= GET_CRC32_TRUNCATED; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, file_path, path_len), path_len); + IO_CHECK(fio_read_all(fio_stdin, &crc, sizeof(crc)), sizeof(crc)); + + return crc; + } + else + { + if (decompress) + return pgFileGetCRCgz(file_path, true, missing_ok); + else if (truncated) + return pgFileGetCRCTruncated(file_path, true, missing_ok); + else + return pgFileGetCRC(file_path, true, missing_ok); + } +} + +pg_crc32 +fio_get_crc32(const char *file_path, fio_location location, + bool decompress, bool missing_ok) +{ + return fio_get_crc32_ex(file_path, location, decompress, missing_ok, false); +} + +pg_crc32 +fio_get_crc32_truncated(const char *file_path, fio_location location, + bool missing_ok) +{ + return fio_get_crc32_ex(file_path, location, false, missing_ok, true); +} + /* Remove file */ -int fio_unlink(char const* path, fio_location location) +int +fio_unlink(char const* path, fio_location location) { if (fio_is_remote(location)) { @@ -788,6 +1452,7 @@ int fio_unlink(char const* path, fio_location location) IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); IO_CHECK(fio_write_all(fio_stdout, path, path_len), path_len); + // TODO: error is swallowed ? return 0; } else @@ -796,8 +1461,11 @@ int fio_unlink(char const* path, fio_location location) } } -/* Create directory */ -int fio_mkdir(char const* path, int mode, fio_location location) +/* Create directory + * TODO: add strict flag + */ +int +fio_mkdir(char const* path, int mode, fio_location location) { if (fio_is_remote(location)) { @@ -818,12 +1486,13 @@ int fio_mkdir(char const* path, int mode, fio_location location) } else { - return dir_create_dir(path, mode); + return dir_create_dir(path, mode, false); } } /* Change file mode */ -int fio_chmod(char const* path, int mode, fio_location location) +int +fio_chmod(char const* path, int mode, fio_location location) { if (fio_is_remote(location)) { @@ -847,10 +1516,14 @@ int fio_chmod(char const* path, int mode, fio_location location) #ifdef HAVE_LIBZ - #define ZLIB_BUFFER_SIZE (64*1024) #define MAX_WBITS 15 /* 32K LZ77 window */ #define DEF_MEM_LEVEL 8 +/* last bit used to differenciate remote gzFile from local gzFile + * TODO: this is insane, we should create our own scructure for this, + * not flip some bits in someone's else and hope that it will not break + * between zlib versions. + */ #define FIO_GZ_REMOTE_MARKER 1 typedef struct fioGZFile @@ -863,6 +1536,33 @@ typedef struct fioGZFile Bytef buf[ZLIB_BUFFER_SIZE]; } fioGZFile; +/* check if remote agent encountered any error during execution of async operations */ +int +fio_check_error_fd_gz(gzFile f, char **errmsg) +{ + if (f && ((size_t)f & FIO_GZ_REMOTE_MARKER)) + { + fio_header hdr; + + hdr.cop = FIO_GET_ASYNC_ERROR; + hdr.size = 0; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + + /* check results */ + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (hdr.size > 0) + { + *errmsg = pgut_malloc(ERRMSG_MAX_LEN); + IO_CHECK(fio_read_all(fio_stdin, *errmsg, hdr.size), hdr.size); + return 1; + } + } + return 0; +} + +/* On error returns NULL and errno should be checked */ gzFile fio_gzopen(char const* path, char const* mode, int level, fio_location location) { @@ -873,6 +1573,7 @@ fio_gzopen(char const* path, char const* mode, int level, fio_location location) memset(&gz->strm, 0, sizeof(gz->strm)); gz->eof = 0; gz->errnum = Z_OK; + /* check if file opened for writing */ if (strcmp(mode, PG_BINARY_W) == 0) /* compress */ { gz->strm.next_out = gz->buf; @@ -885,14 +1586,12 @@ fio_gzopen(char const* path, char const* mode, int level, fio_location location) if (rc == Z_OK) { gz->compress = 1; - if (fio_access(path, F_OK, location) == 0) + gz->fd = fio_open(path, O_WRONLY | O_CREAT | O_EXCL | PG_BINARY, location); + if (gz->fd < 0) { - elog(LOG, "File %s exists", path); free(gz); - errno = EEXIST; return NULL; } - gz->fd = fio_open(path, O_WRONLY | O_CREAT | O_EXCL | PG_BINARY, location); } } else @@ -905,21 +1604,27 @@ fio_gzopen(char const* path, char const* mode, int level, fio_location location) { gz->compress = 0; gz->fd = fio_open(path, O_RDONLY | PG_BINARY, location); + if (gz->fd < 0) + { + free(gz); + return NULL; + } } } if (rc != Z_OK) { - free(gz); - return NULL; + elog(ERROR, "zlib internal error when opening file %s: %s", + path, gz->strm.msg); } return (gzFile)((size_t)gz + FIO_GZ_REMOTE_MARKER); } else { gzFile file; + /* check if file opened for writing */ if (strcmp(mode, PG_BINARY_W) == 0) { - int fd = open(path, O_RDWR | O_CREAT | O_EXCL | PG_BINARY, FILE_PERMISSIONS); + int fd = open(path, O_WRONLY | O_CREAT | O_EXCL | PG_BINARY, FILE_PERMISSIONS); if (fd < 0) return NULL; file = gzdopen(fd, mode); @@ -979,7 +1684,8 @@ fio_gzread(gzFile f, void *buf, unsigned size) { gz->strm.next_in = gz->buf; } - rc = fio_read(gz->fd, gz->strm.next_in + gz->strm.avail_in, gz->buf + ZLIB_BUFFER_SIZE - gz->strm.next_in - gz->strm.avail_in); + rc = fio_read(gz->fd, gz->strm.next_in + gz->strm.avail_in, + gz->buf + ZLIB_BUFFER_SIZE - gz->strm.next_in - gz->strm.avail_in); if (rc > 0) { gz->strm.avail_in += rc; @@ -1028,7 +1734,7 @@ fio_gzwrite(gzFile f, void const* buf, unsigned size) break; } } - rc = fio_write(gz->fd, gz->strm.next_out, ZLIB_BUFFER_SIZE - gz->strm.avail_out); + rc = fio_write_async(gz->fd, gz->strm.next_out, ZLIB_BUFFER_SIZE - gz->strm.avail_out); if (rc >= 0) { gz->strm.next_out += rc; @@ -1053,264 +1759,2027 @@ fio_gzclose(gzFile f) { if ((size_t)f & FIO_GZ_REMOTE_MARKER) { - fioGZFile* gz = (fioGZFile*)((size_t)f - FIO_GZ_REMOTE_MARKER); - int rc; - if (gz->compress) + fioGZFile* gz = (fioGZFile*)((size_t)f - FIO_GZ_REMOTE_MARKER); + int rc; + if (gz->compress) + { + gz->strm.next_out = gz->buf; + rc = deflate(&gz->strm, Z_FINISH); + Assert(rc == Z_STREAM_END && gz->strm.avail_out != ZLIB_BUFFER_SIZE); + deflateEnd(&gz->strm); + rc = fio_write(gz->fd, gz->buf, ZLIB_BUFFER_SIZE - gz->strm.avail_out); + if (rc != ZLIB_BUFFER_SIZE - gz->strm.avail_out) + { + return -1; + } + } + else + { + inflateEnd(&gz->strm); + } + rc = fio_close(gz->fd); + free(gz); + return rc; + } + else + { + return gzclose(f); + } +} + +int +fio_gzeof(gzFile f) +{ + if ((size_t)f & FIO_GZ_REMOTE_MARKER) + { + fioGZFile* gz = (fioGZFile*)((size_t)f - FIO_GZ_REMOTE_MARKER); + return gz->eof; + } + else + { + return gzeof(f); + } +} + +const char* +fio_gzerror(gzFile f, int *errnum) +{ + if ((size_t)f & FIO_GZ_REMOTE_MARKER) + { + fioGZFile* gz = (fioGZFile*)((size_t)f - FIO_GZ_REMOTE_MARKER); + if (errnum) + *errnum = gz->errnum; + return gz->strm.msg; + } + else + { + return gzerror(f, errnum); + } +} + +z_off_t +fio_gzseek(gzFile f, z_off_t offset, int whence) +{ + Assert(!((size_t)f & FIO_GZ_REMOTE_MARKER)); + return gzseek(f, offset, whence); +} + + +#endif + +/* Send file content + * Note: it should not be used for large files. + */ +static void +fio_load_file(int out, char const* path) +{ + int fd = open(path, O_RDONLY); + fio_header hdr; + void* buf = NULL; + + hdr.cop = FIO_SEND; + hdr.size = 0; + + if (fd >= 0) + { + off_t size = lseek(fd, 0, SEEK_END); + buf = pgut_malloc(size); + lseek(fd, 0, SEEK_SET); + IO_CHECK(fio_read_all(fd, buf, size), size); + hdr.size = size; + SYS_CHECK(close(fd)); + } + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + if (buf) + { + IO_CHECK(fio_write_all(out, buf, hdr.size), hdr.size); + free(buf); + } +} + +/* + * Return number of actually(!) readed blocks, attempts or + * half-readed block are not counted. + * Return values in case of error: + * FILE_MISSING + * OPEN_FAILED + * READ_ERROR + * PAGE_CORRUPTION + * WRITE_FAILED + * + * If none of the above, this function return number of blocks + * readed by remote agent. + * + * In case of DELTA mode horizonLsn must be a valid lsn, + * otherwise it should be set to InvalidXLogRecPtr. + */ +int +fio_send_pages(const char *to_fullpath, const char *from_fullpath, pgFile *file, + XLogRecPtr horizonLsn, int calg, int clevel, uint32 checksum_version, + bool use_pagemap, BlockNumber* err_blknum, char **errormsg, + BackupPageHeader2 **headers) +{ + FILE *out = NULL; + char *out_buf = NULL; + struct { + fio_header hdr; + fio_send_request arg; + } req; + BlockNumber n_blocks_read = 0; + BlockNumber blknum = 0; + + /* send message with header + + 16bytes 24bytes var var + -------------------------------------------------------------- + | fio_header | fio_send_request | FILE PATH | BITMAP(if any) | + -------------------------------------------------------------- + */ + + req.hdr.cop = FIO_SEND_PAGES; + + if (use_pagemap) + { + req.hdr.size = sizeof(fio_send_request) + (*file).pagemap.bitmapsize + strlen(from_fullpath) + 1; + req.arg.bitmapsize = (*file).pagemap.bitmapsize; + + /* TODO: add optimization for the case of pagemap + * containing small number of blocks with big serial numbers: + * https://github.com/postgrespro/pg_probackup/blob/remote_page_backup/src/utils/file.c#L1211 + */ + } + else + { + req.hdr.size = sizeof(fio_send_request) + strlen(from_fullpath) + 1; + req.arg.bitmapsize = 0; + } + + req.arg.nblocks = file->size/BLCKSZ; + req.arg.segmentno = file->segno * RELSEG_SIZE; + req.arg.horizonLsn = horizonLsn; + req.arg.checksumVersion = checksum_version; + req.arg.calg = calg; + req.arg.clevel = clevel; + req.arg.path_len = strlen(from_fullpath) + 1; + + file->compress_alg = calg; /* TODO: wtf? why here? */ + +//<----- +// datapagemap_iterator_t *iter; +// BlockNumber blkno; +// iter = datapagemap_iterate(pagemap); +// while (datapagemap_next(iter, &blkno)) +// elog(INFO, "block %u", blkno); +// pg_free(iter); +//<----- + + /* send header */ + IO_CHECK(fio_write_all(fio_stdout, &req, sizeof(req)), sizeof(req)); + + /* send file path */ + IO_CHECK(fio_write_all(fio_stdout, from_fullpath, req.arg.path_len), req.arg.path_len); + + /* send pagemap if any */ + if (use_pagemap) + IO_CHECK(fio_write_all(fio_stdout, (*file).pagemap.bitmap, (*file).pagemap.bitmapsize), (*file).pagemap.bitmapsize); + + while (true) + { + fio_header hdr; + char buf[BLCKSZ + sizeof(BackupPageHeader)]; + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (interrupted) + elog(ERROR, "Interrupted during page reading"); + + if (hdr.cop == FIO_ERROR) + { + /* FILE_MISSING, OPEN_FAILED and READ_FAILED */ + if (hdr.size > 0) + { + IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + *errormsg = pgut_malloc(hdr.size); + snprintf(*errormsg, hdr.size, "%s", buf); + } + + return hdr.arg; + } + else if (hdr.cop == FIO_SEND_FILE_CORRUPTION) + { + *err_blknum = hdr.arg; + + if (hdr.size > 0) + { + IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + *errormsg = pgut_malloc(hdr.size); + snprintf(*errormsg, hdr.size, "%s", buf); + } + return PAGE_CORRUPTION; + } + else if (hdr.cop == FIO_SEND_FILE_EOF) + { + /* n_blocks_read reported by EOF */ + n_blocks_read = hdr.arg; + + /* receive headers if any */ + if (hdr.size > 0) + { + *headers = pgut_malloc(hdr.size); + IO_CHECK(fio_read_all(fio_stdin, *headers, hdr.size), hdr.size); + file->n_headers = (hdr.size / sizeof(BackupPageHeader2)) -1; + } + + break; + } + else if (hdr.cop == FIO_PAGE) + { + blknum = hdr.arg; + + Assert(hdr.size <= sizeof(buf)); + IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + + COMP_FILE_CRC32(true, file->crc, buf, hdr.size); + + /* lazily open backup file */ + if (!out) + out = open_local_file_rw(to_fullpath, &out_buf, STDIO_BUFSIZE); + + if (fio_fwrite(out, buf, hdr.size) != hdr.size) + { + fio_fclose(out); + *err_blknum = blknum; + return WRITE_FAILED; + } + file->write_size += hdr.size; + file->uncompressed_size += BLCKSZ; + } + else + elog(ERROR, "Remote agent returned message of unexpected type: %i", hdr.cop); + } + + if (out) + fclose(out); + pg_free(out_buf); + + return n_blocks_read; +} + +/* + * Return number of actually(!) readed blocks, attempts or + * half-readed block are not counted. + * Return values in case of error: + * FILE_MISSING + * OPEN_FAILED + * READ_ERROR + * PAGE_CORRUPTION + * WRITE_FAILED + * + * If none of the above, this function return number of blocks + * readed by remote agent. + * + * In case of DELTA mode horizonLsn must be a valid lsn, + * otherwise it should be set to InvalidXLogRecPtr. + * Взято из fio_send_pages + */ +int +fio_copy_pages(const char *to_fullpath, const char *from_fullpath, pgFile *file, + XLogRecPtr horizonLsn, int calg, int clevel, uint32 checksum_version, + bool use_pagemap, BlockNumber* err_blknum, char **errormsg) +{ + FILE *out = NULL; + char *out_buf = NULL; + struct { + fio_header hdr; + fio_send_request arg; + } req; + BlockNumber n_blocks_read = 0; + BlockNumber blknum = 0; + + /* send message with header + + 16bytes 24bytes var var + -------------------------------------------------------------- + | fio_header | fio_send_request | FILE PATH | BITMAP(if any) | + -------------------------------------------------------------- + */ + + req.hdr.cop = FIO_SEND_PAGES; + + if (use_pagemap) + { + req.hdr.size = sizeof(fio_send_request) + (*file).pagemap.bitmapsize + strlen(from_fullpath) + 1; + req.arg.bitmapsize = (*file).pagemap.bitmapsize; + + /* TODO: add optimization for the case of pagemap + * containing small number of blocks with big serial numbers: + * https://github.com/postgrespro/pg_probackup/blob/remote_page_backup/src/utils/file.c#L1211 + */ + } + else + { + req.hdr.size = sizeof(fio_send_request) + strlen(from_fullpath) + 1; + req.arg.bitmapsize = 0; + } + + req.arg.nblocks = file->size/BLCKSZ; + req.arg.segmentno = file->segno * RELSEG_SIZE; + req.arg.horizonLsn = horizonLsn; + req.arg.checksumVersion = checksum_version; + req.arg.calg = calg; + req.arg.clevel = clevel; + req.arg.path_len = strlen(from_fullpath) + 1; + + file->compress_alg = calg; /* TODO: wtf? why here? */ + +//<----- +// datapagemap_iterator_t *iter; +// BlockNumber blkno; +// iter = datapagemap_iterate(pagemap); +// while (datapagemap_next(iter, &blkno)) +// elog(INFO, "block %u", blkno); +// pg_free(iter); +//<----- + + /* send header */ + IO_CHECK(fio_write_all(fio_stdout, &req, sizeof(req)), sizeof(req)); + + /* send file path */ + IO_CHECK(fio_write_all(fio_stdout, from_fullpath, req.arg.path_len), req.arg.path_len); + + /* send pagemap if any */ + if (use_pagemap) + IO_CHECK(fio_write_all(fio_stdout, (*file).pagemap.bitmap, (*file).pagemap.bitmapsize), (*file).pagemap.bitmapsize); + + out = fio_fopen(to_fullpath, PG_BINARY_R "+", FIO_BACKUP_HOST); + if (out == NULL) + elog(ERROR, "Cannot open restore target file \"%s\": %s", to_fullpath, strerror(errno)); + + /* update file permission */ + if (fio_chmod(to_fullpath, file->mode, FIO_BACKUP_HOST) == -1) + elog(ERROR, "Cannot change mode of \"%s\": %s", to_fullpath, + strerror(errno)); + + elog(VERBOSE, "ftruncate file \"%s\" to size %lu", + to_fullpath, file->size); + if (fio_ftruncate(out, file->size) == -1) + elog(ERROR, "Cannot ftruncate file \"%s\" to size %lu: %s", + to_fullpath, file->size, strerror(errno)); + + if (!fio_is_remote_file(out)) + { + out_buf = pgut_malloc(STDIO_BUFSIZE); + setvbuf(out, out_buf, _IOFBF, STDIO_BUFSIZE); + } + + while (true) + { + fio_header hdr; + char buf[BLCKSZ + sizeof(BackupPageHeader)]; + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (interrupted) + elog(ERROR, "Interrupted during page reading"); + + if (hdr.cop == FIO_ERROR) + { + /* FILE_MISSING, OPEN_FAILED and READ_FAILED */ + if (hdr.size > 0) + { + IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + *errormsg = pgut_malloc(hdr.size); + snprintf(*errormsg, hdr.size, "%s", buf); + } + + return hdr.arg; + } + else if (hdr.cop == FIO_SEND_FILE_CORRUPTION) + { + *err_blknum = hdr.arg; + + if (hdr.size > 0) + { + IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + *errormsg = pgut_malloc(hdr.size); + snprintf(*errormsg, hdr.size, "%s", buf); + } + return PAGE_CORRUPTION; + } + else if (hdr.cop == FIO_SEND_FILE_EOF) + { + /* n_blocks_read reported by EOF */ + n_blocks_read = hdr.arg; + + /* receive headers if any */ + if (hdr.size > 0) + { + char *tmp = pgut_malloc(hdr.size); + IO_CHECK(fio_read_all(fio_stdin, tmp, hdr.size), hdr.size); + pg_free(tmp); + } + + break; + } + else if (hdr.cop == FIO_PAGE) + { + blknum = hdr.arg; + + Assert(hdr.size <= sizeof(buf)); + IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + + COMP_FILE_CRC32(true, file->crc, buf, hdr.size); + + if (fio_fseek(out, blknum * BLCKSZ) < 0) + { + elog(ERROR, "Cannot seek block %u of \"%s\": %s", + blknum, to_fullpath, strerror(errno)); + } + // должен прилетать некомпрессированный блок с заголовком + // Вставить assert? + if (fio_fwrite(out, buf + sizeof(BackupPageHeader), hdr.size - sizeof(BackupPageHeader)) != BLCKSZ) + { + fio_fclose(out); + *err_blknum = blknum; + return WRITE_FAILED; + } + file->write_size += BLCKSZ; + file->uncompressed_size += BLCKSZ; + } + else + elog(ERROR, "Remote agent returned message of unexpected type: %i", hdr.cop); + } + + if (out) + fclose(out); + pg_free(out_buf); + + return n_blocks_read; +} + +/* TODO: read file using large buffer + * Return codes: + * FIO_ERROR: + * FILE_MISSING (-1) + * OPEN_FAILED (-2) + * READ_FAILED (-3) + + * FIO_SEND_FILE_CORRUPTION + * FIO_SEND_FILE_EOF + */ +static void +fio_send_pages_impl(int out, char* buf) +{ + FILE *in = NULL; + BlockNumber blknum = 0; + int current_pos = 0; + BlockNumber n_blocks_read = 0; + PageState page_st; + char read_buffer[BLCKSZ+1]; + char in_buf[STDIO_BUFSIZE]; + fio_header hdr; + fio_send_request *req = (fio_send_request*) buf; + char *from_fullpath = (char*) buf + sizeof(fio_send_request); + bool with_pagemap = req->bitmapsize > 0 ? true : false; + /* error reporting */ + char *errormsg = NULL; + /* parse buffer */ + datapagemap_t *map = NULL; + datapagemap_iterator_t *iter = NULL; + /* page headers */ + int32 hdr_num = -1; + int32 cur_pos_out = 0; + BackupPageHeader2 *headers = NULL; + + /* open source file */ + in = fopen(from_fullpath, PG_BINARY_R); + if (!in) + { + hdr.cop = FIO_ERROR; + + /* do not send exact wording of ENOENT error message + * because it is a very common error in our case, so + * error code is enough. + */ + if (errno == ENOENT) + { + hdr.arg = FILE_MISSING; + hdr.size = 0; + } + else + { + hdr.arg = OPEN_FAILED; + errormsg = pgut_malloc(ERRMSG_MAX_LEN); + /* Construct the error message */ + snprintf(errormsg, ERRMSG_MAX_LEN, "Cannot open file \"%s\": %s", + from_fullpath, strerror(errno)); + hdr.size = strlen(errormsg) + 1; + } + + /* send header and message */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + if (errormsg) + IO_CHECK(fio_write_all(out, errormsg, hdr.size), hdr.size); + + goto cleanup; + } + + if (with_pagemap) + { + map = pgut_malloc(sizeof(datapagemap_t)); + map->bitmapsize = req->bitmapsize; + map->bitmap = (char*) buf + sizeof(fio_send_request) + req->path_len; + + /* get first block */ + iter = datapagemap_iterate(map); + datapagemap_next(iter, &blknum); + + setvbuf(in, NULL, _IONBF, BUFSIZ); + } + else + setvbuf(in, in_buf, _IOFBF, STDIO_BUFSIZE); + + /* TODO: what is this barrier for? */ + read_buffer[BLCKSZ] = 1; /* barrier */ + + while (blknum < req->nblocks) + { + int rc = 0; + size_t read_len = 0; + int retry_attempts = PAGE_READ_ATTEMPTS; + + /* TODO: handle signals on the agent */ + if (interrupted) + elog(ERROR, "Interrupted during remote page reading"); + + /* read page, check header and validate checksumms */ + for (;;) + { + /* + * Optimize stdio buffer usage, fseek only when current position + * does not match the position of requested block. + */ + if (current_pos != blknum*BLCKSZ) + { + current_pos = blknum*BLCKSZ; + if (fseek(in, current_pos, SEEK_SET) != 0) + elog(ERROR, "fseek to position %u is failed on remote file '%s': %s", + current_pos, from_fullpath, strerror(errno)); + } + + read_len = fread(read_buffer, 1, BLCKSZ, in); + + current_pos += read_len; + + /* report error */ + if (ferror(in)) + { + hdr.cop = FIO_ERROR; + hdr.arg = READ_FAILED; + + errormsg = pgut_malloc(ERRMSG_MAX_LEN); + /* Construct the error message */ + snprintf(errormsg, ERRMSG_MAX_LEN, "Cannot read block %u of '%s': %s", + blknum, from_fullpath, strerror(errno)); + hdr.size = strlen(errormsg) + 1; + + /* send header and message */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(out, errormsg, hdr.size), hdr.size); + goto cleanup; + } + + if (read_len == BLCKSZ) + { + rc = validate_one_page(read_buffer, req->segmentno + blknum, + InvalidXLogRecPtr, &page_st, + req->checksumVersion); + + /* TODO: optimize copy of zeroed page */ + if (rc == PAGE_IS_ZEROED) + break; + else if (rc == PAGE_IS_VALID) + break; + } + + if (feof(in)) + goto eof; +// else /* readed less than BLKSZ bytes, retry */ + + /* File is either has insane header or invalid checksum, + * retry. If retry attempts are exhausted, report corruption. + */ + if (--retry_attempts == 0) + { + hdr.cop = FIO_SEND_FILE_CORRUPTION; + hdr.arg = blknum; + + /* Construct the error message */ + if (rc == PAGE_HEADER_IS_INVALID) + get_header_errormsg(read_buffer, &errormsg); + else if (rc == PAGE_CHECKSUM_MISMATCH) + get_checksum_errormsg(read_buffer, &errormsg, + req->segmentno + blknum); + + /* if error message is not empty, set payload size to its length */ + hdr.size = errormsg ? strlen(errormsg) + 1 : 0; + + /* send header */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + + /* send error message if any */ + if (errormsg) + IO_CHECK(fio_write_all(out, errormsg, hdr.size), hdr.size); + + goto cleanup; + } + } + + n_blocks_read++; + + /* + * horizonLsn is not 0 only in case of delta and ptrack backup. + * As far as unsigned number are always greater or equal than zero, + * there is no sense to add more checks. + */ + if ((req->horizonLsn == InvalidXLogRecPtr) || /* full, page */ + (page_st.lsn == InvalidXLogRecPtr) || /* zeroed page */ + (req->horizonLsn > 0 && page_st.lsn > req->horizonLsn)) /* delta, ptrack */ + { + int compressed_size = 0; + char write_buffer[BLCKSZ*2]; + BackupPageHeader* bph = (BackupPageHeader*)write_buffer; + + /* compress page */ + hdr.cop = FIO_PAGE; + hdr.arg = blknum; + + compressed_size = do_compress(write_buffer + sizeof(BackupPageHeader), + sizeof(write_buffer) - sizeof(BackupPageHeader), + read_buffer, BLCKSZ, req->calg, req->clevel, + NULL); + + if (compressed_size <= 0 || compressed_size >= BLCKSZ) + { + /* Do not compress page */ + memcpy(write_buffer + sizeof(BackupPageHeader), read_buffer, BLCKSZ); + compressed_size = BLCKSZ; + } + bph->block = blknum; + bph->compressed_size = compressed_size; + + hdr.size = compressed_size + sizeof(BackupPageHeader); + + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(out, write_buffer, hdr.size), hdr.size); + + /* set page header for this file */ + hdr_num++; + if (!headers) + headers = (BackupPageHeader2 *) pgut_malloc(sizeof(BackupPageHeader2)); + else + headers = (BackupPageHeader2 *) pgut_realloc(headers, (hdr_num+1) * sizeof(BackupPageHeader2)); + + headers[hdr_num].block = blknum; + headers[hdr_num].lsn = page_st.lsn; + headers[hdr_num].checksum = page_st.checksum; + headers[hdr_num].pos = cur_pos_out; + + cur_pos_out += hdr.size; + } + + /* next block */ + if (with_pagemap) + { + /* exit if pagemap is exhausted */ + if (!datapagemap_next(iter, &blknum)) + break; + } + else + blknum++; + } + +eof: + /* We are done, send eof */ + hdr.cop = FIO_SEND_FILE_EOF; + hdr.arg = n_blocks_read; + hdr.size = 0; + + if (headers) + { + hdr.size = (hdr_num+2) * sizeof(BackupPageHeader2); + + /* add dummy header */ + headers = (BackupPageHeader2 *) pgut_realloc(headers, (hdr_num+2) * sizeof(BackupPageHeader2)); + headers[hdr_num+1].pos = cur_pos_out; + } + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (headers) + IO_CHECK(fio_write_all(out, headers, hdr.size), hdr.size); + +cleanup: + pg_free(map); + pg_free(iter); + pg_free(errormsg); + pg_free(headers); + if (in) + fclose(in); + return; +} + +/* Receive chunks of compressed data, decompress them and write to + * destination file. + * Return codes: + * FILE_MISSING (-1) + * OPEN_FAILED (-2) + * READ_FAILED (-3) + * WRITE_FAILED (-4) + * ZLIB_ERROR (-5) + * REMOTE_ERROR (-6) + */ +int +fio_send_file_gz(const char *from_fullpath, FILE* out, char **errormsg) +{ + fio_header hdr; + int exit_code = SEND_OK; + char *in_buf = pgut_malloc(CHUNK_SIZE); /* buffer for compressed data */ + char *out_buf = pgut_malloc(OUT_BUF_SIZE); /* 1MB buffer for decompressed data */ + size_t path_len = strlen(from_fullpath) + 1; + /* decompressor */ + z_stream *strm = NULL; + + hdr.cop = FIO_SEND_FILE; + hdr.size = path_len; + +// elog(VERBOSE, "Thread [%d]: Attempting to open remote compressed WAL file '%s'", +// thread_num, from_fullpath); + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, from_fullpath, path_len), path_len); + + for (;;) + { + fio_header hdr; + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (hdr.cop == FIO_SEND_FILE_EOF) + { + break; + } + else if (hdr.cop == FIO_ERROR) + { + /* handle error, reported by the agent */ + if (hdr.size > 0) + { + IO_CHECK(fio_read_all(fio_stdin, in_buf, hdr.size), hdr.size); + *errormsg = pgut_malloc(hdr.size); + snprintf(*errormsg, hdr.size, "%s", in_buf); + } + exit_code = hdr.arg; + goto cleanup; + } + else if (hdr.cop == FIO_PAGE || hdr.cop == FIO_PAGE_ZERO) + { + int rc; + unsigned size; + if (hdr.cop == FIO_PAGE) + { + Assert(hdr.size <= CHUNK_SIZE); + size = hdr.size; + IO_CHECK(fio_read_all(fio_stdin, in_buf, hdr.size), hdr.size); + } + else + { + Assert(hdr.arg <= CHUNK_SIZE); + size = hdr.arg; + memset(in_buf, 0, hdr.arg); + } + + /* We have received a chunk of compressed data, lets decompress it */ + if (strm == NULL) + { + /* Initialize decompressor */ + strm = pgut_malloc(sizeof(z_stream)); + memset(strm, 0, sizeof(z_stream)); + + /* The fields next_in, avail_in initialized before init */ + strm->next_in = (Bytef *)in_buf; + strm->avail_in = size; + + rc = inflateInit2(strm, 15 + 16); + + if (rc != Z_OK) + { + *errormsg = pgut_malloc(ERRMSG_MAX_LEN); + snprintf(*errormsg, ERRMSG_MAX_LEN, + "Failed to initialize decompression stream for file '%s': %i: %s", + from_fullpath, rc, strm->msg); + exit_code = ZLIB_ERROR; + goto cleanup; + } + } + else + { + strm->next_in = (Bytef *)in_buf; + strm->avail_in = size; + } + + strm->next_out = (Bytef *)out_buf; /* output buffer */ + strm->avail_out = OUT_BUF_SIZE; /* free space in output buffer */ + + /* + * From zlib documentation: + * The application must update next_in and avail_in when avail_in + * has dropped to zero. It must update next_out and avail_out when + * avail_out has dropped to zero. + */ + while (strm->avail_in != 0) /* while there is data in input buffer, decompress it */ + { + /* decompress until there is no data to decompress, + * or buffer with uncompressed data is full + */ + rc = inflate(strm, Z_NO_FLUSH); + if (rc == Z_STREAM_END) + /* end of stream */ + break; + else if (rc != Z_OK) + { + /* got an error */ + *errormsg = pgut_malloc(ERRMSG_MAX_LEN); + snprintf(*errormsg, ERRMSG_MAX_LEN, + "Decompression failed for file '%s': %i: %s", + from_fullpath, rc, strm->msg); + exit_code = ZLIB_ERROR; + goto cleanup; + } + + if (strm->avail_out == 0) + { + /* Output buffer is full, write it out */ + if (fwrite(out_buf, 1, OUT_BUF_SIZE, out) != OUT_BUF_SIZE) + { + exit_code = WRITE_FAILED; + goto cleanup; + } + + strm->next_out = (Bytef *)out_buf; /* output buffer */ + strm->avail_out = OUT_BUF_SIZE; + } + } + + /* write out leftovers if any */ + if (strm->avail_out != OUT_BUF_SIZE) + { + int len = OUT_BUF_SIZE - strm->avail_out; + + if (fwrite(out_buf, 1, len, out) != len) + { + exit_code = WRITE_FAILED; + goto cleanup; + } + } + } + else + elog(ERROR, "Remote agent returned message of unexpected type: %i", hdr.cop); + } + +cleanup: + if (exit_code < OPEN_FAILED) + fio_disconnect(); /* discard possible pending data in pipe */ + + if (strm) + { + inflateEnd(strm); + pg_free(strm); + } + + pg_free(in_buf); + pg_free(out_buf); + return exit_code; +} + +typedef struct send_file_state { + bool calc_crc; + uint32_t crc; + int64_t read_size; + int64_t write_size; +} send_file_state; + +/* find page border of all-zero tail */ +static size_t +find_zero_tail(char *buf, size_t len) +{ + size_t i, l; + size_t granul = sizeof(zerobuf); + + if (len == 0) + return 0; + + /* fast check for last bytes */ + l = Min(len, PAGE_ZEROSEARCH_FINE_GRANULARITY); + i = len - l; + if (memcmp(buf + i, zerobuf, l) != 0) + return len; + + /* coarse search for zero tail */ + i = (len-1) & ~(granul-1); + l = len - i; + for (;;) + { + if (memcmp(buf+i, zerobuf, l) != 0) + { + i += l; + break; + } + if (i == 0) + break; + i -= granul; + l = granul; + } + + len = i; + /* search zero tail with finer granularity */ + for (granul = sizeof(zerobuf)/2; + len > 0 && granul >= PAGE_ZEROSEARCH_FINE_GRANULARITY; + granul /= 2) + { + if (granul > l) + continue; + i = (len-1) & ~(granul-1); + l = len - i; + if (memcmp(buf+i, zerobuf, l) == 0) + len = i; + } + + return len; +} + +static void +fio_send_file_crc(send_file_state* st, char *buf, size_t len) +{ + int64_t write_size; + + if (!st->calc_crc) + return; + + write_size = st->write_size; + while (st->read_size > write_size) + { + size_t crc_len = Min(st->read_size - write_size, sizeof(zerobuf)); + COMP_FILE_CRC32(true, st->crc, zerobuf, crc_len); + write_size += crc_len; + } + + if (len > 0) + COMP_FILE_CRC32(true, st->crc, buf, len); +} + +static bool +fio_send_file_write(FILE* out, send_file_state* st, char *buf, size_t len) +{ + if (len == 0) + return true; + +#ifdef WIN32 + if (st->read_size > st->write_size && + _chsize_s(fileno(out), st->read_size) != 0) + { + elog(WARNING, "Could not change file size to %lld: %m", st->read_size); + return false; + } +#endif + if (st->read_size > st->write_size && + fseeko(out, st->read_size, SEEK_SET) != 0) + { + return false; + } + + if (fwrite(buf, 1, len, out) != len) + { + return false; + } + + st->read_size += len; + st->write_size = st->read_size; + + return true; +} + +/* Receive chunks of data and write them to destination file. + * Return codes: + * SEND_OK (0) + * FILE_MISSING (-1) + * OPEN_FAILED (-2) + * READ_FAILED (-3) + * WRITE_FAILED (-4) + * + * OPEN_FAILED and READ_FAIL should also set errormsg. + * If pgFile is not NULL then we must calculate crc and read_size for it. + */ +int +fio_send_file(const char *from_fullpath, FILE* out, bool cut_zero_tail, + pgFile *file, char **errormsg) +{ + fio_header hdr; + int exit_code = SEND_OK; + size_t path_len = strlen(from_fullpath) + 1; + char *buf = pgut_malloc(CHUNK_SIZE); /* buffer */ + send_file_state st = {false, 0, 0, 0}; + + memset(&hdr, 0, sizeof(hdr)); + + if (file) + { + st.calc_crc = true; + st.crc = file->crc; + } + + hdr.cop = FIO_SEND_FILE; + hdr.size = path_len; + +// elog(VERBOSE, "Thread [%d]: Attempting to open remote WAL file '%s'", +// thread_num, from_fullpath); + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, from_fullpath, path_len), path_len); + + for (;;) + { + /* receive data */ + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (hdr.cop == FIO_SEND_FILE_EOF) + { + if (st.write_size < st.read_size) + { + if (!cut_zero_tail) + { + /* + * We still need to calc crc for zero tail. + */ + fio_send_file_crc(&st, NULL, 0); + + /* + * Let's write single zero byte to the end of file to restore + * logical size. + * Well, it would be better to use ftruncate here actually, + * but then we need to change interface. + */ + st.read_size -= 1; + buf[0] = 0; + if (!fio_send_file_write(out, &st, buf, 1)) + { + exit_code = WRITE_FAILED; + break; + } + } + } + + if (file) + { + file->crc = st.crc; + file->read_size = st.read_size; + file->write_size = st.write_size; + } + break; + } + else if (hdr.cop == FIO_ERROR) + { + /* handle error, reported by the agent */ + if (hdr.size > 0) + { + IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + *errormsg = pgut_malloc(hdr.size); + snprintf(*errormsg, hdr.size, "%s", buf); + } + exit_code = hdr.arg; + break; + } + else if (hdr.cop == FIO_PAGE) + { + Assert(hdr.size <= CHUNK_SIZE); + IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + + /* We have received a chunk of data data, lets write it out */ + fio_send_file_crc(&st, buf, hdr.size); + if (!fio_send_file_write(out, &st, buf, hdr.size)) + { + exit_code = WRITE_FAILED; + break; + } + } + else if (hdr.cop == FIO_PAGE_ZERO) + { + Assert(hdr.size == 0); + Assert(hdr.arg <= CHUNK_SIZE); + + /* + * We have received a chunk of zero data, lets just think we + * wrote it. + */ + st.read_size += hdr.arg; + } + else + { + /* TODO: fio_disconnect may get assert fail when running after this */ + elog(ERROR, "Remote agent returned message of unexpected type: %i", hdr.cop); + } + } + + if (exit_code < OPEN_FAILED) + fio_disconnect(); /* discard possible pending data in pipe */ + + pg_free(buf); + return exit_code; +} + +int +fio_send_file_local(const char *from_fullpath, FILE* out, bool cut_zero_tail, + pgFile *file, char **errormsg) +{ + FILE* in; + char* buf; + size_t read_len, non_zero_len; + int exit_code = SEND_OK; + send_file_state st = {false, 0, 0, 0}; + + if (file) + { + st.calc_crc = true; + st.crc = file->crc; + } + + /* open source file for read */ + in = fopen(from_fullpath, PG_BINARY_R); + if (in == NULL) + { + /* maybe deleted, it's not error in case of backup */ + if (errno == ENOENT) + return FILE_MISSING; + + + *errormsg = psprintf("Cannot open file \"%s\": %s", from_fullpath, + strerror(errno)); + return OPEN_FAILED; + } + + /* disable stdio buffering for local input/output files to avoid triple buffering */ + setvbuf(in, NULL, _IONBF, BUFSIZ); + setvbuf(out, NULL, _IONBF, BUFSIZ); + + /* allocate 64kB buffer */ + buf = pgut_malloc(CHUNK_SIZE); + + /* copy content and calc CRC */ + for (;;) + { + read_len = fread(buf, 1, CHUNK_SIZE, in); + + if (ferror(in)) + { + *errormsg = psprintf("Cannot read from file \"%s\": %s", + from_fullpath, strerror(errno)); + exit_code = READ_FAILED; + goto cleanup; + } + + if (read_len > 0) + { + non_zero_len = find_zero_tail(buf, read_len); + /* + * It is dirty trick to silence warnings in CFS GC process: + * backup at least cfs header size bytes. + */ + if (st.read_size + non_zero_len < PAGE_ZEROSEARCH_FINE_GRANULARITY && + st.read_size + read_len > 0) + { + non_zero_len = Min(PAGE_ZEROSEARCH_FINE_GRANULARITY, + st.read_size + read_len); + non_zero_len -= st.read_size; + } + if (non_zero_len > 0) + { + fio_send_file_crc(&st, buf, non_zero_len); + if (!fio_send_file_write(out, &st, buf, non_zero_len)) + { + exit_code = WRITE_FAILED; + goto cleanup; + } + } + if (non_zero_len < read_len) + { + /* Just pretend we wrote it. */ + st.read_size += read_len - non_zero_len; + } + } + + if (feof(in)) + break; + } + + if (st.write_size < st.read_size) + { + if (!cut_zero_tail) + { + /* + * We still need to calc crc for zero tail. + */ + fio_send_file_crc(&st, NULL, 0); + + /* + * Let's write single zero byte to the end of file to restore + * logical size. + * Well, it would be better to use ftruncate here actually, + * but then we need to change interface. + */ + st.read_size -= 1; + buf[0] = 0; + if (!fio_send_file_write(out, &st, buf, 1)) + { + exit_code = WRITE_FAILED; + goto cleanup; + } + } + } + + if (file) + { + file->crc = st.crc; + file->read_size = st.read_size; + file->write_size = st.write_size; + } + +cleanup: + free(buf); + fclose(in); + return exit_code; +} + +/* Send file content + * On error we return FIO_ERROR message with following codes + * FIO_ERROR: + * FILE_MISSING (-1) + * OPEN_FAILED (-2) + * READ_FAILED (-3) + * + * FIO_PAGE + * FIO_SEND_FILE_EOF + * + */ +static void +fio_send_file_impl(int out, char const* path) +{ + FILE *fp; + fio_header hdr; + char *buf = pgut_malloc(CHUNK_SIZE); + size_t read_len = 0; + int64_t read_size = 0; + char *errormsg = NULL; + + /* open source file for read */ + /* TODO: check that file is regular file */ + fp = fopen(path, PG_BINARY_R); + if (!fp) + { + hdr.cop = FIO_ERROR; + + /* do not send exact wording of ENOENT error message + * because it is a very common error in our case, so + * error code is enough. + */ + if (errno == ENOENT) + { + hdr.arg = FILE_MISSING; + hdr.size = 0; + } + else + { + hdr.arg = OPEN_FAILED; + errormsg = pgut_malloc(ERRMSG_MAX_LEN); + /* Construct the error message */ + snprintf(errormsg, ERRMSG_MAX_LEN, "Cannot open file '%s': %s", path, strerror(errno)); + hdr.size = strlen(errormsg) + 1; + } + + /* send header and message */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + if (errormsg) + IO_CHECK(fio_write_all(out, errormsg, hdr.size), hdr.size); + + goto cleanup; + } + + /* disable stdio buffering */ + setvbuf(fp, NULL, _IONBF, BUFSIZ); + + /* copy content */ + for (;;) + { + read_len = fread(buf, 1, CHUNK_SIZE, fp); + memset(&hdr, 0, sizeof(hdr)); + + /* report error */ + if (ferror(fp)) + { + hdr.cop = FIO_ERROR; + errormsg = pgut_malloc(ERRMSG_MAX_LEN); + hdr.arg = READ_FAILED; + /* Construct the error message */ + snprintf(errormsg, ERRMSG_MAX_LEN, "Cannot read from file '%s': %s", path, strerror(errno)); + hdr.size = strlen(errormsg) + 1; + /* send header and message */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(out, errormsg, hdr.size), hdr.size); + + goto cleanup; + } + + if (read_len > 0) + { + /* send chunk */ + int64_t non_zero_len = find_zero_tail(buf, read_len); + /* + * It is dirty trick to silence warnings in CFS GC process: + * backup at least cfs header size bytes. + */ + if (read_size + non_zero_len < PAGE_ZEROSEARCH_FINE_GRANULARITY && + read_size + read_len > 0) + { + non_zero_len = Min(PAGE_ZEROSEARCH_FINE_GRANULARITY, + read_size + read_len); + non_zero_len -= read_size; + } + + if (non_zero_len > 0) + { + hdr.cop = FIO_PAGE; + hdr.size = non_zero_len; + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(out, buf, non_zero_len), non_zero_len); + } + + if (non_zero_len < read_len) + { + hdr.cop = FIO_PAGE_ZERO; + hdr.size = 0; + hdr.arg = read_len - non_zero_len; + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + } + + read_size += read_len; + } + + if (feof(fp)) + break; + } + + /* we are done, send eof */ + hdr.cop = FIO_SEND_FILE_EOF; + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + +cleanup: + if (fp) + fclose(fp); + pg_free(buf); + pg_free(errormsg); + return; +} + +/* + * Read the local file to compute its CRC. + * We cannot make decision about file decompression because + * user may ask to backup already compressed files and we should be + * obvious about it. + */ +pg_crc32 +pgFileGetCRC(const char *file_path, bool use_crc32c, bool missing_ok) +{ + FILE *fp; + pg_crc32 crc = 0; + char *buf; + size_t len = 0; + + INIT_FILE_CRC32(use_crc32c, crc); + + /* open file in binary read mode */ + fp = fopen(file_path, PG_BINARY_R); + if (fp == NULL) + { + if (errno == ENOENT) + { + if (missing_ok) + { + FIN_FILE_CRC32(use_crc32c, crc); + return crc; + } + } + + elog(ERROR, "Cannot open file \"%s\": %s", + file_path, strerror(errno)); + } + + /* disable stdio buffering */ + setvbuf(fp, NULL, _IONBF, BUFSIZ); + buf = pgut_malloc(STDIO_BUFSIZE); + + /* calc CRC of file */ + for (;;) + { + if (interrupted) + elog(ERROR, "interrupted during CRC calculation"); + + len = fread(buf, 1, STDIO_BUFSIZE, fp); + + if (ferror(fp)) + elog(ERROR, "Cannot read \"%s\": %s", file_path, strerror(errno)); + + /* update CRC */ + COMP_FILE_CRC32(use_crc32c, crc, buf, len); + + if (feof(fp)) + break; + } + + FIN_FILE_CRC32(use_crc32c, crc); + fclose(fp); + pg_free(buf); + + return crc; +} + +/* + * Read the local file to compute CRC for it extened to real_size. + */ +pg_crc32 +pgFileGetCRCTruncated(const char *file_path, bool use_crc32c, bool missing_ok) +{ + FILE *fp; + char *buf; + size_t len = 0; + size_t non_zero_len; + send_file_state st = {true, 0, 0, 0}; + + INIT_FILE_CRC32(use_crc32c, st.crc); + + /* open file in binary read mode */ + fp = fopen(file_path, PG_BINARY_R); + if (fp == NULL) + { + if (errno == ENOENT) + { + if (missing_ok) + { + FIN_FILE_CRC32(use_crc32c, st.crc); + return st.crc; + } + } + + elog(ERROR, "Cannot open file \"%s\": %s", + file_path, strerror(errno)); + } + + /* disable stdio buffering */ + setvbuf(fp, NULL, _IONBF, BUFSIZ); + buf = pgut_malloc(CHUNK_SIZE); + + /* calc CRC of file */ + for (;;) + { + if (interrupted) + elog(ERROR, "interrupted during CRC calculation"); + + len = fread(buf, 1, STDIO_BUFSIZE, fp); + + if (ferror(fp)) + elog(ERROR, "Cannot read \"%s\": %s", file_path, strerror(errno)); + + non_zero_len = find_zero_tail(buf, len); + /* same trick as in fio_send_file */ + if (st.read_size + non_zero_len < PAGE_ZEROSEARCH_FINE_GRANULARITY && + st.read_size + len > 0) + { + non_zero_len = Min(PAGE_ZEROSEARCH_FINE_GRANULARITY, + st.read_size + len); + non_zero_len -= st.read_size; + } + if (non_zero_len) + { + fio_send_file_crc(&st, buf, non_zero_len); + st.write_size += st.read_size + non_zero_len; + } + st.read_size += len; + + if (feof(fp)) + break; + } + + FIN_FILE_CRC32(use_crc32c, st.crc); + fclose(fp); + pg_free(buf); + + return st.crc; +} + +/* + * Read the local file to compute its CRC. + * We cannot make decision about file decompression because + * user may ask to backup already compressed files and we should be + * obvious about it. + */ +pg_crc32 +pgFileGetCRCgz(const char *file_path, bool use_crc32c, bool missing_ok) +{ + gzFile fp; + pg_crc32 crc = 0; + int len = 0; + int err; + char *buf; + + INIT_FILE_CRC32(use_crc32c, crc); + + /* open file in binary read mode */ + fp = gzopen(file_path, PG_BINARY_R); + if (fp == NULL) + { + if (errno == ENOENT) + { + if (missing_ok) + { + FIN_FILE_CRC32(use_crc32c, crc); + return crc; + } + } + + elog(ERROR, "Cannot open file \"%s\": %s", + file_path, strerror(errno)); + } + + buf = pgut_malloc(STDIO_BUFSIZE); + + /* calc CRC of file */ + for (;;) + { + if (interrupted) + elog(ERROR, "interrupted during CRC calculation"); + + len = gzread(fp, buf, STDIO_BUFSIZE); + + if (len <= 0) + { + /* we either run into eof or error */ + if (gzeof(fp)) + break; + else + { + const char *err_str = NULL; + + err_str = gzerror(fp, &err); + elog(ERROR, "Cannot read from compressed file %s", err_str); + } + } + + /* update CRC */ + COMP_FILE_CRC32(use_crc32c, crc, buf, len); + } + + FIN_FILE_CRC32(use_crc32c, crc); + gzclose(fp); + pg_free(buf); + + return crc; +} + +/* Compile the array of files located on remote machine in directory root */ +static void +fio_list_dir_internal(parray *files, const char *root, bool exclude, + bool follow_symlink, bool add_root, bool backup_logs, + bool skip_hidden, int external_dir_num) +{ + fio_header hdr; + fio_list_dir_request req; + char *buf = pgut_malloc(CHUNK_SIZE); + + /* Send to the agent message with parameters for directory listing */ + snprintf(req.path, MAXPGPATH, "%s", root); + req.exclude = exclude; + req.follow_symlink = follow_symlink; + req.add_root = add_root; + req.backup_logs = backup_logs; + req.exclusive_backup = exclusive_backup; + req.skip_hidden = skip_hidden; + req.external_dir_num = external_dir_num; + + hdr.cop = FIO_LIST_DIR; + hdr.size = sizeof(req); + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, &req, hdr.size), hdr.size); + + for (;;) + { + /* receive data */ + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (hdr.cop == FIO_SEND_FILE_EOF) { - gz->strm.next_out = gz->buf; - rc = deflate(&gz->strm, Z_FINISH); - Assert(rc == Z_STREAM_END && gz->strm.avail_out != ZLIB_BUFFER_SIZE); - deflateEnd(&gz->strm); - rc = fio_write(gz->fd, gz->buf, ZLIB_BUFFER_SIZE - gz->strm.avail_out); - if (rc != ZLIB_BUFFER_SIZE - gz->strm.avail_out) + /* the work is done */ + break; + } + else if (hdr.cop == FIO_SEND_FILE) + { + pgFile *file = NULL; + fio_pgFile fio_file; + + /* receive rel_path */ + IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); + file = pgFileInit(buf); + + /* receive metainformation */ + IO_CHECK(fio_read_all(fio_stdin, &fio_file, sizeof(fio_file)), sizeof(fio_file)); + + file->mode = fio_file.mode; + file->size = fio_file.size; + file->mtime = fio_file.mtime; + file->is_datafile = fio_file.is_datafile; + file->tblspcOid = fio_file.tblspcOid; + file->dbOid = fio_file.dbOid; + file->relOid = fio_file.relOid; + file->forkName = fio_file.forkName; + file->segno = fio_file.segno; + file->external_dir_num = fio_file.external_dir_num; + + if (fio_file.linked_len > 0) { - return -1; + IO_CHECK(fio_read_all(fio_stdin, buf, fio_file.linked_len), fio_file.linked_len); + + file->linked = pgut_malloc(fio_file.linked_len); + snprintf(file->linked, fio_file.linked_len, "%s", buf); } + +// elog(INFO, "Received file: %s, mode: %u, size: %lu, mtime: %lu", +// file->rel_path, file->mode, file->size, file->mtime); + + parray_append(files, file); } else { - inflateEnd(&gz->strm); + /* TODO: fio_disconnect may get assert fail when running after this */ + elog(ERROR, "Remote agent returned message of unexpected type: %i", hdr.cop); } - rc = fio_close(gz->fd); - free(gz); - return rc; - } - else - { - return gzclose(f); } + + pg_free(buf); } -int fio_gzeof(gzFile f) + +/* + * To get the arrays of files we use the same function dir_list_file(), + * that is used for local backup. + * After that we iterate over arrays and for every file send at least + * two messages to main process: + * 1. rel_path + * 2. metainformation (size, mtime, etc) + * 3. link path (optional) + * + * TODO: replace FIO_SEND_FILE and FIO_SEND_FILE_EOF with dedicated messages + */ +static void +fio_list_dir_impl(int out, char* buf) { - if ((size_t)f & FIO_GZ_REMOTE_MARKER) + int i; + fio_header hdr; + fio_list_dir_request *req = (fio_list_dir_request*) buf; + parray *file_files = parray_new(); + + /* + * Disable logging into console any messages with exception of ERROR messages, + * because currently we have no mechanism to notify the main process + * about then message been sent. + * TODO: correctly send elog messages from agent to main process. + */ + instance_config.logger.log_level_console = ERROR; + exclusive_backup = req->exclusive_backup; + + dir_list_file(file_files, req->path, req->exclude, req->follow_symlink, + req->add_root, req->backup_logs, req->skip_hidden, + req->external_dir_num, FIO_LOCAL_HOST); + + /* send information about files to the main process */ + for (i = 0; i < parray_num(file_files); i++) { - fioGZFile* gz = (fioGZFile*)((size_t)f - FIO_GZ_REMOTE_MARKER); - return gz->eof; + fio_pgFile fio_file; + pgFile *file = (pgFile *) parray_get(file_files, i); + + fio_file.mode = file->mode; + fio_file.size = file->size; + fio_file.mtime = file->mtime; + fio_file.is_datafile = file->is_datafile; + fio_file.tblspcOid = file->tblspcOid; + fio_file.dbOid = file->dbOid; + fio_file.relOid = file->relOid; + fio_file.forkName = file->forkName; + fio_file.segno = file->segno; + fio_file.external_dir_num = file->external_dir_num; + + if (file->linked) + fio_file.linked_len = strlen(file->linked) + 1; + else + fio_file.linked_len = 0; + + hdr.cop = FIO_SEND_FILE; + hdr.size = strlen(file->rel_path) + 1; + + /* send rel_path first */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(out, file->rel_path, hdr.size), hdr.size); + + /* now send file metainformation */ + IO_CHECK(fio_write_all(out, &fio_file, sizeof(fio_file)), sizeof(fio_file)); + + /* If file is a symlink, then send link path */ + if (file->linked) + IO_CHECK(fio_write_all(out, file->linked, fio_file.linked_len), fio_file.linked_len); + + pgFileFree(file); } + + parray_free(file_files); + hdr.cop = FIO_SEND_FILE_EOF; + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); +} + +/* Wrapper for directory listing */ +void +fio_list_dir(parray *files, const char *root, bool exclude, + bool follow_symlink, bool add_root, bool backup_logs, + bool skip_hidden, int external_dir_num) +{ + if (fio_is_remote(FIO_DB_HOST)) + fio_list_dir_internal(files, root, exclude, follow_symlink, add_root, + backup_logs, skip_hidden, external_dir_num); else - { - return gzeof(f); - } + dir_list_file(files, root, exclude, follow_symlink, add_root, + backup_logs, skip_hidden, external_dir_num, FIO_LOCAL_HOST); } -const char* fio_gzerror(gzFile f, int *errnum) +PageState * +fio_get_checksum_map(const char *fullpath, uint32 checksum_version, int n_blocks, + XLogRecPtr dest_stop_lsn, BlockNumber segmentno, fio_location location) { - if ((size_t)f & FIO_GZ_REMOTE_MARKER) + if (fio_is_remote(location)) { - fioGZFile* gz = (fioGZFile*)((size_t)f - FIO_GZ_REMOTE_MARKER); - if (errnum) - *errnum = gz->errnum; - return gz->strm.msg; + fio_header hdr; + fio_checksum_map_request req_hdr; + PageState *checksum_map = NULL; + size_t path_len = strlen(fullpath) + 1; + + req_hdr.n_blocks = n_blocks; + req_hdr.segmentno = segmentno; + req_hdr.stop_lsn = dest_stop_lsn; + req_hdr.checksumVersion = checksum_version; + + hdr.cop = FIO_GET_CHECKSUM_MAP; + hdr.size = sizeof(req_hdr) + path_len; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, &req_hdr, sizeof(req_hdr)), sizeof(req_hdr)); + IO_CHECK(fio_write_all(fio_stdout, fullpath, path_len), path_len); + + /* receive data */ + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (hdr.size > 0) + { + checksum_map = pgut_malloc(n_blocks * sizeof(PageState)); + memset(checksum_map, 0, n_blocks * sizeof(PageState)); + IO_CHECK(fio_read_all(fio_stdin, checksum_map, hdr.size * sizeof(PageState)), hdr.size * sizeof(PageState)); + } + + return checksum_map; } else { - return gzerror(f, errnum); + + return get_checksum_map(fullpath, checksum_version, + n_blocks, dest_stop_lsn, segmentno); } } -z_off_t fio_gzseek(gzFile f, z_off_t offset, int whence) +static void +fio_get_checksum_map_impl(int out, char *buf) { - Assert(!((size_t)f & FIO_GZ_REMOTE_MARKER)); - return gzseek(f, offset, whence); -} + fio_header hdr; + PageState *checksum_map = NULL; + char *fullpath = (char*) buf + sizeof(fio_checksum_map_request); + fio_checksum_map_request *req = (fio_checksum_map_request*) buf; + checksum_map = get_checksum_map(fullpath, req->checksumVersion, + req->n_blocks, req->stop_lsn, req->segmentno); + hdr.size = req->n_blocks; -#endif + /* send array of PageState`s to main process */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + if (hdr.size > 0) + IO_CHECK(fio_write_all(out, checksum_map, hdr.size * sizeof(PageState)), hdr.size * sizeof(PageState)); -/* Send file content */ -static void fio_send_file(int out, char const* path) -{ - int fd = open(path, O_RDONLY); - fio_header hdr; - void* buf = NULL; + pg_free(checksum_map); +} - hdr.cop = FIO_SEND; - hdr.size = 0; +datapagemap_t * +fio_get_lsn_map(const char *fullpath, uint32 checksum_version, + int n_blocks, XLogRecPtr shift_lsn, BlockNumber segmentno, + fio_location location) +{ + datapagemap_t* lsn_map = NULL; - if (fd >= 0) + if (fio_is_remote(location)) { - off_t size = lseek(fd, 0, SEEK_END); - buf = pgut_malloc(size); - lseek(fd, 0, SEEK_SET); - IO_CHECK(fio_read_all(fd, buf, size), size); - hdr.size = size; - SYS_CHECK(close(fd)); + fio_header hdr; + fio_lsn_map_request req_hdr; + size_t path_len = strlen(fullpath) + 1; + + req_hdr.n_blocks = n_blocks; + req_hdr.segmentno = segmentno; + req_hdr.shift_lsn = shift_lsn; + req_hdr.checksumVersion = checksum_version; + + hdr.cop = FIO_GET_LSN_MAP; + hdr.size = sizeof(req_hdr) + path_len; + + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, &req_hdr, sizeof(req_hdr)), sizeof(req_hdr)); + IO_CHECK(fio_write_all(fio_stdout, fullpath, path_len), path_len); + + /* receive data */ + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + + if (hdr.size > 0) + { + lsn_map = pgut_malloc(sizeof(datapagemap_t)); + memset(lsn_map, 0, sizeof(datapagemap_t)); + + lsn_map->bitmap = pgut_malloc(hdr.size); + lsn_map->bitmapsize = hdr.size; + + IO_CHECK(fio_read_all(fio_stdin, lsn_map->bitmap, hdr.size), hdr.size); + } } - IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); - if (buf) + else { - IO_CHECK(fio_write_all(out, buf, hdr.size), hdr.size); - free(buf); + lsn_map = get_lsn_map(fullpath, checksum_version, n_blocks, + shift_lsn, segmentno); } + + return lsn_map; } -int fio_send_pages(FILE* in, FILE* out, pgFile *file, - XLogRecPtr horizonLsn, BlockNumber* nBlocksSkipped, int calg, int clevel) +static void +fio_get_lsn_map_impl(int out, char *buf) { - struct { - fio_header hdr; - fio_send_request arg; - } req; - BlockNumber n_blocks_read = 0; - BlockNumber blknum = 0; + fio_header hdr; + datapagemap_t *lsn_map = NULL; + char *fullpath = (char*) buf + sizeof(fio_lsn_map_request); + fio_lsn_map_request *req = (fio_lsn_map_request*) buf; + + lsn_map = get_lsn_map(fullpath, req->checksumVersion, req->n_blocks, + req->shift_lsn, req->segmentno); + if (lsn_map) + hdr.size = lsn_map->bitmapsize; + else + hdr.size = 0; - Assert(fio_is_remote_file(in)); + /* send bitmap to main process */ + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + if (hdr.size > 0) + IO_CHECK(fio_write_all(out, lsn_map->bitmap, hdr.size), hdr.size); - req.hdr.cop = FIO_SEND_PAGES; - req.hdr.size = sizeof(fio_send_request); - req.hdr.handle = fio_fileno(in) & ~FIO_PIPE_MARKER; + if (lsn_map) + { + pg_free(lsn_map->bitmap); + pg_free(lsn_map); + } +} - req.arg.nblocks = file->size/BLCKSZ; - req.arg.segBlockNum = file->segno * RELSEG_SIZE; - req.arg.horizonLsn = horizonLsn; - req.arg.checksumVersion = current.checksum_version; - req.arg.calg = calg; - req.arg.clevel = clevel; +/* + * Return pid of postmaster process running in given pgdata on local machine. + * Return 0 if there is none. + * Return 1 if postmaster.pid is mangled. + */ +static pid_t +local_check_postmaster(const char *pgdata) +{ + FILE *fp; + pid_t pid; + char pid_file[MAXPGPATH]; - file->compress_alg = calg; + join_path_components(pid_file, pgdata, "postmaster.pid"); - IO_CHECK(fio_write_all(fio_stdout, &req, sizeof(req)), sizeof(req)); + fp = fopen(pid_file, "r"); + if (fp == NULL) + { + /* No pid file, acceptable*/ + if (errno == ENOENT) + return 0; + else + elog(ERROR, "Cannot open file \"%s\": %s", + pid_file, strerror(errno)); + } - while (true) + if (fscanf(fp, "%i", &pid) != 1) { - fio_header hdr; - char buf[BLCKSZ + sizeof(BackupPageHeader)]; - IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); - Assert(hdr.cop == FIO_PAGE); + /* something is wrong with the file content */ + pid = 1; + } - if ((int)hdr.arg < 0) /* read error */ + if (pid > 1) + { + if (kill(pid, 0) != 0) { - return (int)hdr.arg; + /* process no longer exists */ + if (errno == ESRCH) + pid = 0; + else + elog(ERROR, "Failed to send signal 0 to a process %d: %s", + pid, strerror(errno)); } + } - blknum = hdr.arg; - if (hdr.size == 0) /* end of segment */ - break; + fclose(fp); + return pid; +} - Assert(hdr.size <= sizeof(buf)); - IO_CHECK(fio_read_all(fio_stdin, buf, hdr.size), hdr.size); +/* + * Go to the remote host and get postmaster pid from file postmaster.pid + * and check that process is running, if process is running, return its pid number. + */ +pid_t +fio_check_postmaster(const char *pgdata, fio_location location) +{ + if (fio_is_remote(location)) + { + fio_header hdr; - COMP_FILE_CRC32(true, file->crc, buf, hdr.size); + hdr.cop = FIO_CHECK_POSTMASTER; + hdr.size = strlen(pgdata) + 1; - if (fio_fwrite(out, buf, hdr.size) != hdr.size) - { - int errno_tmp = errno; - fio_fclose(out); - elog(ERROR, "File: %s, cannot write backup at block %u: %s", - file->path, blknum, strerror(errno_tmp)); - } - file->write_size += hdr.size; - n_blocks_read++; + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, pgdata, hdr.size), hdr.size); - if (((BackupPageHeader*)buf)->compressed_size == PageIsTruncated) - { - blknum += 1; - break; - } - file->read_size += BLCKSZ; + /* receive result */ + IO_CHECK(fio_read_all(fio_stdin, &hdr, sizeof(hdr)), sizeof(hdr)); + return hdr.arg; } - *nBlocksSkipped = blknum - n_blocks_read; - return blknum; + else + return local_check_postmaster(pgdata); } -static void fio_send_pages_impl(int fd, int out, fio_send_request* req) +static void +fio_check_postmaster_impl(int out, char *buf) { - BlockNumber blknum; - char read_buffer[BLCKSZ+1]; - fio_header hdr; - - hdr.cop = FIO_PAGE; - read_buffer[BLCKSZ] = 1; /* barrier */ + fio_header hdr; + pid_t postmaster_pid; + char *pgdata = (char*) buf; - for (blknum = 0; blknum < req->nblocks; blknum++) - { - int retry_attempts = PAGE_READ_ATTEMPTS; - XLogRecPtr page_lsn = InvalidXLogRecPtr; + postmaster_pid = local_check_postmaster(pgdata); - while (true) - { - ssize_t rc = pread(fd, read_buffer, BLCKSZ, blknum*BLCKSZ); + /* send arrays of checksums to main process */ + hdr.arg = postmaster_pid; + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); +} - if (rc <= 0) - { - if (rc < 0) - { - hdr.arg = -errno; - hdr.size = 0; - Assert((int)hdr.arg < 0); - IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); - } - else - { - BackupPageHeader bph; - bph.block = blknum; - bph.compressed_size = PageIsTruncated; - hdr.arg = blknum; - hdr.size = sizeof(bph); - IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); - IO_CHECK(fio_write_all(out, &bph, sizeof(bph)), sizeof(bph)); - } - return; - } - else if (rc == BLCKSZ) - { - if (!parse_page((Page)read_buffer, &page_lsn)) - { - int i; - for (i = 0; read_buffer[i] == 0; i++); +/* + * Delete file pointed by the pgFile. + * If the pgFile points directory, the directory must be empty. + */ +void +fio_delete(mode_t mode, const char *fullpath, fio_location location) +{ + if (fio_is_remote(location)) + { + fio_header hdr; - /* Page is zeroed. No need to check header and checksum. */ - if (i == BLCKSZ) - break; - } - else if (!req->checksumVersion - || pg_checksum_page(read_buffer, req->segBlockNum + blknum) == ((PageHeader)read_buffer)->pd_checksum) - { - break; - } - } + hdr.cop = FIO_DELETE; + hdr.size = strlen(fullpath) + 1; + hdr.arg = mode; - if (--retry_attempts == 0) - { - hdr.size = 0; - hdr.arg = PAGE_CHECKSUM_MISMATCH; - IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); - return; - } - } - /* horizonLsn is not 0 for delta backup. As far as unsigned number are always greater or equal than zero, there is no sense to add more checks */ - if (page_lsn >= req->horizonLsn || page_lsn == InvalidXLogRecPtr) - { - char write_buffer[BLCKSZ*2]; - BackupPageHeader* bph = (BackupPageHeader*)write_buffer; - const char *errormsg = NULL; + IO_CHECK(fio_write_all(fio_stdout, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(fio_stdout, fullpath, hdr.size), hdr.size); - hdr.arg = bph->block = blknum; - hdr.size = sizeof(BackupPageHeader); + } + else + pgFileDelete(mode, fullpath); +} - bph->compressed_size = do_compress(write_buffer + sizeof(BackupPageHeader), sizeof(write_buffer) - sizeof(BackupPageHeader), - read_buffer, BLCKSZ, req->calg, req->clevel, - &errormsg); - if (bph->compressed_size <= 0 || bph->compressed_size >= BLCKSZ) - { - /* Do not compress page */ - memcpy(write_buffer + sizeof(BackupPageHeader), read_buffer, BLCKSZ); - bph->compressed_size = BLCKSZ; - } - hdr.size += MAXALIGN(bph->compressed_size); +static void +fio_delete_impl(mode_t mode, char *buf) +{ + char *fullpath = (char*) buf; - IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); - IO_CHECK(fio_write_all(out, write_buffer, hdr.size), hdr.size); - } - } - hdr.size = 0; - hdr.arg = blknum; - IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + pgFileDelete(mode, fullpath); } /* Execute commands at remote host */ -void fio_communicate(int in, int out) +void +fio_communicate(int in, int out) { /* * Map of file and directory descriptors. @@ -1325,13 +3794,15 @@ void fio_communicate(int in, int out) fio_header hdr; struct stat st; int rc; + int tmp_fd; + pg_crc32 crc; #ifdef WIN32 SYS_CHECK(setmode(in, _O_BINARY)); SYS_CHECK(setmode(out, _O_BINARY)); #endif - /* Main loop until command of processing master command */ + /* Main loop until end of processing all master commands */ while ((rc = fio_read_all(in, &hdr, sizeof hdr)) == sizeof(hdr)) { if (hdr.size != 0) { if (hdr.size > buf_size) { @@ -1341,9 +3812,10 @@ void fio_communicate(int in, int out) } IO_CHECK(fio_read_all(in, buf, hdr.size), hdr.size); } + errno = 0; /* reset errno */ switch (hdr.cop) { case FIO_LOAD: /* Send file content */ - fio_send_file(out, buf); + fio_load_file(out, buf); break; case FIO_OPENDIR: /* Open directory for traversal */ dir[hdr.handle] = opendir(buf); @@ -1376,10 +3848,17 @@ void fio_communicate(int in, int out) IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); break; case FIO_CLOSE: /* Close file */ - SYS_CHECK(close(fd[hdr.handle])); + fio_close_impl(fd[hdr.handle], out); break; case FIO_WRITE: /* Write to the current position in file */ - IO_CHECK(fio_write_all(fd[hdr.handle], buf, hdr.size), hdr.size); +// IO_CHECK(fio_write_all(fd[hdr.handle], buf, hdr.size), hdr.size); + fio_write_impl(fd[hdr.handle], buf, hdr.size, out); + break; + case FIO_WRITE_ASYNC: /* Write to the current position in file */ + fio_write_async_impl(fd[hdr.handle], buf, hdr.size, out); + break; + case FIO_WRITE_COMPRESSED_ASYNC: /* Write to the current position in file */ + fio_write_compressed_impl(fd[hdr.handle], buf, hdr.size, hdr.arg); break; case FIO_READ: /* Read from the current position in file */ if ((size_t)hdr.arg > buf_size) { @@ -1402,12 +3881,17 @@ void fio_communicate(int in, int out) if (hdr.size != 0) IO_CHECK(fio_write_all(out, buf, hdr.size), hdr.size); break; - case FIO_FSTAT: /* Get information about opened file */ - hdr.size = sizeof(st); - hdr.arg = fstat(fd[hdr.handle], &st) < 0 ? errno : 0; - IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); - IO_CHECK(fio_write_all(out, &st, sizeof(st)), sizeof(st)); - break; + case FIO_AGENT_VERSION: + { + size_t payload_size = prepare_compatibility_str(buf, buf_size); + + hdr.arg = AGENT_PROTOCOL_VERSION; + hdr.size = payload_size; + + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + IO_CHECK(fio_write_all(out, buf, payload_size), payload_size); + break; + } case FIO_STAT: /* Get information about file with specified path */ hdr.size = sizeof(st); rc = hdr.arg ? stat(buf, &st) : lstat(buf, &st); @@ -1424,28 +3908,106 @@ void fio_communicate(int in, int out) SYS_CHECK(rename(buf, buf + strlen(buf) + 1)); break; case FIO_SYMLINK: /* Create symbolic link */ - SYS_CHECK(symlink(buf, buf + strlen(buf) + 1)); + fio_symlink_impl(out, buf, hdr.arg > 0 ? true : false); break; case FIO_UNLINK: /* Remove file or directory (TODO: Win32) */ SYS_CHECK(remove_file_or_dir(buf)); break; case FIO_MKDIR: /* Create directory */ hdr.size = 0; - hdr.arg = dir_create_dir(buf, hdr.arg); + hdr.arg = dir_create_dir(buf, hdr.arg, false); IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); break; case FIO_CHMOD: /* Change file mode */ SYS_CHECK(chmod(buf, hdr.arg)); break; case FIO_SEEK: /* Set current position in file */ - SYS_CHECK(lseek(fd[hdr.handle], hdr.arg, SEEK_SET)); + fio_seek_impl(fd[hdr.handle], hdr.arg); break; case FIO_TRUNCATE: /* Truncate file */ SYS_CHECK(ftruncate(fd[hdr.handle], hdr.arg)); break; + case FIO_LIST_DIR: + fio_list_dir_impl(out, buf); + break; case FIO_SEND_PAGES: - Assert(hdr.size == sizeof(fio_send_request)); - fio_send_pages_impl(fd[hdr.handle], out, (fio_send_request*)buf); + // buf contain fio_send_request header and bitmap. + fio_send_pages_impl(out, buf); + break; + case FIO_SEND_FILE: + fio_send_file_impl(out, buf); + break; + case FIO_SYNC: + /* open file and fsync it */ + tmp_fd = open(buf, O_WRONLY | PG_BINARY, FILE_PERMISSIONS); + if (tmp_fd < 0) + hdr.arg = errno; + else + { + if (fsync(tmp_fd) == 0) + hdr.arg = 0; + else + hdr.arg = errno; + } + close(tmp_fd); + + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + break; + case FIO_GET_CRC32: + Assert((hdr.arg & GET_CRC32_TRUNCATED) == 0 || + (hdr.arg & (GET_CRC32_TRUNCATED|GET_CRC32_DECOMPRESS)) == GET_CRC32_TRUNCATED); + /* calculate crc32 for a file */ + if ((hdr.arg & GET_CRC32_DECOMPRESS)) + crc = pgFileGetCRCgz(buf, true, (hdr.arg & GET_CRC32_MISSING_OK) != 0); + else if ((hdr.arg & GET_CRC32_TRUNCATED)) + crc = pgFileGetCRCTruncated(buf, true, (hdr.arg & GET_CRC32_MISSING_OK) != 0); + else + crc = pgFileGetCRC(buf, true, (hdr.arg & GET_CRC32_MISSING_OK) != 0); + IO_CHECK(fio_write_all(out, &crc, sizeof(crc)), sizeof(crc)); + break; + case FIO_GET_CHECKSUM_MAP: + /* calculate crc32 for a file */ + fio_get_checksum_map_impl(out, buf); + break; + case FIO_GET_LSN_MAP: + /* calculate crc32 for a file */ + fio_get_lsn_map_impl(out, buf); + break; + case FIO_CHECK_POSTMASTER: + /* calculate crc32 for a file */ + fio_check_postmaster_impl(out, buf); + break; + case FIO_DELETE: + /* delete file */ + fio_delete_impl(hdr.arg, buf); + break; + case FIO_DISCONNECT: + hdr.cop = FIO_DISCONNECTED; + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + free(buf); + return; + case FIO_GET_ASYNC_ERROR: + fio_get_async_error_impl(out); + break; + case FIO_READLINK: /* Read content of a symbolic link */ + { + /* + * We need a buf for a arguments and for a result at the same time + * hdr.size = strlen(symlink_name) + 1 + * hdr.arg = bufsize for a answer (symlink content) + */ + size_t filename_size = (size_t)hdr.size; + if (filename_size + hdr.arg > buf_size) { + buf_size = hdr.arg; + buf = (char*)realloc(buf, buf_size); + } + rc = readlink(buf, buf + filename_size, hdr.arg); + hdr.cop = FIO_READLINK; + hdr.size = rc > 0 ? rc : 0; + IO_CHECK(fio_write_all(out, &hdr, sizeof(hdr)), sizeof(hdr)); + if (hdr.size != 0) + IO_CHECK(fio_write_all(out, buf + filename_size, hdr.size), hdr.size); + } break; default: Assert(false); @@ -1457,4 +4019,3 @@ void fio_communicate(int in, int out) exit(EXIT_FAILURE); } } - diff --git a/src/utils/file.h b/src/utils/file.h index bb6101015..01e5a24f4 100644 --- a/src/utils/file.h +++ b/src/utils/file.h @@ -12,9 +12,12 @@ typedef enum { + /* message for compatibility check */ + FIO_AGENT_VERSION, /* never move this */ FIO_OPEN, FIO_CLOSE, FIO_WRITE, + FIO_SYNC, FIO_RENAME, FIO_SYMLINK, FIO_UNLINK, @@ -22,18 +25,39 @@ typedef enum FIO_CHMOD, FIO_SEEK, FIO_TRUNCATE, + FIO_DELETE, FIO_PREAD, FIO_READ, FIO_LOAD, FIO_STAT, - FIO_FSTAT, FIO_SEND, FIO_ACCESS, FIO_OPENDIR, FIO_READDIR, FIO_CLOSEDIR, + FIO_PAGE, + FIO_WRITE_COMPRESSED_ASYNC, + FIO_GET_CRC32, + /* used for incremental restore */ + FIO_GET_CHECKSUM_MAP, + FIO_GET_LSN_MAP, + /* used in fio_send_pages */ FIO_SEND_PAGES, - FIO_PAGE + FIO_ERROR, + FIO_SEND_FILE, +// FIO_CHUNK, + FIO_SEND_FILE_EOF, + FIO_SEND_FILE_CORRUPTION, + FIO_SEND_FILE_HEADERS, + /* messages for closing connection */ + FIO_DISCONNECT, + FIO_DISCONNECTED, + FIO_LIST_DIR, + FIO_CHECK_POSTMASTER, + FIO_GET_ASYNC_ERROR, + FIO_WRITE_ASYNC, + FIO_READLINK, + FIO_PAGE_ZERO } fio_operations; typedef enum @@ -46,16 +70,17 @@ typedef enum #define FIO_FDMAX 64 #define FIO_PIPE_MARKER 0x40000000 -#define PAGE_CHECKSUM_MISMATCH (-256) #define SYS_CHECK(cmd) do if ((cmd) < 0) { fprintf(stderr, "%s:%d: (%s) %s\n", __FILE__, __LINE__, #cmd, strerror(errno)); exit(EXIT_FAILURE); } while (0) -#define IO_CHECK(cmd, size) do { int _rc = (cmd); if (_rc != (size)) { if (remote_agent) { fprintf(stderr, "%s:%d: proceeds %d bytes instead of %d: %s\n", __FILE__, __LINE__, _rc, (int)(size), _rc >= 0 ? "end of data" : strerror(errno)); exit(EXIT_FAILURE); } else elog(ERROR, "Communication error: %s", _rc >= 0 ? "end of data" : strerror(errno)); } } while (0) +#define IO_CHECK(cmd, size) do { int _rc = (cmd); if (_rc != (size)) fio_error(_rc, size, __FILE__, __LINE__); } while (0) typedef struct { - unsigned cop : 5; - unsigned handle : 7; - unsigned size : 20; +// fio_operations cop; +// 16 + unsigned cop : 32; + unsigned handle : 32; + unsigned size : 32; unsigned arg; } fio_header; @@ -64,11 +89,17 @@ extern fio_location MyLocation; /* Check if FILE handle is local or remote (created by FIO) */ #define fio_is_remote_file(file) ((size_t)(file) <= FIO_FDMAX) -extern void fio_redirect(int in, int out); +extern void fio_redirect(int in, int out, int err); extern void fio_communicate(int in, int out); +extern void fio_get_agent_version(int* protocol, char* payload_buf, size_t payload_buf_size); extern FILE* fio_fopen(char const* name, char const* mode, fio_location location); extern size_t fio_fwrite(FILE* f, void const* buf, size_t size); +extern ssize_t fio_fwrite_async_compressed(FILE* f, void const* buf, size_t size, int compress_alg); +extern size_t fio_fwrite_async(FILE* f, void const* buf, size_t size); +extern int fio_check_error_file(FILE* f, char **errmsg); +extern int fio_check_error_fd(int fd, char **errmsg); +extern int fio_check_error_fd_gz(gzFile f, char **errmsg); extern ssize_t fio_fread(FILE* f, void* buf, size_t size); extern int fio_pread(FILE* f, void* buf, off_t offs); extern int fio_fprintf(FILE* f, char const* arg, ...) pg_attribute_printf(2, 3); @@ -77,13 +108,11 @@ extern int fio_fseek(FILE* f, off_t offs); extern int fio_ftruncate(FILE* f, off_t size); extern int fio_fclose(FILE* f); extern int fio_ffstat(FILE* f, struct stat* st); - -struct pgFile; -extern int fio_send_pages(FILE* in, FILE* out, struct pgFile *file, XLogRecPtr horizonLsn, - BlockNumber* nBlocksSkipped, int calg, int clevel); +extern void fio_error(int rc, int size, char const* file, int line); extern int fio_open(char const* name, int mode, fio_location location); extern ssize_t fio_write(int fd, void const* buf, size_t size); +extern ssize_t fio_write_async(int fd, void const* buf, size_t size); extern ssize_t fio_read(int fd, void* buf, size_t size); extern int fio_flush(int fd); extern int fio_seek(int fd, off_t offs); @@ -91,14 +120,21 @@ extern int fio_fstat(int fd, struct stat* st); extern int fio_truncate(int fd, off_t size); extern int fio_close(int fd); extern void fio_disconnect(void); +extern int fio_sync(char const* path, fio_location location); +extern pg_crc32 fio_get_crc32(const char *file_path, fio_location location, + bool decompress, bool missing_ok); +extern pg_crc32 fio_get_crc32_truncated(const char *file_path, fio_location location, + bool missing_ok); extern int fio_rename(char const* old_path, char const* new_path, fio_location location); -extern int fio_symlink(char const* target, char const* link_path, fio_location location); +extern int fio_symlink(char const* target, char const* link_path, bool overwrite, fio_location location); extern int fio_unlink(char const* path, fio_location location); extern int fio_mkdir(char const* path, int mode, fio_location location); extern int fio_chmod(char const* path, int mode, fio_location location); extern int fio_access(char const* path, int mode, fio_location location); extern int fio_stat(char const* path, struct stat* st, bool follow_symlinks, fio_location location); +extern bool fio_is_same_file(char const* filename1, char const* filename2, bool follow_symlink, fio_location location); +extern ssize_t fio_readlink(const char *path, char *value, size_t valsiz, fio_location location); extern DIR* fio_opendir(char const* path, fio_location location); extern struct dirent * fio_readdir(DIR *dirp); extern int fio_closedir(DIR *dirp); @@ -116,4 +152,3 @@ extern const char* fio_gzerror(gzFile file, int *errnum); #endif #endif - diff --git a/src/utils/json.c b/src/utils/json.c index 9f13a958f..2c8e0fe9b 100644 --- a/src/utils/json.c +++ b/src/utils/json.c @@ -144,3 +144,21 @@ json_add_escaped(PQExpBuffer buf, const char *str) } appendPQExpBufferChar(buf, '"'); } + +void +json_add_min(PQExpBuffer buf, JsonToken type) +{ + switch (type) + { + case JT_BEGIN_OBJECT: + appendPQExpBufferChar(buf, '{'); + add_comma = false; + break; + case JT_END_OBJECT: + appendPQExpBufferStr(buf, "}\n"); + add_comma = true; + break; + default: + break; + } +} diff --git a/src/utils/json.h b/src/utils/json.h index cc9f1168d..f80832e69 100644 --- a/src/utils/json.h +++ b/src/utils/json.h @@ -25,6 +25,7 @@ typedef enum } JsonToken; extern void json_add(PQExpBuffer buf, JsonToken type, int32 *level); +extern void json_add_min(PQExpBuffer buf, JsonToken type); extern void json_add_key(PQExpBuffer buf, const char *name, int32 level); extern void json_add_value(PQExpBuffer buf, const char *name, const char *value, int32 level, bool escaped); diff --git a/src/utils/logger.c b/src/utils/logger.c index 0e0cba29d..7ea41f74e 100644 --- a/src/utils/logger.c +++ b/src/utils/logger.c @@ -19,14 +19,19 @@ #include "utils/configuration.h" +#include "json.h" + /* Logger parameters */ LoggerConfig logger_config = { LOG_LEVEL_CONSOLE_DEFAULT, LOG_LEVEL_FILE_DEFAULT, LOG_FILENAME_DEFAULT, NULL, + NULL, LOG_ROTATION_SIZE_DEFAULT, - LOG_ROTATION_AGE_DEFAULT + LOG_ROTATION_AGE_DEFAULT, + LOG_FORMAT_CONSOLE_DEFAULT, + LOG_FORMAT_FILE_DEFAULT }; /* Implementation for logging.h */ @@ -39,6 +44,10 @@ typedef enum PG_FATAL } eLogType; +#ifndef ENABLE_VIRTUAL_TERMINAL_PROCESSING +#define ENABLE_VIRTUAL_TERMINAL_PROCESSING 0x0004 +#endif + void pg_log(eLogType type, const char *fmt,...) pg_attribute_printf(2, 3); static void elog_internal(int elevel, bool file_only, const char *message); @@ -48,7 +57,7 @@ static char *get_log_message(const char *fmt, va_list args) pg_attribute_printf( /* Functions to work with log files */ static void open_logfile(FILE **file, const char *filename_format); -static void release_logfile(void); +static void release_logfile(bool fatal, void *userdata); static char *logfile_getname(const char *format, time_t timestamp); static FILE *logfile_open(const char *filename, const char *mode); @@ -88,6 +97,110 @@ init_logger(const char *root_path, LoggerConfig *config) canonicalize_path(config->log_directory); logger_config = *config; + +#if PG_VERSION_NUM >= 120000 + /* Setup logging for functions from other modules called by pg_probackup */ + pg_logging_init(PROGRAM_NAME); + errno = 0; /* sometimes pg_logging_init sets errno */ + + switch (logger_config.log_level_console) + { + case VERBOSE: + pg_logging_set_level(PG_LOG_DEBUG); + break; + case INFO: + case NOTICE: + case LOG: + pg_logging_set_level(PG_LOG_INFO); + break; + case WARNING: + pg_logging_set_level(PG_LOG_WARNING); + break; + case ERROR: + pg_logging_set_level(PG_LOG_ERROR); + break; + default: + break; + }; +#endif +} + +/* + * Check that we are connected to terminal and + * enable ANSI escape codes for Windows if possible + */ +void +init_console(void) +{ + + /* no point in tex coloring if we do not connected to terminal */ + if (!isatty(fileno(stderr)) || + !isatty(fileno(stdout))) + { + show_color = false; + return; + } + +#ifdef WIN32 + HANDLE hOut = INVALID_HANDLE_VALUE; + HANDLE hErr = INVALID_HANDLE_VALUE; + DWORD dwMode_out = 0; + DWORD dwMode_err = 0; + + hOut = GetStdHandle(STD_OUTPUT_HANDLE); + if (hOut == INVALID_HANDLE_VALUE || !hOut) + { + show_color = false; + _dosmaperr(GetLastError()); + elog(WARNING, "Failed to get terminal stdout handle: %s", strerror(errno)); + return; + } + + hErr = GetStdHandle(STD_ERROR_HANDLE); + if (hErr == INVALID_HANDLE_VALUE || !hErr) + { + show_color = false; + _dosmaperr(GetLastError()); + elog(WARNING, "Failed to get terminal stderror handle: %s", strerror(errno)); + return; + } + + if (!GetConsoleMode(hOut, &dwMode_out)) + { + show_color = false; + _dosmaperr(GetLastError()); + elog(WARNING, "Failed to get console mode for stdout: %s", strerror(errno)); + return; + } + + if (!GetConsoleMode(hErr, &dwMode_err)) + { + show_color = false; + _dosmaperr(GetLastError()); + elog(WARNING, "Failed to get console mode for stderr: %s", strerror(errno)); + return; + } + + /* Add ANSI codes support */ + dwMode_out |= ENABLE_VIRTUAL_TERMINAL_PROCESSING; + dwMode_err |= ENABLE_VIRTUAL_TERMINAL_PROCESSING; + + if (!SetConsoleMode(hOut, dwMode_out)) + { + show_color = false; + _dosmaperr(GetLastError()); + elog(WARNING, "Cannot set console mode for stdout: %s", strerror(errno)); + return; + } + + if (!SetConsoleMode(hErr, dwMode_err)) + { + show_color = false; + _dosmaperr(GetLastError()); + elog(WARNING, "Cannot set console mode for stderr: %s", strerror(errno)); + return; + } +#endif } static void @@ -119,6 +232,35 @@ write_elevel(FILE *stream, int elevel) } } +static void +write_elevel_for_json(PQExpBuffer buf, int elevel) +{ + switch (elevel) + { + case VERBOSE: + appendPQExpBufferStr(buf, "\"VERBOSE\""); + break; + case LOG: + appendPQExpBufferStr(buf, "\"LOG\""); + break; + case INFO: + appendPQExpBufferStr(buf, "\"INFO\""); + break; + case NOTICE: + appendPQExpBufferStr(buf, "\"NOTICE\""); + break; + case WARNING: + appendPQExpBufferStr(buf, "\"WARNING\""); + break; + case ERROR: + appendPQExpBufferStr(buf, "\"ERROR\""); + break; + default: + elog_stderr(ERROR, "invalid logging level: %d", elevel); + break; + } +} + /* * Exit with code if it is an error. * Check for in_cleanup flag to avoid deadlock in case of ERROR in cleanup @@ -167,6 +309,13 @@ elog_internal(int elevel, bool file_only, const char *message) write_to_stderr; time_t log_time = (time_t) time(NULL); char strfbuf[128]; + char str_pid[128]; + char str_pid_json[128]; + char str_thread_json[64]; + PQExpBufferData show_buf; + PQExpBuffer buf_json = &show_buf; + int8 format_console, + format_file; write_to_file = elevel >= logger_config.log_level_file && logger_config.log_directory @@ -174,6 +323,8 @@ elog_internal(int elevel, bool file_only, const char *message) write_to_error_log = elevel >= ERROR && logger_config.error_log_filename && logger_config.log_directory && logger_config.log_directory[0] != '\0'; write_to_stderr = elevel >= logger_config.log_level_console && !file_only; + format_console = logger_config.log_format_console; + format_file = logger_config.log_format_file; if (remote_agent) { @@ -183,10 +334,29 @@ elog_internal(int elevel, bool file_only, const char *message) pthread_lock(&log_file_mutex); loggin_in_progress = true; - if (write_to_file || write_to_error_log) + if (write_to_file || write_to_error_log || is_archive_cmd || + format_console == JSON) strftime(strfbuf, sizeof(strfbuf), "%Y-%m-%d %H:%M:%S %Z", localtime(&log_time)); + if (format_file == JSON || format_console == JSON) + { + snprintf(str_pid_json, sizeof(str_pid_json), "%d", my_pid); + snprintf(str_thread_json, sizeof(str_thread_json), "[%d-1]", my_thread_num); + + initPQExpBuffer(&show_buf); + json_add_min(buf_json, JT_BEGIN_OBJECT); + json_add_value(buf_json, "ts", strfbuf, 0, true); + json_add_value(buf_json, "pid", str_pid_json, 0, true); + json_add_key(buf_json, "level", 0); + write_elevel_for_json(buf_json, elevel); + json_add_value(buf_json, "msg", message, 0, true); + json_add_value(buf_json, "my_thread_num", str_thread_json, 0, true); + json_add_min(buf_json, JT_END_OBJECT); + } + + snprintf(str_pid, sizeof(str_pid), "[%d]:", my_pid); + /* * Write message to log file. * Do not write to file if this error was raised during write previous @@ -195,17 +365,19 @@ elog_internal(int elevel, bool file_only, const char *message) if (write_to_file) { if (log_file == NULL) + open_logfile(&log_file, logger_config.log_filename ? logger_config.log_filename : LOG_FILENAME_DEFAULT); + if (format_file == JSON) { - if (logger_config.log_filename == NULL) - open_logfile(&log_file, LOG_FILENAME_DEFAULT); - else - open_logfile(&log_file, logger_config.log_filename); + fputs(buf_json->data, log_file); } + else + { + fprintf(log_file, "%s ", strfbuf); + fprintf(log_file, "%s ", str_pid); + write_elevel(log_file, elevel); - fprintf(log_file, "%s: ", strfbuf); - write_elevel(log_file, elevel); - - fprintf(log_file, "%s\n", message); + fprintf(log_file, "%s\n", message); + } fflush(log_file); } @@ -219,10 +391,18 @@ elog_internal(int elevel, bool file_only, const char *message) if (error_log_file == NULL) open_logfile(&error_log_file, logger_config.error_log_filename); - fprintf(error_log_file, "%s: ", strfbuf); - write_elevel(error_log_file, elevel); + if (format_file == JSON) + { + fputs(buf_json->data, error_log_file); + } + else + { + fprintf(error_log_file, "%s ", strfbuf); + fprintf(error_log_file, "%s ", str_pid); + write_elevel(error_log_file, elevel); - fprintf(error_log_file, "%s\n", message); + fprintf(error_log_file, "%s\n", message); + } fflush(error_log_file); } @@ -232,9 +412,48 @@ elog_internal(int elevel, bool file_only, const char *message) */ if (write_to_stderr) { - write_elevel(stderr, elevel); + if (format_console == JSON) + { + fprintf(stderr, "%s", buf_json->data); + } + else + { + if (is_archive_cmd) + { + char str_thread[64]; + /* [Issue #213] fix pgbadger parsing */ + snprintf(str_thread, sizeof(str_thread), "[%d-1]:", my_thread_num); + + fprintf(stderr, "%s ", strfbuf); + fprintf(stderr, "%s ", str_pid); + fprintf(stderr, "%s ", str_thread); + } + else if (show_color) + { + /* color WARNING and ERROR messages */ + if (elevel == WARNING) + fprintf(stderr, "%s", TC_YELLOW_BOLD); + else if (elevel == ERROR) + fprintf(stderr, "%s", TC_RED_BOLD); + } + + write_elevel(stderr, elevel); + + /* main payload */ + fprintf(stderr, "%s", message); + + /* reset color to default */ + if (show_color && (elevel == WARNING || elevel == ERROR)) + fprintf(stderr, "%s", TC_RESET); + + fprintf(stderr, "\n"); + } + + if (format_file == JSON || format_console == JSON) + { + termPQExpBuffer(buf_json); + } - fprintf(stderr, "%s\n", message); fflush(stderr); } @@ -251,7 +470,15 @@ elog_internal(int elevel, bool file_only, const char *message) static void elog_stderr(int elevel, const char *fmt, ...) { - va_list args; + va_list args; + PQExpBufferData show_buf; + PQExpBuffer buf_json = &show_buf; + time_t log_time = (time_t) time(NULL); + char strfbuf[128]; + char str_pid[128]; + char str_thread_json[64]; + char *message; + int8 format_console; /* * Do not log message if severity level is less than log_level. @@ -262,11 +489,37 @@ elog_stderr(int elevel, const char *fmt, ...) va_start(args, fmt); - write_elevel(stderr, elevel); - vfprintf(stderr, fmt, args); - fputc('\n', stderr); - fflush(stderr); + format_console = logger_config.log_format_console; + if (format_console == JSON) + { + strftime(strfbuf, sizeof(strfbuf), "%Y-%m-%d %H:%M:%S %Z", + localtime(&log_time)); + snprintf(str_pid, sizeof(str_pid), "%d", my_pid); + snprintf(str_thread_json, sizeof(str_thread_json), "[%d-1]", my_thread_num); + + initPQExpBuffer(&show_buf); + json_add_min(buf_json, JT_BEGIN_OBJECT); + json_add_value(buf_json, "ts", strfbuf, 0, true); + json_add_value(buf_json, "pid", str_pid, 0, true); + json_add_key(buf_json, "level", 0); + write_elevel_for_json(buf_json, elevel); + message = get_log_message(fmt, args); + json_add_value(buf_json, "msg", message, 0, true); + json_add_value(buf_json, "my_thread_num", str_thread_json, 0, true); + json_add_min(buf_json, JT_END_OBJECT); + fputs(buf_json->data, stderr); + pfree(message); + termPQExpBuffer(buf_json); + } + else + { + write_elevel(stderr, elevel); + vfprintf(stderr, fmt, args); + fputc('\n', stderr); + } + + fflush(stderr); va_end(args); exit_if_necessary(elevel); @@ -435,6 +688,36 @@ parse_log_level(const char *level) return 0; } +int +parse_log_format(const char *format) +{ + const char *v = format; + size_t len; + + if (v == NULL) + { + elog(ERROR, "log-format got invalid value"); + return 0; + } + + /* Skip all spaces detected */ + while (isspace((unsigned char)*v)) + v++; + len = strlen(v); + + if (len == 0) + elog(ERROR, "log-format is empty"); + + if (pg_strncasecmp("plain", v, len) == 0) + return PLAIN; + else if (pg_strncasecmp("json", v, len) == 0) + return JSON; + + /* Log format is invalid */ + elog(ERROR, "invalid log-format \"%s\"", format); + return 0; +} + /* * Converts integer representation of log level to string. */ @@ -464,6 +747,22 @@ deparse_log_level(int level) return NULL; } +const char * +deparse_log_format(int format) +{ + switch (format) + { + case PLAIN: + return "PLAIN"; + case JSON: + return "JSON"; + default: + elog(ERROR, "invalid log-format %d", format); + } + + return NULL; +} + /* * Construct logfile name using timestamp information. * @@ -657,7 +956,7 @@ open_logfile(FILE **file, const char *filename_format) */ if (!exit_hook_registered) { - atexit(release_logfile); + pgut_atexit_push(release_logfile, NULL); exit_hook_registered = true; } } @@ -666,7 +965,7 @@ open_logfile(FILE **file, const char *filename_format) * Closes opened file. */ static void -release_logfile(void) +release_logfile(bool fatal, void *userdata) { if (log_file) { diff --git a/src/utils/logger.h b/src/utils/logger.h index 37b6ff095..adc5061e0 100644 --- a/src/utils/logger.h +++ b/src/utils/logger.h @@ -21,6 +21,9 @@ #define ERROR 1 #define LOG_OFF 10 +#define PLAIN 0 +#define JSON 1 + typedef struct LoggerConfig { int log_level_console; @@ -32,6 +35,8 @@ typedef struct LoggerConfig uint64 log_rotation_size; /* Maximum lifetime of an individual log file in minutes */ uint64 log_rotation_age; + int8 log_format_console; + int8 log_format_file; } LoggerConfig; /* Logger parameters */ @@ -43,6 +48,9 @@ extern LoggerConfig logger_config; #define LOG_LEVEL_CONSOLE_DEFAULT INFO #define LOG_LEVEL_FILE_DEFAULT LOG_OFF +#define LOG_FORMAT_CONSOLE_DEFAULT PLAIN +#define LOG_FORMAT_FILE_DEFAULT PLAIN + #define LOG_FILENAME_DEFAULT "pg_probackup.log" #define LOG_DIRECTORY_DEFAULT "log" @@ -51,8 +59,11 @@ extern void elog(int elevel, const char *fmt, ...) pg_attribute_printf(2, 3); extern void elog_file(int elevel, const char *fmt, ...) pg_attribute_printf(2, 3); extern void init_logger(const char *root_path, LoggerConfig *config); +extern void init_console(void); extern int parse_log_level(const char *level); extern const char *deparse_log_level(int level); +extern int parse_log_format(const char *format); +extern const char *deparse_log_format(int format); #endif /* LOGGER_H */ diff --git a/src/utils/parray.c b/src/utils/parray.c index 23c227b19..65377c001 100644 --- a/src/utils/parray.c +++ b/src/utils/parray.c @@ -175,7 +175,7 @@ parray_rm(parray *array, const void *key, int(*compare)(const void *, const void size_t parray_num(const parray *array) { - return array->used; + return array!= NULL ? array->used : (size_t) 0; } void @@ -197,3 +197,50 @@ parray_bsearch(parray *array, const void *key, int(*compare)(const void *, const { return bsearch(&key, array->data, array->used, sizeof(void *), compare); } + +int +parray_bsearch_index(parray *array, const void *key, int(*compare)(const void *, const void *)) +{ + void **elem = parray_bsearch(array, key, compare); + return elem != NULL ? elem - array->data : -1; +} + +/* checks that parray contains element */ +bool parray_contains(parray *array, void *elem) +{ + int i; + + for (i = 0; i < parray_num(array); i++) + { + if (parray_get(array, i) == elem) + return true; + } + return false; +} + +/* effectively remove elements that satisfy certain criterion */ +void +parray_remove_if(parray *array, criterion_fn criterion, void *args, cleanup_fn clean) { + int i = 0; + int j = 0; + + /* removing certain elements */ + while(j < parray_num(array)) { + void *value = array->data[j]; + // if the value satisfies the criterion, clean it up + if(criterion(value, args)) { + clean(value); + j++; + continue; + } + + if(i != j) + array->data[i] = array->data[j]; + + i++; + j++; + } + + /* adjust the number of used elements */ + array->used -= j - i; +} diff --git a/src/utils/parray.h b/src/utils/parray.h index 833a6961b..08846f252 100644 --- a/src/utils/parray.h +++ b/src/utils/parray.h @@ -16,6 +16,9 @@ */ typedef struct parray parray; +typedef bool (*criterion_fn)(void *value, void *args); +typedef void (*cleanup_fn)(void *ref); + extern parray *parray_new(void); extern void parray_expand(parray *array, size_t newnum); extern void parray_free(parray *array); @@ -29,7 +32,10 @@ extern bool parray_rm(parray *array, const void *key, int(*compare)(const void * extern size_t parray_num(const parray *array); extern void parray_qsort(parray *array, int(*compare)(const void *, const void *)); extern void *parray_bsearch(parray *array, const void *key, int(*compare)(const void *, const void *)); +extern int parray_bsearch_index(parray *array, const void *key, int(*compare)(const void *, const void *)); extern void parray_walk(parray *array, void (*action)(void *)); +extern bool parray_contains(parray *array, void *elem); +extern void parray_remove_if(parray *array, criterion_fn criterion, void *args, cleanup_fn clean); #endif /* PARRAY_H */ diff --git a/src/utils/pgut.c b/src/utils/pgut.c index 92a8028a2..9559fa644 100644 --- a/src/utils/pgut.c +++ b/src/utils/pgut.c @@ -3,7 +3,7 @@ * pgut.c * * Portions Copyright (c) 2009-2013, NIPPON TELEGRAPH AND TELEPHONE CORPORATION - * Portions Copyright (c) 2017-2019, Postgres Professional + * Portions Copyright (c) 2017-2021, Postgres Professional * *------------------------------------------------------------------------- */ @@ -16,6 +16,16 @@ #include "libpq/pqsignal.h" #include "pqexpbuffer.h" +#if PG_VERSION_NUM >= 140000 +#include "common/string.h" +#endif + +#if PG_VERSION_NUM >= 100000 +#include "common/connect.h" +#else +#include "fe_utils/connect.h" +#endif + #include #include "pgut.h" @@ -35,6 +45,9 @@ bool interrupted = false; bool in_cleanup = false; bool in_password = false; +/* critical section when adding disconnect callbackups */ +static pthread_mutex_t atexit_callback_disconnect_mutex = PTHREAD_MUTEX_INITIALIZER; + /* Connection routines */ static void init_cancel_handler(void); static void on_before_exec(PGconn *conn, PGcancel *thread_cancel_conn); @@ -43,8 +56,12 @@ static void on_interrupt(void); static void on_cleanup(void); static pqsigfunc oldhandler = NULL; +static char ** pgut_pgfnames(const char *path, bool strict); +static void pgut_pgfnames_cleanup(char **filenames); + void discard_response(PGconn *conn); +/* Note that atexit handlers always called on the main thread */ void pgut_init(void) { @@ -68,7 +85,16 @@ prompt_for_password(const char *username) password = NULL; } -#if PG_VERSION_NUM >= 100000 +#if PG_VERSION_NUM >= 140000 + if (username == NULL) + password = simple_prompt("Password: ", false); + else + { + char message[256]; + snprintf(message, lengthof(message), "Password for user %s: ", username); + password = simple_prompt(message , false); + } +#elif PG_VERSION_NUM >= 100000 password = (char *) pgut_malloc(sizeof(char) * 100 + 1); if (username == NULL) simple_prompt("Password: ", password, 100, false); @@ -187,8 +213,10 @@ pgut_get_conninfo_string(PGconn *conn) (option->val != NULL && option->val[0] == '\0')) continue; - /* do not print password into the file */ - if (strcmp(option->keyword, "password") == 0) + /* do not print password, passfile and options into the file */ + if (strcmp(option->keyword, "password") == 0 || + strcmp(option->keyword, "passfile") == 0 || + strcmp(option->keyword, "options") == 0) continue; if (!firstkeyword) @@ -232,8 +260,10 @@ pgut_connect(const char *host, const char *port, if (PQstatus(conn) == CONNECTION_OK) { + pthread_lock(&atexit_callback_disconnect_mutex); pgut_atexit_push(pgut_disconnect_callback, conn); - return conn; + pthread_mutex_unlock(&atexit_callback_disconnect_mutex); + break; } if (conn && PQconnectionNeedsPassword(conn) && prompt_password) @@ -255,11 +285,34 @@ pgut_connect(const char *host, const char *port, PQfinish(conn); return NULL; } + + /* + * Fix for CVE-2018-1058. This code was taken with small modification from + * src/bin/pg_basebackup/streamutil.c:GetConnection() + */ + if (dbname != NULL) + { + PGresult *res; + + res = PQexec(conn, ALWAYS_SECURE_SEARCH_PATH_SQL); + if (PQresultStatus(res) != PGRES_TUPLES_OK) + { + elog(ERROR, "could not clear search_path: %s", + PQerrorMessage(conn)); + PQclear(res); + PQfinish(conn); + return NULL; + } + PQclear(res); + } + + return conn; } PGconn * pgut_connect_replication(const char *host, const char *port, - const char *dbname, const char *username) + const char *dbname, const char *username, + bool strict) { PGconn *tmpconn; int argcount = 7; /* dbname, replication, fallback_app_name, @@ -345,7 +398,7 @@ pgut_connect_replication(const char *host, const char *port, continue; } - elog(ERROR, "could not connect to database %s: %s", + elog(strict ? ERROR : WARNING, "could not connect to database %s: %s", dbname, PQerrorMessage(tmpconn)); PQfinish(tmpconn); free(values); @@ -360,7 +413,10 @@ pgut_disconnect(PGconn *conn) { if (conn) PQfinish(conn); + + pthread_lock(&atexit_callback_disconnect_mutex); pgut_atexit_pop(pgut_disconnect_callback, conn); + pthread_mutex_unlock(&atexit_callback_disconnect_mutex); } @@ -674,7 +730,7 @@ on_before_exec(PGconn *conn, PGcancel *thread_cancel_conn) //elog(WARNING, "Handle tread_cancel_conn. on_before_exec"); old = thread_cancel_conn; - /* be sure handle_sigint doesn't use pointer while freeing */ + /* be sure handle_interrupt doesn't use pointer while freeing */ thread_cancel_conn = NULL; if (old != NULL) @@ -687,7 +743,7 @@ on_before_exec(PGconn *conn, PGcancel *thread_cancel_conn) /* Free the old one if we have one */ old = cancel_conn; - /* be sure handle_sigint doesn't use pointer while freeing */ + /* be sure handle_interrupt doesn't use pointer while freeing */ cancel_conn = NULL; if (old != NULL) @@ -723,7 +779,7 @@ on_after_exec(PGcancel *thread_cancel_conn) //elog(WARNING, "Handle tread_cancel_conn. on_after_exec"); old = thread_cancel_conn; - /* be sure handle_sigint doesn't use pointer while freeing */ + /* be sure handle_interrupt doesn't use pointer while freeing */ thread_cancel_conn = NULL; if (old != NULL) @@ -733,7 +789,7 @@ on_after_exec(PGcancel *thread_cancel_conn) { old = cancel_conn; - /* be sure handle_sigint doesn't use pointer while freeing */ + /* be sure handle_interrupt doesn't use pointer while freeing */ cancel_conn = NULL; if (old != NULL) @@ -834,9 +890,13 @@ static void call_atexit_callbacks(bool fatal) { pgut_atexit_item *item; + pgut_atexit_item *next; - for (item = pgut_atexit_stack; item; item = item->next) + for (item = pgut_atexit_stack; item; item = next) + { + next = item->next; item->callback(fatal, item->userdata); + } } static void @@ -858,6 +918,17 @@ pgut_malloc(size_t size) return ret; } +void * +pgut_malloc0(size_t size) +{ + char *ret; + + ret = pgut_malloc(size); + memset(ret, 0, size); + + return ret; +} + void * pgut_realloc(void *p, size_t size) { @@ -883,6 +954,51 @@ pgut_strdup(const char *str) return ret; } +char * +pgut_strndup(const char *str, size_t n) +{ + char *ret; + + if (str == NULL) + return NULL; + +#if _POSIX_C_SOURCE >= 200809L + if ((ret = strndup(str, n)) == NULL) + elog(ERROR, "could not duplicate string \"%s\": %s", + str, strerror(errno)); +#else /* WINDOWS doesn't have strndup() */ + if ((ret = malloc(n + 1)) == NULL) + elog(ERROR, "could not duplicate string \"%s\": %s", + str, strerror(errno)); + + memcpy(ret, str, n); + ret[n] = '\0'; +#endif + return ret; +} + +/* + * Allocates new string, that contains part of filepath string minus trailing filename string + * If trailing filename string not found, returns copy of filepath. + * Result must be free by caller. + */ +char * +pgut_str_strip_trailing_filename(const char *filepath, const char *filename) +{ + size_t fp_len = strlen(filepath); + size_t fn_len = strlen(filename); + if (strncmp(filepath + fp_len - fn_len, filename, fn_len) == 0) + return pgut_strndup(filepath, fp_len - fn_len); + else + return pgut_strndup(filepath, fp_len); +} + +void +pgut_free(void *p) +{ + free(p); +} + FILE * pgut_fopen(const char *path, const char *mode, bool missing_ok) { @@ -937,15 +1053,19 @@ wait_for_sockets(int nfds, fd_set *fds, struct timeval *timeout) #ifndef WIN32 static void -handle_sigint(SIGNAL_ARGS) +handle_interrupt(SIGNAL_ARGS) { on_interrupt(); } +/* Handle various inrerruptions in the same way */ static void init_cancel_handler(void) { - oldhandler = pqsignal(SIGINT, handle_sigint); + oldhandler = pqsignal(SIGINT, handle_interrupt); + pqsignal(SIGQUIT, handle_interrupt); + pqsignal(SIGTERM, handle_interrupt); + pqsignal(SIGPIPE, handle_interrupt); } #else /* WIN32 */ @@ -1055,3 +1175,201 @@ discard_response(PGconn *conn) PQclear(res); } while (res); } + +/* + * pgfnames + * + * return a list of the names of objects in the argument directory. Caller + * must call pgfnames_cleanup later to free the memory allocated by this + * function. + */ +char ** +pgut_pgfnames(const char *path, bool strict) +{ + DIR *dir; + struct dirent *file; + char **filenames; + int numnames = 0; + int fnsize = 200; /* enough for many small dbs */ + + dir = opendir(path); + if (dir == NULL) + { + elog(strict ? ERROR : WARNING, "could not open directory \"%s\": %m", path); + return NULL; + } + + filenames = (char **) palloc(fnsize * sizeof(char *)); + + while (errno = 0, (file = readdir(dir)) != NULL) + { + if (strcmp(file->d_name, ".") != 0 && strcmp(file->d_name, "..") != 0) + { + if (numnames + 1 >= fnsize) + { + fnsize *= 2; + filenames = (char **) repalloc(filenames, + fnsize * sizeof(char *)); + } + filenames[numnames++] = pstrdup(file->d_name); + } + } + + filenames[numnames] = NULL; + + if (errno) + { + elog(strict ? ERROR : WARNING, "could not read directory \"%s\": %m", path); + pgut_pgfnames_cleanup(filenames); + closedir(dir); + return NULL; + } + + + if (closedir(dir)) + { + elog(strict ? ERROR : WARNING, "could not close directory \"%s\": %m", path); + return NULL; + } + + return filenames; +} + +/* + * pgfnames_cleanup + * + * deallocate memory used for filenames + */ +void +pgut_pgfnames_cleanup(char **filenames) +{ + char **fn; + + for (fn = filenames; *fn; fn++) + pfree(*fn); + + pfree(filenames); +} + +/* Shamelessly stolen from commom/rmtree.c */ +bool +pgut_rmtree(const char *path, bool rmtopdir, bool strict) +{ + bool result = true; + char pathbuf[MAXPGPATH]; + char **filenames; + char **filename; + struct stat statbuf; + + /* + * we copy all the names out of the directory before we start modifying + * it. + */ + filenames = pgut_pgfnames(path, strict); + + if (filenames == NULL) + return false; + + /* now we have the names we can start removing things */ + for (filename = filenames; *filename; filename++) + { + join_path_components(pathbuf, path, *filename); + + if (lstat(pathbuf, &statbuf) != 0) + { + elog(strict ? ERROR : WARNING, "could not stat file or directory \"%s\": %m", pathbuf); + result = false; + break; + } + + if (S_ISDIR(statbuf.st_mode)) + { + /* call ourselves recursively for a directory */ + if (!pgut_rmtree(pathbuf, true, strict)) + { + result = false; + break; + } + } + else + { + if (unlink(pathbuf) != 0) + { + elog(strict ? ERROR : WARNING, "could not remove file or directory \"%s\": %m", pathbuf); + result = false; + break; + } + } + } + + if (rmtopdir) + { + if (rmdir(path) != 0) + { + elog(strict ? ERROR : WARNING, "could not remove file or directory \"%s\": %m", path); + result = false; + } + } + + pgut_pgfnames_cleanup(filenames); + + return result; +} + +/* cross-platform setenv */ +void +pgut_setenv(const char *key, const char *val) +{ +#ifdef WIN32 + char *envstr = NULL; + envstr = psprintf("%s=%s", key, val); + putenv(envstr); +#else + setenv(key, val, 1); +#endif +} + +/* stolen from unsetenv.c */ +void +pgut_unsetenv(const char *key) +{ +#ifdef WIN32 + char *envstr = NULL; + + if (getenv(key) == NULL) + return; /* no work */ + + /* + * The technique embodied here works if libc follows the Single Unix Spec + * and actually uses the storage passed to putenv() to hold the environ + * entry. When we clobber the entry in the second step we are ensuring + * that we zap the actual environ member. However, there are some libc + * implementations (notably recent BSDs) that do not obey SUS but copy the + * presented string. This method fails on such platforms. Hopefully all + * such platforms have unsetenv() and thus won't be using this hack. See: + * http://www.greenend.org.uk/rjk/2008/putenv.html + * + * Note that repeatedly setting and unsetting a var using this code will + * leak memory. + */ + + envstr = (char *) pgut_malloc(strlen(key) + 2); + if (!envstr) /* not much we can do if no memory */ + return; + + /* Override the existing setting by forcibly defining the var */ + sprintf(envstr, "%s=", key); + putenv(envstr); + + /* Now we can clobber the variable definition this way: */ + strcpy(envstr, "="); + + /* + * This last putenv cleans up if we have multiple zero-length names as a + * result of unsetting multiple things. + */ + putenv(envstr); +#else + unsetenv(key); +#endif +} diff --git a/src/utils/pgut.h b/src/utils/pgut.h index d196aad3d..1b7b7864c 100644 --- a/src/utils/pgut.h +++ b/src/utils/pgut.h @@ -3,7 +3,7 @@ * pgut.h * * Portions Copyright (c) 2009-2013, NIPPON TELEGRAPH AND TELEPHONE CORPORATION - * Portions Copyright (c) 2017-2019, Postgres Professional + * Portions Copyright (c) 2017-2021, Postgres Professional * *------------------------------------------------------------------------- */ @@ -40,8 +40,8 @@ extern char *pgut_get_conninfo_string(PGconn *conn); extern PGconn *pgut_connect(const char *host, const char *port, const char *dbname, const char *username); extern PGconn *pgut_connect_replication(const char *host, const char *port, - const char *dbname, - const char *username); + const char *dbname, const char *username, + bool strict); extern void pgut_disconnect(PGconn *conn); extern void pgut_disconnect_callback(bool fatal, void *userdata); extern PGresult *pgut_execute(PGconn* conn, const char *query, int nParams, @@ -59,10 +59,15 @@ extern int pgut_wait(int num, PGconn *connections[], struct timeval *timeout); * memory allocators */ extern void *pgut_malloc(size_t size); +extern void *pgut_malloc0(size_t size); extern void *pgut_realloc(void *p, size_t size); extern char *pgut_strdup(const char *str); +extern char *pgut_strndup(const char *str, size_t n); +extern char *pgut_str_strip_trailing_filename(const char *filepath, const char *filename); +extern void pgut_free(void *p); #define pgut_new(type) ((type *) pgut_malloc(sizeof(type))) +#define pgut_new0(type) ((type *) pgut_malloc0(sizeof(type))) #define pgut_newarray(type, n) ((type *) pgut_malloc(sizeof(type) * (n))) /* @@ -104,4 +109,20 @@ extern int sleep(unsigned int seconds); extern int usleep(unsigned int usec); #endif +#ifdef _MSC_VER +#define ARG_SIZE_HINT +#else +#define ARG_SIZE_HINT static +#endif + +static inline uint32 hash_mix32_2(uint32 a, uint32 b) +{ + b ^= (a<<7)|(a>>25); + a *= 0xdeadbeef; + b *= 0xcafeabed; + a ^= a >> 16; + b ^= b >> 15; + return a^b; +} + #endif /* PGUT_H */ diff --git a/src/utils/remote.c b/src/utils/remote.c index 36b1a4188..7ef8d3239 100644 --- a/src/utils/remote.c +++ b/src/utils/remote.c @@ -82,6 +82,11 @@ void wait_ssh(void) #endif } +/* + * On windows we launch a new pbk process via 'pg_probackup ssh ...' + * so this process would new that it should exec ssh, because + * there is no fork on Windows. + */ #ifdef WIN32 void launch_ssh(char* argv[]) { @@ -110,6 +115,8 @@ bool launch_agent(void) int ssh_argc; int outfd[2]; int infd[2]; + int errfd[2]; + int agent_version; ssh_argc = 0; #ifdef WIN32 @@ -140,6 +147,9 @@ bool launch_agent(void) ssh_argv[ssh_argc++] = "-o"; ssh_argv[ssh_argc++] = "Compression=no"; + ssh_argv[ssh_argc++] = "-o"; + ssh_argv[ssh_argc++] = "ControlMaster=no"; + ssh_argv[ssh_argc++] = "-o"; ssh_argv[ssh_argc++] = "LogLevel=error"; @@ -162,24 +172,24 @@ bool launch_agent(void) } } if (needs_quotes(instance_config.remote.path) || needs_quotes(PROGRAM_NAME_FULL)) - snprintf(cmd, sizeof(cmd), "\"%s\\%s\" agent %s", - instance_config.remote.path, probackup, PROGRAM_VERSION); + snprintf(cmd, sizeof(cmd), "\"%s\\%s\" agent", + instance_config.remote.path, probackup); else - snprintf(cmd, sizeof(cmd), "%s\\%s agent %s", - instance_config.remote.path, probackup, PROGRAM_VERSION); + snprintf(cmd, sizeof(cmd), "%s\\%s agent", + instance_config.remote.path, probackup); #else if (needs_quotes(instance_config.remote.path) || needs_quotes(PROGRAM_NAME_FULL)) - snprintf(cmd, sizeof(cmd), "\"%s/%s\" agent %s", - instance_config.remote.path, probackup, PROGRAM_VERSION); + snprintf(cmd, sizeof(cmd), "\"%s/%s\" agent", + instance_config.remote.path, probackup); else - snprintf(cmd, sizeof(cmd), "%s/%s agent %s", - instance_config.remote.path, probackup, PROGRAM_VERSION); + snprintf(cmd, sizeof(cmd), "%s/%s agent", + instance_config.remote.path, probackup); #endif } else { if (needs_quotes(PROGRAM_NAME_FULL)) - snprintf(cmd, sizeof(cmd), "\"%s\" agent %s", PROGRAM_NAME_FULL, PROGRAM_VERSION); + snprintf(cmd, sizeof(cmd), "\"%s\" agent", PROGRAM_NAME_FULL); else - snprintf(cmd, sizeof(cmd), "%s agent %s", PROGRAM_NAME_FULL, PROGRAM_VERSION); + snprintf(cmd, sizeof(cmd), "%s agent", PROGRAM_NAME_FULL); } #ifdef WIN32 @@ -195,31 +205,175 @@ bool launch_agent(void) #else SYS_CHECK(pipe(infd)); SYS_CHECK(pipe(outfd)); + SYS_CHECK(pipe(errfd)); SYS_CHECK(child_pid = fork()); if (child_pid == 0) { /* child */ SYS_CHECK(close(STDIN_FILENO)); SYS_CHECK(close(STDOUT_FILENO)); + SYS_CHECK(close(STDERR_FILENO)); SYS_CHECK(dup2(outfd[0], STDIN_FILENO)); SYS_CHECK(dup2(infd[1], STDOUT_FILENO)); + SYS_CHECK(dup2(errfd[1], STDERR_FILENO)); SYS_CHECK(close(infd[0])); SYS_CHECK(close(infd[1])); SYS_CHECK(close(outfd[0])); SYS_CHECK(close(outfd[1])); + SYS_CHECK(close(errfd[0])); + SYS_CHECK(close(errfd[1])); if (execvp(ssh_argv[0], ssh_argv) < 0) return false; } else { #endif - elog(LOG, "Spawn agent %d version %s", child_pid, PROGRAM_VERSION); + elog(LOG, "Start SSH client process, pid %d, cmd \"%s\"", child_pid, cmd); SYS_CHECK(close(infd[1])); /* These are being used by the child */ SYS_CHECK(close(outfd[0])); + SYS_CHECK(close(errfd[1])); /*atexit(kill_child);*/ - fio_redirect(infd[0], outfd[1]); /* write to stdout */ + fio_redirect(infd[0], outfd[1], errfd[0]); /* write to stdout */ + } + + + /* Make sure that remote agent has the same version, fork and other features to be binary compatible */ + { + char payload_buf[1024]; + fio_get_agent_version(&agent_version, payload_buf, sizeof payload_buf); + check_remote_agent_compatibility(agent_version, payload_buf, sizeof payload_buf); } + return true; } + +#ifdef PGPRO_EDITION +/* PGPRO 10-13 checks to be "(certified)", with exceptional case PGPRO_11 conforming to "(standard certified)" */ +static bool check_certified() +{ + return strstr(PGPRO_VERSION_STR, "(certified)") || + strstr(PGPRO_VERSION_STR, "(standard certified)"); +} +#endif + +static char* extract_pg_edition_str() +{ + static char *vanilla = "vanilla"; +#ifdef PGPRO_EDITION + static char *_1C = "1C"; + static char *std = "standard"; + static char *ent = "enterprise"; + static char *std_cert = "standard-certified"; + static char *ent_cert = "enterprise-certified"; + + if (strcmp(PGPRO_EDITION, _1C) == 0) + return vanilla; + + if (PG_VERSION_NUM < 100000) + return PGPRO_EDITION; + + /* these "certified" checks are applicable to PGPRO from 10 up to 12 versions. + * 13+ certified versions are compatible to non-certified ones */ + if (PG_VERSION_NUM < 130000 && check_certified()) + { + if (strcmp(PGPRO_EDITION, std) == 0) + return std_cert; + else if (strcmp(PGPRO_EDITION, ent) == 0) + return ent_cert; + else + Assert("Bad #define PGPRO_EDITION value" == 0); + } + + return PGPRO_EDITION; +#else + return vanilla; +#endif +} + +#define COMPATIBILITY_VAL_STR(macro) { #macro, macro, 0 } +#define COMPATIBILITY_VAL_INT(macro) { #macro, NULL, macro } + +#define COMPATIBILITY_VAL_SEPARATOR "=" +#define COMPATIBILITY_LINE_SEPARATOR "\n" + +/* + * Compose compatibility string to be sent by pg_probackup agent + * through ssh and to be verified by pg_probackup peer. + * Compatibility string contains postgres essential vars as strings + * in format "var_name" + COMPATIBILITY_VAL_SEPARATOR + "var_value" + COMPATIBILITY_LINE_SEPARATOR + */ +size_t prepare_compatibility_str(char* compatibility_buf, size_t compatibility_buf_size) +{ + typedef struct compatibility_param_tag { + const char* name; + const char* strval; + int intval; + } compatibility_param; + + compatibility_param compatibility_params[] = { + COMPATIBILITY_VAL_STR(PG_MAJORVERSION), + { "edition", extract_pg_edition_str(), 0 }, + COMPATIBILITY_VAL_INT(SIZEOF_VOID_P), + }; + + size_t result_size = 0; + int i; + *compatibility_buf = '\0'; + + for (i = 0; i < (sizeof compatibility_params / sizeof(compatibility_param)); i++) + { + if (compatibility_params[i].strval != NULL) + result_size += snprintf(compatibility_buf + result_size, compatibility_buf_size - result_size, + "%s" COMPATIBILITY_VAL_SEPARATOR "%s" COMPATIBILITY_LINE_SEPARATOR, + compatibility_params[i].name, + compatibility_params[i].strval); + else + result_size += snprintf(compatibility_buf + result_size, compatibility_buf_size - result_size, + "%s" COMPATIBILITY_VAL_SEPARATOR "%d" COMPATIBILITY_LINE_SEPARATOR, + compatibility_params[i].name, + compatibility_params[i].intval); + Assert(result_size < compatibility_buf_size); + } + return result_size + 1; +} + +/* + * Check incoming remote agent's compatibility params for equality to local ones. + */ +void check_remote_agent_compatibility(int agent_version, char *compatibility_str, size_t compatibility_str_max_size) +{ + elog(LOG, "Agent version=%d\n", agent_version); + + if (agent_version != AGENT_PROTOCOL_VERSION) + { + char agent_version_str[1024]; + sprintf(agent_version_str, "%d.%d.%d", + agent_version / 10000, + (agent_version / 100) % 100, + agent_version % 100); + + elog(ERROR, "Remote agent protocol version %s does not match local program protocol version %s, " + "consider to upgrade pg_probackup binary", + agent_version_str, AGENT_PROTOCOL_VERSION_STR); + } + + /* checking compatibility params */ + if (strnlen(compatibility_str, compatibility_str_max_size) == compatibility_str_max_size) + { + elog(ERROR, "Corrupted remote compatibility protocol: compatibility string has no terminating \\0"); + } + + elog(LOG, "Agent compatibility params:\n%s", compatibility_str); + + { + char buf[1024]; + + prepare_compatibility_str(buf, sizeof buf); + if(strcmp(compatibility_str, buf)) + { + elog(ERROR, "Incompatible remote agent params, expected:\n%s, actual:\n:%s", buf, compatibility_str); + } + } +} diff --git a/src/utils/thread.c b/src/utils/thread.c index 5ceee068d..1c469bd29 100644 --- a/src/utils/thread.c +++ b/src/utils/thread.c @@ -11,6 +11,10 @@ #include "thread.h" +/* + * Global var used to detect error condition (not signal interrupt!) in threads, + * so if one thread errored out, then others may abort + */ bool thread_interrupted = false; #ifdef WIN32 diff --git a/src/validate.c b/src/validate.c index 9ccc5ba8a..0887b2e7a 100644 --- a/src/validate.c +++ b/src/validate.c @@ -16,7 +16,7 @@ #include "utils/thread.h" static void *pgBackupValidateFiles(void *arg); -static void do_validate_instance(void); +static void do_validate_instance(InstanceState *instanceState); static bool corrupted_backup_found = false; static bool skipped_due_to_lock = false; @@ -30,6 +30,9 @@ typedef struct uint32 checksum_version; uint32 backup_version; BackupMode backup_mode; + parray *dbOid_exclude_list; + const char *external_prefix; + HeaderMap *hdr_map; /* * Return value from the thread. @@ -40,33 +43,39 @@ typedef struct /* * Validate backup files. + * TODO: partial validation. */ void -pgBackupValidate(pgBackup *backup) +pgBackupValidate(pgBackup *backup, pgRestoreParams *params) { - char base_path[MAXPGPATH]; char external_prefix[MAXPGPATH]; - char path[MAXPGPATH]; - parray *files; + parray *files = NULL; bool corrupted = false; bool validation_isok = true; /* arrays with meta info for multi threaded validate */ pthread_t *threads; validate_files_arg *threads_args; int i; +// parray *dbOid_exclude_list = NULL; - /* Check backup version */ + /* Check backup program version */ if (parse_program_version(backup->program_version) > parse_program_version(PROGRAM_VERSION)) elog(ERROR, "pg_probackup binary version is %s, but backup %s version is %s. " "pg_probackup do not guarantee to be forward compatible. " "Please upgrade pg_probackup binary.", - PROGRAM_VERSION, base36enc(backup->start_time), backup->program_version); + PROGRAM_VERSION, backup_id_of(backup), backup->program_version); + + /* Check backup server version */ + if (strcmp(backup->server_version, PG_MAJORVERSION) != 0) + elog(ERROR, "Backup %s has server version %s, but current pg_probackup binary " + "compiled with server version %s", + backup_id_of(backup), backup->server_version, PG_MAJORVERSION); if (backup->status == BACKUP_STATUS_RUNNING) { elog(WARNING, "Backup %s has status %s, change it to ERROR and skip validation", - base36enc(backup->start_time), status2str(backup->status)); - write_backup_status(backup, BACKUP_STATUS_ERROR); + backup_id_of(backup), status2str(backup->status)); + write_backup_status(backup, BACKUP_STATUS_ERROR, true); corrupted_backup_found = true; return; } @@ -75,43 +84,54 @@ pgBackupValidate(pgBackup *backup) if (backup->status != BACKUP_STATUS_OK && backup->status != BACKUP_STATUS_DONE && backup->status != BACKUP_STATUS_ORPHAN && + backup->status != BACKUP_STATUS_MERGING && backup->status != BACKUP_STATUS_CORRUPT) { elog(WARNING, "Backup %s has status %s. Skip validation.", - base36enc(backup->start_time), status2str(backup->status)); + backup_id_of(backup), status2str(backup->status)); corrupted_backup_found = true; return; } - if (backup->status == BACKUP_STATUS_OK || backup->status == BACKUP_STATUS_DONE) - elog(INFO, "Validating backup %s", base36enc(backup->start_time)); - /* backups in MERGING status must have an option of revalidation without losing MERGING status - else if (backup->status == BACKUP_STATUS_MERGING) + /* additional sanity */ + if (backup->backup_mode == BACKUP_MODE_FULL && + backup->status == BACKUP_STATUS_MERGING) { - some message here + elog(WARNING, "Full backup %s has status %s, skip validation", + backup_id_of(backup), status2str(backup->status)); + return; } - */ + + if (backup->status == BACKUP_STATUS_OK || backup->status == BACKUP_STATUS_DONE || + backup->status == BACKUP_STATUS_MERGING) + elog(INFO, "Validating backup %s", backup_id_of(backup)); else - elog(INFO, "Revalidating backup %s", base36enc(backup->start_time)); + elog(INFO, "Revalidating backup %s", backup_id_of(backup)); if (backup->backup_mode != BACKUP_MODE_FULL && backup->backup_mode != BACKUP_MODE_DIFF_PAGE && backup->backup_mode != BACKUP_MODE_DIFF_PTRACK && backup->backup_mode != BACKUP_MODE_DIFF_DELTA) - elog(WARNING, "Invalid backup_mode of backup %s", base36enc(backup->start_time)); + elog(WARNING, "Invalid backup_mode of backup %s", backup_id_of(backup)); - pgBackupGetPath(backup, base_path, lengthof(base_path), DATABASE_DIR); - pgBackupGetPath(backup, external_prefix, lengthof(external_prefix), EXTERNAL_DIR); - pgBackupGetPath(backup, path, lengthof(path), DATABASE_FILE_LIST); - files = dir_read_file_list(base_path, external_prefix, path, FIO_BACKUP_HOST); + join_path_components(external_prefix, backup->root_dir, EXTERNAL_DIR); + files = get_backup_filelist(backup, false); - /* setup threads */ - for (i = 0; i < parray_num(files); i++) + if (!files) { - pgFile *file = (pgFile *) parray_get(files, i); - pg_atomic_clear_flag(&file->lock); + elog(WARNING, "Backup %s file list is corrupted", backup_id_of(backup)); + backup->status = BACKUP_STATUS_CORRUPT; + write_backup_status(backup, BACKUP_STATUS_CORRUPT, true); + return; } +// if (params && params->partial_db_list) +// dbOid_exclude_list = get_dbOid_exclude_list(backup, files, params->partial_db_list, +// params->partial_restore_type); + + /* setup threads */ + pfilearray_clear_locks(files); + /* init thread args with own file lists */ threads = (pthread_t *) palloc(sizeof(pthread_t) * num_threads); threads_args = (validate_files_arg *) @@ -123,13 +143,16 @@ pgBackupValidate(pgBackup *backup) { validate_files_arg *arg = &(threads_args[i]); - arg->base_path = base_path; + arg->base_path = backup->database_dir; arg->files = files; arg->corrupted = false; arg->backup_mode = backup->backup_mode; arg->stop_lsn = backup->stop_lsn; arg->checksum_version = backup->checksum_version; arg->backup_version = parse_program_version(backup->program_version); + arg->external_prefix = external_prefix; + arg->hdr_map = &(backup->hdr_map); +// arg->dbOid_exclude_list = dbOid_exclude_list; /* By default there are some error */ threads_args[i].ret = 1; @@ -156,15 +179,40 @@ pgBackupValidate(pgBackup *backup) /* cleanup */ parray_walk(files, pgFileFree); parray_free(files); + cleanup_header_map(&(backup->hdr_map)); /* Update backup status */ + if (corrupted) + backup->status = BACKUP_STATUS_CORRUPT; + write_backup_status(backup, corrupted ? BACKUP_STATUS_CORRUPT : - BACKUP_STATUS_OK); + BACKUP_STATUS_OK, true); if (corrupted) - elog(WARNING, "Backup %s data files are corrupted", base36enc(backup->start_time)); + elog(WARNING, "Backup %s data files are corrupted", backup_id_of(backup)); else - elog(INFO, "Backup %s data files are valid", base36enc(backup->start_time)); + elog(INFO, "Backup %s data files are valid", backup_id_of(backup)); + + /* Issue #132 kludge */ + if (!corrupted && + ((parse_program_version(backup->program_version) == 20104)|| + (parse_program_version(backup->program_version) == 20105)|| + (parse_program_version(backup->program_version) == 20201))) + { + char path[MAXPGPATH]; + + join_path_components(path, backup->root_dir, DATABASE_FILE_LIST); + + if (pgFileSize(path) >= (BLCKSZ*500)) + { + elog(WARNING, "Backup %s is a victim of metadata corruption. " + "Additional information can be found here: " + "https://github.com/postgrespro/pg_probackup/issues/132", + backup_id_of(backup)); + backup->status = BACKUP_STATUS_CORRUPT; + write_backup_status(backup, BACKUP_STATUS_CORRUPT, true); + } + } } /* @@ -185,6 +233,7 @@ pgBackupValidateFiles(void *arg) { struct stat st; pgFile *file = (pgFile *) parray_get(arguments->files, i); + char file_fullpath[MAXPGPATH]; if (interrupted || thread_interrupted) elog(ERROR, "Interrupted during validate"); @@ -194,19 +243,24 @@ pgBackupValidateFiles(void *arg) continue; /* - * Currently we don't compute checksums for - * cfs_compressed data files, so skip them. - * TODO: investigate + * If in partial validate, check if the file belongs to the database + * we exclude. Only files from pgdata can be skipped. */ - if (file->is_cfs) - continue; + //if (arguments->dbOid_exclude_list && file->external_dir_num == 0 + // && parray_bsearch(arguments->dbOid_exclude_list, + // &file->dbOid, pgCompareOid)) + //{ + // elog(VERBOSE, "Skip file validation due to partial restore: \"%s\"", + // file->rel_path); + // continue; + //} if (!pg_atomic_test_set_flag(&file->lock)) continue; if (progress) - elog(INFO, "Progress: (%d/%d). Process file \"%s\"", - i + 1, num_files, file->path); + elog(INFO, "Progress: (%d/%d). Validate file \"%s\"", + i + 1, num_files, file->rel_path); /* * Skip files which has no data, because they @@ -214,11 +268,12 @@ pgBackupValidateFiles(void *arg) */ if (file->write_size == BYTES_INVALID) { + /* TODO: lookup corresponding merge bug */ if (arguments->backup_mode == BACKUP_MODE_FULL) { /* It is illegal for file in FULL backup to have BYTES_INVALID */ elog(WARNING, "Backup file \"%s\" has invalid size. Possible metadata corruption.", - file->path); + file->rel_path); arguments->corrupted = true; break; } @@ -226,13 +281,28 @@ pgBackupValidateFiles(void *arg) continue; } - if (stat(file->path, &st) == -1) + /* no point in trying to open empty file */ + if (file->write_size == 0) + continue; + + if (file->external_dir_num) + { + char temp[MAXPGPATH]; + + makeExternalDirPathByNum(temp, arguments->external_prefix, file->external_dir_num); + join_path_components(file_fullpath, temp, file->rel_path); + } + else + join_path_components(file_fullpath, arguments->base_path, file->rel_path); + + /* TODO: it is redundant to check file existence using stat */ + if (stat(file_fullpath, &st) == -1) { if (errno == ENOENT) - elog(WARNING, "Backup file \"%s\" is not found", file->path); + elog(WARNING, "Backup file \"%s\" is not found", file_fullpath); else elog(WARNING, "Cannot stat backup file \"%s\": %s", - file->path, strerror(errno)); + file_fullpath, strerror(errno)); arguments->corrupted = true; break; } @@ -240,7 +310,7 @@ pgBackupValidateFiles(void *arg) if (file->write_size != st.st_size) { elog(WARNING, "Invalid size of backup file \"%s\" : " INT64_FORMAT ". Expected %lu", - file->path, (unsigned long) st.st_size, file->write_size); + file_fullpath, (unsigned long) st.st_size, file->write_size); arguments->corrupted = true; break; } @@ -248,8 +318,10 @@ pgBackupValidateFiles(void *arg) /* * If option skip-block-validation is set, compute only file-level CRC for * datafiles, otherwise check them block by block. + * Currently we don't compute checksums for + * cfs_compressed data files, so skip block validation for them. */ - if (!file->is_datafile || skip_block_validation) + if (!file->is_datafile || skip_block_validation || file->is_cfs) { /* * Pre 2.0.22 we use CRC-32C, but in newer version of pg_probackup we @@ -269,14 +341,14 @@ pgBackupValidateFiles(void *arg) !file->external_dir_num) crc = get_pgcontrol_checksum(arguments->base_path); else - crc = pgFileGetCRC(file->path, + crc = pgFileGetCRC(file_fullpath, arguments->backup_version <= 20021 || arguments->backup_version >= 20025, - true, NULL, FIO_LOCAL_HOST); + false); if (crc != file->crc) { elog(WARNING, "Invalid CRC of backup file \"%s\" : %X. Expected %X", - file->path, crc, file->crc); + file_fullpath, crc, file->crc); arguments->corrupted = true; } } @@ -287,9 +359,10 @@ pgBackupValidateFiles(void *arg) * check page headers, checksums (if enabled) * and compute checksum of the file */ - if (!check_file_pages(file, arguments->stop_lsn, + if (!validate_file_pages(file, file_fullpath, arguments->stop_lsn, arguments->checksum_version, - arguments->backup_version)) + arguments->backup_version, + arguments->hdr_map)) arguments->corrupted = true; } } @@ -303,30 +376,29 @@ pgBackupValidateFiles(void *arg) /* * Validate all backups in the backup catalog. * If --instance option was provided, validate only backups of this instance. + * + * TODO: split into two functions: do_validate_catalog and do_validate_instance. */ int -do_validate_all(void) +do_validate_all(CatalogState *catalogState, InstanceState *instanceState) { corrupted_backup_found = false; skipped_due_to_lock = false; - if (instance_name == NULL) + if (instanceState == NULL) { /* Show list of instances */ - char path[MAXPGPATH]; DIR *dir; struct dirent *dent; /* open directory and list contents */ - join_path_components(path, backup_path, BACKUPS_DIR); - dir = opendir(path); + dir = opendir(catalogState->backup_subdir_path); if (dir == NULL) - elog(ERROR, "cannot open directory \"%s\": %s", path, strerror(errno)); + elog(ERROR, "Cannot open directory \"%s\": %s", catalogState->backup_subdir_path, strerror(errno)); errno = 0; while ((dent = readdir(dir))) { - char conf_path[MAXPGPATH]; char child[MAXPGPATH]; struct stat st; @@ -335,10 +407,10 @@ do_validate_all(void) strcmp(dent->d_name, "..") == 0) continue; - join_path_components(child, path, dent->d_name); + join_path_components(child, catalogState->backup_subdir_path, dent->d_name); if (lstat(child, &st) == -1) - elog(ERROR, "cannot stat file \"%s\": %s", child, strerror(errno)); + elog(ERROR, "Cannot stat file \"%s\": %s", child, strerror(errno)); if (!S_ISDIR(st.st_mode)) continue; @@ -346,26 +418,30 @@ do_validate_all(void) /* * Initialize instance configuration. */ - instance_name = dent->d_name; - sprintf(backup_instance_path, "%s/%s/%s", - backup_path, BACKUPS_DIR, instance_name); - sprintf(arclog_path, "%s/%s/%s", backup_path, "wal", instance_name); - join_path_components(conf_path, backup_instance_path, - BACKUP_CATALOG_CONF_FILE); - if (config_read_opt(conf_path, instance_options, ERROR, false, + instanceState = pgut_new(InstanceState); /* memory leak */ + strncpy(instanceState->instance_name, dent->d_name, MAXPGPATH); + + join_path_components(instanceState->instance_backup_subdir_path, + catalogState->backup_subdir_path, instanceState->instance_name); + join_path_components(instanceState->instance_wal_subdir_path, + catalogState->wal_subdir_path, instanceState->instance_name); + join_path_components(instanceState->instance_config_path, + instanceState->instance_backup_subdir_path, BACKUP_CATALOG_CONF_FILE); + + if (config_read_opt(instanceState->instance_config_path, instance_options, ERROR, false, true) == 0) { - elog(WARNING, "Configuration file \"%s\" is empty", conf_path); + elog(WARNING, "Configuration file \"%s\" is empty", instanceState->instance_config_path); corrupted_backup_found = true; continue; } - do_validate_instance(); + do_validate_instance(instanceState); } } else { - do_validate_instance(); + do_validate_instance(instanceState); } /* TODO: Probably we should have different exit code for every condition @@ -395,17 +471,17 @@ do_validate_all(void) * Validate all backups in the given instance of the backup catalog. */ static void -do_validate_instance(void) +do_validate_instance(InstanceState *instanceState) { int i; int j; parray *backups; pgBackup *current_backup = NULL; - elog(INFO, "Validate backups of the instance '%s'", instance_name); + elog(INFO, "Validate backups of the instance '%s'", instanceState->instance_name); /* Get list of all backups sorted in order of descending start time */ - backups = catalog_get_backup_list(INVALID_BACKUP_ID); + backups = catalog_get_backup_list(instanceState, INVALID_BACKUP_ID); /* Examine backups one by one and validate them */ for (i = 0; i < parray_num(backups); i++) @@ -423,54 +499,54 @@ do_validate_instance(void) result = scan_parent_chain(current_backup, &tmp_backup); /* chain is broken */ - if (result == 0) + if (result == ChainIsBroken) { - char *parent_backup_id; + const char *parent_backup_id; + const char *current_backup_id; /* determine missing backup ID */ - parent_backup_id = base36enc_dup(tmp_backup->parent_backup); + parent_backup_id = base36enc(tmp_backup->parent_backup); + current_backup_id = backup_id_of(current_backup); corrupted_backup_found = true; /* orphanize current_backup */ if (current_backup->status == BACKUP_STATUS_OK || current_backup->status == BACKUP_STATUS_DONE) { - write_backup_status(current_backup, BACKUP_STATUS_ORPHAN); + write_backup_status(current_backup, BACKUP_STATUS_ORPHAN, true); elog(WARNING, "Backup %s is orphaned because his parent %s is missing", - base36enc(current_backup->start_time), - parent_backup_id); + current_backup_id, parent_backup_id); } else { elog(WARNING, "Backup %s has missing parent %s", - base36enc(current_backup->start_time), parent_backup_id); + current_backup_id, parent_backup_id); } - pg_free(parent_backup_id); continue; } /* chain is whole, but at least one parent is invalid */ - else if (result == 1) + else if (result == ChainIsInvalid) { /* Oldest corrupt backup has a chance for revalidation */ if (current_backup->start_time != tmp_backup->start_time) { - char *backup_id = base36enc_dup(tmp_backup->start_time); /* orphanize current_backup */ if (current_backup->status == BACKUP_STATUS_OK || current_backup->status == BACKUP_STATUS_DONE) { - write_backup_status(current_backup, BACKUP_STATUS_ORPHAN); + write_backup_status(current_backup, BACKUP_STATUS_ORPHAN, true); elog(WARNING, "Backup %s is orphaned because his parent %s has status: %s", - base36enc(current_backup->start_time), backup_id, + backup_id_of(current_backup), + backup_id_of(tmp_backup), status2str(tmp_backup->status)); } else { elog(WARNING, "Backup %s has parent %s with status: %s", - base36enc(current_backup->start_time), backup_id, + backup_id_of(current_backup), + backup_id_of(tmp_backup), status2str(tmp_backup->status)); } - pg_free(backup_id); continue; } base_full_backup = find_parent_full_backup(current_backup); @@ -478,7 +554,7 @@ do_validate_instance(void) /* sanity */ if (!base_full_backup) elog(ERROR, "Parent full backup for the given backup %s was not found", - base36enc(current_backup->start_time)); + backup_id_of(current_backup)); } /* chain is whole, all parents are valid at first glance, * current backup validation can proceed @@ -490,20 +566,20 @@ do_validate_instance(void) base_full_backup = current_backup; /* Do not interrupt, validate the next backup */ - if (!lock_backup(current_backup)) + if (!lock_backup(current_backup, true, false)) { elog(WARNING, "Cannot lock backup %s directory, skip validation", - base36enc(current_backup->start_time)); + backup_id_of(current_backup)); skipped_due_to_lock = true; continue; } /* Valiate backup files*/ - pgBackupValidate(current_backup); + pgBackupValidate(current_backup, NULL); /* Validate corresponding WAL files */ if (current_backup->status == BACKUP_STATUS_OK) - validate_wal(current_backup, arclog_path, 0, - 0, 0, base_full_backup->tli, + validate_wal(current_backup, instanceState->instance_wal_subdir_path, 0, + 0, 0, current_backup->tli, instance_config.xlog_seg_size); /* @@ -511,7 +587,6 @@ do_validate_instance(void) */ if (current_backup->status != BACKUP_STATUS_OK) { - char *current_backup_id; /* This is ridiculous but legal. * PAGE_b2 <- OK * PAGE_a2 <- OK @@ -521,7 +596,6 @@ do_validate_instance(void) */ corrupted_backup_found = true; - current_backup_id = base36enc_dup(current_backup->start_time); for (j = i - 1; j >= 0; j--) { @@ -532,16 +606,15 @@ do_validate_instance(void) if (backup->status == BACKUP_STATUS_OK || backup->status == BACKUP_STATUS_DONE) { - write_backup_status(backup, BACKUP_STATUS_ORPHAN); + write_backup_status(backup, BACKUP_STATUS_ORPHAN, true); elog(WARNING, "Backup %s is orphaned because his parent %s has status: %s", - base36enc(backup->start_time), - current_backup_id, + backup_id_of(backup), + backup_id_of(current_backup), status2str(current_backup->status)); } } } - free(current_backup_id); } /* For every OK backup we try to revalidate all his ORPHAN descendants. */ @@ -574,7 +647,7 @@ do_validate_instance(void) */ result = scan_parent_chain(backup, &tmp_backup); - if (result == 1) + if (result == ChainIsInvalid) { /* revalidation make sense only if oldest invalid backup is current_backup */ @@ -585,22 +658,22 @@ do_validate_instance(void) if (backup->status == BACKUP_STATUS_ORPHAN) { /* Do not interrupt, validate the next backup */ - if (!lock_backup(backup)) + if (!lock_backup(backup, true, false)) { elog(WARNING, "Cannot lock backup %s directory, skip validation", - base36enc(backup->start_time)); + backup_id_of(backup)); skipped_due_to_lock = true; continue; } /* Revalidate backup files*/ - pgBackupValidate(backup); + pgBackupValidate(backup, NULL); if (backup->status == BACKUP_STATUS_OK) { /* Revalidation successful, validate corresponding WAL files */ - validate_wal(backup, arclog_path, 0, - 0, 0, current_backup->tli, + validate_wal(backup, instanceState->instance_wal_subdir_path, 0, + 0, 0, backup->tli, instance_config.xlog_seg_size); } } @@ -620,3 +693,60 @@ do_validate_instance(void) parray_walk(backups, pgBackupFree); parray_free(backups); } + +/* + * Validate tablespace_map checksum. + * Error out in case of checksum mismatch. + * Return 'false' if there are no tablespaces in backup. + * + * TODO: it is a bad, that we read the whole filelist just for + * the sake of tablespace_map. Probably pgBackup should come with + * already filled pgBackup.files + */ +bool +validate_tablespace_map(pgBackup *backup, bool no_validate) +{ + char map_path[MAXPGPATH]; + pgFile *dummy = NULL; + pgFile **tablespace_map = NULL; + pg_crc32 crc; + parray *files = get_backup_filelist(backup, true); + bool use_crc32c = parse_program_version(backup->program_version) <= 20021 || + parse_program_version(backup->program_version) >= 20025; + + parray_qsort(files, pgFileCompareRelPathWithExternal); + join_path_components(map_path, backup->database_dir, PG_TABLESPACE_MAP_FILE); + + dummy = pgFileInit(PG_TABLESPACE_MAP_FILE); + tablespace_map = (pgFile **) parray_bsearch(files, dummy, pgFileCompareRelPathWithExternal); + + if (!tablespace_map) + { + elog(LOG, "there is no file tablespace_map"); + parray_walk(files, pgFileFree); + parray_free(files); + return false; + } + + /* Exit if database/tablespace_map doesn't exist */ + if (!fileExists(map_path, FIO_BACKUP_HOST)) + elog(ERROR, "Tablespace map is missing: \"%s\", " + "probably backup %s is corrupt, validate it", + map_path, backup_id_of(backup)); + + /* check tablespace map checksumms */ + if (!no_validate) + { + crc = pgFileGetCRC(map_path, use_crc32c, false); + + if ((*tablespace_map)->crc != crc) + elog(ERROR, "Invalid CRC of tablespace map file \"%s\" : %X. Expected %X, " + "probably backup %s is corrupt, validate it", + map_path, crc, (*tablespace_map)->crc, backup_id_of(backup)); + } + + pgFileFree(dummy); + parray_walk(files, pgFileFree); + parray_free(files); + return true; +} diff --git a/tests/CVE_2018_1058_test.py b/tests/CVE_2018_1058_test.py new file mode 100644 index 000000000..cfd55cc60 --- /dev/null +++ b/tests/CVE_2018_1058_test.py @@ -0,0 +1,129 @@ +import os +import unittest +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException + +class CVE_2018_1058(ProbackupTest, unittest.TestCase): + + # @unittest.skip("skip") + def test_basic_default_search_path(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + "CREATE FUNCTION public.pgpro_edition() " + "RETURNS text " + "AS $$ " + "BEGIN " + " RAISE 'pg_probackup vulnerable!'; " + "END " + "$$ LANGUAGE plpgsql") + + self.backup_node(backup_dir, 'node', node, backup_type='full', options=['--stream']) + + # @unittest.skip("skip") + def test_basic_backup_modified_search_path(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True) + self.set_auto_conf(node, options={'search_path': 'public,pg_catalog'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + "CREATE FUNCTION public.pg_control_checkpoint(OUT timeline_id integer, OUT dummy integer) " + "RETURNS record " + "AS $$ " + "BEGIN " + " RAISE '% vulnerable!', 'pg_probackup'; " + "END " + "$$ LANGUAGE plpgsql") + + node.safe_psql( + 'postgres', + "CREATE FUNCTION public.pg_proc(OUT proname name, OUT dummy integer) " + "RETURNS record " + "AS $$ " + "BEGIN " + " RAISE '% vulnerable!', 'pg_probackup'; " + "END " + "$$ LANGUAGE plpgsql; " + "CREATE VIEW public.pg_proc AS SELECT proname FROM public.pg_proc()") + + self.backup_node(backup_dir, 'node', node, backup_type='full', options=['--stream']) + + log_file = os.path.join(node.logs_dir, 'postgresql.log') + with open(log_file, 'r') as f: + log_content = f.read() + self.assertFalse( + 'pg_probackup vulnerable!' in log_content) + + # @unittest.skip("skip") + def test_basic_checkdb_modified_search_path(self): + """""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + self.set_auto_conf(node, options={'search_path': 'public,pg_catalog'}) + node.slow_start() + + node.safe_psql( + 'postgres', + "CREATE FUNCTION public.pg_database(OUT datname name, OUT oid oid, OUT dattablespace oid) " + "RETURNS record " + "AS $$ " + "BEGIN " + " RAISE 'pg_probackup vulnerable!'; " + "END " + "$$ LANGUAGE plpgsql; " + "CREATE VIEW public.pg_database AS SELECT * FROM public.pg_database()") + + node.safe_psql( + 'postgres', + "CREATE FUNCTION public.pg_extension(OUT extname name, OUT extnamespace oid, OUT extversion text) " + "RETURNS record " + "AS $$ " + "BEGIN " + " RAISE 'pg_probackup vulnerable!'; " + "END " + "$$ LANGUAGE plpgsql; " + "CREATE FUNCTION public.pg_namespace(OUT oid oid, OUT nspname name) " + "RETURNS record " + "AS $$ " + "BEGIN " + " RAISE 'pg_probackup vulnerable!'; " + "END " + "$$ LANGUAGE plpgsql; " + "CREATE VIEW public.pg_extension AS SELECT * FROM public.pg_extension();" + "CREATE VIEW public.pg_namespace AS SELECT * FROM public.pg_namespace();" + ) + + try: + self.checkdb_node( + options=[ + '--amcheck', + '--skip-block-validation', + '-d', 'postgres', '-p', str(node.port)]) + self.assertEqual( + 1, 0, + "Expecting Error because amcheck{,_next} not installed\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "WARNING: Extension 'amcheck' or 'amcheck_next' are not installed in database postgres", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) diff --git a/tests/Readme.md b/tests/Readme.md index 3adf0c019..11c5272f9 100644 --- a/tests/Readme.md +++ b/tests/Readme.md @@ -1,7 +1,11 @@ -[см wiki](https://confluence.postgrespro.ru/display/DEV/pg_probackup) +****[see wiki](https://confluence.postgrespro.ru/display/DEV/pg_probackup) ``` -Note: For now these are works on Linux and "kinda" works on Windows +Note: For now these tests work on Linux and "kinda" work on Windows +``` + +``` +Note: tests require python3 to work properly. ``` ``` @@ -26,15 +30,40 @@ Specify path to pg_probackup binary file. By default tests use /proc/sys/kernel/yama/ptrace_scope pip install testgres export PG_CONFIG=/path/to/pg_config python -m unittest [-v] tests[.specific_module][.class.test] ``` + +# Troubleshooting FAQ + +## Python tests failure +### 1. Could not open extension "..." +``` +testgres.exceptions.QueryException ERROR: could not open extension control file "/share/extension/amcheck.control": No such file or directory +``` + +#### Solution: + +You have no `/contrib/...` extension installed, please do + +```commandline +cd +make install-world +``` diff --git a/tests/__init__.py b/tests/__init__.py index e844d29b8..c8d2c70c3 100644 --- a/tests/__init__.py +++ b/tests/__init__.py @@ -1,12 +1,13 @@ import unittest import os -from . import init, merge, option, show, compatibility, \ - backup, delete, delta, restore, validate, \ - retention, pgpro560, pgpro589, pgpro2068, false_positive, replica, \ - compression, page, ptrack, archive, exclude, cfs_backup, cfs_restore, \ - cfs_validate_backup, auth_test, time_stamp, snapfs, logging, \ - locking, remote, external, config, checkdb +from . import init_test, merge_test, option_test, show_test, compatibility_test, \ + backup_test, delete_test, delta_test, restore_test, validate_test, \ + retention_test, pgpro560_test, pgpro589_test, pgpro2068_test, false_positive_test, replica_test, \ + compression_test, page_test, ptrack_test, archive_test, exclude_test, cfs_backup_test, cfs_restore_test, \ + cfs_validate_backup_test, auth_test, time_stamp_test, logging_test, \ + locking_test, remote_test, external_test, config_test, checkdb_test, set_backup_test, incr_restore_test, \ + catchup_test, CVE_2018_1058_test, time_consuming_test def load_tests(loader, tests, pattern): @@ -18,40 +19,50 @@ def load_tests(loader, tests, pattern): if 'PG_PROBACKUP_PTRACK' in os.environ: if os.environ['PG_PROBACKUP_PTRACK'] == 'ON': - suite.addTests(loader.loadTestsFromModule(ptrack)) - -# suite.addTests(loader.loadTestsFromModule(auth_test)) - suite.addTests(loader.loadTestsFromModule(archive)) - suite.addTests(loader.loadTestsFromModule(backup)) - suite.addTests(loader.loadTestsFromModule(compatibility)) - suite.addTests(loader.loadTestsFromModule(checkdb)) - suite.addTests(loader.loadTestsFromModule(config)) -# suite.addTests(loader.loadTestsFromModule(cfs_backup)) -# suite.addTests(loader.loadTestsFromModule(cfs_restore)) -# suite.addTests(loader.loadTestsFromModule(cfs_validate_backup)) - suite.addTests(loader.loadTestsFromModule(compression)) - suite.addTests(loader.loadTestsFromModule(delete)) - suite.addTests(loader.loadTestsFromModule(delta)) - suite.addTests(loader.loadTestsFromModule(exclude)) - suite.addTests(loader.loadTestsFromModule(external)) - suite.addTests(loader.loadTestsFromModule(false_positive)) - suite.addTests(loader.loadTestsFromModule(init)) - suite.addTests(loader.loadTestsFromModule(locking)) - suite.addTests(loader.loadTestsFromModule(logging)) - suite.addTests(loader.loadTestsFromModule(merge)) - suite.addTests(loader.loadTestsFromModule(option)) - suite.addTests(loader.loadTestsFromModule(page)) - suite.addTests(loader.loadTestsFromModule(pgpro560)) - suite.addTests(loader.loadTestsFromModule(pgpro589)) - suite.addTests(loader.loadTestsFromModule(pgpro2068)) - suite.addTests(loader.loadTestsFromModule(remote)) - suite.addTests(loader.loadTestsFromModule(replica)) - suite.addTests(loader.loadTestsFromModule(restore)) - suite.addTests(loader.loadTestsFromModule(retention)) - suite.addTests(loader.loadTestsFromModule(show)) - suite.addTests(loader.loadTestsFromModule(snapfs)) - suite.addTests(loader.loadTestsFromModule(time_stamp)) - suite.addTests(loader.loadTestsFromModule(validate)) + suite.addTests(loader.loadTestsFromModule(ptrack_test)) + + # PG_PROBACKUP_LONG section for tests that are long + # by design e.g. they contain loops, sleeps and so on + if 'PG_PROBACKUP_LONG' in os.environ: + if os.environ['PG_PROBACKUP_LONG'] == 'ON': + suite.addTests(loader.loadTestsFromModule(time_consuming_test)) + + suite.addTests(loader.loadTestsFromModule(auth_test)) + suite.addTests(loader.loadTestsFromModule(archive_test)) + suite.addTests(loader.loadTestsFromModule(backup_test)) + suite.addTests(loader.loadTestsFromModule(catchup_test)) + if 'PGPROBACKUPBIN_OLD' in os.environ and os.environ['PGPROBACKUPBIN_OLD']: + suite.addTests(loader.loadTestsFromModule(compatibility_test)) + suite.addTests(loader.loadTestsFromModule(checkdb_test)) + suite.addTests(loader.loadTestsFromModule(config_test)) + suite.addTests(loader.loadTestsFromModule(cfs_backup_test)) + suite.addTests(loader.loadTestsFromModule(cfs_restore_test)) + suite.addTests(loader.loadTestsFromModule(cfs_validate_backup_test)) + suite.addTests(loader.loadTestsFromModule(compression_test)) + suite.addTests(loader.loadTestsFromModule(delete_test)) + suite.addTests(loader.loadTestsFromModule(delta_test)) + suite.addTests(loader.loadTestsFromModule(exclude_test)) + suite.addTests(loader.loadTestsFromModule(external_test)) + suite.addTests(loader.loadTestsFromModule(false_positive_test)) + suite.addTests(loader.loadTestsFromModule(init_test)) + suite.addTests(loader.loadTestsFromModule(incr_restore_test)) + suite.addTests(loader.loadTestsFromModule(locking_test)) + suite.addTests(loader.loadTestsFromModule(logging_test)) + suite.addTests(loader.loadTestsFromModule(merge_test)) + suite.addTests(loader.loadTestsFromModule(option_test)) + suite.addTests(loader.loadTestsFromModule(page_test)) + suite.addTests(loader.loadTestsFromModule(pgpro560_test)) + suite.addTests(loader.loadTestsFromModule(pgpro589_test)) + suite.addTests(loader.loadTestsFromModule(pgpro2068_test)) + suite.addTests(loader.loadTestsFromModule(remote_test)) + suite.addTests(loader.loadTestsFromModule(replica_test)) + suite.addTests(loader.loadTestsFromModule(restore_test)) + suite.addTests(loader.loadTestsFromModule(retention_test)) + suite.addTests(loader.loadTestsFromModule(set_backup_test)) + suite.addTests(loader.loadTestsFromModule(show_test)) + suite.addTests(loader.loadTestsFromModule(time_stamp_test)) + suite.addTests(loader.loadTestsFromModule(validate_test)) + suite.addTests(loader.loadTestsFromModule(CVE_2018_1058_test)) return suite diff --git a/tests/archive.py b/tests/archive.py deleted file mode 100644 index cf3441960..000000000 --- a/tests/archive.py +++ /dev/null @@ -1,1117 +0,0 @@ -import os -import shutil -import gzip -import unittest -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException, GdbException -from datetime import datetime, timedelta -import subprocess -from sys import exit -from time import sleep - - -module_name = 'archive' - - -class ArchiveTest(ProbackupTest, unittest.TestCase): - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_pgpro434_1(self): - """Description in jira issue PGPRO-434""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '30s'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "create table t_heap as select 1 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector from " - "generate_series(0,100) i") - - result = node.safe_psql("postgres", "SELECT * FROM t_heap") - self.backup_node( - backup_dir, 'node', node) - node.cleanup() - - self.restore_node( - backup_dir, 'node', node) - node.slow_start() - - # Recreate backup catalog - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - - # Make backup - self.backup_node( - backup_dir, 'node', node) - node.cleanup() - - # Restore Database - self.restore_node( - backup_dir, 'node', node, - options=["--recovery-target-action=promote"]) - node.slow_start() - - self.assertEqual( - result, node.safe_psql("postgres", "SELECT * FROM t_heap"), - 'data after restore not equal to original data') - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_pgpro434_2(self): - """ - Check that timelines are correct. - WAITING PGPRO-1053 for --immediate - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '30s'} - ) - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # FIRST TIMELINE - node.safe_psql( - "postgres", - "create table t_heap as select 1 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,100) i") - backup_id = self.backup_node(backup_dir, 'node', node) - node.safe_psql( - "postgres", - "insert into t_heap select 100501 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1) i") - - # SECOND TIMELIN - node.cleanup() - self.restore_node( - backup_dir, 'node', node, - options=['--immediate', '--recovery-target-action=promote']) - node.slow_start() - - if self.verbose: - print(node.safe_psql( - "postgres", - "select redo_wal_file from pg_control_checkpoint()")) - self.assertFalse( - node.execute( - "postgres", - "select exists(select 1 " - "from t_heap where id = 100501)")[0][0], - 'data after restore not equal to original data') - - node.safe_psql( - "postgres", - "insert into t_heap select 2 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(100,200) i") - - backup_id = self.backup_node(backup_dir, 'node', node) - - node.safe_psql( - "postgres", - "insert into t_heap select 100502 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,256) i") - - # THIRD TIMELINE - node.cleanup() - self.restore_node( - backup_dir, 'node', node, - options=['--immediate', '--recovery-target-action=promote']) - node.slow_start() - - if self.verbose: - print( - node.safe_psql( - "postgres", - "select redo_wal_file from pg_control_checkpoint()")) - - node.safe_psql( - "postgres", - "insert into t_heap select 3 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(200,300) i") - - backup_id = self.backup_node(backup_dir, 'node', node) - - result = node.safe_psql("postgres", "SELECT * FROM t_heap") - node.safe_psql( - "postgres", - "insert into t_heap select 100503 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,256) i") - - # FOURTH TIMELINE - node.cleanup() - self.restore_node( - backup_dir, 'node', node, - options=['--immediate', '--recovery-target-action=promote']) - node.slow_start() - - if self.verbose: - print('Fourth timeline') - print(node.safe_psql( - "postgres", - "select redo_wal_file from pg_control_checkpoint()")) - - # FIFTH TIMELINE - node.cleanup() - self.restore_node( - backup_dir, 'node', node, - options=['--immediate', '--recovery-target-action=promote']) - node.slow_start() - - if self.verbose: - print('Fifth timeline') - print(node.safe_psql( - "postgres", - "select redo_wal_file from pg_control_checkpoint()")) - - # SIXTH TIMELINE - node.cleanup() - self.restore_node( - backup_dir, 'node', node, - options=['--immediate', '--recovery-target-action=promote']) - node.slow_start() - - if self.verbose: - print('Sixth timeline') - print(node.safe_psql( - "postgres", - "select redo_wal_file from pg_control_checkpoint()")) - - self.assertFalse( - node.execute( - "postgres", - "select exists(select 1 from t_heap where id > 100500)")[0][0], - 'data after restore not equal to original data') - - self.assertEqual( - result, - node.safe_psql( - "postgres", - "SELECT * FROM t_heap"), - 'data after restore not equal to original data') - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_pgpro434_3(self): - """ - Check pg_stop_backup_timeout, needed backup_timeout - Fixed in commit d84d79668b0c139 and assert fixed by ptrack 1.7 - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - - node.slow_start() - - gdb = self.backup_node( - backup_dir, 'node', node, - options=[ - "--archive-timeout=60", - "--log-level-file=info"], - gdb=True) - - gdb.set_breakpoint('pg_stop_backup') - gdb.run_until_break() - - node.append_conf( - 'postgresql.auto.conf', "archive_command = 'exit 1'") - node.reload() - - gdb.continue_execution_until_exit() - - log_file = os.path.join(backup_dir, 'log', 'pg_probackup.log') - with open(log_file, 'r') as f: - log_content = f.read() - - # in PG =< 9.6 pg_stop_backup always wait - if self.get_version(node) < 100000: - self.assertIn( - "ERROR: pg_stop_backup doesn't answer in 60 seconds, cancel it", - log_content) - else: - self.assertIn( - "ERROR: Switched WAL segment 000000010000000000000002 " - "could not be archived in 60 seconds", - log_content) - - log_file = os.path.join(node.logs_dir, 'postgresql.log') - with open(log_file, 'r') as f: - log_content = f.read() - - self.assertNotIn( - 'FailedAssertion', - log_content, - 'PostgreSQL crashed because of a failed assert') - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_pgpro434_4(self): - """ - Check pg_stop_backup_timeout, needed backup_timeout - Fixed in commit d84d79668b0c139 and assert fixed by ptrack 1.7 - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - - node.slow_start() - - gdb = self.backup_node( - backup_dir, 'node', node, - options=[ - "--archive-timeout=60", - "--log-level-file=info"], - gdb=True) - - gdb.set_breakpoint('pg_stop_backup') - gdb.run_until_break() - - node.append_conf( - 'postgresql.auto.conf', "archive_command = 'exit 1'") - node.reload() - - os.environ["PGAPPNAME"] = "foo" - - pid = node.safe_psql( - "postgres", - "SELECT pid " - "FROM pg_stat_activity " - "WHERE application_name = 'pg_probackup'").rstrip() - - os.environ["PGAPPNAME"] = "pg_probackup" - - postgres_gdb = self.gdb_attach(pid) - postgres_gdb.set_breakpoint('do_pg_stop_backup') - postgres_gdb.continue_execution_until_running() - - gdb.continue_execution_until_exit() - # gdb._execute('detach') - - log_file = os.path.join(backup_dir, 'log', 'pg_probackup.log') - with open(log_file, 'r') as f: - log_content = f.read() - - self.assertIn( - "ERROR: pg_stop_backup doesn't answer in 60 seconds, cancel it", - log_content) - - log_file = os.path.join(node.logs_dir, 'postgresql.log') - with open(log_file, 'r') as f: - log_content = f.read() - - self.assertNotIn( - 'FailedAssertion', - log_content, - 'PostgreSQL crashed because of a failed assert') - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_archive_push_file_exists(self): - """Archive-push if file exists""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '30s'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - - wals_dir = os.path.join(backup_dir, 'wal', 'node') - if self.archive_compress: - filename = '000000010000000000000001.gz' - file = os.path.join(wals_dir, filename) - else: - filename = '000000010000000000000001' - file = os.path.join(wals_dir, filename) - - with open(file, 'a') as f: - f.write(b"blablablaadssaaaaaaaaaaaaaaa") - f.flush() - f.close() - - node.slow_start() - node.safe_psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,100500) i") - log_file = os.path.join(node.logs_dir, 'postgresql.log') - - self.switch_wal_segment(node) - sleep(1) - - with open(log_file, 'r') as f: - log_content = f.read() - self.assertIn( - 'LOG: archive command failed with exit code 1', - log_content) - - self.assertIn( - 'DETAIL: The failed archive command was:', - log_content) - - self.assertIn( - 'INFO: pg_probackup archive-push from', - log_content) - - self.assertIn( - 'ERROR: WAL segment ', - log_content) - - self.assertIn( - 'already exists.', - log_content) - - self.assertNotIn( - 'pg_probackup archive-push completed successfully', log_content) - - if self.get_version(node) < 100000: - wal_src = os.path.join( - node.data_dir, 'pg_xlog', '000000010000000000000001') - else: - wal_src = os.path.join( - node.data_dir, 'pg_wal', '000000010000000000000001') - - if self.archive_compress: - with open(wal_src, 'rb') as f_in, gzip.open( - file, 'wb', compresslevel=1) as f_out: - shutil.copyfileobj(f_in, f_out) - else: - shutil.copyfile(wal_src, file) - - self.switch_wal_segment(node) - sleep(5) - - with open(log_file, 'r') as f: - log_content = f.read() - - self.assertIn( - 'pg_probackup archive-push completed successfully', - log_content) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_archive_push_file_exists_overwrite(self): - """Archive-push if file exists""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '30s'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - - wals_dir = os.path.join(backup_dir, 'wal', 'node') - if self.archive_compress: - filename = '000000010000000000000001.gz' - file = os.path.join(wals_dir, filename) - else: - filename = '000000010000000000000001' - file = os.path.join(wals_dir, filename) - - with open(file, 'a') as f: - f.write(b"blablablaadssaaaaaaaaaaaaaaa") - f.flush() - f.close() - - node.slow_start() - node.safe_psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,100500) i") - log_file = os.path.join(node.logs_dir, 'postgresql.log') - - self.switch_wal_segment(node) - sleep(1) - - with open(log_file, 'r') as f: - log_content = f.read() - - self.assertIn( - 'LOG: archive command failed with exit code 1', log_content) - self.assertIn( - 'DETAIL: The failed archive command was:', log_content) - self.assertIn( - 'INFO: pg_probackup archive-push from', log_content) - self.assertIn( - '{0}" already exists.'.format(filename), log_content) - - self.assertNotIn( - 'pg_probackup archive-push completed successfully', log_content) - - self.set_archiving(backup_dir, 'node', node, overwrite=True) - node.reload() - self.switch_wal_segment(node) - sleep(2) - - with open(log_file, 'r') as f: - log_content = f.read() - self.assertTrue( - 'pg_probackup archive-push completed successfully' in log_content, - 'Expecting messages about successfull execution archive_command') - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_archive_push_partial_file_exists(self): - """Archive-push if stale '.partial' file exists""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - - node.slow_start() - - # this backup is needed only for validation to xid - self.backup_node(backup_dir, 'node', node) - - node.safe_psql( - "postgres", - "create table t1(a int)") - - xid = node.safe_psql( - "postgres", - "INSERT INTO t1 VALUES (1) RETURNING (xmin)").rstrip() - - if self.get_version(node) < 100000: - filename_orig = node.safe_psql( - "postgres", - "SELECT file_name " - "FROM pg_xlogfile_name_offset(pg_current_xlog_location());").rstrip() - else: - filename_orig = node.safe_psql( - "postgres", - "SELECT file_name " - "FROM pg_walfile_name_offset(pg_current_wal_flush_lsn());").rstrip() - - # form up path to next .partial WAL segment - wals_dir = os.path.join(backup_dir, 'wal', 'node') - if self.archive_compress: - filename = filename_orig + '.gz' + '.partial' - file = os.path.join(wals_dir, filename) - else: - filename = filename_orig + '.partial' - file = os.path.join(wals_dir, filename) - - # emulate stale .partial file - with open(file, 'a') as f: - f.write(b"blahblah") - f.flush() - f.close() - - self.switch_wal_segment(node) - sleep(20) - - # check that segment is archived - if self.archive_compress: - filename_orig = filename_orig + '.gz' - - file = os.path.join(wals_dir, filename_orig) - self.assertTrue(os.path.isfile(file)) - - # successful validate means that archive-push reused stale wal segment - self.validate_pb( - backup_dir, 'node', - options=['--recovery-target-xid={0}'.format(xid)]) - - # log_file = os.path.join(node.logs_dir, 'postgresql.log') - # with open(log_file, 'r') as f: - # log_content = f.read() - # self.assertIn( - # 'Reusing stale destination temporary WAL file', - # log_content) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_archive_push_partial_file_exists_not_stale(self): - """Archive-push if .partial file exists and it is not stale""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - - node.slow_start() - - node.safe_psql( - "postgres", - "create table t1()") - self.switch_wal_segment(node) - - node.safe_psql( - "postgres", - "create table t2()") - - if self.get_version(node) < 100000: - filename_orig = node.safe_psql( - "postgres", - "SELECT file_name " - "FROM pg_xlogfile_name_offset(pg_current_xlog_location());").rstrip() - else: - filename_orig = node.safe_psql( - "postgres", - "SELECT file_name " - "FROM pg_walfile_name_offset(pg_current_wal_flush_lsn());").rstrip() - - # form up path to next .partial WAL segment - wals_dir = os.path.join(backup_dir, 'wal', 'node') - if self.archive_compress: - filename = filename_orig + '.gz' + '.partial' - file = os.path.join(wals_dir, filename) - else: - filename = filename_orig + '.partial' - file = os.path.join(wals_dir, filename) - - with open(file, 'a') as f: - f.write(b"blahblah") - f.flush() - f.close() - - self.switch_wal_segment(node) - sleep(4) - - with open(file, 'a') as f: - f.write(b"blahblahblahblah") - f.flush() - f.close() - - sleep(10) - - # check that segment is NOT archived - if self.archive_compress: - filename_orig = filename_orig + '.gz' - - file = os.path.join(wals_dir, filename_orig) - - self.assertFalse(os.path.isfile(file)) - - # log_file = os.path.join(node.logs_dir, 'postgresql.log') - # with open(log_file, 'r') as f: - # log_content = f.read() - # self.assertIn( - # 'is not stale', - # log_content) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_replica_archive(self): - """ - make node without archiving, take stream backup and - turn it into replica, set replica with archiving, - make archive backup from replica - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'archive_timeout': '10s', - 'checkpoint_timeout': '30s', - 'max_wal_size': '32MB'}) - - self.init_pb(backup_dir) - # ADD INSTANCE 'MASTER' - self.add_instance(backup_dir, 'master', master) - master.slow_start() - - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() - - master.psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,2560) i") - - self.backup_node(backup_dir, 'master', master, options=['--stream']) - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - - # Settings for Replica - self.restore_node(backup_dir, 'master', replica) - self.set_replica(master, replica, synchronous=True) - - self.add_instance(backup_dir, 'replica', replica) - self.set_archiving(backup_dir, 'replica', replica, replica=True) - replica.slow_start(replica=True) - - # Check data correctness on replica - after = replica.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(before, after) - - # Change data on master, take FULL backup from replica, - # restore taken backup and check that restored data equal - # to original data - master.psql( - "postgres", - "insert into t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(256,512) i") - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - - backup_id = self.backup_node( - backup_dir, 'replica', replica, - options=[ - '--archive-timeout=30', - '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port), - '--stream']) - - self.validate_pb(backup_dir, 'replica') - self.assertEqual( - 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) - - # RESTORE FULL BACKUP TAKEN FROM replica - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node')) - node.cleanup() - self.restore_node(backup_dir, 'replica', data_dir=node.data_dir) - node.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node.port)) - node.slow_start() - # CHECK DATA CORRECTNESS - after = node.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(before, after) - - # Change data on master, make PAGE backup from replica, - # restore taken backup and check that restored data equal - # to original data - master.psql( - "postgres", - "insert into t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(512,80680) i") - - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - - master.safe_psql( - "postgres", - "CHECKPOINT") - - self.wait_until_replica_catch_with_master(master, replica) - - backup_id = self.backup_node( - backup_dir, 'replica', - replica, backup_type='page', - options=[ - '--archive-timeout=60', - '--master-db=postgres', - '--master-host=localhost', - '--master-port={0}'.format(master.port), - '--stream']) - - self.validate_pb(backup_dir, 'replica') - self.assertEqual( - 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) - - # RESTORE PAGE BACKUP TAKEN FROM replica - node.cleanup() - self.restore_node( - backup_dir, 'replica', data_dir=node.data_dir, backup_id=backup_id) - - node.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node.port)) - - node.slow_start() - # CHECK DATA CORRECTNESS - after = node.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(before, after) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_master_and_replica_parallel_archiving(self): - """ - make node 'master 'with archiving, - take archive backup and turn it into replica, - set replica with archiving, make archive backup from replica, - make archive backup from master - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'archive_timeout': '10s'} - ) - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() - - self.init_pb(backup_dir) - # ADD INSTANCE 'MASTER' - self.add_instance(backup_dir, 'master', master) - self.set_archiving(backup_dir, 'master', master) - master.slow_start() - - master.psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,10000) i") - - # TAKE FULL ARCHIVE BACKUP FROM MASTER - self.backup_node(backup_dir, 'master', master) - # GET LOGICAL CONTENT FROM MASTER - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - # GET PHYSICAL CONTENT FROM MASTER - pgdata_master = self.pgdata_content(master.data_dir) - - # Settings for Replica - self.restore_node(backup_dir, 'master', replica) - # CHECK PHYSICAL CORRECTNESS on REPLICA - pgdata_replica = self.pgdata_content(replica.data_dir) - self.compare_pgdata(pgdata_master, pgdata_replica) - - self.set_replica(master, replica) - # ADD INSTANCE REPLICA - self.add_instance(backup_dir, 'replica', replica) - # SET ARCHIVING FOR REPLICA - self.set_archiving(backup_dir, 'replica', replica, replica=True) - replica.slow_start(replica=True) - - # CHECK LOGICAL CORRECTNESS on REPLICA - after = replica.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(before, after) - - master.psql( - "postgres", - "insert into t_heap select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0, 60000) i") - - master.psql( - "postgres", - "CHECKPOINT") - - backup_id = self.backup_node( - backup_dir, 'replica', replica, - options=[ - '--archive-timeout=30', - '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port), - '--stream']) - - self.validate_pb(backup_dir, 'replica') - self.assertEqual( - 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) - - # TAKE FULL ARCHIVE BACKUP FROM MASTER - backup_id = self.backup_node(backup_dir, 'master', master) - self.validate_pb(backup_dir, 'master') - self.assertEqual( - 'OK', self.show_pb(backup_dir, 'master', backup_id)['status']) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_basic_master_and_replica_concurrent_archiving(self): - """ - make node 'master 'with archiving, - take archive backup and turn it into replica, - set replica with archiving, make archive backup from replica, - make archive backup from master - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '30s', - 'archive_timeout': '10s'} - ) - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() - - self.init_pb(backup_dir) - # ADD INSTANCE 'MASTER' - self.add_instance(backup_dir, 'master', master) - self.set_archiving(backup_dir, 'master', master) - master.slow_start() - - master.psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,10000) i") - - # TAKE FULL ARCHIVE BACKUP FROM MASTER - self.backup_node(backup_dir, 'master', master) - # GET LOGICAL CONTENT FROM MASTER - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - # GET PHYSICAL CONTENT FROM MASTER - pgdata_master = self.pgdata_content(master.data_dir) - - # Settings for Replica - self.restore_node( - backup_dir, 'master', replica) - # CHECK PHYSICAL CORRECTNESS on REPLICA - pgdata_replica = self.pgdata_content(replica.data_dir) - self.compare_pgdata(pgdata_master, pgdata_replica) - - self.set_replica(master, replica, synchronous=True) - # ADD INSTANCE REPLICA - # self.add_instance(backup_dir, 'replica', replica) - # SET ARCHIVING FOR REPLICA - # self.set_archiving(backup_dir, 'replica', replica, replica=True) - replica.slow_start(replica=True) - - # CHECK LOGICAL CORRECTNESS on REPLICA - after = replica.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(before, after) - - master.psql( - "postgres", - "insert into t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,10000) i") - - # TAKE FULL ARCHIVE BACKUP FROM REPLICA - backup_id = self.backup_node( - backup_dir, 'master', replica, - options=[ - '--archive-timeout=30', - '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port)]) - - self.validate_pb(backup_dir, 'master') - self.assertEqual( - 'OK', self.show_pb(backup_dir, 'master', backup_id)['status']) - - # TAKE FULL ARCHIVE BACKUP FROM MASTER - backup_id = self.backup_node(backup_dir, 'master', master) - self.validate_pb(backup_dir, 'master') - self.assertEqual( - 'OK', self.show_pb(backup_dir, 'master', backup_id)['status']) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_archive_pg_receivexlog(self): - """Test backup with pg_receivexlog wal delivary method""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '30s'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - if self.get_version(node) < 100000: - pg_receivexlog_path = self.get_bin_path('pg_receivexlog') - else: - pg_receivexlog_path = self.get_bin_path('pg_receivewal') - - pg_receivexlog = self.run_binary( - [ - pg_receivexlog_path, '-p', str(node.port), '--synchronous', - '-D', os.path.join(backup_dir, 'wal', 'node') - ], asynchronous=True) - - if pg_receivexlog.returncode: - self.assertFalse( - True, - 'Failed to start pg_receivexlog: {0}'.format( - pg_receivexlog.communicate()[1])) - - node.safe_psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,10000) i") - - self.backup_node(backup_dir, 'node', node) - - # PAGE - node.safe_psql( - "postgres", - "insert into t_heap select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(10000,20000) i") - - self.backup_node( - backup_dir, - 'node', - node, - backup_type='page' - ) - result = node.safe_psql("postgres", "SELECT * FROM t_heap") - self.validate_pb(backup_dir) - - # Check data correctness - node.cleanup() - self.restore_node(backup_dir, 'node', node) - node.slow_start() - - self.assertEqual( - result, - node.safe_psql( - "postgres", "SELECT * FROM t_heap" - ), - 'data after restore not equal to original data') - - # Clean after yourself - pg_receivexlog.kill() - self.del_test_dir(module_name, fname) - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_archive_pg_receivexlog_compression_pg10(self): - """Test backup with pg_receivewal compressed wal delivary method""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '30s'} - ) - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - if self.get_version(node) < self.version_to_num('10.0'): - return unittest.skip('You need PostgreSQL >= 10 for this test') - else: - pg_receivexlog_path = self.get_bin_path('pg_receivewal') - - pg_receivexlog = self.run_binary( - [ - pg_receivexlog_path, '-p', str(node.port), '--synchronous', - '-Z', '9', '-D', os.path.join(backup_dir, 'wal', 'node') - ], asynchronous=True) - - if pg_receivexlog.returncode: - self.assertFalse( - True, - 'Failed to start pg_receivexlog: {0}'.format( - pg_receivexlog.communicate()[1])) - - node.safe_psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,10000) i") - - self.backup_node(backup_dir, 'node', node) - - # PAGE - node.safe_psql( - "postgres", - "insert into t_heap select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(10000,20000) i") - - self.backup_node( - backup_dir, 'node', node, - backup_type='page' - ) - result = node.safe_psql("postgres", "SELECT * FROM t_heap") - self.validate_pb(backup_dir) - - # Check data correctness - node.cleanup() - self.restore_node(backup_dir, 'node', node) - node.slow_start() - - self.assertEqual( - result, node.safe_psql("postgres", "SELECT * FROM t_heap"), - 'data after restore not equal to original data') - - # Clean after yourself - pg_receivexlog.kill() - self.del_test_dir(module_name, fname) diff --git a/tests/archive_test.py b/tests/archive_test.py new file mode 100644 index 000000000..00fd1f592 --- /dev/null +++ b/tests/archive_test.py @@ -0,0 +1,2699 @@ +import os +import shutil +import gzip +import unittest +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException, GdbException +from .helpers.data_helpers import tail_file +from datetime import datetime, timedelta +import subprocess +from sys import exit +from time import sleep +from distutils.dir_util import copy_tree + + +class ArchiveTest(ProbackupTest, unittest.TestCase): + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_pgpro434_1(self): + """Description in jira issue PGPRO-434""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector from " + "generate_series(0,100) i") + + result = node.table_checksum("t_heap") + self.backup_node( + backup_dir, 'node', node) + node.cleanup() + + self.restore_node( + backup_dir, 'node', node) + node.slow_start() + + # Recreate backup catalog + self.clean_pb(backup_dir) + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + # Make backup + self.backup_node(backup_dir, 'node', node) + node.cleanup() + + # Restore Database + self.restore_node(backup_dir, 'node', node) + node.slow_start() + + self.assertEqual( + result, node.table_checksum("t_heap"), + 'data after restore not equal to original data') + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_pgpro434_2(self): + """ + Check that timelines are correct. + WAITING PGPRO-1053 for --immediate + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s'} + ) + + if self.get_version(node) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because pg_control_checkpoint() is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FIRST TIMELINE + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,100) i") + backup_id = self.backup_node(backup_dir, 'node', node) + node.safe_psql( + "postgres", + "insert into t_heap select 100501 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1) i") + + # SECOND TIMELIN + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=['--immediate', '--recovery-target-action=promote']) + node.slow_start() + + if self.verbose: + print(node.safe_psql( + "postgres", + "select redo_wal_file from pg_control_checkpoint()")) + self.assertFalse( + node.execute( + "postgres", + "select exists(select 1 " + "from t_heap where id = 100501)")[0][0], + 'data after restore not equal to original data') + + node.safe_psql( + "postgres", + "insert into t_heap select 2 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(100,200) i") + + backup_id = self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + "postgres", + "insert into t_heap select 100502 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,256) i") + + # THIRD TIMELINE + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=['--immediate', '--recovery-target-action=promote']) + node.slow_start() + + if self.verbose: + print( + node.safe_psql( + "postgres", + "select redo_wal_file from pg_control_checkpoint()")) + + node.safe_psql( + "postgres", + "insert into t_heap select 3 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(200,300) i") + + backup_id = self.backup_node(backup_dir, 'node', node) + + result = node.table_checksum("t_heap") + node.safe_psql( + "postgres", + "insert into t_heap select 100503 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,256) i") + + # FOURTH TIMELINE + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=['--immediate', '--recovery-target-action=promote']) + node.slow_start() + + if self.verbose: + print('Fourth timeline') + print(node.safe_psql( + "postgres", + "select redo_wal_file from pg_control_checkpoint()")) + + # FIFTH TIMELINE + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=['--immediate', '--recovery-target-action=promote']) + node.slow_start() + + if self.verbose: + print('Fifth timeline') + print(node.safe_psql( + "postgres", + "select redo_wal_file from pg_control_checkpoint()")) + + # SIXTH TIMELINE + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=['--immediate', '--recovery-target-action=promote']) + node.slow_start() + + if self.verbose: + print('Sixth timeline') + print(node.safe_psql( + "postgres", + "select redo_wal_file from pg_control_checkpoint()")) + + self.assertFalse( + node.execute( + "postgres", + "select exists(select 1 from t_heap where id > 100500)")[0][0], + 'data after restore not equal to original data') + + self.assertEqual(result, node.table_checksum("t_heap"), + 'data after restore not equal to original data') + + # @unittest.skip("skip") + def test_pgpro434_3(self): + """ + Check pg_stop_backup_timeout, needed backup_timeout + Fixed in commit d84d79668b0c139 and assert fixed by ptrack 1.7 + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + node.slow_start() + + gdb = self.backup_node( + backup_dir, 'node', node, + options=[ + "--archive-timeout=60", + "--log-level-file=LOG"], + gdb=True) + + # Attention! this breakpoint has been set on internal probackup function, not on a postgres core one + gdb.set_breakpoint('pg_stop_backup') + gdb.run_until_break() + + self.set_auto_conf(node, {'archive_command': 'exit 1'}) + node.reload() + + gdb.continue_execution_until_exit() + + sleep(1) + + log_file = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(log_file, 'r') as f: + log_content = f.read() + + # in PG =< 9.6 pg_stop_backup always wait + if self.get_version(node) < 100000: + self.assertIn( + "ERROR: pg_stop_backup doesn't answer in 60 seconds, cancel it", + log_content) + else: + self.assertIn( + "ERROR: WAL segment 000000010000000000000003 could not be archived in 60 seconds", + log_content) + + log_file = os.path.join(node.logs_dir, 'postgresql.log') + with open(log_file, 'r') as f: + log_content = f.read() + + self.assertNotIn( + 'FailedAssertion', + log_content, + 'PostgreSQL crashed because of a failed assert') + + # @unittest.skip("skip") + def test_pgpro434_4(self): + """ + Check pg_stop_backup_timeout, libpq-timeout requested. + Fixed in commit d84d79668b0c139 and assert fixed by ptrack 1.7 + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + node.slow_start() + + gdb = self.backup_node( + backup_dir, 'node', node, + options=[ + "--archive-timeout=60", + "--log-level-file=info"], + gdb=True) + + # Attention! this breakpoint has been set on internal probackup function, not on a postgres core one + gdb.set_breakpoint('pg_stop_backup') + gdb.run_until_break() + + self.set_auto_conf(node, {'archive_command': 'exit 1'}) + node.reload() + + os.environ["PGAPPNAME"] = "foo" + + pid = node.safe_psql( + "postgres", + "SELECT pid " + "FROM pg_stat_activity " + "WHERE application_name = 'pg_probackup'").decode('utf-8').rstrip() + + os.environ["PGAPPNAME"] = "pg_probackup" + + postgres_gdb = self.gdb_attach(pid) + if self.get_version(node) < 150000: + postgres_gdb.set_breakpoint('do_pg_stop_backup') + else: + postgres_gdb.set_breakpoint('do_pg_backup_stop') + postgres_gdb.continue_execution_until_running() + + gdb.continue_execution_until_exit() + # gdb._execute('detach') + + log_file = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(log_file, 'r') as f: + log_content = f.read() + + if self.get_version(node) < 150000: + self.assertIn( + "ERROR: pg_stop_backup doesn't answer in 60 seconds, cancel it", + log_content) + else: + self.assertIn( + "ERROR: pg_backup_stop doesn't answer in 60 seconds, cancel it", + log_content) + + log_file = os.path.join(node.logs_dir, 'postgresql.log') + with open(log_file, 'r') as f: + log_content = f.read() + + self.assertNotIn( + 'FailedAssertion', + log_content, + 'PostgreSQL crashed because of a failed assert') + + # @unittest.skip("skip") + def test_archive_push_file_exists(self): + """Archive-push if file exists""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + wals_dir = os.path.join(backup_dir, 'wal', 'node') + if self.archive_compress: + filename = '000000010000000000000001.gz' + file = os.path.join(wals_dir, filename) + else: + filename = '000000010000000000000001' + file = os.path.join(wals_dir, filename) + + with open(file, 'a+b') as f: + f.write(b"blablablaadssaaaaaaaaaaaaaaa") + f.flush() + f.close() + + node.slow_start() + node.safe_psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,100500) i") + log_file = os.path.join(node.logs_dir, 'postgresql.log') + + self.switch_wal_segment(node) + sleep(1) + + log = tail_file(log_file, linetimeout=30, totaltimeout=120, + collect=True) + log.wait(contains = 'The failed archive command was') + + self.assertIn( + 'LOG: archive command failed with exit code 1', + log.content) + + self.assertIn( + 'DETAIL: The failed archive command was:', + log.content) + + self.assertIn( + 'pg_probackup archive-push WAL file', + log.content) + + self.assertIn( + 'WAL file already exists in archive with different checksum', + log.content) + + self.assertNotIn( + 'pg_probackup archive-push completed successfully', log.content) + + # btw check that console coloring codes are not slipped into log file + self.assertNotIn('[0m', log.content) + + if self.get_version(node) < 100000: + wal_src = os.path.join( + node.data_dir, 'pg_xlog', '000000010000000000000001') + else: + wal_src = os.path.join( + node.data_dir, 'pg_wal', '000000010000000000000001') + + if self.archive_compress: + with open(wal_src, 'rb') as f_in, gzip.open( + file, 'wb', compresslevel=1) as f_out: + shutil.copyfileobj(f_in, f_out) + else: + shutil.copyfile(wal_src, file) + + self.switch_wal_segment(node) + + log.stop_collect() + log.wait(contains = 'pg_probackup archive-push completed successfully') + + # @unittest.skip("skip") + def test_archive_push_file_exists_overwrite(self): + """Archive-push if file exists""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={'checkpoint_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + wals_dir = os.path.join(backup_dir, 'wal', 'node') + if self.archive_compress: + filename = '000000010000000000000001.gz' + file = os.path.join(wals_dir, filename) + else: + filename = '000000010000000000000001' + file = os.path.join(wals_dir, filename) + + with open(file, 'a+b') as f: + f.write(b"blablablaadssaaaaaaaaaaaaaaa") + f.flush() + f.close() + + node.slow_start() + node.safe_psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,100500) i") + log_file = os.path.join(node.logs_dir, 'postgresql.log') + + self.switch_wal_segment(node) + sleep(1) + + log = tail_file(log_file, linetimeout=30, collect=True) + log.wait(contains = 'The failed archive command was') + + self.assertIn( + 'LOG: archive command failed with exit code 1', log.content) + self.assertIn( + 'DETAIL: The failed archive command was:', log.content) + self.assertIn( + 'pg_probackup archive-push WAL file', log.content) + self.assertNotIn( + 'WAL file already exists in archive with ' + 'different checksum, overwriting', log.content) + self.assertIn( + 'WAL file already exists in archive with ' + 'different checksum', log.content) + + self.assertNotIn( + 'pg_probackup archive-push completed successfully', log.content) + + self.set_archiving(backup_dir, 'node', node, overwrite=True) + node.reload() + self.switch_wal_segment(node) + + log.drop_content() + log.wait(contains = 'pg_probackup archive-push completed successfully') + + self.assertIn( + 'WAL file already exists in archive with ' + 'different checksum, overwriting', log.content) + + # @unittest.skip("skip") + def test_archive_push_partial_file_exists(self): + """Archive-push if stale '.part' file exists""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving( + backup_dir, 'node', node, + log_level='verbose', archive_timeout=60) + + node.slow_start() + + # this backup is needed only for validation to xid + self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + "postgres", + "create table t1(a int)") + + xid = node.safe_psql( + "postgres", + "INSERT INTO t1 VALUES (1) RETURNING (xmin)").decode('utf-8').rstrip() + + if self.get_version(node) < 100000: + filename_orig = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_xlogfile_name_offset(pg_current_xlog_location());").rstrip() + else: + filename_orig = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_walfile_name_offset(pg_current_wal_flush_lsn());").rstrip() + + filename_orig = filename_orig.decode('utf-8') + + # form up path to next .part WAL segment + wals_dir = os.path.join(backup_dir, 'wal', 'node') + if self.archive_compress: + filename = filename_orig + '.gz' + '.part' + file = os.path.join(wals_dir, filename) + else: + filename = filename_orig + '.part' + file = os.path.join(wals_dir, filename) + + # emulate stale .part file + with open(file, 'a+b') as f: + f.write(b"blahblah") + f.flush() + f.close() + + self.switch_wal_segment(node) + sleep(70) + + # check that segment is archived + if self.archive_compress: + filename_orig = filename_orig + '.gz' + + file = os.path.join(wals_dir, filename_orig) + self.assertTrue(os.path.isfile(file)) + + # successful validate means that archive-push reused stale wal segment + self.validate_pb( + backup_dir, 'node', + options=['--recovery-target-xid={0}'.format(xid)]) + + log_file = os.path.join(node.logs_dir, 'postgresql.log') + with open(log_file, 'r') as f: + log_content = f.read() + + self.assertIn( + 'Reusing stale temp WAL file', + log_content) + + # @unittest.skip("skip") + def test_archive_push_part_file_exists_not_stale(self): + """Archive-push if .part file exists and it is not stale""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node, archive_timeout=60) + + node.slow_start() + + node.safe_psql( + "postgres", + "create table t1()") + self.switch_wal_segment(node) + + node.safe_psql( + "postgres", + "create table t2()") + + if self.get_version(node) < 100000: + filename_orig = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_xlogfile_name_offset(pg_current_xlog_location());").rstrip() + else: + filename_orig = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_walfile_name_offset(pg_current_wal_flush_lsn());").rstrip() + + filename_orig = filename_orig.decode('utf-8') + + # form up path to next .part WAL segment + wals_dir = os.path.join(backup_dir, 'wal', 'node') + if self.archive_compress: + filename = filename_orig + '.gz' + '.part' + file = os.path.join(wals_dir, filename) + else: + filename = filename_orig + '.part' + file = os.path.join(wals_dir, filename) + + with open(file, 'a+b') as f: + f.write(b"blahblah") + f.flush() + f.close() + + self.switch_wal_segment(node) + sleep(30) + + with open(file, 'a+b') as f: + f.write(b"blahblahblahblah") + f.flush() + f.close() + + sleep(40) + + # check that segment is NOT archived + if self.archive_compress: + filename_orig = filename_orig + '.gz' + + file = os.path.join(wals_dir, filename_orig) + + self.assertFalse(os.path.isfile(file)) + + # log_file = os.path.join(node.logs_dir, 'postgresql.log') + # with open(log_file, 'r') as f: + # log_content = f.read() + # self.assertIn( + # 'is not stale', + # log_content) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_replica_archive(self): + """ + make node without archiving, take stream backup and + turn it into replica, set replica with archiving, + make archive backup from replica + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_timeout': '10s', + 'checkpoint_timeout': '30s', + 'max_wal_size': '32MB'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + # ADD INSTANCE 'MASTER' + self.add_instance(backup_dir, 'master', master) + master.slow_start() + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + master.psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,2560) i") + + self.backup_node(backup_dir, 'master', master, options=['--stream']) + before = master.table_checksum("t_heap") + + # Settings for Replica + self.restore_node(backup_dir, 'master', replica) + self.set_replica(master, replica, synchronous=True) + + self.add_instance(backup_dir, 'replica', replica) + self.set_archiving(backup_dir, 'replica', replica, replica=True) + replica.slow_start(replica=True) + + # Check data correctness on replica + after = replica.table_checksum("t_heap") + self.assertEqual(before, after) + + # Change data on master, take FULL backup from replica, + # restore taken backup and check that restored data equal + # to original data + master.psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(256,512) i") + before = master.table_checksum("t_heap") + + backup_id = self.backup_node( + backup_dir, 'replica', replica, + options=[ + '--archive-timeout=30', + '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port), + '--stream']) + + self.validate_pb(backup_dir, 'replica') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) + + # RESTORE FULL BACKUP TAKEN FROM replica + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node')) + node.cleanup() + self.restore_node(backup_dir, 'replica', data_dir=node.data_dir) + + self.set_auto_conf(node, {'port': node.port}) + node.slow_start() + # CHECK DATA CORRECTNESS + after = node.table_checksum("t_heap") + self.assertEqual(before, after) + + # Change data on master, make PAGE backup from replica, + # restore taken backup and check that restored data equal + # to original data + master.psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(512,80680) i") + + before = master.table_checksum("t_heap") + + self.wait_until_replica_catch_with_master(master, replica) + + backup_id = self.backup_node( + backup_dir, 'replica', + replica, backup_type='page', + options=[ + '--archive-timeout=60', + '--master-db=postgres', + '--master-host=localhost', + '--master-port={0}'.format(master.port), + '--stream']) + + self.validate_pb(backup_dir, 'replica') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) + + # RESTORE PAGE BACKUP TAKEN FROM replica + node.cleanup() + self.restore_node( + backup_dir, 'replica', data_dir=node.data_dir, backup_id=backup_id) + + self.set_auto_conf(node, {'port': node.port}) + + node.slow_start() + # CHECK DATA CORRECTNESS + after = node.table_checksum("t_heap") + self.assertEqual(before, after) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_master_and_replica_parallel_archiving(self): + """ + make node 'master 'with archiving, + take archive backup and turn it into replica, + set replica with archiving, make archive backup from replica, + make archive backup from master + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_timeout': '10s'} + ) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.init_pb(backup_dir) + # ADD INSTANCE 'MASTER' + self.add_instance(backup_dir, 'master', master) + self.set_archiving(backup_dir, 'master', master) + master.slow_start() + + master.psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + # TAKE FULL ARCHIVE BACKUP FROM MASTER + self.backup_node(backup_dir, 'master', master) + # GET LOGICAL CONTENT FROM MASTER + before = master.table_checksum("t_heap") + # GET PHYSICAL CONTENT FROM MASTER + pgdata_master = self.pgdata_content(master.data_dir) + + # Settings for Replica + self.restore_node(backup_dir, 'master', replica) + # CHECK PHYSICAL CORRECTNESS on REPLICA + pgdata_replica = self.pgdata_content(replica.data_dir) + self.compare_pgdata(pgdata_master, pgdata_replica) + + self.set_replica(master, replica) + # ADD INSTANCE REPLICA + self.add_instance(backup_dir, 'replica', replica) + # SET ARCHIVING FOR REPLICA + self.set_archiving(backup_dir, 'replica', replica, replica=True) + replica.slow_start(replica=True) + + # CHECK LOGICAL CORRECTNESS on REPLICA + after = replica.table_checksum("t_heap") + self.assertEqual(before, after) + + master.psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0, 60000) i") + + backup_id = self.backup_node( + backup_dir, 'replica', replica, + options=[ + '--archive-timeout=30', + '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port), + '--stream']) + + self.validate_pb(backup_dir, 'replica') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) + + # TAKE FULL ARCHIVE BACKUP FROM MASTER + backup_id = self.backup_node(backup_dir, 'master', master) + self.validate_pb(backup_dir, 'master') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'master', backup_id)['status']) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_basic_master_and_replica_concurrent_archiving(self): + """ + make node 'master 'with archiving, + take archive backup and turn it into replica, + set replica with archiving, + make sure that archiving on both node is working. + """ + if self.pg_config_version < self.version_to_num('9.6.0'): + self.skipTest('You need PostgreSQL >= 9.6 for this test') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s', + 'archive_timeout': '10s'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.init_pb(backup_dir) + # ADD INSTANCE 'MASTER' + self.add_instance(backup_dir, 'master', master) + self.set_archiving(backup_dir, 'master', master) + master.slow_start() + + master.psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + master.pgbench_init(scale=5) + + # TAKE FULL ARCHIVE BACKUP FROM MASTER + self.backup_node(backup_dir, 'master', master) + # GET LOGICAL CONTENT FROM MASTER + before = master.table_checksum("t_heap") + # GET PHYSICAL CONTENT FROM MASTER + pgdata_master = self.pgdata_content(master.data_dir) + + # Settings for Replica + self.restore_node( + backup_dir, 'master', replica) + # CHECK PHYSICAL CORRECTNESS on REPLICA + pgdata_replica = self.pgdata_content(replica.data_dir) + self.compare_pgdata(pgdata_master, pgdata_replica) + + self.set_replica(master, replica, synchronous=False) + # ADD INSTANCE REPLICA + # self.add_instance(backup_dir, 'replica', replica) + # SET ARCHIVING FOR REPLICA + self.set_archiving(backup_dir, 'master', replica, replica=True) + replica.slow_start(replica=True) + + # CHECK LOGICAL CORRECTNESS on REPLICA + after = replica.table_checksum("t_heap") + self.assertEqual(before, after) + + master.psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + # TAKE FULL ARCHIVE BACKUP FROM REPLICA + backup_id = self.backup_node(backup_dir, 'master', replica) + + self.validate_pb(backup_dir, 'master') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'master', backup_id)['status']) + + # TAKE FULL ARCHIVE BACKUP FROM MASTER + backup_id = self.backup_node(backup_dir, 'master', master) + self.validate_pb(backup_dir, 'master') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'master', backup_id)['status']) + + master.pgbench_init(scale=10) + + sleep(10) + + replica.promote() + + master.pgbench_init(scale=10) + replica.pgbench_init(scale=10) + + self.backup_node(backup_dir, 'master', master) + self.backup_node(backup_dir, 'master', replica) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_concurrent_archiving(self): + """ + Concurrent archiving from master, replica and cascade replica + https://github.com/postgrespro/pg_probackup/issues/327 + + For PG >= 11 it is expected to pass this test + """ + + if self.pg_config_version < self.version_to_num('11.0'): + self.skipTest('You need PostgreSQL >= 11 for this test') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', master) + self.set_archiving(backup_dir, 'node', master, replica=True) + master.slow_start() + + master.pgbench_init(scale=10) + + # TAKE FULL ARCHIVE BACKUP FROM MASTER + self.backup_node(backup_dir, 'node', master) + + # Settings for Replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'node', replica) + + self.set_replica(master, replica, synchronous=True) + self.set_archiving(backup_dir, 'node', replica, replica=True) + self.set_auto_conf(replica, {'port': replica.port}) + replica.slow_start(replica=True) + + # create cascade replicas + replica1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica1')) + replica1.cleanup() + + # Settings for casaced replica + self.restore_node(backup_dir, 'node', replica1) + self.set_replica(replica, replica1, synchronous=False) + self.set_auto_conf(replica1, {'port': replica1.port}) + replica1.slow_start(replica=True) + + # Take full backup from master + self.backup_node(backup_dir, 'node', master) + + pgbench = master.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '30', '-c', '1']) + + # Take several incremental backups from master + self.backup_node(backup_dir, 'node', master, backup_type='page', options=['--no-validate']) + + self.backup_node(backup_dir, 'node', master, backup_type='page', options=['--no-validate']) + + pgbench.wait() + pgbench.stdout.close() + + with open(os.path.join(master.logs_dir, 'postgresql.log'), 'r') as f: + log_content = f.read() + self.assertNotIn('different checksum', log_content) + + with open(os.path.join(replica.logs_dir, 'postgresql.log'), 'r') as f: + log_content = f.read() + self.assertNotIn('different checksum', log_content) + + with open(os.path.join(replica1.logs_dir, 'postgresql.log'), 'r') as f: + log_content = f.read() + self.assertNotIn('different checksum', log_content) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_archive_pg_receivexlog(self): + """Test backup with pg_receivexlog wal delivary method""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + if self.get_version(node) < 100000: + pg_receivexlog_path = self.get_bin_path('pg_receivexlog') + else: + pg_receivexlog_path = self.get_bin_path('pg_receivewal') + + pg_receivexlog = self.run_binary( + [ + pg_receivexlog_path, '-p', str(node.port), '--synchronous', + '-D', os.path.join(backup_dir, 'wal', 'node') + ], asynchronous=True) + + if pg_receivexlog.returncode: + self.assertFalse( + True, + 'Failed to start pg_receivexlog: {0}'.format( + pg_receivexlog.communicate()[1])) + + node.safe_psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + self.backup_node(backup_dir, 'node', node) + + # PAGE + node.safe_psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(10000,20000) i") + + self.backup_node( + backup_dir, + 'node', + node, + backup_type='page' + ) + result = node.table_checksum("t_heap") + self.validate_pb(backup_dir) + + # Check data correctness + node.cleanup() + self.restore_node(backup_dir, 'node', node) + node.slow_start() + + self.assertEqual( + result, + node.table_checksum("t_heap"), + 'data after restore not equal to original data') + + # Clean after yourself + pg_receivexlog.kill() + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_archive_pg_receivexlog_compression_pg10(self): + """Test backup with pg_receivewal compressed wal delivary method""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s'} + ) + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + if self.get_version(node) < self.version_to_num('10.0'): + self.skipTest('You need PostgreSQL >= 10 for this test') + else: + pg_receivexlog_path = self.get_bin_path('pg_receivewal') + + pg_receivexlog = self.run_binary( + [ + pg_receivexlog_path, '-p', str(node.port), '--synchronous', + '-Z', '9', '-D', os.path.join(backup_dir, 'wal', 'node') + ], asynchronous=True) + + if pg_receivexlog.returncode: + self.assertFalse( + True, + 'Failed to start pg_receivexlog: {0}'.format( + pg_receivexlog.communicate()[1])) + + node.safe_psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + self.backup_node(backup_dir, 'node', node) + + # PAGE + node.safe_psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(10000,20000) i") + + self.backup_node( + backup_dir, 'node', node, + backup_type='page' + ) + result = node.table_checksum("t_heap") + self.validate_pb(backup_dir) + + # Check data correctness + node.cleanup() + self.restore_node(backup_dir, 'node', node) + node.slow_start() + + self.assertEqual( + result, node.table_checksum("t_heap"), + 'data after restore not equal to original data') + + # Clean after yourself + pg_receivexlog.kill() + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_archive_catalog(self): + """ + ARCHIVE replica: + + t6 |----------------------- + t5 | |------- + | | + t4 | |-------------- + | | + t3 | |--B1--|/|--B2-|/|-B3--- + | | + t2 |--A1--------A2--- + t1 ---------Y1--Y2-- + + ARCHIVE master: + t1 -Z1--Z2--- + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_timeout': '30s', + 'checkpoint_timeout': '30s'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + self.set_archiving(backup_dir, 'master', master) + + master.slow_start() + + # FULL + master.safe_psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + self.backup_node(backup_dir, 'master', master) + + # PAGE + master.safe_psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(10000,20000) i") + + self.backup_node( + backup_dir, 'master', master, backup_type='page') + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'master', replica) + self.set_replica(master, replica) + + self.add_instance(backup_dir, 'replica', replica) + self.set_archiving(backup_dir, 'replica', replica, replica=True) + + copy_tree( + os.path.join(backup_dir, 'wal', 'master'), + os.path.join(backup_dir, 'wal', 'replica')) + + replica.slow_start(replica=True) + + # FULL backup replica + Y1 = self.backup_node( + backup_dir, 'replica', replica, + options=['--stream', '--archive-timeout=60s']) + + master.pgbench_init(scale=5) + + # PAGE backup replica + Y2 = self.backup_node( + backup_dir, 'replica', replica, + backup_type='page', options=['--stream', '--archive-timeout=60s']) + + # create timeline t2 + replica.promote() + + # FULL backup replica + A1 = self.backup_node( + backup_dir, 'replica', replica) + + replica.pgbench_init(scale=5) + + replica.safe_psql( + 'postgres', + "CREATE TABLE t1 (a text)") + + target_xid = None + with replica.connect("postgres") as con: + res = con.execute( + "INSERT INTO t1 VALUES ('inserted') RETURNING (xmin)") + con.commit() + target_xid = res[0][0] + + # DELTA backup replica + A2 = self.backup_node( + backup_dir, 'replica', replica, backup_type='delta') + + # create timeline t3 + replica.cleanup() + self.restore_node( + backup_dir, 'replica', replica, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-timeline=2', + '--recovery-target-action=promote']) + + replica.slow_start() + + B1 = self.backup_node( + backup_dir, 'replica', replica) + + replica.pgbench_init(scale=2) + + B2 = self.backup_node( + backup_dir, 'replica', replica, backup_type='page') + + replica.pgbench_init(scale=2) + + target_xid = None + with replica.connect("postgres") as con: + res = con.execute( + "INSERT INTO t1 VALUES ('inserted') RETURNING (xmin)") + con.commit() + target_xid = res[0][0] + + B3 = self.backup_node( + backup_dir, 'replica', replica, backup_type='page') + + replica.pgbench_init(scale=2) + + # create timeline t4 + replica.cleanup() + self.restore_node( + backup_dir, 'replica', replica, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-timeline=3', + '--recovery-target-action=promote']) + + replica.slow_start() + + replica.safe_psql( + 'postgres', + 'CREATE TABLE ' + 't2 as select i, ' + 'repeat(md5(i::text),5006056) as fat_attr ' + 'from generate_series(0,6) i') + + target_xid = None + with replica.connect("postgres") as con: + res = con.execute( + "INSERT INTO t1 VALUES ('inserted') RETURNING (xmin)") + con.commit() + target_xid = res[0][0] + + replica.safe_psql( + 'postgres', + 'CREATE TABLE ' + 't3 as select i, ' + 'repeat(md5(i::text),5006056) as fat_attr ' + 'from generate_series(0,10) i') + + # create timeline t5 + replica.cleanup() + self.restore_node( + backup_dir, 'replica', replica, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-timeline=4', + '--recovery-target-action=promote']) + + replica.slow_start() + + replica.safe_psql( + 'postgres', + 'CREATE TABLE ' + 't4 as select i, ' + 'repeat(md5(i::text),5006056) as fat_attr ' + 'from generate_series(0,6) i') + + # create timeline t6 + replica.cleanup() + + self.restore_node( + backup_dir, 'replica', replica, backup_id=A1, + options=[ + '--recovery-target=immediate', + '--recovery-target-action=promote']) + replica.slow_start() + + replica.pgbench_init(scale=2) + + sleep(5) + + show = self.show_archive(backup_dir, as_text=True) + show = self.show_archive(backup_dir) + + for instance in show: + if instance['instance'] == 'replica': + replica_timelines = instance['timelines'] + + if instance['instance'] == 'master': + master_timelines = instance['timelines'] + + # check that all timelines are ok + for timeline in replica_timelines: + self.assertTrue(timeline['status'], 'OK') + + # check that all timelines are ok + for timeline in master_timelines: + self.assertTrue(timeline['status'], 'OK') + + # create holes in t3 + wals_dir = os.path.join(backup_dir, 'wal', 'replica') + wals = [ + f for f in os.listdir(wals_dir) if os.path.isfile(os.path.join(wals_dir, f)) + and not f.endswith('.backup') and not f.endswith('.history') and f.startswith('00000003') + ] + wals.sort() + + # check that t3 is ok + self.show_archive(backup_dir) + + file = os.path.join(backup_dir, 'wal', 'replica', '000000030000000000000017') + if self.archive_compress: + file = file + '.gz' + os.remove(file) + + file = os.path.join(backup_dir, 'wal', 'replica', '000000030000000000000012') + if self.archive_compress: + file = file + '.gz' + os.remove(file) + + file = os.path.join(backup_dir, 'wal', 'replica', '000000030000000000000013') + if self.archive_compress: + file = file + '.gz' + os.remove(file) + + # check that t3 is not OK + show = self.show_archive(backup_dir) + + show = self.show_archive(backup_dir) + + for instance in show: + if instance['instance'] == 'replica': + replica_timelines = instance['timelines'] + + # sanity + for timeline in replica_timelines: + if timeline['tli'] == 1: + timeline_1 = timeline + continue + + if timeline['tli'] == 2: + timeline_2 = timeline + continue + + if timeline['tli'] == 3: + timeline_3 = timeline + continue + + if timeline['tli'] == 4: + timeline_4 = timeline + continue + + if timeline['tli'] == 5: + timeline_5 = timeline + continue + + if timeline['tli'] == 6: + timeline_6 = timeline + continue + + self.assertEqual(timeline_6['status'], "OK") + self.assertEqual(timeline_5['status'], "OK") + self.assertEqual(timeline_4['status'], "OK") + self.assertEqual(timeline_3['status'], "DEGRADED") + self.assertEqual(timeline_2['status'], "OK") + self.assertEqual(timeline_1['status'], "OK") + + self.assertEqual(len(timeline_3['lost-segments']), 2) + self.assertEqual( + timeline_3['lost-segments'][0]['begin-segno'], + '000000030000000000000012') + self.assertEqual( + timeline_3['lost-segments'][0]['end-segno'], + '000000030000000000000013') + self.assertEqual( + timeline_3['lost-segments'][1]['begin-segno'], + '000000030000000000000017') + self.assertEqual( + timeline_3['lost-segments'][1]['end-segno'], + '000000030000000000000017') + + self.assertEqual(len(timeline_6['backups']), 0) + self.assertEqual(len(timeline_5['backups']), 0) + self.assertEqual(len(timeline_4['backups']), 0) + self.assertEqual(len(timeline_3['backups']), 3) + self.assertEqual(len(timeline_2['backups']), 2) + self.assertEqual(len(timeline_1['backups']), 2) + + # check closest backup correctness + self.assertEqual(timeline_6['closest-backup-id'], A1) + self.assertEqual(timeline_5['closest-backup-id'], B2) + self.assertEqual(timeline_4['closest-backup-id'], B2) + self.assertEqual(timeline_3['closest-backup-id'], A1) + self.assertEqual(timeline_2['closest-backup-id'], Y2) + + # check parent tli correctness + self.assertEqual(timeline_6['parent-tli'], 2) + self.assertEqual(timeline_5['parent-tli'], 4) + self.assertEqual(timeline_4['parent-tli'], 3) + self.assertEqual(timeline_3['parent-tli'], 2) + self.assertEqual(timeline_2['parent-tli'], 1) + self.assertEqual(timeline_1['parent-tli'], 0) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_archive_catalog_1(self): + """ + double segment - compressed and not + """ + if not self.archive_compress: + self.skipTest('You need to enable ARCHIVE_COMPRESSION ' + 'for this test to run') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_timeout': '30s', + 'checkpoint_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node, compress=True) + + node.slow_start() + + # FULL + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=2) + + wals_dir = os.path.join(backup_dir, 'wal', 'node') + original_file = os.path.join(wals_dir, '000000010000000000000001.gz') + tmp_file = os.path.join(wals_dir, '000000010000000000000001') + + with gzip.open(original_file, 'rb') as f_in, open(tmp_file, 'wb') as f_out: + shutil.copyfileobj(f_in, f_out) + + os.rename( + os.path.join(wals_dir, '000000010000000000000001'), + os.path.join(wals_dir, '000000010000000000000002')) + + show = self.show_archive(backup_dir) + + for instance in show: + timelines = instance['timelines'] + + # sanity + for timeline in timelines: + self.assertEqual( + timeline['min-segno'], + '000000010000000000000001') + self.assertEqual(timeline['status'], 'OK') + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_archive_catalog_2(self): + """ + double segment - compressed and not + """ + if not self.archive_compress: + self.skipTest('You need to enable ARCHIVE_COMPRESSION ' + 'for this test to run') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_timeout': '30s', + 'checkpoint_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node, compress=True) + + node.slow_start() + + # FULL + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=2) + + wals_dir = os.path.join(backup_dir, 'wal', 'node') + original_file = os.path.join(wals_dir, '000000010000000000000001.gz') + tmp_file = os.path.join(wals_dir, '000000010000000000000001') + + with gzip.open(original_file, 'rb') as f_in, open(tmp_file, 'wb') as f_out: + shutil.copyfileobj(f_in, f_out) + + os.rename( + os.path.join(wals_dir, '000000010000000000000001'), + os.path.join(wals_dir, '000000010000000000000002')) + + os.remove(original_file) + + show = self.show_archive(backup_dir) + + for instance in show: + timelines = instance['timelines'] + + # sanity + for timeline in timelines: + self.assertEqual( + timeline['min-segno'], + '000000010000000000000002') + self.assertEqual(timeline['status'], 'OK') + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_archive_options(self): + """ + check that '--archive-host', '--archive-user', '--archiver-port' + and '--restore-command' are working as expected. + """ + if not self.remote: + self.skipTest("You must enable PGPROBACKUP_SSH_REMOTE" + " for run this test") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node, compress=True) + + node.slow_start() + + # FULL + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=1) + + node.cleanup() + + wal_dir = os.path.join(backup_dir, 'wal', 'node') + self.restore_node( + backup_dir, 'node', node, + options=[ + '--restore-command="cp {0}/%f %p"'.format(wal_dir), + '--archive-host=localhost', + '--archive-port=22', + '--archive-user={0}'.format(self.user) + ]) + + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_conf = os.path.join(node.data_dir, 'postgresql.auto.conf') + else: + recovery_conf = os.path.join(node.data_dir, 'recovery.conf') + + with open(recovery_conf, 'r') as f: + recovery_content = f.read() + + self.assertIn( + 'restore_command = \'"cp {0}/%f %p"\''.format(wal_dir), + recovery_content) + + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=[ + '--archive-host=localhost', + '--archive-port=22', + '--archive-user={0}'.format(self.user)]) + + with open(recovery_conf, 'r') as f: + recovery_content = f.read() + + self.assertIn( + "restore_command = '\"{0}\" archive-get -B \"{1}\" --instance \"{2}\" " + "--wal-file-path=%p --wal-file-name=%f --remote-host=localhost " + "--remote-port=22 --remote-user={3}'".format( + self.probackup_path, backup_dir, 'node', self.user), + recovery_content) + + node.slow_start() + + node.safe_psql( + 'postgres', + 'select 1') + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_archive_options_1(self): + """ + check that '--archive-host', '--archive-user', '--archiver-port' + and '--restore-command' are working as expected with set-config + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node, compress=True) + + node.slow_start() + + # FULL + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=1) + + node.cleanup() + + wal_dir = os.path.join(backup_dir, 'wal', 'node') + self.set_config( + backup_dir, 'node', + options=[ + '--restore-command="cp {0}/%f %p"'.format(wal_dir), + '--archive-host=localhost', + '--archive-port=22', + '--archive-user={0}'.format(self.user)]) + self.restore_node(backup_dir, 'node', node) + + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_conf = os.path.join(node.data_dir, 'postgresql.auto.conf') + else: + recovery_conf = os.path.join(node.data_dir, 'recovery.conf') + + with open(recovery_conf, 'r') as f: + recovery_content = f.read() + + self.assertIn( + 'restore_command = \'"cp {0}/%f %p"\''.format(wal_dir), + recovery_content) + + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=[ + '--restore-command=none', + '--archive-host=localhost1', + '--archive-port=23', + '--archive-user={0}'.format(self.user) + ]) + + with open(recovery_conf, 'r') as f: + recovery_content = f.read() + + self.assertIn( + "restore_command = '\"{0}\" archive-get -B \"{1}\" --instance \"{2}\" " + "--wal-file-path=%p --wal-file-name=%f --remote-host=localhost1 " + "--remote-port=23 --remote-user={3}'".format( + self.probackup_path, backup_dir, 'node', self.user), + recovery_content) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_undefined_wal_file_path(self): + """ + check that archive-push works correct with undefined + --wal-file-path + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + if os.name == 'posix': + archive_command = '\"{0}\" archive-push -B \"{1}\" --instance \"{2}\" --wal-file-name=%f'.format( + self.probackup_path, backup_dir, 'node') + elif os.name == 'nt': + archive_command = '\"{0}\" archive-push -B \"{1}\" --instance \"{2}\" --wal-file-name=%f'.format( + self.probackup_path, backup_dir, 'node').replace("\\","\\\\") + else: + self.assertTrue(False, 'Unexpected os family') + + self.set_auto_conf( + node, + {'archive_command': archive_command}) + + node.slow_start() + node.safe_psql( + "postgres", + "create table t_heap as select i" + " as id from generate_series(0, 10) i") + self.switch_wal_segment(node) + + # check + self.assertEqual(self.show_archive(backup_dir, instance='node', tli=1)['min-segno'], '000000010000000000000001') + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_intermediate_archiving(self): + """ + check that archive-push works correct with --wal-file-path setting by user + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + node_pg_options = {} + if node.major_version >= 13: + node_pg_options['wal_keep_size'] = '0MB' + else: + node_pg_options['wal_keep_segments'] = '0' + self.set_auto_conf(node, node_pg_options) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + wal_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'intermediate_dir') + shutil.rmtree(wal_dir, ignore_errors=True) + os.makedirs(wal_dir) + if os.name == 'posix': + self.set_archiving(backup_dir, 'node', node, custom_archive_command='cp -v %p {0}/%f'.format(wal_dir)) + elif os.name == 'nt': + self.set_archiving(backup_dir, 'node', node, custom_archive_command='copy /Y "%p" "{0}\\\\%f"'.format(wal_dir.replace("\\","\\\\"))) + else: + self.assertTrue(False, 'Unexpected os family') + + node.slow_start() + node.safe_psql( + "postgres", + "create table t_heap as select i" + " as id from generate_series(0, 10) i") + self.switch_wal_segment(node) + + wal_segment = '000000010000000000000001' + + self.run_pb(["archive-push", "-B", backup_dir, + "--instance=node", "-D", node.data_dir, + "--wal-file-path", "{0}/{1}".format(wal_dir, wal_segment), "--wal-file-name", wal_segment]) + + self.assertEqual(self.show_archive(backup_dir, instance='node', tli=1)['min-segno'], wal_segment) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_waldir_outside_pgdata_archiving(self): + """ + check that archive-push works correct with symlinked waldir + """ + if self.pg_config_version < self.version_to_num('10.0'): + self.skipTest( + 'Skipped because waldir outside pgdata is supported since PG 10') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + external_wal_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'ext_wal_dir') + shutil.rmtree(external_wal_dir, ignore_errors=True) + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums', '--waldir={0}'.format(external_wal_dir)]) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + node.slow_start() + node.safe_psql( + "postgres", + "create table t_heap as select i" + " as id from generate_series(0, 10) i") + self.switch_wal_segment(node) + + # check + self.assertEqual(self.show_archive(backup_dir, instance='node', tli=1)['min-segno'], '000000010000000000000001') + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_hexadecimal_timeline(self): + """ + Check that timelines are correct. + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node, log_level='verbose') + node.slow_start() + + backup_id = self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=2) + + # create timelines + for i in range(1, 13): + # print(i) + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=['--recovery-target-timeline={0}'.format(i)]) + node.slow_start() + node.pgbench_init(scale=2) + + sleep(5) + + show = self.show_archive(backup_dir) + + timelines = show[0]['timelines'] + + print(timelines[0]) + + tli13 = timelines[0] + + self.assertEqual( + 13, + tli13['tli']) + + self.assertEqual( + 12, + tli13['parent-tli']) + + self.assertEqual( + backup_id, + tli13['closest-backup-id']) + + self.assertEqual( + '0000000D000000000000001C', + tli13['max-segno']) + + @unittest.skip("skip") + # @unittest.expectedFailure + def test_archiving_and_slots(self): + """ + Check that archiving don`t break slot + guarantee. + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s', + 'max_wal_size': '64MB'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node, log_level='verbose') + node.slow_start() + + if self.get_version(node) < 100000: + pg_receivexlog_path = self.get_bin_path('pg_receivexlog') + else: + pg_receivexlog_path = self.get_bin_path('pg_receivewal') + + # "pg_receivewal --create-slot --slot archive_slot --if-not-exists " + # "&& pg_receivewal --synchronous -Z 1 /tmp/wal --slot archive_slot --no-loop" + + self.run_binary( + [ + pg_receivexlog_path, '-p', str(node.port), '--synchronous', + '--create-slot', '--slot', 'archive_slot', '--if-not-exists' + ]) + + node.pgbench_init(scale=10) + + pg_receivexlog = self.run_binary( + [ + pg_receivexlog_path, '-p', str(node.port), '--synchronous', + '-D', os.path.join(backup_dir, 'wal', 'node'), + '--no-loop', '--slot', 'archive_slot', + '-Z', '1' + ], asynchronous=True) + + if pg_receivexlog.returncode: + self.assertFalse( + True, + 'Failed to start pg_receivexlog: {0}'.format( + pg_receivexlog.communicate()[1])) + + sleep(2) + + pg_receivexlog.kill() + + backup_id = self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=20) + + exit(1) + + def test_archive_push_sanity(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_mode': 'on', + 'archive_command': 'exit 1'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + node.slow_start() + + node.pgbench_init(scale=50) + node.stop() + + self.set_archiving(backup_dir, 'node', node) + os.remove(os.path.join(node.logs_dir, 'postgresql.log')) + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + + with open(os.path.join(node.logs_dir, 'postgresql.log'), 'r') as f: + postgres_log_content = cleanup_ptrack(f.read()) + + # print(postgres_log_content) + # make sure that .backup file is not compressed + self.assertNotIn('.backup.gz', postgres_log_content) + self.assertNotIn('WARNING', postgres_log_content) + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.restore_node( + backup_dir, 'node', replica, + data_dir=replica.data_dir, options=['-R']) + + # self.set_archiving(backup_dir, 'replica', replica, replica=True) + self.set_auto_conf(replica, {'port': replica.port}) + self.set_auto_conf(replica, {'archive_mode': 'always'}) + self.set_auto_conf(replica, {'hot_standby': 'on'}) + replica.slow_start(replica=True) + + self.wait_until_replica_catch_with_master(node, replica) + + node.pgbench_init(scale=5) + + replica.promote() + replica.pgbench_init(scale=10) + + log = tail_file(os.path.join(replica.logs_dir, 'postgresql.log'), + collect=True) + log.wait(regex=r"pushing file.*history") + log.wait(contains='archive-push completed successfully') + log.wait(regex=r"pushing file.*partial") + log.wait(contains='archive-push completed successfully') + + # make sure that .partial file is not compressed + self.assertNotIn('.partial.gz', log.content) + # make sure that .history file is not compressed + self.assertNotIn('.history.gz', log.content) + + replica.stop() + log.wait_shutdown() + + self.assertNotIn('WARNING', cleanup_ptrack(log.content)) + + output = self.show_archive( + backup_dir, 'node', as_json=False, as_text=True, + options=['--log-level-console=INFO']) + + self.assertNotIn('WARNING', output) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_archive_pg_receivexlog_partial_handling(self): + """check that archive-get delivers .partial and .gz.partial files""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + if self.get_version(node) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + node.slow_start() + + if self.get_version(node) < 100000: + app_name = 'pg_receivexlog' + pg_receivexlog_path = self.get_bin_path('pg_receivexlog') + else: + app_name = 'pg_receivewal' + pg_receivexlog_path = self.get_bin_path('pg_receivewal') + + cmdline = [ + pg_receivexlog_path, '-p', str(node.port), '--synchronous', + '-D', os.path.join(backup_dir, 'wal', 'node')] + + if self.archive_compress and node.major_version >= 10: + cmdline += ['-Z', '1'] + + env = self.test_env + env["PGAPPNAME"] = app_name + pg_receivexlog = self.run_binary(cmdline, asynchronous=True, env=env) + + if pg_receivexlog.returncode: + self.assertFalse( + True, + 'Failed to start pg_receivexlog: {0}'.format( + pg_receivexlog.communicate()[1])) + + self.set_auto_conf(node, {'synchronous_standby_names': app_name}) + self.set_auto_conf(node, {'synchronous_commit': 'on'}) + node.reload() + + # FULL + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + node.safe_psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1000000) i") + + # PAGE + self.backup_node( + backup_dir, 'node', node, backup_type='page', options=['--stream']) + + node.safe_psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(1000000,2000000) i") + + pg_receivexlog.kill() + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, node_restored.data_dir, + options=['--recovery-target=latest', '--recovery-target-action=promote']) + self.set_auto_conf(node_restored, {'port': node_restored.port}) + self.set_auto_conf(node_restored, {'hot_standby': 'off'}) + + node_restored.slow_start() + + result = node.table_checksum("t_heap") + result_new = node_restored.table_checksum("t_heap") + + self.assertEqual(result, result_new) + + @unittest.skip("skip") + def test_multi_timeline_recovery_prefetching(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + + node.pgbench_init(scale=50) + + target_xid = node.safe_psql( + 'postgres', + 'select txid_current()').rstrip() + + node.pgbench_init(scale=20) + + node.stop() + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-action=promote']) + + node.slow_start() + + node.pgbench_init(scale=20) + + target_xid = node.safe_psql( + 'postgres', + 'select txid_current()').rstrip() + + node.stop(['-m', 'immediate', '-D', node.data_dir]) + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=[ +# '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-timeline=2', +# '--recovery-target-action=promote', + '--no-validate']) + node.slow_start() + + node.pgbench_init(scale=20) + result = node.table_checksum("pgbench_accounts") + node.stop() + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=[ +# '--recovery-target-xid=100500', + '--recovery-target-timeline=3', +# '--recovery-target-action=promote', + '--no-validate']) + os.remove(os.path.join(node.logs_dir, 'postgresql.log')) + + restore_command = self.get_restore_command(backup_dir, 'node', node) + restore_command += ' -j 2 --batch-size=10 --log-level-console=VERBOSE' + + if node.major_version >= 12: + node.append_conf( + 'postgresql.auto.conf', "restore_command = '{0}'".format(restore_command)) + else: + node.append_conf( + 'recovery.conf', "restore_command = '{0}'".format(restore_command)) + + node.slow_start() + + result_new = node.table_checksum("pgbench_accounts") + + self.assertEqual(result, result_new) + + with open(os.path.join(node.logs_dir, 'postgresql.log'), 'r') as f: + postgres_log_content = f.read() + + # check that requesting of non-existing segment do not + # throwns aways prefetch + self.assertIn( + 'pg_probackup archive-get failed to ' + 'deliver WAL file: 000000030000000000000006', + postgres_log_content) + + self.assertIn( + 'pg_probackup archive-get failed to ' + 'deliver WAL file: 000000020000000000000006', + postgres_log_content) + + self.assertIn( + 'pg_probackup archive-get used prefetched ' + 'WAL segment 000000010000000000000006, prefetch state: 5/10', + postgres_log_content) + + def test_archive_get_batching_sanity(self): + """ + Make sure that batching works. + .gz file is corrupted and uncompressed is not, check that both + corruption detected and uncompressed file is used. + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + if self.get_version(node) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + node.slow_start() + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + node.pgbench_init(scale=50) + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.restore_node( + backup_dir, 'node', replica, replica.data_dir) + self.set_replica(node, replica, log_shipping=True) + + if node.major_version >= 12: + self.set_auto_conf(replica, {'restore_command': 'exit 1'}) + else: + replica.append_conf('recovery.conf', "restore_command = 'exit 1'") + + replica.slow_start(replica=True) + + # at this point replica is consistent + restore_command = self.get_restore_command(backup_dir, 'node', replica) + + restore_command += ' -j 2 --batch-size=10' + + # print(restore_command) + + if node.major_version >= 12: + self.set_auto_conf(replica, {'restore_command': restore_command}) + else: + replica.append_conf( + 'recovery.conf', "restore_command = '{0}'".format(restore_command)) + + replica.restart() + + sleep(5) + + with open(os.path.join(replica.logs_dir, 'postgresql.log'), 'r') as f: + postgres_log_content = f.read() + + self.assertIn( + 'pg_probackup archive-get completed successfully, fetched: 10/10', + postgres_log_content) + self.assertIn('used prefetched WAL segment', postgres_log_content) + self.assertIn('prefetch state: 9/10', postgres_log_content) + self.assertIn('prefetch state: 8/10', postgres_log_content) + + def test_archive_get_prefetch_corruption(self): + """ + Make sure that WAL corruption is detected. + And --prefetch-dir is honored. + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + node.slow_start() + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + node.pgbench_init(scale=50) + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.restore_node( + backup_dir, 'node', replica, replica.data_dir) + self.set_replica(node, replica, log_shipping=True) + + if node.major_version >= 12: + self.set_auto_conf(replica, {'restore_command': 'exit 1'}) + else: + replica.append_conf('recovery.conf', "restore_command = 'exit 1'") + + replica.slow_start(replica=True) + + # at this point replica is consistent + restore_command = self.get_restore_command(backup_dir, 'node', replica) + + restore_command += ' -j5 --batch-size=10 --log-level-console=VERBOSE' + #restore_command += ' --batch-size=2 --log-level-console=VERBOSE' + + if node.major_version >= 12: + self.set_auto_conf(replica, {'restore_command': restore_command}) + else: + replica.append_conf( + 'recovery.conf', "restore_command = '{0}'".format(restore_command)) + + replica.restart() + + sleep(5) + + with open(os.path.join(replica.logs_dir, 'postgresql.log'), 'r') as f: + postgres_log_content = f.read() + + self.assertIn( + 'pg_probackup archive-get completed successfully, fetched: 10/10', + postgres_log_content) + self.assertIn('used prefetched WAL segment', postgres_log_content) + self.assertIn('prefetch state: 9/10', postgres_log_content) + self.assertIn('prefetch state: 8/10', postgres_log_content) + + replica.stop() + + # generate WAL, copy it into prefetch directory, then corrupt + # some segment + node.pgbench_init(scale=20) + sleep(20) + + # now copy WAL files into prefetch directory and corrupt some of them + archive_dir = os.path.join(backup_dir, 'wal', 'node') + files = os.listdir(archive_dir) + files.sort() + + for filename in [files[-4], files[-3], files[-2], files[-1]]: + src_file = os.path.join(archive_dir, filename) + + if node.major_version >= 10: + wal_dir = 'pg_wal' + else: + wal_dir = 'pg_xlog' + + if filename.endswith('.gz'): + dst_file = os.path.join(replica.data_dir, wal_dir, 'pbk_prefetch', filename[:-3]) + with gzip.open(src_file, 'rb') as f_in, open(dst_file, 'wb') as f_out: + shutil.copyfileobj(f_in, f_out) + else: + dst_file = os.path.join(replica.data_dir, wal_dir, 'pbk_prefetch', filename) + shutil.copyfile(src_file, dst_file) + + # print(dst_file) + + # corrupt file + if files[-2].endswith('.gz'): + filename = files[-2][:-3] + else: + filename = files[-2] + + prefetched_file = os.path.join(replica.data_dir, wal_dir, 'pbk_prefetch', filename) + + with open(prefetched_file, "rb+", 0) as f: + f.seek(8192*2) + f.write(b"SURIKEN") + f.flush() + f.close + + # enable restore_command + restore_command = self.get_restore_command(backup_dir, 'node', replica) + restore_command += ' --batch-size=2 --log-level-console=VERBOSE' + + if node.major_version >= 12: + self.set_auto_conf(replica, {'restore_command': restore_command}) + else: + replica.append_conf( + 'recovery.conf', "restore_command = '{0}'".format(restore_command)) + + os.remove(os.path.join(replica.logs_dir, 'postgresql.log')) + replica.slow_start(replica=True) + + prefetch_line = 'Prefetched WAL segment {0} is invalid, cannot use it'.format(filename) + restored_line = 'LOG: restored log file "{0}" from archive'.format(filename) + tailer = tail_file(os.path.join(replica.logs_dir, 'postgresql.log')) + tailer.wait(contains=prefetch_line) + tailer.wait(contains=restored_line) + + # @unittest.skip("skip") + def test_archive_show_partial_files_handling(self): + """ + check that files with '.part', '.part.gz', '.partial' and '.partial.gz' + siffixes are handled correctly + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node, compress=False) + + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + + wals_dir = os.path.join(backup_dir, 'wal', 'node') + + # .part file + node.safe_psql( + "postgres", + "create table t1()") + + if self.get_version(node) < 100000: + filename = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_xlogfile_name_offset(pg_current_xlog_location())").rstrip() + else: + filename = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_walfile_name_offset(pg_current_wal_flush_lsn())").rstrip() + + filename = filename.decode('utf-8') + + self.switch_wal_segment(node) + + os.rename( + os.path.join(wals_dir, filename), + os.path.join(wals_dir, '{0}.part'.format(filename))) + + # .gz.part file + node.safe_psql( + "postgres", + "create table t2()") + + if self.get_version(node) < 100000: + filename = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_xlogfile_name_offset(pg_current_xlog_location())").rstrip() + else: + filename = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_walfile_name_offset(pg_current_wal_flush_lsn())").rstrip() + + filename = filename.decode('utf-8') + + self.switch_wal_segment(node) + + os.rename( + os.path.join(wals_dir, filename), + os.path.join(wals_dir, '{0}.gz.part'.format(filename))) + + # .partial file + node.safe_psql( + "postgres", + "create table t3()") + + if self.get_version(node) < 100000: + filename = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_xlogfile_name_offset(pg_current_xlog_location())").rstrip() + else: + filename = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_walfile_name_offset(pg_current_wal_flush_lsn())").rstrip() + + filename = filename.decode('utf-8') + + self.switch_wal_segment(node) + + os.rename( + os.path.join(wals_dir, filename), + os.path.join(wals_dir, '{0}.partial'.format(filename))) + + # .gz.partial file + node.safe_psql( + "postgres", + "create table t4()") + + if self.get_version(node) < 100000: + filename = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_xlogfile_name_offset(pg_current_xlog_location())").rstrip() + else: + filename = node.safe_psql( + "postgres", + "SELECT file_name " + "FROM pg_walfile_name_offset(pg_current_wal_flush_lsn())").rstrip() + + filename = filename.decode('utf-8') + + self.switch_wal_segment(node) + + os.rename( + os.path.join(wals_dir, filename), + os.path.join(wals_dir, '{0}.gz.partial'.format(filename))) + + self.show_archive(backup_dir, 'node', options=['--log-level-file=VERBOSE']) + + with open(os.path.join(backup_dir, 'log', 'pg_probackup.log'), 'r') as f: + log_content = f.read() + + self.assertNotIn( + 'WARNING', + log_content) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_archive_empty_history_file(self): + """ + https://github.com/postgrespro/pg_probackup/issues/326 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + node.slow_start() + node.pgbench_init(scale=5) + + # FULL + self.backup_node(backup_dir, 'node', node) + + node.pgbench_init(scale=5) + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target=latest', + '--recovery-target-action=promote']) + + # Node in timeline 2 + node.slow_start() + + node.pgbench_init(scale=5) + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target=latest', + '--recovery-target-timeline=2', + '--recovery-target-action=promote']) + + # Node in timeline 3 + node.slow_start() + + node.pgbench_init(scale=5) + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target=latest', + '--recovery-target-timeline=3', + '--recovery-target-action=promote']) + + # Node in timeline 4 + node.slow_start() + node.pgbench_init(scale=5) + + # Truncate history files + for tli in range(2, 5): + file = os.path.join( + backup_dir, 'wal', 'node', '0000000{0}.history'.format(tli)) + with open(file, "w+") as f: + f.truncate() + + timelines = self.show_archive(backup_dir, 'node', options=['--log-level-file=INFO']) + + # check that all timelines has zero switchpoint + for timeline in timelines: + self.assertEqual(timeline['switchpoint'], '0/0') + + log_file = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(log_file, 'r') as f: + log_content = f.read() + wal_dir = os.path.join(backup_dir, 'wal', 'node') + + self.assertIn( + 'WARNING: History file is corrupted or missing: "{0}"'.format(os.path.join(wal_dir, '00000002.history')), + log_content) + self.assertIn( + 'WARNING: History file is corrupted or missing: "{0}"'.format(os.path.join(wal_dir, '00000003.history')), + log_content) + self.assertIn( + 'WARNING: History file is corrupted or missing: "{0}"'.format(os.path.join(wal_dir, '00000004.history')), + log_content) + + +def cleanup_ptrack(log_content): + # PBCKP-423 - need to clean ptrack warning + ptrack_is_not = 'Ptrack 1.X is not supported anymore' + if ptrack_is_not in log_content: + lines = [line for line in log_content.splitlines() + if ptrack_is_not not in line] + log_content = "".join(lines) + return log_content + + +# TODO test with multiple not archived segments. +# TODO corrupted file in archive. + +# important - switchpoint may be NullOffset LSN and not actually existing in archive to boot. +# so write WAL validation code accordingly + +# change wal-seg-size +# +# +#t3 ---------------- +# / +#t2 ---------------- +# / +#t1 -A-------- +# +# + + +#t3 ---------------- +# / +#t2 ---------------- +# / +#t1 -A-------- +# diff --git a/tests/auth_test.py b/tests/auth_test.py index d9b5c283e..32cabc4a1 100644 --- a/tests/auth_test.py +++ b/tests/auth_test.py @@ -30,20 +30,23 @@ def test_backup_via_unprivileged_user(self): run a backups without EXECUTE rights on certain functions """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'max_wal_senders': '2'} - ) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() + if self.ptrack: + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + node.safe_psql("postgres", "CREATE ROLE backup with LOGIN") try: @@ -53,18 +56,39 @@ def test_backup_via_unprivileged_user(self): 1, 0, "Expecting Error due to missing grant on EXECUTE.") except ProbackupException as e: - self.assertIn( - "ERROR: query failed: ERROR: permission denied " - "for function pg_start_backup", e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) + if self.get_version(node) < 150000: + self.assertIn( + "ERROR: query failed: ERROR: permission denied " + "for function pg_start_backup", e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + else: + self.assertIn( + "ERROR: query failed: ERROR: permission denied " + "for function pg_backup_start", e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) - node.safe_psql( - "postgres", - "GRANT EXECUTE ON FUNCTION" - " pg_start_backup(text, boolean, boolean) TO backup;") + if self.get_version(node) < 150000: + node.safe_psql( + "postgres", + "GRANT EXECUTE ON FUNCTION" + " pg_start_backup(text, boolean, boolean) TO backup;") + else: + node.safe_psql( + "postgres", + "GRANT EXECUTE ON FUNCTION" + " pg_backup_start(text, boolean) TO backup;") + + if self.get_version(node) < 100000: + node.safe_psql( + 'postgres', + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup") + else: + node.safe_psql( + 'postgres', + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup") - time.sleep(1) try: self.backup_node( backup_dir, 'node', node, options=['-U', 'backup']) @@ -84,8 +108,6 @@ def test_backup_via_unprivileged_user(self): "GRANT EXECUTE ON FUNCTION" " pg_create_restore_point(text) TO backup;") - time.sleep(1) - try: self.backup_node( backup_dir, 'node', node, options=['-U', 'backup']) @@ -93,25 +115,32 @@ def test_backup_via_unprivileged_user(self): 1, 0, "Expecting Error due to missing grant on EXECUTE.") except ProbackupException as e: - self.assertIn( - "ERROR: query failed: ERROR: permission denied " - "for function pg_stop_backup", e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) + if self.get_version(node) < 150000: + self.assertIn( + "ERROR: Query failed: ERROR: permission denied " + "for function pg_stop_backup", e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + else: + self.assertIn( + "ERROR: Query failed: ERROR: permission denied " + "for function pg_backup_stop", e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) if self.get_version(node) < self.version_to_num('10.0'): node.safe_psql( "postgres", "GRANT EXECUTE ON FUNCTION pg_stop_backup(boolean) TO backup") - else: + elif self.get_version(node) < self.version_to_num('15.0'): node.safe_psql( "postgres", - "GRANT EXECUTE ON FUNCTION " - "pg_stop_backup(boolean, boolean) TO backup") - # Do this for ptrack backups + "GRANT EXECUTE ON FUNCTION pg_stop_backup() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_stop_backup(boolean, boolean) TO backup;") + else: node.safe_psql( "postgres", - "GRANT EXECUTE ON FUNCTION pg_stop_backup() TO backup") + "GRANT EXECUTE ON FUNCTION pg_backup_stop(boolean) TO backup;") self.backup_node( backup_dir, 'node', node, options=['-U', 'backup']) @@ -124,63 +153,29 @@ def test_backup_via_unprivileged_user(self): node.safe_psql( "test1", "create table t1 as select generate_series(0,100)") - node.append_conf("postgresql.auto.conf", "ptrack_enable = 'on'") node.stop() node.slow_start() - try: - self.backup_node( - backup_dir, 'node', node, options=['-U', 'backup']) - self.assertEqual( - 1, 0, - "Expecting Error due to missing grant on clearing ptrack_files.") - except ProbackupException as e: - self.assertIn( - "ERROR: must be superuser or replication role to clear ptrack files\n" - "query was: SELECT pg_catalog.pg_ptrack_clear()", e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - time.sleep(1) - - try: - self.backup_node( - backup_dir, 'node', node, - backup_type='ptrack', options=['-U', 'backup']) - self.assertEqual( - 1, 0, - "Expecting Error due to missing grant on clearing ptrack_files.") - except ProbackupException as e: - self.assertIn( - "ERROR: must be superuser or replication role read ptrack files\n" - "query was: select pg_catalog.pg_ptrack_control_lsn()", e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - node.safe_psql( "postgres", "ALTER ROLE backup REPLICATION") - time.sleep(1) - # FULL self.backup_node( - backup_dir, 'node', node, - options=['-U', 'backup']) + backup_dir, 'node', node, options=['-U', 'backup']) # PTRACK - self.backup_node( - backup_dir, 'node', node, - backup_type='ptrack', options=['-U', 'backup']) - - # Clean after yourself - self.del_test_dir(module_name, fname) + if self.ptrack: + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['-U', 'backup']) class AuthTest(unittest.TestCase): pb = None node = None + # TODO move to object scope, replace module_name @classmethod def setUpClass(cls): @@ -194,7 +189,10 @@ def setUpClass(cls): set_replication=True, initdb_params=['--data-checksums', '--auth-host=md5'] ) - modify_pg_hba(cls.node) + + cls.username = cls.pb.get_username() + + cls.modify_pg_hba(cls.node) cls.pb.init_pb(cls.backup_dir) cls.pb.add_instance(cls.backup_dir, cls.node.name, cls.node) @@ -204,24 +202,54 @@ def setUpClass(cls): except StartNodeException: raise unittest.skip("Node hasn't started") - cls.node.safe_psql( - "postgres", - "CREATE ROLE backup WITH LOGIN PASSWORD 'password'; " - "GRANT USAGE ON SCHEMA pg_catalog TO backup; " - "GRANT EXECUTE ON FUNCTION current_setting(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_is_in_recovery() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_start_backup(text, boolean, boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_stop_backup() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_stop_backup(boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_create_restore_point(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_switch_xlog() TO backup; " - "GRANT EXECUTE ON FUNCTION txid_current() TO backup; " - "GRANT EXECUTE ON FUNCTION txid_current_snapshot() TO backup; " - "GRANT EXECUTE ON FUNCTION txid_snapshot_xmax(txid_snapshot) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_ptrack_clear() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_ptrack_get_and_clear(oid, oid) TO backup;") + if cls.pb.get_version(cls.node) < 100000: + cls.node.safe_psql( + "postgres", + "CREATE ROLE backup WITH LOGIN PASSWORD 'password'; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT EXECUTE ON FUNCTION current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_stop_backup() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_stop_backup(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_switch_xlog() TO backup; " + "GRANT EXECUTE ON FUNCTION txid_current() TO backup; " + "GRANT EXECUTE ON FUNCTION txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION txid_snapshot_xmax(txid_snapshot) TO backup;") + elif cls.pb.get_version(cls.node) < 150000: + cls.node.safe_psql( + "postgres", + "CREATE ROLE backup WITH LOGIN PASSWORD 'password'; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT EXECUTE ON FUNCTION current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_stop_backup() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_stop_backup(boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION txid_current() TO backup; " + "GRANT EXECUTE ON FUNCTION txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION txid_snapshot_xmax(txid_snapshot) TO backup;") + else: + cls.node.safe_psql( + "postgres", + "CREATE ROLE backup WITH LOGIN PASSWORD 'password'; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT EXECUTE ON FUNCTION current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_backup_start(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_backup_stop(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION txid_current() TO backup; " + "GRANT EXECUTE ON FUNCTION txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION txid_snapshot_xmax(txid_snapshot) TO backup;") + cls.pgpass_file = os.path.join(os.path.expanduser('~'), '.pgpass') + # TODO move to object scope, replace module_name @classmethod def tearDownClass(cls): cls.node.cleanup() @@ -229,12 +257,13 @@ def tearDownClass(cls): @unittest.skipIf(skip_test, "Module pexpect isn't installed. You need to install it.") def setUp(self): - self.cmd = ['backup', + self.pb_cmd = ['backup', '-B', self.backup_dir, '--instance', self.node.name, '-h', '127.0.0.1', '-p', str(self.node.port), '-U', 'backup', + '-d', 'postgres', '-b', 'FULL' ] @@ -254,44 +283,31 @@ def test_empty_password(self): """ Test case: PGPB_AUTH03 - zero password length """ try: self.assertIn("ERROR: no password supplied", - str(run_pb_with_auth([self.pb.probackup_path] + self.cmd, '\0\r\n')) - ) + self.run_pb_with_auth('\0\r\n')) except (TIMEOUT, ExceptionPexpect) as e: self.fail(e.value) def test_wrong_password(self): """ Test case: PGPB_AUTH04 - incorrect password """ - try: - self.assertIn("password authentication failed", - str(run_pb_with_auth([self.pb.probackup_path] + self.cmd, 'wrong_password\r\n')) - ) - except (TIMEOUT, ExceptionPexpect) as e: - self.fail(e.value) + self.assertIn("password authentication failed", + self.run_pb_with_auth('wrong_password\r\n')) def test_right_password(self): """ Test case: PGPB_AUTH01 - correct password """ - try: - self.assertIn("completed", - str(run_pb_with_auth([self.pb.probackup_path] + self.cmd, 'password\r\n')) - ) - except (TIMEOUT, ExceptionPexpect) as e: - self.fail(e.value) + self.assertIn("completed", + self.run_pb_with_auth('password\r\n')) def test_right_password_and_wrong_pgpass(self): """ Test case: PGPB_AUTH05 - correct password and incorrect .pgpass (-W)""" line = ":".join(['127.0.0.1', str(self.node.port), 'postgres', 'backup', 'wrong_password']) - create_pgpass(self.pgpass_file, line) - try: - self.assertIn("completed", - str(run_pb_with_auth([self.pb.probackup_path] + self.cmd + ['-W'], 'password\r\n')) - ) - except (TIMEOUT, ExceptionPexpect) as e: - self.fail(e.value) + self.create_pgpass(self.pgpass_file, line) + self.assertIn("completed", + self.run_pb_with_auth('password\r\n', add_args=["-W"])) def test_ctrl_c_event(self): """ Test case: PGPB_AUTH02 - send interrupt signal """ try: - run_pb_with_auth([self.pb.probackup_path] + self.cmd, kill=True) + self.run_pb_with_auth(kill=True) except TIMEOUT: self.fail("Error: CTRL+C event ignored") @@ -299,91 +315,74 @@ def test_pgpassfile_env(self): """ Test case: PGPB_AUTH06 - set environment var PGPASSFILE """ path = os.path.join(self.pb.tmp_path, module_name, 'pgpass.conf') line = ":".join(['127.0.0.1', str(self.node.port), 'postgres', 'backup', 'password']) - create_pgpass(path, line) + self.create_pgpass(path, line) self.pb.test_env["PGPASSFILE"] = path - try: - self.assertEqual( - "OK", - self.pb.show_pb(self.backup_dir, self.node.name, self.pb.run_pb(self.cmd + ['-w']))["status"], - "ERROR: Full backup status is not valid." - ) - except ProbackupException as e: - self.fail(e) + self.assertEqual( + "OK", + self.pb.show_pb(self.backup_dir, self.node.name, self.pb.run_pb(self.pb_cmd + ['-w']))["status"], + "ERROR: Full backup status is not valid." + ) def test_pgpass(self): """ Test case: PGPB_AUTH07 - Create file .pgpass in home dir. """ line = ":".join(['127.0.0.1', str(self.node.port), 'postgres', 'backup', 'password']) - create_pgpass(self.pgpass_file, line) - try: - self.assertEqual( - "OK", - self.pb.show_pb(self.backup_dir, self.node.name, self.pb.run_pb(self.cmd + ['-w']))["status"], - "ERROR: Full backup status is not valid." - ) - except ProbackupException as e: - self.fail(e) + self.create_pgpass(self.pgpass_file, line) + self.assertEqual( + "OK", + self.pb.show_pb(self.backup_dir, self.node.name, self.pb.run_pb(self.pb_cmd + ['-w']))["status"], + "ERROR: Full backup status is not valid." + ) def test_pgpassword(self): """ Test case: PGPB_AUTH08 - set environment var PGPASSWORD """ self.pb.test_env["PGPASSWORD"] = "password" - try: - self.assertEqual( - "OK", - self.pb.show_pb(self.backup_dir, self.node.name, self.pb.run_pb(self.cmd + ['-w']))["status"], - "ERROR: Full backup status is not valid." - ) - except ProbackupException as e: - self.fail(e) + self.assertEqual( + "OK", + self.pb.show_pb(self.backup_dir, self.node.name, self.pb.run_pb(self.pb_cmd + ['-w']))["status"], + "ERROR: Full backup status is not valid." + ) def test_pgpassword_and_wrong_pgpass(self): """ Test case: PGPB_AUTH09 - Check priority between PGPASSWORD and .pgpass file""" line = ":".join(['127.0.0.1', str(self.node.port), 'postgres', 'backup', 'wrong_password']) - create_pgpass(self.pgpass_file, line) + self.create_pgpass(self.pgpass_file, line) self.pb.test_env["PGPASSWORD"] = "password" - try: - self.assertEqual( - "OK", - self.pb.show_pb(self.backup_dir, self.node.name, self.pb.run_pb(self.cmd + ['-w']))["status"], - "ERROR: Full backup status is not valid." - ) - except ProbackupException as e: - self.fail(e) - + self.assertEqual( + "OK", + self.pb.show_pb(self.backup_dir, self.node.name, self.pb.run_pb(self.pb_cmd + ['-w']))["status"], + "ERROR: Full backup status is not valid." + ) -def run_pb_with_auth(cmd, password=None, kill=False): - try: - with spawn(" ".join(cmd), encoding='utf-8', timeout=10) as probackup: + def run_pb_with_auth(self, password=None, add_args = [], kill=False): + with spawn(self.pb.probackup_path, self.pb_cmd + add_args, encoding='utf-8', timeout=10) as probackup: result = probackup.expect(u"Password for user .*:", 5) if kill: probackup.kill(signal.SIGINT) elif result == 0: probackup.sendline(password) probackup.expect(EOF) - return probackup.before + return str(probackup.before) else: raise ExceptionPexpect("Other pexpect errors.") - except TIMEOUT: - raise TIMEOUT("Timeout error.") - except ExceptionPexpect: - raise ExceptionPexpect("Pexpect error.") - - -def modify_pg_hba(node): - """ - Description: - Add trust authentication for user postgres. Need for add new role and set grant. - :param node: - :return None: - """ - hba_conf = os.path.join(node.data_dir, "pg_hba.conf") - with open(hba_conf, 'r+') as fio: - data = fio.read() - fio.seek(0) - fio.write('host\tall\tpostgres\t127.0.0.1/0\ttrust\n' + data) - - -def create_pgpass(path, line): - with open(path, 'w') as passfile: - # host:port:db:username:password - passfile.write(line) - os.chmod(path, 0o600) + + + @classmethod + def modify_pg_hba(cls, node): + """ + Description: + Add trust authentication for user postgres. Need for add new role and set grant. + :param node: + :return None: + """ + hba_conf = os.path.join(node.data_dir, "pg_hba.conf") + with open(hba_conf, 'r+') as fio: + data = fio.read() + fio.seek(0) + fio.write('host\tall\t%s\t127.0.0.1/0\ttrust\n%s' % (cls.username, data)) + + + def create_pgpass(self, path, line): + with open(path, 'w') as passfile: + # host:port:db:username:password + passfile.write(line) + os.chmod(path, 0o600) diff --git a/tests/backup.py b/tests/backup.py deleted file mode 100644 index 81370a173..000000000 --- a/tests/backup.py +++ /dev/null @@ -1,2039 +0,0 @@ -import unittest -import os -from time import sleep -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException -import shutil - - -module_name = 'backup' - - -class BackupTest(ProbackupTest, unittest.TestCase): - - # @unittest.skip("skip") - # @unittest.expectedFailure - # PGPRO-707 - def test_backup_modes_archive(self): - """standart backup modes with ARCHIVE WAL method""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node(backup_dir, 'node', node) - show_backup = self.show_pb(backup_dir, 'node')[0] - - self.assertEqual(show_backup['status'], "OK") - self.assertEqual(show_backup['backup-mode'], "FULL") - - # postmaster.pid and postmaster.opts shouldn't be copied - excluded = True - db_dir = os.path.join( - backup_dir, "backups", 'node', backup_id, "database") - - for f in os.listdir(db_dir): - if ( - os.path.isfile(os.path.join(db_dir, f)) and - ( - f == "postmaster.pid" or - f == "postmaster.opts" - ) - ): - excluded = False - self.assertEqual(excluded, True) - - # page backup mode - page_backup_id = self.backup_node( - backup_dir, 'node', node, backup_type="page") - - # print self.show_pb(node) - show_backup = self.show_pb(backup_dir, 'node')[1] - self.assertEqual(show_backup['status'], "OK") - self.assertEqual(show_backup['backup-mode'], "PAGE") - - # Check parent backup - self.assertEqual( - backup_id, - self.show_pb( - backup_dir, 'node', - backup_id=show_backup['id'])["parent-backup-id"]) - - # ptrack backup mode - self.backup_node(backup_dir, 'node', node, backup_type="ptrack") - - show_backup = self.show_pb(backup_dir, 'node')[2] - self.assertEqual(show_backup['status'], "OK") - self.assertEqual(show_backup['backup-mode'], "PTRACK") - - # Check parent backup - self.assertEqual( - page_backup_id, - self.show_pb( - backup_dir, 'node', - backup_id=show_backup['id'])["parent-backup-id"]) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_smooth_checkpoint(self): - """full backup with smooth checkpoint""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.backup_node( - backup_dir, 'node', node, - options=["-C"]) - self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") - node.stop() - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_incremental_backup_without_full(self): - """page-level backup without validated full backup""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums'], - pg_options={'ptrack_enable': 'on'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - try: - self.backup_node(backup_dir, 'node', node, backup_type="page") - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because page backup should not be possible " - "without valid full backup.\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "ERROR: Valid backup on current timeline 1 is not found. " - "Create new FULL backup before an incremental one.", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - sleep(1) - - try: - self.backup_node(backup_dir, 'node', node, backup_type="ptrack") - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because page backup should not be possible " - "without valid full backup.\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "ERROR: Valid backup on current timeline 1 is not found. " - "Create new FULL backup before an incremental one.", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - self.assertEqual( - self.show_pb(backup_dir, 'node')[0]['status'], - "ERROR") - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_incremental_backup_corrupt_full(self): - """page-level backup with corrupted full backup""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node(backup_dir, 'node', node) - file = os.path.join( - backup_dir, "backups", "node", backup_id, - "database", "postgresql.conf") - os.remove(file) - - try: - self.validate_pb(backup_dir, 'node') - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because of validation of corrupted backup.\n" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertTrue( - "INFO: Validate backups of the instance 'node'" in e.message and - "WARNING: Backup file".format( - file) in e.message and - "is not found".format(file) in e.message and - "WARNING: Backup {0} data files are corrupted".format( - backup_id) in e.message and - "WARNING: Some backups are not valid" in e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - try: - self.backup_node(backup_dir, 'node', node, backup_type="page") - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because page backup should not be possible " - "without valid full backup.\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "ERROR: Valid backup on current timeline 1 is not found. " - "Create new FULL backup before an incremental one.", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - self.assertEqual( - self.show_pb(backup_dir, 'node', backup_id)['status'], "CORRUPT") - self.assertEqual( - self.show_pb(backup_dir, 'node')[1]['status'], "ERROR") - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_ptrack_threads(self): - """ptrack multi thread backup mode""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums'], - pg_options={'ptrack_enable': 'on'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.backup_node( - backup_dir, 'node', node, - backup_type="full", options=["-j", "4"]) - self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") - - self.backup_node( - backup_dir, 'node', node, - backup_type="ptrack", options=["-j", "4"]) - self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_ptrack_threads_stream(self): - """ptrack multi thread backup mode and stream""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream"]) - - self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") - self.backup_node( - backup_dir, 'node', node, - backup_type="ptrack", options=["-j", "4", "--stream"]) - self.assertEqual(self.show_pb(backup_dir, 'node')[1]['status'], "OK") - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_page_corruption_heal_via_ptrack_1(self): - """make node, corrupt some page, check that backup failed""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - self.backup_node( - backup_dir, 'node', node, - backup_type="full", options=["-j", "4", "--stream"]) - - node.safe_psql( - "postgres", - "create table t_heap as select 1 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i") - node.safe_psql( - "postgres", - "CHECKPOINT;") - - heap_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('t_heap')").rstrip() - - with open(os.path.join(node.data_dir, heap_path), "rb+", 0) as f: - f.seek(9000) - f.write(b"bla") - f.flush() - f.close - - self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream", "--log-level-file=verbose"]) - - # open log file and check - with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: - log_content = f.read() - self.assertIn('block 1, try to fetch via SQL', log_content) - self.assertIn('SELECT pg_catalog.pg_ptrack_get_block', log_content) - f.close - - self.assertTrue( - self.show_pb(backup_dir, 'node')[1]['status'] == 'OK', - "Backup Status should be OK") - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_page_corruption_heal_via_ptrack_2(self): - """make node, corrupt some page, check that backup failed""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream"]) - - node.safe_psql( - "postgres", - "create table t_heap as select 1 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i") - node.safe_psql( - "postgres", - "CHECKPOINT;") - - heap_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('t_heap')").rstrip() - node.stop() - - with open(os.path.join(node.data_dir, heap_path), "rb+", 0) as f: - f.seek(9000) - f.write(b"bla") - f.flush() - f.close - node.slow_start() - - try: - self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream", '--log-level-console=LOG']) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because of page " - "corruption in PostgreSQL instance.\n" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - if self.remote: - self.assertTrue( - "WARNING: File" in e.message and - "try to fetch via SQL" in e.message and - "WARNING: page verification failed, " - "calculated checksum" in e.message and - "ERROR: query failed: " - "ERROR: invalid page in block" in e.message and - "query was: SELECT pg_catalog.pg_ptrack_get_block_2" in e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - else: - self.assertTrue( - "LOG: File" in e.message and - "blknum" in e.message and - "have wrong checksum" in e.message and - "try to fetch via SQL" in e.message and - "WARNING: page verification failed, " - "calculated checksum" in e.message and - "ERROR: query failed: " - "ERROR: invalid page in block" in e.message and - "query was: SELECT pg_catalog.pg_ptrack_get_block_2" in e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - self.assertTrue( - self.show_pb(backup_dir, 'node')[1]['status'] == 'ERROR', - "Backup Status should be ERROR") - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_backup_detect_corruption(self): - """make node, corrupt some page, check that backup failed""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "create table t_heap as select 1 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i") - node.safe_psql( - "postgres", - "CHECKPOINT;") - - heap_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('t_heap')").rstrip() - - node.stop() - - with open(os.path.join(node.data_dir, heap_path), "rb+", 0) as f: - f.seek(9000) - f.write(b"bla") - f.flush() - f.close - - node.slow_start() - - try: - self.backup_node( - backup_dir, 'node', node, - backup_type="full", options=["-j", "4", "--stream"]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because tablespace mapping is incorrect" - "\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - if self.ptrack: - self.assertTrue( - 'WARNING: page verification failed, ' - 'calculated checksum' in e.message and - 'ERROR: query failed: ERROR: ' - 'invalid page in block 1 of relation' in e.message and - 'ERROR: Data files transferring failed' in e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - else: - if self.remote: - self.assertTrue( - "ERROR: Failed to read file" in e.message and - "data file checksum mismatch" in e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - else: - self.assertIn( - 'WARNING: Corruption detected in file', - e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - self.assertIn( - 'ERROR: Data file corruption', - e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_backup_truncate_misaligned(self): - """ - make node, truncate file to size not even to BLCKSIZE, - take backup - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "create table t_heap as select 1 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,100000) i") - - node.safe_psql( - "postgres", - "CHECKPOINT;") - - heap_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('t_heap')").rstrip() - - heap_size = node.safe_psql( - "postgres", - "select pg_relation_size('t_heap')") - - with open(os.path.join(node.data_dir, heap_path), "rb+", 0) as f: - f.truncate(int(heap_size) - 4096) - f.flush() - f.close - - output = self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream"], return_id=False) - - self.assertIn("WARNING: File", output) - self.assertIn("invalid file size", output) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_tablespace_in_pgdata_pgpro_1376(self): - """PGPRO-1376 """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - self.create_tblspace_in_node( - node, 'tblspace1', - tblspc_path=( - os.path.join( - node.data_dir, 'somedirectory', '100500')) - ) - - self.create_tblspace_in_node( - node, 'tblspace2', - tblspc_path=(os.path.join(node.data_dir)) - ) - - node.safe_psql( - "postgres", - "create table t_heap1 tablespace tblspace1 as select 1 as id, " - "md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i") - - node.safe_psql( - "postgres", - "create table t_heap2 tablespace tblspace2 as select 1 as id, " - "md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i") - - backup_id_1 = self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream"]) - - node.safe_psql( - "postgres", - "drop table t_heap2") - node.safe_psql( - "postgres", - "drop tablespace tblspace2") - - self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream"]) - - pgdata = self.pgdata_content(node.data_dir) - - relfilenode = node.safe_psql( - "postgres", - "select 't_heap1'::regclass::oid" - ).rstrip() - - list = [] - for root, dirs, files in os.walk(os.path.join( - backup_dir, 'backups', 'node', backup_id_1)): - for file in files: - if file == relfilenode: - path = os.path.join(root, file) - list = list + [path] - - # We expect that relfilenode can be encountered only once - if len(list) > 1: - message = "" - for string in list: - message = message + string + "\n" - self.assertEqual( - 1, 0, - "Following file copied twice by backup:\n {0}".format( - message) - ) - - node.cleanup() - - self.restore_node( - backup_dir, 'node', node, options=["-j", "4"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_basic_tablespace_handling(self): - """ - make node, take full backup, check that restore with - tablespace mapping will end with error, take page backup, - check that restore with tablespace mapping will end with - success - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream"]) - - tblspace1_old_path = self.get_tblspace_path(node, 'tblspace1_old') - tblspace2_old_path = self.get_tblspace_path(node, 'tblspace2_old') - - self.create_tblspace_in_node( - node, 'some_lame_tablespace') - - self.create_tblspace_in_node( - node, 'tblspace1', - tblspc_path=tblspace1_old_path) - - self.create_tblspace_in_node( - node, 'tblspace2', - tblspc_path=tblspace2_old_path) - - node.safe_psql( - "postgres", - "create table t_heap_lame tablespace some_lame_tablespace " - "as select 1 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i") - - node.safe_psql( - "postgres", - "create table t_heap2 tablespace tblspace2 as select 1 as id, " - "md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i") - - tblspace1_new_path = self.get_tblspace_path(node, 'tblspace1_new') - tblspace2_new_path = self.get_tblspace_path(node, 'tblspace2_new') - - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - node_restored.cleanup() - - try: - self.restore_node( - backup_dir, 'node', node_restored, - options=[ - "-j", "4", - "-T", "{0}={1}".format( - tblspace1_old_path, tblspace1_new_path), - "-T", "{0}={1}".format( - tblspace2_old_path, tblspace2_new_path)]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because tablespace mapping is incorrect" - "\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertTrue( - 'ERROR: --tablespace-mapping option' in e.message and - 'have an entry in tablespace_map file' in e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - node.safe_psql( - "postgres", - "drop table t_heap_lame") - - node.safe_psql( - "postgres", - "drop tablespace some_lame_tablespace") - - self.backup_node( - backup_dir, 'node', node, backup_type="delta", - options=["-j", "4", "--stream"]) - - self.restore_node( - backup_dir, 'node', node_restored, - options=[ - "-j", "4", - "-T", "{0}={1}".format( - tblspace1_old_path, tblspace1_new_path), - "-T", "{0}={1}".format( - tblspace2_old_path, tblspace2_new_path)]) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_tablespace_handling_1(self): - """ - make node with tablespace A, take full backup, check that restore with - tablespace mapping of tablespace B will end with error - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - tblspace1_old_path = self.get_tblspace_path(node, 'tblspace1_old') - tblspace2_old_path = self.get_tblspace_path(node, 'tblspace2_old') - - tblspace_new_path = self.get_tblspace_path(node, 'tblspace_new') - - self.create_tblspace_in_node( - node, 'tblspace1', - tblspc_path=tblspace1_old_path) - - self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream"]) - - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - node_restored.cleanup() - - try: - self.restore_node( - backup_dir, 'node', node_restored, - options=[ - "-j", "4", - "-T", "{0}={1}".format( - tblspace2_old_path, tblspace_new_path)]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because tablespace mapping is incorrect" - "\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertTrue( - 'ERROR: --tablespace-mapping option' in e.message and - 'have an entry in tablespace_map file' in e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_tablespace_handling_2(self): - """ - make node without tablespaces, take full backup, check that restore with - tablespace mapping will end with error - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - tblspace1_old_path = self.get_tblspace_path(node, 'tblspace1_old') - tblspace_new_path = self.get_tblspace_path(node, 'tblspace_new') - - self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream"]) - - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - node_restored.cleanup() - - try: - self.restore_node( - backup_dir, 'node', node_restored, - options=[ - "-j", "4", - "-T", "{0}={1}".format( - tblspace1_old_path, tblspace_new_path)]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because tablespace mapping is incorrect" - "\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertTrue( - 'ERROR: --tablespace-mapping option' in e.message and - 'have an entry in tablespace_map file' in e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_drop_rel_during_backup_delta(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "create table t_heap as select i" - " as id from generate_series(0,100) i") - - relative_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('t_heap')").rstrip() - - absolute_path = os.path.join(node.data_dir, relative_path) - - # FULL backup - self.backup_node(backup_dir, 'node', node, options=['--stream']) - - # DELTA backup - gdb = self.backup_node( - backup_dir, 'node', node, backup_type='delta', - gdb=True, options=['--log-level-file=verbose']) - - gdb.set_breakpoint('backup_files') - gdb.run_until_break() - - # REMOVE file - node.safe_psql( - "postgres", - "DROP TABLE t_heap") - - node.safe_psql( - "postgres", - "CHECKPOINT") - - # File removed, we can proceed with backup - gdb.continue_execution_until_exit() - - pgdata = self.pgdata_content(node.data_dir) - - with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: - log_content = f.read() - self.assertTrue( - 'LOG: File "{0}" is not found'.format(absolute_path) in log_content, - 'File "{0}" should be deleted but it`s not'.format(absolute_path)) - - node.cleanup() - self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) - - # Physical comparison - pgdata_restored = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_drop_rel_during_backup_page(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "create table t_heap as select i" - " as id from generate_series(0,100) i") - - relative_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('t_heap')").rstrip() - - absolute_path = os.path.join(node.data_dir, relative_path) - - # FULL backup - self.backup_node(backup_dir, 'node', node, options=['--stream']) - - # PAGE backup - gdb = self.backup_node( - backup_dir, 'node', node, backup_type='page', - gdb=True, options=['--log-level-file=verbose']) - - gdb.set_breakpoint('backup_files') - gdb.run_until_break() - - # REMOVE file - os.remove(absolute_path) - - # File removed, we can proceed with backup - gdb.continue_execution_until_exit() - - pgdata = self.pgdata_content(node.data_dir) - - with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: - log_content = f.read() - self.assertTrue( - 'LOG: File "{0}" is not found'.format(absolute_path) in log_content, - 'File "{0}" should be deleted but it`s not'.format(absolute_path)) - - node.cleanup() - self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) - - # Physical comparison - pgdata_restored = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_drop_rel_during_backup_ptrack(self): - """""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "create table t_heap as select i" - " as id from generate_series(0,100) i") - - relative_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('t_heap')").rstrip() - - absolute_path = os.path.join(node.data_dir, relative_path) - - # FULL backup - self.backup_node(backup_dir, 'node', node, options=['--stream']) - - # PTRACK backup - gdb = self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - gdb=True, options=['--log-level-file=verbose']) - - gdb.set_breakpoint('backup_files') - gdb.run_until_break() - - # REMOVE file - os.remove(absolute_path) - - # File removed, we can proceed with backup - gdb.continue_execution_until_exit() - - pgdata = self.pgdata_content(node.data_dir) - - with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: - log_content = f.read() - self.assertTrue( - 'LOG: File "{0}" is not found'.format(absolute_path) in log_content, - 'File "{0}" should be deleted but it`s not'.format(absolute_path)) - - node.cleanup() - self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) - - # Physical comparison - pgdata_restored = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_persistent_slot_for_stream_backup(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'max_wal_size': '40MB'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "SELECT pg_create_physical_replication_slot('slot_1')") - - # FULL backup - self.backup_node( - backup_dir, 'node', node, - options=['--stream', '--slot=slot_1']) - - # FULL backup - self.backup_node( - backup_dir, 'node', node, - options=['--stream', '--slot=slot_1']) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_basic_temp_slot_for_stream_backup(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'max_wal_size': '40MB'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # FULL backup - self.backup_node( - backup_dir, 'node', node, - options=['--stream', '--temp-slot']) - - if self.get_version(node) < self.version_to_num('10.0'): - return unittest.skip('You need PostgreSQL >= 10 for this test') - else: - pg_receivexlog_path = self.get_bin_path('pg_receivewal') - - # FULL backup - self.backup_node( - backup_dir, 'node', node, - options=['--stream', '--slot=slot_1', '--temp-slot']) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_backup_concurrent_drop_table(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=1) - - # FULL backup - gdb = self.backup_node( - backup_dir, 'node', node, - options=['--stream', '--compress'], - gdb=True) - - gdb.set_breakpoint('backup_data_file') - gdb.run_until_break() - - node.safe_psql( - 'postgres', - 'DROP TABLE pgbench_accounts') - - # do checkpoint to guarantee filenode removal - node.safe_psql( - 'postgres', - 'CHECKPOINT') - - gdb.remove_all_breakpoints() - gdb.continue_execution_until_exit() - - show_backup = self.show_pb(backup_dir, 'node')[0] - - self.assertEqual(show_backup['status'], "OK") - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_pg_11_adjusted_wal_segment_size(self): - """""" - if self.pg_config_version < self.version_to_num('11.0'): - return unittest.skip('You need PostgreSQL >= 11 for this test') - - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=[ - '--data-checksums', - '--wal-segsize=64'], - pg_options={ - 'min_wal_size': '128MB', - 'autovacuum': 'off'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=5) - - # FULL STREAM backup - self.backup_node( - backup_dir, 'node', node, options=['--stream']) - - pgbench = node.pgbench(options=['-T', '5', '-c', '2']) - pgbench.wait() - - # PAGE STREAM backup - self.backup_node( - backup_dir, 'node', node, - backup_type='page', options=['--stream']) - - pgbench = node.pgbench(options=['-T', '5', '-c', '2']) - pgbench.wait() - - # DELTA STREAM backup - self.backup_node( - backup_dir, 'node', node, - backup_type='delta', options=['--stream']) - - pgbench = node.pgbench(options=['-T', '5', '-c', '2']) - pgbench.wait() - - # FULL ARCHIVE backup - self.backup_node(backup_dir, 'node', node) - - pgbench = node.pgbench(options=['-T', '5', '-c', '2']) - pgbench.wait() - - # PAGE ARCHIVE backup - self.backup_node(backup_dir, 'node', node, backup_type='page') - - pgbench = node.pgbench(options=['-T', '5', '-c', '2']) - pgbench.wait() - - # DELTA ARCHIVE backup - backup_id = self.backup_node(backup_dir, 'node', node, backup_type='delta') - pgdata = self.pgdata_content(node.data_dir) - - # delete - output = self.delete_pb( - backup_dir, 'node', - options=[ - '--expired', - '--delete-wal', - '--retention-redundancy=1']) - - # validate - self.validate_pb(backup_dir) - - # merge - self.merge_backup(backup_dir, 'node', backup_id=backup_id) - - # restore - node.cleanup() - self.restore_node( - backup_dir, 'node', node, backup_id=backup_id) - - pgdata_restored = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_sigint_handling(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - # FULL backup - gdb = self.backup_node( - backup_dir, 'node', node, gdb=True, - options=['--stream', '--log-level-file=verbose']) - - gdb.set_breakpoint('copy_file') - gdb.run_until_break() - - gdb.continue_execution_until_break(20) - gdb.remove_all_breakpoints() - - gdb._execute('signal SIGINT') - gdb.continue_execution_until_error() - - backup_id = self.show_pb(backup_dir, 'node')[0]['id'] - - self.assertEqual( - 'ERROR', - self.show_pb(backup_dir, 'node', backup_id)['status'], - 'Backup STATUS should be "ERROR"') - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_sigterm_handling(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - # FULL backup - gdb = self.backup_node( - backup_dir, 'node', node, gdb=True, - options=['--stream', '--log-level-file=verbose']) - - gdb.set_breakpoint('copy_file') - gdb.run_until_break() - - gdb.continue_execution_until_break(20) - gdb.remove_all_breakpoints() - - gdb._execute('signal SIGTERM') - gdb.continue_execution_until_error() - - backup_id = self.show_pb(backup_dir, 'node')[0]['id'] - - self.assertEqual( - 'ERROR', - self.show_pb(backup_dir, 'node', backup_id)['status'], - 'Backup STATUS should be "ERROR"') - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_sigquit_handling(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - # FULL backup - gdb = self.backup_node( - backup_dir, 'node', node, gdb=True, - options=['--stream', '--log-level-file=verbose']) - - gdb.set_breakpoint('copy_file') - gdb.run_until_break() - - gdb.continue_execution_until_break(20) - gdb.remove_all_breakpoints() - - gdb._execute('signal SIGQUIT') - gdb.continue_execution_until_error() - - backup_id = self.show_pb(backup_dir, 'node')[0]['id'] - - self.assertEqual( - 'ERROR', - self.show_pb(backup_dir, 'node', backup_id)['status'], - 'Backup STATUS should be "ERROR"') - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_drop_table(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - connect_1 = node.connect("postgres") - connect_1.execute( - "create table t_heap as select i" - " as id from generate_series(0,100) i") - connect_1.commit() - - connect_2 = node.connect("postgres") - connect_2.execute("SELECT * FROM t_heap") - connect_2.commit() - - # DROP table - connect_2.execute("DROP TABLE t_heap") - connect_2.commit() - - # FULL backup - self.backup_node( - backup_dir, 'node', node, options=['--stream']) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_basic_missing_file_permissions(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - relative_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('pg_class')").rstrip() - - full_path = os.path.join(node.data_dir, relative_path) - - os.chmod(full_path, 000) - - try: - # FULL backup - self.backup_node( - backup_dir, 'node', node, options=['--stream']) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because of missing permissions" - "\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - 'ERROR: cannot open file', - e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - os.chmod(full_path, 700) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_basic_missing_dir_permissions(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - full_path = os.path.join(node.data_dir, 'pg_twophase') - - os.chmod(full_path, 000) - - try: - # FULL backup - self.backup_node( - backup_dir, 'node', node, options=['--stream']) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because of missing permissions" - "\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - 'ERROR: Cannot open directory', - e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - os.chmod(full_path, 700) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_backup_with_least_privileges_role(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'archive_timeout': '30s'}) - - if self.ptrack: - node.append_conf('postgresql.auto.conf', 'ptrack_enable = on') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - 'postgres', - 'CREATE DATABASE backupdb') - - # PG 9.5 - if self.get_version(node) < 90600: - node.safe_psql( - 'backupdb', - "REVOKE ALL ON DATABASE backupdb from PUBLIC; " - "REVOKE ALL ON SCHEMA public from PUBLIC; " - "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " - "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " - "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " - "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " - "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " - "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " - "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " - "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " - "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " - "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " - "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " - "CREATE ROLE backup WITH LOGIN REPLICATION; " - "GRANT CONNECT ON DATABASE backupdb to backup; " - "GRANT USAGE ON SCHEMA pg_catalog TO backup; " - "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " - "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack - "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" - ) - # PG 9.6 - elif self.get_version(node) > 90600 and self.get_version(node) < 100000: - node.safe_psql( - 'backupdb', - "REVOKE ALL ON DATABASE backupdb from PUBLIC; " - "REVOKE ALL ON SCHEMA public from PUBLIC; " - "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " - "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " - "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " - "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " - "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " - "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " - "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " - "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " - "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " - "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " - "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " - "CREATE ROLE backup WITH LOGIN REPLICATION; " - "GRANT CONNECT ON DATABASE backupdb to backup; " - "GRANT USAGE ON SCHEMA pg_catalog TO backup; " - "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " - "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack - "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_xlog_replay_location() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" - ) - # >= 10 - else: - node.safe_psql( - 'backupdb', - "REVOKE ALL ON DATABASE backupdb from PUBLIC; " - "REVOKE ALL ON SCHEMA public from PUBLIC; " - "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " - "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " - "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " - "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " - "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " - "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " - "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " - "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " - "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " - "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " - "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " - "CREATE ROLE backup WITH LOGIN REPLICATION; " - "GRANT CONNECT ON DATABASE backupdb to backup; " - "GRANT USAGE ON SCHEMA pg_catalog TO backup; " - "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " - "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack - "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" - ) - - if self.ptrack: - for fname in [ - 'pg_catalog.oideq(oid, oid)', - 'pg_catalog.ptrack_version()', - 'pg_catalog.pg_ptrack_clear()', - 'pg_catalog.pg_ptrack_control_lsn()', - 'pg_catalog.pg_ptrack_get_and_clear_db(oid, oid)', - 'pg_catalog.pg_ptrack_get_and_clear(oid, oid)', - 'pg_catalog.pg_ptrack_get_block_2(oid, oid, oid, bigint)', - 'pg_catalog.pg_stop_backup()']: - # try: - node.safe_psql( - "backupdb", - "GRANT EXECUTE ON FUNCTION {0} " - "TO backup".format(fname)) - # except: - # pass - - # FULL backup - self.backup_node( - backup_dir, 'node', node, - datname='backupdb', options=['--stream', '-U', 'backup']) - self.backup_node( - backup_dir, 'node', node, - datname='backupdb', options=['-U', 'backup']) - - # PAGE - self.backup_node( - backup_dir, 'node', node, backup_type='page', - datname='backupdb', options=['-U', 'backup']) - self.backup_node( - backup_dir, 'node', node, backup_type='page', datname='backupdb', - options=['--stream', '-U', 'backup']) - - # DELTA - self.backup_node( - backup_dir, 'node', node, backup_type='delta', - datname='backupdb', options=['-U', 'backup']) - self.backup_node( - backup_dir, 'node', node, backup_type='delta', - datname='backupdb', options=['--stream', '-U', 'backup']) - - # PTRACK - if self.ptrack: - self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - datname='backupdb', options=['-U', 'backup']) - self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - datname='backupdb', options=['--stream', '-U', 'backup']) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_parent_choosing(self): - """ - PAGE3 <- RUNNING(parent should be FULL) - PAGE2 <- OK - PAGE1 <- CORRUPT - FULL - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - full_id = self.backup_node(backup_dir, 'node', node) - - # PAGE1 - page1_id = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGE2 - page2_id = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # Change PAGE1 to ERROR - self.change_backup_status(backup_dir, 'node', page1_id, 'ERROR') - - # PAGE3 - page3_id = self.backup_node( - backup_dir, 'node', node, - backup_type='page', options=['--log-level-file=LOG']) - - log_file_path = os.path.join(backup_dir, 'log', 'pg_probackup.log') - with open(log_file_path) as f: - log_file_content = f.read() - - self.assertIn( - "WARNING: Backup {0} has invalid parent: {1}. " - "Cannot be a parent".format(page2_id, page1_id), - log_file_content) - - self.assertIn( - "WARNING: Backup {0} has status: ERROR. " - "Cannot be a parent".format(page1_id), - log_file_content) - - self.assertIn( - "Parent backup: {0}".format(full_id), - log_file_content) - - self.assertEqual( - self.show_pb( - backup_dir, 'node', backup_id=page3_id)['parent-backup-id'], - full_id) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_parent_choosing_1(self): - """ - PAGE3 <- RUNNING(parent should be FULL) - PAGE2 <- OK - PAGE1 <- (missing) - FULL - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - full_id = self.backup_node(backup_dir, 'node', node) - - # PAGE1 - page1_id = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGE2 - page2_id = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # Delete PAGE1 - shutil.rmtree( - os.path.join(backup_dir, 'backups', 'node', page1_id)) - - # PAGE3 - page3_id = self.backup_node( - backup_dir, 'node', node, - backup_type='page', options=['--log-level-file=LOG']) - - log_file_path = os.path.join(backup_dir, 'log', 'pg_probackup.log') - with open(log_file_path) as f: - log_file_content = f.read() - - self.assertIn( - "WARNING: Backup {0} has missing parent: {1}. " - "Cannot be a parent".format(page2_id, page1_id), - log_file_content) - - self.assertIn( - "Parent backup: {0}".format(full_id), - log_file_content) - - self.assertEqual( - self.show_pb( - backup_dir, 'node', backup_id=page3_id)['parent-backup-id'], - full_id) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_parent_choosing_2(self): - """ - PAGE3 <- RUNNING(backup should fail) - PAGE2 <- OK - PAGE1 <- OK - FULL <- (missing) - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - full_id = self.backup_node(backup_dir, 'node', node) - - # PAGE1 - page1_id = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGE2 - page2_id = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # Delete FULL - shutil.rmtree( - os.path.join(backup_dir, 'backups', 'node', full_id)) - - # PAGE3 - try: - self.backup_node( - backup_dir, 'node', node, - backup_type='page', options=['--log-level-file=LOG']) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because FULL backup is missing" - "\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - 'ERROR: Valid backup on current timeline 1 is not found. ' - 'Create new FULL backup before an incremental one.', - e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - self.assertEqual( - self.show_pb( - backup_dir, 'node')[2]['status'], - 'ERROR') - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_backup_with_less_privileges_role(self): - """ - check permissions correctness from documentation: - https://github.com/postgrespro/pg_probackup/blob/master/Documentation.md#configuring-the-database-cluster - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'archive_timeout': '30s', - 'checkpoint_timeout': '30s'}) - - if self.ptrack: - node.append_conf('postgresql.auto.conf', 'ptrack_enable = on') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - 'postgres', - 'CREATE DATABASE backupdb') - - # PG 9.5 - if self.get_version(node) < 90600: - node.safe_psql( - 'backupdb', - "BEGIN; " - "CREATE ROLE backup WITH LOGIN; " - "GRANT USAGE ON SCHEMA pg_catalog TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; " - "COMMIT;" - ) - # PG 9.6 - elif self.get_version(node) > 90600 and self.get_version(node) < 100000: - node.safe_psql( - 'backupdb', - "BEGIN; " - "CREATE ROLE backup WITH LOGIN; " - "GRANT USAGE ON SCHEMA pg_catalog TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_xlog_replay_location() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; " - "COMMIT;" - ) - # >= 10 - else: - node.safe_psql( - 'backupdb', - "BEGIN; " - "CREATE ROLE backup WITH LOGIN; " - "GRANT USAGE ON SCHEMA pg_catalog TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " - "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; " - "COMMIT;" - ) - - # enable STREAM backup - node.safe_psql( - 'backupdb', - 'ALTER ROLE backup WITH REPLICATION;') - - # FULL backup - self.backup_node( - backup_dir, 'node', node, - datname='backupdb', options=['--stream', '-U', 'backup']) - self.backup_node( - backup_dir, 'node', node, - datname='backupdb', options=['-U', 'backup']) - - # PAGE - self.backup_node( - backup_dir, 'node', node, backup_type='page', - datname='backupdb', options=['-U', 'backup']) - self.backup_node( - backup_dir, 'node', node, backup_type='page', datname='backupdb', - options=['--stream', '-U', 'backup']) - - # DELTA - self.backup_node( - backup_dir, 'node', node, backup_type='delta', - datname='backupdb', options=['-U', 'backup']) - self.backup_node( - backup_dir, 'node', node, backup_type='delta', - datname='backupdb', options=['--stream', '-U', 'backup']) - - # Restore as replica - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() - - self.restore_node(backup_dir, 'node', replica) - self.set_replica(node, replica) - self.add_instance(backup_dir, 'replica', replica) - self.set_archiving(backup_dir, 'replica', replica, replica=True) - - replica.slow_start(replica=True) - - # FULL backup from replica - self.backup_node( - backup_dir, 'replica', replica, - datname='backupdb', options=['--stream', '-U', 'backup']) - - self.backup_node( - backup_dir, 'replica', replica, datname='backupdb', - options=['-U', 'backup', '--log-level-file=verbose']) - - # PAGE backup from replica - self.backup_node( - backup_dir, 'replica', replica, backup_type='page', - datname='backupdb', options=['-U', 'backup']) - self.backup_node( - backup_dir, 'replica', replica, backup_type='page', - datname='backupdb', options=['--stream', '-U', 'backup']) - - # DELTA backup from replica - self.backup_node( - backup_dir, 'replica', replica, backup_type='delta', - datname='backupdb', options=['-U', 'backup']) - self.backup_node( - backup_dir, 'replica', replica, backup_type='delta', - datname='backupdb', options=['--stream', '-U', 'backup']) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/backup_test.py b/tests/backup_test.py new file mode 100644 index 000000000..dc60228b5 --- /dev/null +++ b/tests/backup_test.py @@ -0,0 +1,3658 @@ +import unittest +import os +import re +from time import sleep, time +from .helpers.ptrack_helpers import base36enc, ProbackupTest, ProbackupException +import shutil +from distutils.dir_util import copy_tree +from testgres import ProcessType, QueryException +import subprocess + + +class BackupTest(ProbackupTest, unittest.TestCase): + + def test_full_backup(self): + """ + Just test full backup with at least two segments + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + # we need to write a lot. Lets speedup a bit. + pg_options={"fsync": "off", "synchronous_commit": "off"}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Fill with data + # Have to use scale=100 to create second segment. + node.pgbench_init(scale=100, no_vacuum=True) + + # FULL + backup_id = self.backup_node(backup_dir, 'node', node) + + out = self.validate_pb(backup_dir, 'node', backup_id) + self.assertIn( + "INFO: Backup {0} is valid".format(backup_id), + out) + + def test_full_backup_stream(self): + """ + Just test full backup with at least two segments in stream mode + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + # we need to write a lot. Lets speedup a bit. + pg_options={"fsync": "off", "synchronous_commit": "off"}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # Fill with data + # Have to use scale=100 to create second segment. + node.pgbench_init(scale=100, no_vacuum=True) + + # FULL + backup_id = self.backup_node(backup_dir, 'node', node, + options=["--stream"]) + + out = self.validate_pb(backup_dir, 'node', backup_id) + self.assertIn( + "INFO: Backup {0} is valid".format(backup_id), + out) + + # @unittest.skip("skip") + # @unittest.expectedFailure + # PGPRO-707 + def test_backup_modes_archive(self): + """standart backup modes with ARCHIVE WAL method""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + full_backup_id = self.backup_node(backup_dir, 'node', node) + show_backup = self.show_pb(backup_dir, 'node')[0] + + self.assertEqual(show_backup['status'], "OK") + self.assertEqual(show_backup['backup-mode'], "FULL") + + # postmaster.pid and postmaster.opts shouldn't be copied + excluded = True + db_dir = os.path.join( + backup_dir, "backups", 'node', full_backup_id, "database") + + for f in os.listdir(db_dir): + if ( + os.path.isfile(os.path.join(db_dir, f)) and + ( + f == "postmaster.pid" or + f == "postmaster.opts" + ) + ): + excluded = False + self.assertEqual(excluded, True) + + # page backup mode + page_backup_id = self.backup_node( + backup_dir, 'node', node, backup_type="page") + + show_backup_1 = self.show_pb(backup_dir, 'node')[1] + self.assertEqual(show_backup_1['status'], "OK") + self.assertEqual(show_backup_1['backup-mode'], "PAGE") + + # delta backup mode + delta_backup_id = self.backup_node( + backup_dir, 'node', node, backup_type="delta") + + show_backup_2 = self.show_pb(backup_dir, 'node')[2] + self.assertEqual(show_backup_2['status'], "OK") + self.assertEqual(show_backup_2['backup-mode'], "DELTA") + + # Check parent backup + self.assertEqual( + full_backup_id, + self.show_pb( + backup_dir, 'node', + backup_id=show_backup_1['id'])["parent-backup-id"]) + + self.assertEqual( + page_backup_id, + self.show_pb( + backup_dir, 'node', + backup_id=show_backup_2['id'])["parent-backup-id"]) + + # @unittest.skip("skip") + def test_smooth_checkpoint(self): + """full backup with smooth checkpoint""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.backup_node( + backup_dir, 'node', node, + options=["-C"]) + self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") + node.stop() + + # @unittest.skip("skip") + def test_incremental_backup_without_full(self): + """page backup without validated full backup""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + try: + self.backup_node(backup_dir, 'node', node, backup_type="page") + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because page backup should not be possible " + "without valid full backup.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertTrue( + "WARNING: Valid full backup on current timeline 1 is not found" in e.message and + "ERROR: Create new full backup before an incremental one" in e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['status'], + "ERROR") + + # @unittest.skip("skip") + def test_incremental_backup_corrupt_full(self): + """page-level backup with corrupted full backup""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node(backup_dir, 'node', node) + file = os.path.join( + backup_dir, "backups", "node", backup_id, + "database", "postgresql.conf") + os.remove(file) + + try: + self.validate_pb(backup_dir, 'node') + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of validation of corrupted backup.\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertTrue( + "INFO: Validate backups of the instance 'node'" in e.message and + "WARNING: Backup file" in e.message and "is not found" in e.message and + "WARNING: Backup {0} data files are corrupted".format( + backup_id) in e.message and + "WARNING: Some backups are not valid" in e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + try: + self.backup_node(backup_dir, 'node', node, backup_type="page") + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because page backup should not be possible " + "without valid full backup.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertTrue( + "WARNING: Valid full backup on current timeline 1 is not found" in e.message and + "ERROR: Create new full backup before an incremental one" in e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertEqual( + self.show_pb(backup_dir, 'node', backup_id)['status'], "CORRUPT") + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['status'], "ERROR") + + # @unittest.skip("skip") + def test_delta_threads_stream(self): + """delta multi thread backup mode and stream""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.backup_node( + backup_dir, 'node', node, backup_type="full", + options=["-j", "4", "--stream"]) + + self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") + self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + self.assertEqual(self.show_pb(backup_dir, 'node')[1]['status'], "OK") + + # @unittest.skip("skip") + def test_page_detect_corruption(self): + """make node, corrupt some page, check that backup failed""" + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1000) i") + + node.safe_psql( + "postgres", + "CHECKPOINT") + + heap_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + path = os.path.join(node.data_dir, heap_path) + with open(path, "rb+", 0) as f: + f.seek(9000) + f.write(b"bla") + f.flush() + f.close + + try: + self.backup_node( + backup_dir, 'node', node, backup_type="full", + options=["-j", "4", "--stream", "--log-level-file=VERBOSE"]) + self.assertEqual( + 1, 0, + "Expecting Error because data file is corrupted" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertTrue( + 'ERROR: Corruption detected in file "{0}", ' + 'block 1: page verification failed, calculated checksum'.format(path), + e.message) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['status'], + 'ERROR', + "Backup Status should be ERROR") + + # @unittest.skip("skip") + def test_backup_detect_corruption(self): + """make node, corrupt some page, check that backup failed""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + if self.ptrack: + node.safe_psql( + "postgres", + "create extension ptrack") + + self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + heap_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + + node.safe_psql( + "postgres", + "select count(*) from t_heap") + + node.safe_psql( + "postgres", + "update t_heap set id = id + 10000") + + node.stop() + + heap_fullpath = os.path.join(node.data_dir, heap_path) + + with open(heap_fullpath, "rb+", 0) as f: + f.seek(9000) + f.write(b"bla") + f.flush() + f.close + + node.slow_start() + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page verification failed, calculated checksum'.format( + heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + sleep(1) + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page verification failed, calculated checksum'.format( + heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + sleep(1) + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="page", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page verification failed, calculated checksum'.format( + heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + sleep(1) + + if self.ptrack: + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="ptrack", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page verification failed, calculated checksum'.format( + heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_backup_detect_invalid_block_header(self): + """make node, corrupt some page, check that backup failed""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + if self.ptrack: + node.safe_psql( + "postgres", + "create extension ptrack") + + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + heap_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + + node.safe_psql( + "postgres", + "select count(*) from t_heap") + + node.safe_psql( + "postgres", + "update t_heap set id = id + 10000") + + node.stop() + + heap_fullpath = os.path.join(node.data_dir, heap_path) + with open(heap_fullpath, "rb+", 0) as f: + f.seek(8193) + f.write(b"blahblahblahblah") + f.flush() + f.close + + node.slow_start() + +# self.backup_node( +# backup_dir, 'node', node, +# backup_type="full", options=["-j", "4", "--stream"]) + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page header invalid, pd_lower'.format(heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + sleep(1) + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page header invalid, pd_lower'.format(heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + sleep(1) + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="page", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page header invalid, pd_lower'.format(heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + sleep(1) + + if self.ptrack: + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="ptrack", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page header invalid, pd_lower'.format(heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_backup_detect_missing_permissions(self): + """make node, corrupt some page, check that backup failed""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + if self.ptrack: + node.safe_psql( + "postgres", + "create extension ptrack") + + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + heap_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + + node.safe_psql( + "postgres", + "select count(*) from t_heap") + + node.safe_psql( + "postgres", + "update t_heap set id = id + 10000") + + node.stop() + + heap_fullpath = os.path.join(node.data_dir, heap_path) + with open(heap_fullpath, "rb+", 0) as f: + f.seek(8193) + f.write(b"blahblahblahblah") + f.flush() + f.close + + node.slow_start() + +# self.backup_node( +# backup_dir, 'node', node, +# backup_type="full", options=["-j", "4", "--stream"]) + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page header invalid, pd_lower'.format(heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + sleep(1) + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page header invalid, pd_lower'.format(heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + sleep(1) + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="page", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page header invalid, pd_lower'.format(heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + sleep(1) + + if self.ptrack: + try: + self.backup_node( + backup_dir, 'node', node, + backup_type="ptrack", options=["-j", "4", "--stream"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of block corruption" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Corruption detected in file "{0}", block 1: ' + 'page header invalid, pd_lower'.format(heap_fullpath), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_backup_truncate_misaligned(self): + """ + make node, truncate file to size not even to BLCKSIZE, + take backup + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,100000) i") + + node.safe_psql( + "postgres", + "CHECKPOINT;") + + heap_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + heap_size = node.safe_psql( + "postgres", + "select pg_relation_size('t_heap')") + + with open(os.path.join(node.data_dir, heap_path), "rb+", 0) as f: + f.truncate(int(heap_size) - 4096) + f.flush() + f.close + + output = self.backup_node( + backup_dir, 'node', node, backup_type="full", + options=["-j", "4", "--stream"], return_id=False) + + self.assertIn("WARNING: File", output) + self.assertIn("invalid file size", output) + + # @unittest.skip("skip") + def test_tablespace_in_pgdata_pgpro_1376(self): + """PGPRO-1376 """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.create_tblspace_in_node( + node, 'tblspace1', + tblspc_path=( + os.path.join( + node.data_dir, 'somedirectory', '100500')) + ) + + self.create_tblspace_in_node( + node, 'tblspace2', + tblspc_path=(os.path.join(node.data_dir)) + ) + + node.safe_psql( + "postgres", + "create table t_heap1 tablespace tblspace1 as select 1 as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1000) i") + + node.safe_psql( + "postgres", + "create table t_heap2 tablespace tblspace2 as select 1 as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1000) i") + + backup_id_1 = self.backup_node( + backup_dir, 'node', node, backup_type="full", + options=["-j", "4", "--stream"]) + + node.safe_psql( + "postgres", + "drop table t_heap2") + node.safe_psql( + "postgres", + "drop tablespace tblspace2") + + self.backup_node( + backup_dir, 'node', node, backup_type="full", + options=["-j", "4", "--stream"]) + + pgdata = self.pgdata_content(node.data_dir) + + relfilenode = node.safe_psql( + "postgres", + "select 't_heap1'::regclass::oid" + ).decode('utf-8').rstrip() + + list = [] + for root, dirs, files in os.walk(os.path.join( + backup_dir, 'backups', 'node', backup_id_1)): + for file in files: + if file == relfilenode: + path = os.path.join(root, file) + list = list + [path] + + # We expect that relfilenode can be encountered only once + if len(list) > 1: + message = "" + for string in list: + message = message + string + "\n" + self.assertEqual( + 1, 0, + "Following file copied twice by backup:\n {0}".format( + message) + ) + + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_basic_tablespace_handling(self): + """ + make node, take full backup, check that restore with + tablespace mapping will end with error, take page backup, + check that restore with tablespace mapping will end with + success + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type="full", + options=["-j", "4", "--stream"]) + + tblspace1_old_path = self.get_tblspace_path(node, 'tblspace1_old') + tblspace2_old_path = self.get_tblspace_path(node, 'tblspace2_old') + + self.create_tblspace_in_node( + node, 'some_lame_tablespace') + + self.create_tblspace_in_node( + node, 'tblspace1', + tblspc_path=tblspace1_old_path) + + self.create_tblspace_in_node( + node, 'tblspace2', + tblspc_path=tblspace2_old_path) + + node.safe_psql( + "postgres", + "create table t_heap_lame tablespace some_lame_tablespace " + "as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1000) i") + + node.safe_psql( + "postgres", + "create table t_heap2 tablespace tblspace2 as select 1 as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1000) i") + + tblspace1_new_path = self.get_tblspace_path(node, 'tblspace1_new') + tblspace2_new_path = self.get_tblspace_path(node, 'tblspace2_new') + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + try: + self.restore_node( + backup_dir, 'node', node_restored, + options=[ + "-j", "4", + "-T", "{0}={1}".format( + tblspace1_old_path, tblspace1_new_path), + "-T", "{0}={1}".format( + tblspace2_old_path, tblspace2_new_path)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because tablespace mapping is incorrect" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Backup {0} has no tablespaceses, ' + 'nothing to remap'.format(backup_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + node.safe_psql( + "postgres", + "drop table t_heap_lame") + + node.safe_psql( + "postgres", + "drop tablespace some_lame_tablespace") + + self.backup_node( + backup_dir, 'node', node, backup_type="delta", + options=["-j", "4", "--stream"]) + + self.restore_node( + backup_dir, 'node', node_restored, + options=[ + "-j", "4", + "-T", "{0}={1}".format( + tblspace1_old_path, tblspace1_new_path), + "-T", "{0}={1}".format( + tblspace2_old_path, tblspace2_new_path)]) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_tablespace_handling_1(self): + """ + make node with tablespace A, take full backup, check that restore with + tablespace mapping of tablespace B will end with error + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + tblspace1_old_path = self.get_tblspace_path(node, 'tblspace1_old') + tblspace2_old_path = self.get_tblspace_path(node, 'tblspace2_old') + + tblspace_new_path = self.get_tblspace_path(node, 'tblspace_new') + + self.create_tblspace_in_node( + node, 'tblspace1', + tblspc_path=tblspace1_old_path) + + self.backup_node( + backup_dir, 'node', node, backup_type="full", + options=["-j", "4", "--stream"]) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + try: + self.restore_node( + backup_dir, 'node', node_restored, + options=[ + "-j", "4", + "-T", "{0}={1}".format( + tblspace2_old_path, tblspace_new_path)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because tablespace mapping is incorrect" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertTrue( + 'ERROR: --tablespace-mapping option' in e.message and + 'have an entry in tablespace_map file' in e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_tablespace_handling_2(self): + """ + make node without tablespaces, take full backup, check that restore with + tablespace mapping will end with error + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + tblspace1_old_path = self.get_tblspace_path(node, 'tblspace1_old') + tblspace_new_path = self.get_tblspace_path(node, 'tblspace_new') + + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type="full", + options=["-j", "4", "--stream"]) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + try: + self.restore_node( + backup_dir, 'node', node_restored, + options=[ + "-j", "4", + "-T", "{0}={1}".format( + tblspace1_old_path, tblspace_new_path)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because tablespace mapping is incorrect" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Backup {0} has no tablespaceses, ' + 'nothing to remap'.format(backup_id), e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_drop_rel_during_full_backup(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + for i in range(1, 512): + node.safe_psql( + "postgres", + "create table t_heap_{0} as select i" + " as id from generate_series(0,100) i".format(i)) + + node.safe_psql( + "postgres", + "VACUUM") + + node.pgbench_init(scale=10) + + relative_path_1 = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap_1')").decode('utf-8').rstrip() + + relative_path_2 = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap_1')").decode('utf-8').rstrip() + + absolute_path_1 = os.path.join(node.data_dir, relative_path_1) + absolute_path_2 = os.path.join(node.data_dir, relative_path_2) + + # FULL backup + gdb = self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--log-level-file=LOG', '--log-level-console=LOG', '--progress'], + gdb=True) + + gdb.set_breakpoint('backup_files') + gdb.run_until_break() + + # REMOVE file + for i in range(1, 512): + node.safe_psql( + "postgres", + "drop table t_heap_{0}".format(i)) + + node.safe_psql( + "postgres", + "CHECKPOINT") + + node.safe_psql( + "postgres", + "CHECKPOINT") + + # File removed, we can proceed with backup + gdb.continue_execution_until_exit() + + pgdata = self.pgdata_content(node.data_dir) + + #with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: + # log_content = f.read() + # self.assertTrue( + # 'LOG: File "{0}" is not found'.format(absolute_path) in log_content, + # 'File "{0}" should be deleted but it`s not'.format(absolute_path)) + + node.cleanup() + self.restore_node(backup_dir, 'node', node) + + # Physical comparison + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + @unittest.skip("skip") + def test_drop_db_during_full_backup(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + for i in range(1, 2): + node.safe_psql( + "postgres", + "create database t_heap_{0}".format(i)) + + node.safe_psql( + "postgres", + "VACUUM") + + # FULL backup + gdb = self.backup_node( + backup_dir, 'node', node, gdb=True, + options=[ + '--stream', '--log-level-file=LOG', + '--log-level-console=LOG', '--progress']) + + gdb.set_breakpoint('backup_files') + gdb.run_until_break() + + # REMOVE file + for i in range(1, 2): + node.safe_psql( + "postgres", + "drop database t_heap_{0}".format(i)) + + node.safe_psql( + "postgres", + "CHECKPOINT") + + node.safe_psql( + "postgres", + "CHECKPOINT") + + # File removed, we can proceed with backup + gdb.continue_execution_until_exit() + + pgdata = self.pgdata_content(node.data_dir) + + #with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: + # log_content = f.read() + # self.assertTrue( + # 'LOG: File "{0}" is not found'.format(absolute_path) in log_content, + # 'File "{0}" should be deleted but it`s not'.format(absolute_path)) + + node.cleanup() + self.restore_node(backup_dir, 'node', node) + + # Physical comparison + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_drop_rel_during_backup_delta(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=10) + + node.safe_psql( + "postgres", + "create table t_heap as select i" + " as id from generate_series(0,100) i") + + relative_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + absolute_path = os.path.join(node.data_dir, relative_path) + + # FULL backup + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + # DELTA backup + gdb = self.backup_node( + backup_dir, 'node', node, backup_type='delta', + gdb=True, options=['--log-level-file=LOG']) + + gdb.set_breakpoint('backup_files') + gdb.run_until_break() + + # REMOVE file + node.safe_psql( + "postgres", + "DROP TABLE t_heap") + + node.safe_psql( + "postgres", + "CHECKPOINT") + + # File removed, we can proceed with backup + gdb.continue_execution_until_exit() + + pgdata = self.pgdata_content(node.data_dir) + + with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: + log_content = f.read() + self.assertTrue( + 'LOG: File not found: "{0}"'.format(absolute_path) in log_content, + 'File "{0}" should be deleted but it`s not'.format(absolute_path)) + + node.cleanup() + self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) + + # Physical comparison + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_drop_rel_during_backup_page(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "create table t_heap as select i" + " as id from generate_series(0,100) i") + + relative_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + absolute_path = os.path.join(node.data_dir, relative_path) + + # FULL backup + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + node.safe_psql( + "postgres", + "insert into t_heap select i" + " as id from generate_series(101,102) i") + + # PAGE backup + gdb = self.backup_node( + backup_dir, 'node', node, backup_type='page', + gdb=True, options=['--log-level-file=LOG']) + + gdb.set_breakpoint('backup_files') + gdb.run_until_break() + + # REMOVE file + os.remove(absolute_path) + + # File removed, we can proceed with backup + gdb.continue_execution_until_exit() + gdb.kill() + + pgdata = self.pgdata_content(node.data_dir) + + backup_id = self.show_pb(backup_dir, 'node')[1]['id'] + + filelist = self.get_backup_filelist(backup_dir, 'node', backup_id) + self.assertNotIn(relative_path, filelist) + + node.cleanup() + self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) + + # Physical comparison + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_persistent_slot_for_stream_backup(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'max_wal_size': '40MB'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "SELECT pg_create_physical_replication_slot('slot_1')") + + # FULL backup + self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--slot=slot_1']) + + # FULL backup + self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--slot=slot_1']) + + # @unittest.skip("skip") + def test_basic_temp_slot_for_stream_backup(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={'max_wal_size': '40MB'}) + + if self.get_version(node) < self.version_to_num('10.0'): + self.skipTest('You need PostgreSQL >= 10 for this test') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--temp-slot']) + + # FULL backup + self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--slot=slot_1', '--temp-slot']) + + # @unittest.skip("skip") + def test_backup_concurrent_drop_table(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=1) + + # FULL backup + gdb = self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--compress'], + gdb=True) + + gdb.set_breakpoint('backup_data_file') + gdb.run_until_break() + + node.safe_psql( + 'postgres', + 'DROP TABLE pgbench_accounts') + + # do checkpoint to guarantee filenode removal + node.safe_psql( + 'postgres', + 'CHECKPOINT') + + gdb.remove_all_breakpoints() + gdb.continue_execution_until_exit() + gdb.kill() + + show_backup = self.show_pb(backup_dir, 'node')[0] + + self.assertEqual(show_backup['status'], "OK") + + # @unittest.skip("skip") + def test_pg_11_adjusted_wal_segment_size(self): + """""" + if self.pg_config_version < self.version_to_num('11.0'): + self.skipTest('You need PostgreSQL >= 11 for this test') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=[ + '--data-checksums', + '--wal-segsize=64'], + pg_options={ + 'min_wal_size': '128MB'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=5) + + # FULL STREAM backup + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + pgbench = node.pgbench(options=['-T', '5', '-c', '2']) + pgbench.wait() + + # PAGE STREAM backup + self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=['--stream']) + + pgbench = node.pgbench(options=['-T', '5', '-c', '2']) + pgbench.wait() + + # DELTA STREAM backup + self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['--stream']) + + pgbench = node.pgbench(options=['-T', '5', '-c', '2']) + pgbench.wait() + + # FULL ARCHIVE backup + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench(options=['-T', '5', '-c', '2']) + pgbench.wait() + + # PAGE ARCHIVE backup + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench(options=['-T', '5', '-c', '2']) + pgbench.wait() + + # DELTA ARCHIVE backup + backup_id = self.backup_node(backup_dir, 'node', node, backup_type='delta') + pgdata = self.pgdata_content(node.data_dir) + + # delete + output = self.delete_pb( + backup_dir, 'node', + options=[ + '--expired', + '--delete-wal', + '--retention-redundancy=1']) + + # validate + self.validate_pb(backup_dir) + + # merge + self.merge_backup(backup_dir, 'node', backup_id=backup_id) + + # restore + node.cleanup() + self.restore_node( + backup_dir, 'node', node, backup_id=backup_id) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_sigint_handling(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + gdb = self.backup_node( + backup_dir, 'node', node, gdb=True, + options=['--stream', '--log-level-file=LOG']) + + gdb.set_breakpoint('backup_non_data_file') + gdb.run_until_break() + + gdb.continue_execution_until_break(20) + gdb.remove_all_breakpoints() + + gdb._execute('signal SIGINT') + gdb.continue_execution_until_error() + gdb.kill() + + backup_id = self.show_pb(backup_dir, 'node')[0]['id'] + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node', backup_id)['status'], + 'Backup STATUS should be "ERROR"') + + # @unittest.skip("skip") + def test_sigterm_handling(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + gdb = self.backup_node( + backup_dir, 'node', node, gdb=True, + options=['--stream', '--log-level-file=LOG']) + + gdb.set_breakpoint('backup_non_data_file') + gdb.run_until_break() + + gdb.continue_execution_until_break(20) + gdb.remove_all_breakpoints() + + gdb._execute('signal SIGTERM') + gdb.continue_execution_until_error() + + backup_id = self.show_pb(backup_dir, 'node')[0]['id'] + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node', backup_id)['status'], + 'Backup STATUS should be "ERROR"') + + # @unittest.skip("skip") + def test_sigquit_handling(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + gdb = self.backup_node( + backup_dir, 'node', node, gdb=True, options=['--stream']) + + gdb.set_breakpoint('backup_non_data_file') + gdb.run_until_break() + + gdb.continue_execution_until_break(20) + gdb.remove_all_breakpoints() + + gdb._execute('signal SIGQUIT') + gdb.continue_execution_until_error() + + backup_id = self.show_pb(backup_dir, 'node')[0]['id'] + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node', backup_id)['status'], + 'Backup STATUS should be "ERROR"') + + # @unittest.skip("skip") + def test_drop_table(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + connect_1 = node.connect("postgres") + connect_1.execute( + "create table t_heap as select i" + " as id from generate_series(0,100) i") + connect_1.commit() + + connect_2 = node.connect("postgres") + connect_2.execute("SELECT * FROM t_heap") + connect_2.commit() + + # DROP table + connect_2.execute("DROP TABLE t_heap") + connect_2.commit() + + # FULL backup + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + # @unittest.skip("skip") + def test_basic_missing_file_permissions(self): + """""" + if os.name == 'nt': + self.skipTest('Skipped because it is POSIX only test') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + relative_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('pg_class')").decode('utf-8').rstrip() + + full_path = os.path.join(node.data_dir, relative_path) + + os.chmod(full_path, 000) + + try: + # FULL backup + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of missing permissions" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Cannot open file', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + os.chmod(full_path, 700) + + # @unittest.skip("skip") + def test_basic_missing_dir_permissions(self): + """""" + if os.name == 'nt': + self.skipTest('Skipped because it is POSIX only test') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + full_path = os.path.join(node.data_dir, 'pg_twophase') + + os.chmod(full_path, 000) + + try: + # FULL backup + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of missing permissions" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Cannot open directory', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + os.rmdir(full_path) + + # @unittest.skip("skip") + def test_backup_with_least_privileges_role(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums'], + pg_options={'archive_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + 'CREATE DATABASE backupdb') + + if self.ptrack: + node.safe_psql( + "backupdb", + "CREATE SCHEMA ptrack; " + "CREATE EXTENSION ptrack WITH SCHEMA ptrack") + + # PG 9.5 + if self.get_version(node) < 90600: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;") + # PG 9.6 + elif self.get_version(node) > 90600 and self.get_version(node) < 100000: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_xlog_replay_location() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + # >= 10 && < 15 + elif self.get_version(node) >= 100000 and self.get_version(node) < 150000: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + # >= 15 + else: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_start(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_stop(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + + if self.ptrack: + node.safe_psql( + "backupdb", + "GRANT USAGE ON SCHEMA ptrack TO backup") + + node.safe_psql( + "backupdb", + "GRANT EXECUTE ON FUNCTION ptrack.ptrack_get_pagemapset(pg_lsn) TO backup; " + "GRANT EXECUTE ON FUNCTION ptrack.ptrack_init_lsn() TO backup;") + + if ProbackupTest.enterprise: + node.safe_psql( + "backupdb", + "GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_version() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_edition() TO backup;") + + # FULL backup + self.backup_node( + backup_dir, 'node', node, + datname='backupdb', options=['--stream', '-U', 'backup']) + self.backup_node( + backup_dir, 'node', node, + datname='backupdb', options=['-U', 'backup']) + + # PAGE + self.backup_node( + backup_dir, 'node', node, backup_type='page', + datname='backupdb', options=['-U', 'backup']) + self.backup_node( + backup_dir, 'node', node, backup_type='page', datname='backupdb', + options=['--stream', '-U', 'backup']) + + # DELTA + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + datname='backupdb', options=['-U', 'backup']) + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + datname='backupdb', options=['--stream', '-U', 'backup']) + + # PTRACK + if self.ptrack: + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + datname='backupdb', options=['-U', 'backup']) + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + datname='backupdb', options=['--stream', '-U', 'backup']) + + # @unittest.skip("skip") + def test_parent_choosing(self): + """ + PAGE3 <- RUNNING(parent should be FULL) + PAGE2 <- OK + PAGE1 <- CORRUPT + FULL + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + full_id = self.backup_node(backup_dir, 'node', node) + + # PAGE1 + page1_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGE2 + page2_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Change PAGE1 to ERROR + self.change_backup_status(backup_dir, 'node', page1_id, 'ERROR') + + # PAGE3 + page3_id = self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=['--log-level-file=LOG']) + + log_file_path = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(log_file_path) as f: + log_file_content = f.read() + + self.assertIn( + "WARNING: Backup {0} has invalid parent: {1}. " + "Cannot be a parent".format(page2_id, page1_id), + log_file_content) + + self.assertIn( + "WARNING: Backup {0} has status: ERROR. " + "Cannot be a parent".format(page1_id), + log_file_content) + + self.assertIn( + "Parent backup: {0}".format(full_id), + log_file_content) + + self.assertEqual( + self.show_pb( + backup_dir, 'node', backup_id=page3_id)['parent-backup-id'], + full_id) + + # @unittest.skip("skip") + def test_parent_choosing_1(self): + """ + PAGE3 <- RUNNING(parent should be FULL) + PAGE2 <- OK + PAGE1 <- (missing) + FULL + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + full_id = self.backup_node(backup_dir, 'node', node) + + # PAGE1 + page1_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGE2 + page2_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Delete PAGE1 + shutil.rmtree( + os.path.join(backup_dir, 'backups', 'node', page1_id)) + + # PAGE3 + page3_id = self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=['--log-level-file=LOG']) + + log_file_path = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(log_file_path) as f: + log_file_content = f.read() + + self.assertIn( + "WARNING: Backup {0} has missing parent: {1}. " + "Cannot be a parent".format(page2_id, page1_id), + log_file_content) + + self.assertIn( + "Parent backup: {0}".format(full_id), + log_file_content) + + self.assertEqual( + self.show_pb( + backup_dir, 'node', backup_id=page3_id)['parent-backup-id'], + full_id) + + # @unittest.skip("skip") + def test_parent_choosing_2(self): + """ + PAGE3 <- RUNNING(backup should fail) + PAGE2 <- OK + PAGE1 <- OK + FULL <- (missing) + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + full_id = self.backup_node(backup_dir, 'node', node) + + # PAGE1 + page1_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGE2 + page2_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Delete FULL + shutil.rmtree( + os.path.join(backup_dir, 'backups', 'node', full_id)) + + # PAGE3 + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=['--log-level-file=LOG']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because FULL backup is missing" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertTrue( + 'WARNING: Valid full backup on current timeline 1 is not found' in e.message and + 'ERROR: Create new full backup before an incremental one' in e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.assertEqual( + self.show_pb( + backup_dir, 'node')[2]['status'], + 'ERROR') + + # @unittest.skip("skip") + def test_backup_with_less_privileges_role(self): + """ + check permissions correctness from documentation: + https://github.com/postgrespro/pg_probackup/blob/master/Documentation.md#configuring-the-database-cluster + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_timeout': '30s', + 'archive_mode': 'always', + 'checkpoint_timeout': '60s', + 'wal_level': 'logical'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_config(backup_dir, 'node', options=['--archive-timeout=60s']) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + 'CREATE DATABASE backupdb') + + if self.ptrack: + node.safe_psql( + 'backupdb', + 'CREATE EXTENSION ptrack') + + # PG 9.5 + if self.get_version(node) < 90600: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;") + # PG 9.6 + elif self.get_version(node) > 90600 and self.get_version(node) < 100000: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_xlog_replay_location() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; " + "COMMIT;" + ) + # >= 10 && < 15 + elif self.get_version(node) >= 100000 and self.get_version(node) < 150000: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; " + "COMMIT;" + ) + # >= 15 + else: + node.safe_psql( + 'backupdb', + "BEGIN; " + "CREATE ROLE backup WITH LOGIN; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_start(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_stop(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup; " + "COMMIT;" + ) + + # enable STREAM backup + node.safe_psql( + 'backupdb', + 'ALTER ROLE backup WITH REPLICATION;') + + # FULL backup + self.backup_node( + backup_dir, 'node', node, + datname='backupdb', options=['--stream', '-U', 'backup']) + self.backup_node( + backup_dir, 'node', node, + datname='backupdb', options=['-U', 'backup']) + + # PAGE + self.backup_node( + backup_dir, 'node', node, backup_type='page', + datname='backupdb', options=['-U', 'backup']) + self.backup_node( + backup_dir, 'node', node, backup_type='page', datname='backupdb', + options=['--stream', '-U', 'backup']) + + # DELTA + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + datname='backupdb', options=['-U', 'backup']) + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + datname='backupdb', options=['--stream', '-U', 'backup']) + + # PTRACK + if self.ptrack: + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + datname='backupdb', options=['-U', 'backup']) + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + datname='backupdb', options=['--stream', '-U', 'backup']) + + if self.get_version(node) < 90600: + return + + # Restore as replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.restore_node(backup_dir, 'node', replica) + self.set_replica(node, replica) + self.add_instance(backup_dir, 'replica', replica) + self.set_config( + backup_dir, 'replica', + options=['--archive-timeout=120s', '--log-level-console=LOG']) + self.set_archiving(backup_dir, 'replica', replica, replica=True) + self.set_auto_conf(replica, {'hot_standby': 'on'}) + + # freeze bgwriter to get rid of RUNNING XACTS records + # bgwriter_pid = node.auxiliary_pids[ProcessType.BackgroundWriter][0] + # gdb_checkpointer = self.gdb_attach(bgwriter_pid) + + copy_tree( + os.path.join(backup_dir, 'wal', 'node'), + os.path.join(backup_dir, 'wal', 'replica')) + + replica.slow_start(replica=True) + + # self.switch_wal_segment(node) + # self.switch_wal_segment(node) + + self.backup_node( + backup_dir, 'replica', replica, + datname='backupdb', options=['-U', 'backup']) + + # stream full backup from replica + self.backup_node( + backup_dir, 'replica', replica, + datname='backupdb', options=['--stream', '-U', 'backup']) + +# self.switch_wal_segment(node) + + # PAGE backup from replica + self.switch_wal_segment(node) + self.backup_node( + backup_dir, 'replica', replica, backup_type='page', + datname='backupdb', options=['-U', 'backup', '--archive-timeout=30s']) + + self.backup_node( + backup_dir, 'replica', replica, backup_type='page', + datname='backupdb', options=['--stream', '-U', 'backup']) + + # DELTA backup from replica + self.switch_wal_segment(node) + self.backup_node( + backup_dir, 'replica', replica, backup_type='delta', + datname='backupdb', options=['-U', 'backup']) + self.backup_node( + backup_dir, 'replica', replica, backup_type='delta', + datname='backupdb', options=['--stream', '-U', 'backup']) + + # PTRACK backup from replica + if self.ptrack: + self.switch_wal_segment(node) + self.backup_node( + backup_dir, 'replica', replica, backup_type='ptrack', + datname='backupdb', options=['-U', 'backup']) + self.backup_node( + backup_dir, 'replica', replica, backup_type='ptrack', + datname='backupdb', options=['--stream', '-U', 'backup']) + + @unittest.skip("skip") + def test_issue_132(self): + """ + https://github.com/postgrespro/pg_probackup/issues/132 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + with node.connect("postgres") as conn: + for i in range(50000): + conn.execute( + "CREATE TABLE t_{0} as select 1".format(i)) + conn.commit() + + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + node.cleanup() + self.restore_node(backup_dir, 'node', node) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + exit(1) + + @unittest.skip("skip") + def test_issue_132_1(self): + """ + https://github.com/postgrespro/pg_probackup/issues/132 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + # TODO: check version of old binary, it should be 2.1.4, 2.1.5 or 2.2.1 + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + with node.connect("postgres") as conn: + for i in range(30000): + conn.execute( + "CREATE TABLE t_{0} as select 1".format(i)) + conn.commit() + + full_id = self.backup_node( + backup_dir, 'node', node, options=['--stream'], old_binary=True) + + delta_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--stream'], old_binary=True) + + node.cleanup() + + # make sure that new binary can detect corruption + try: + self.validate_pb(backup_dir, 'node', backup_id=full_id) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because FULL backup is CORRUPT" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'WARNING: Backup {0} is a victim of metadata corruption'.format(full_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + try: + self.validate_pb(backup_dir, 'node', backup_id=delta_id) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because FULL backup is CORRUPT" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'WARNING: Backup {0} is a victim of metadata corruption'.format(full_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.assertEqual( + 'CORRUPT', self.show_pb(backup_dir, 'node', full_id)['status'], + 'Backup STATUS should be "CORRUPT"') + + self.assertEqual( + 'ORPHAN', self.show_pb(backup_dir, 'node', delta_id)['status'], + 'Backup STATUS should be "ORPHAN"') + + # check that revalidation is working correctly + try: + self.restore_node( + backup_dir, 'node', node, backup_id=delta_id) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because FULL backup is CORRUPT" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'WARNING: Backup {0} is a victim of metadata corruption'.format(full_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.assertEqual( + 'CORRUPT', self.show_pb(backup_dir, 'node', full_id)['status'], + 'Backup STATUS should be "CORRUPT"') + + self.assertEqual( + 'ORPHAN', self.show_pb(backup_dir, 'node', delta_id)['status'], + 'Backup STATUS should be "ORPHAN"') + + # check that '--no-validate' do not allow to restore ORPHAN backup +# try: +# self.restore_node( +# backup_dir, 'node', node, backup_id=delta_id, +# options=['--no-validate']) +# # we should die here because exception is what we expect to happen +# self.assertEqual( +# 1, 0, +# "Expecting Error because FULL backup is CORRUPT" +# "\n Output: {0} \n CMD: {1}".format( +# repr(self.output), self.cmd)) +# except ProbackupException as e: +# self.assertIn( +# 'Insert data', +# e.message, +# '\n Unexpected Error Message: {0}\n CMD: {1}'.format( +# repr(e.message), self.cmd)) + + node.cleanup() + + output = self.restore_node( + backup_dir, 'node', node, backup_id=full_id, options=['--force']) + + self.assertIn( + 'WARNING: Backup {0} has status: CORRUPT'.format(full_id), + output) + + self.assertIn( + 'WARNING: Backup {0} is corrupt.'.format(full_id), + output) + + self.assertIn( + 'WARNING: Backup {0} is not valid, restore is forced'.format(full_id), + output) + + self.assertIn( + 'INFO: Restore of backup {0} completed.'.format(full_id), + output) + + node.cleanup() + + output = self.restore_node( + backup_dir, 'node', node, backup_id=delta_id, options=['--force']) + + self.assertIn( + 'WARNING: Backup {0} is orphan.'.format(delta_id), + output) + + self.assertIn( + 'WARNING: Backup {0} is not valid, restore is forced'.format(full_id), + output) + + self.assertIn( + 'WARNING: Backup {0} is not valid, restore is forced'.format(delta_id), + output) + + self.assertIn( + 'INFO: Restore of backup {0} completed.'.format(delta_id), + output) + + def test_note_sanity(self): + """ + test that adding note to backup works as expected + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + backup_id = self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--log-level-file=LOG', '--note=test_note']) + + show_backups = self.show_pb(backup_dir, 'node') + + print(self.show_pb(backup_dir, as_text=True, as_json=True)) + + self.assertEqual(show_backups[0]['note'], "test_note") + + self.set_backup(backup_dir, 'node', backup_id, options=['--note=none']) + + backup_meta = self.show_pb(backup_dir, 'node', backup_id) + + self.assertNotIn( + 'note', + backup_meta) + + # @unittest.skip("skip") + def test_parent_backup_made_by_newer_version(self): + """incremental backup with parent made by newer version""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node(backup_dir, 'node', node) + + control_file = os.path.join( + backup_dir, "backups", "node", backup_id, + "backup.control") + + version = self.probackup_version + fake_new_version = str(int(version.split('.')[0]) + 1) + '.0.0' + + with open(control_file, 'r') as f: + data = f.read(); + + data = data.replace(version, fake_new_version) + + with open(control_file, 'w') as f: + f.write(data); + + try: + self.backup_node(backup_dir, 'node', node, backup_type="page") + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because incremental backup should not be possible " + "if parent made by newer version.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "pg_probackup do not guarantee to be forward compatible. " + "Please upgrade pg_probackup binary.", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['status'], "ERROR") + + # @unittest.skip("skip") + def test_issue_289(self): + """ + https://github.com/postgrespro/pg_probackup/issues/289 + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + node.slow_start() + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=['--archive-timeout=10s']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because full backup is missing" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertNotIn( + "INFO: Wait for WAL segment", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertIn( + "ERROR: Create new full backup before an incremental one", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['status'], "ERROR") + + # @unittest.skip("skip") + def test_issue_290(self): + """ + https://github.com/postgrespro/pg_probackup/issues/290 + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + os.rmdir( + os.path.join(backup_dir, "wal", "node")) + + node.slow_start() + + try: + self.backup_node( + backup_dir, 'node', node, + options=['--archive-timeout=10s']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because full backup is missing" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertNotIn( + "INFO: Wait for WAL segment", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertIn( + "WAL archive directory is not accessible", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['status'], "ERROR") + + @unittest.skip("skip") + def test_issue_203(self): + """ + https://github.com/postgrespro/pg_probackup/issues/203 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + with node.connect("postgres") as conn: + for i in range(1000000): + conn.execute( + "CREATE TABLE t_{0} as select 1".format(i)) + conn.commit() + + full_id = self.backup_node( + backup_dir, 'node', node, options=['--stream', '-j2']) + + pgdata = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node(backup_dir, 'node', + node_restored, data_dir=node_restored.data_dir) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_issue_231(self): + """ + https://github.com/postgrespro/pg_probackup/issues/231 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node')) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + datadir = os.path.join(node.data_dir, '123') + + t0 = time() + while True: + with self.assertRaises(ProbackupException) as ctx: + self.backup_node(backup_dir, 'node', node) + pb1 = re.search(r' backup ID: ([^\s,]+),', ctx.exception.message).groups()[0] + + t = time() + if int(pb1, 36) == int(t) and t % 1 < 0.5: + # ok, we have a chance to start next backup in same second + break + elif t - t0 > 20: + # Oops, we are waiting for too long. Looks like this runner + # is too slow. Lets skip the test. + self.skipTest("runner is too slow") + # sleep to the second's end so backup will not sleep for a second. + sleep(1 - t % 1) + + with self.assertRaises(ProbackupException) as ctx: + self.backup_node(backup_dir, 'node', node) + pb2 = re.search(r' backup ID: ([^\s,]+),', ctx.exception.message).groups()[0] + + self.assertNotEqual(pb1, pb2) + + def test_incr_backup_filenode_map(self): + """ + https://github.com/postgrespro/pg_probackup/issues/320 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1'), + initdb_params=['--data-checksums']) + node1.cleanup() + + node.pgbench_init(scale=5) + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1']) + + backup_id = self.backup_node(backup_dir, 'node', node, backup_type='delta') + + node.safe_psql( + 'postgres', + 'reindex index pg_type_oid_index') + + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # incremental restore into node1 + node.cleanup() + + self.restore_node(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + 'select 1') + + # @unittest.skip("skip") + def test_missing_wal_segment(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums'], + pg_options={'archive_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=10) + + node.safe_psql( + 'postgres', + 'CREATE DATABASE backupdb') + + # get segments in pg_wal, sort then and remove all but the latest + pg_wal_dir = os.path.join(node.data_dir, 'pg_wal') + + if node.major_version >= 10: + pg_wal_dir = os.path.join(node.data_dir, 'pg_wal') + else: + pg_wal_dir = os.path.join(node.data_dir, 'pg_xlog') + + # Full backup in streaming mode + gdb = self.backup_node( + backup_dir, 'node', node, datname='backupdb', + options=['--stream', '--log-level-file=INFO'], gdb=True) + + # break at streaming start + gdb.set_breakpoint('start_WAL_streaming') + gdb.run_until_break() + + # generate some more data + node.pgbench_init(scale=3) + + # remove redundant WAL segments in pg_wal + files = os.listdir(pg_wal_dir) + files.sort(reverse=True) + + # leave first two files in list + del files[:2] + for filename in files: + os.remove(os.path.join(pg_wal_dir, filename)) + + gdb.continue_execution_until_exit() + + self.assertIn( + 'unexpected termination of replication stream: ERROR: requested WAL segment', + gdb.output) + + self.assertIn( + 'has already been removed', + gdb.output) + + self.assertIn( + 'ERROR: Interrupted during waiting for WAL streaming', + gdb.output) + + self.assertIn( + 'WARNING: A backup is in progress, stopping it', + gdb.output) + + # TODO: check the same for PAGE backup + + # @unittest.skip("skip") + def test_missing_replication_permission(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) +# self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'node', replica) + + # Settings for Replica + self.set_replica(node, replica) + replica.slow_start(replica=True) + + node.safe_psql( + 'postgres', + 'CREATE DATABASE backupdb') + + # PG 9.5 + if self.get_version(node) < 90600: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;") + # PG 9.6 + elif self.get_version(node) > 90600 and self.get_version(node) < 100000: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_xlog_replay_location() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;") + # >= 10 && < 15 + elif self.get_version(node) >= 100000 and self.get_version(node) < 150000: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + # >= 15 + else: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_start(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_stop(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + + if ProbackupTest.enterprise: + node.safe_psql( + "backupdb", + "GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_version() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_edition() TO backup;") + + sleep(2) + replica.promote() + + # Delta backup + try: + self.backup_node( + backup_dir, 'node', replica, backup_type='delta', + data_dir=replica.data_dir, datname='backupdb', options=['--stream', '-U', 'backup']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because incremental backup should not be possible " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + # 9.5: ERROR: must be superuser or replication role to run a backup + # >=9.6: FATAL: must be superuser or replication role to start walsender + if self.pg_config_version < 160000: + self.assertRegex( + e.message, + "ERROR: must be superuser or replication role to run a backup|" + "FATAL: must be superuser or replication role to start walsender", + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + else: + self.assertRegex( + e.message, + "FATAL: permission denied to start WAL sender\n" + "DETAIL: Only roles with the REPLICATION", + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_missing_replication_permission_1(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'node', replica) + + # Settings for Replica + self.set_replica(node, replica) + replica.slow_start(replica=True) + + node.safe_psql( + 'postgres', + 'CREATE DATABASE backupdb') + + # PG 9.5 + if self.get_version(node) < 90600: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;") + # PG 9.6 + elif self.get_version(node) > 90600 and self.get_version(node) < 100000: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_xlog_replay_location() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + # >= 10 && < 15 + elif self.get_version(node) >= 100000 and self.get_version(node) < 150000: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + # > 15 + else: + node.safe_psql( + 'backupdb', + "CREATE ROLE backup WITH LOGIN; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_start(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_stop(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + + if ProbackupTest.enterprise: + node.safe_psql( + "backupdb", + "GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_version() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_edition() TO backup;") + + replica.promote() + + # PAGE + output = self.backup_node( + backup_dir, 'node', replica, backup_type='page', + data_dir=replica.data_dir, datname='backupdb', options=['-U', 'backup'], + return_id=False) + + self.assertIn( + 'WARNING: Valid full backup on current timeline 2 is not found, trying to look up on previous timelines', + output) + + # Messages before 14 + # 'WARNING: could not connect to database backupdb: FATAL: must be superuser or replication role to start walsender' + # Messages for >=14 + # 'WARNING: could not connect to database backupdb: connection to server on socket "/tmp/.s.PGSQL.30983" failed: FATAL: must be superuser or replication role to start walsender' + # 'WARNING: could not connect to database backupdb: connection to server at "localhost" (127.0.0.1), port 29732 failed: FATAL: must be superuser or replication role to start walsender' + # OS-dependant messages: + # 'WARNING: could not connect to database backupdb: connection to server at "localhost" (::1), port 12101 failed: Connection refused\n\tIs the server running on that host and accepting TCP/IP connections?\nconnection to server at "localhost" (127.0.0.1), port 12101 failed: FATAL: must be superuser or replication role to start walsender' + + if self.pg_config_version < 160000: + self.assertRegex( + output, + r'WARNING: could not connect to database backupdb:[\s\S]*?' + r'FATAL: must be superuser or replication role to start walsender') + else: + self.assertRegex( + output, + r'WARNING: could not connect to database backupdb:[\s\S]*?' + r'FATAL: permission denied to start WAL sender') + + # @unittest.skip("skip") + def test_basic_backup_default_transaction_read_only(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={'default_transaction_read_only': 'on'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + try: + node.safe_psql( + 'postgres', + 'create temp table t1()') + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because incremental backup should not be possible " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except QueryException as e: + self.assertIn( + "cannot execute CREATE TABLE in a read-only transaction", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # FULL backup + self.backup_node( + backup_dir, 'node', node, + options=['--stream']) + + # DELTA backup + self.backup_node( + backup_dir, 'node', node, backup_type='delta', options=['--stream']) + + # PAGE backup + self.backup_node(backup_dir, 'node', node, backup_type='page') + + # @unittest.skip("skip") + def test_backup_atexit(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=5) + + # Full backup in streaming mode + gdb = self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--log-level-file=VERBOSE'], gdb=True) + + # break at streaming start + gdb.set_breakpoint('backup_data_file') + gdb.run_until_break() + + gdb.remove_all_breakpoints() + gdb._execute('signal SIGINT') + sleep(1) + + self.assertEqual( + self.show_pb( + backup_dir, 'node')[0]['status'], 'ERROR') + + with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: + log_content = f.read() + #print(log_content) + self.assertIn( + 'WARNING: A backup is in progress, stopping it.', + log_content) + + if self.get_version(node) < 150000: + self.assertIn( + 'FROM pg_catalog.pg_stop_backup', + log_content) + else: + self.assertIn( + 'FROM pg_catalog.pg_backup_stop', + log_content) + + self.assertIn( + 'setting its status to ERROR', + log_content) + + # @unittest.skip("skip") + def test_pg_stop_backup_missing_permissions(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=5) + + self.simple_bootstrap(node, 'backup') + + if self.get_version(node) < 90600: + node.safe_psql( + 'postgres', + 'REVOKE EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() FROM backup') + elif self.get_version(node) > 90600 and self.get_version(node) < 100000: + node.safe_psql( + 'postgres', + 'REVOKE EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) FROM backup') + elif self.get_version(node) < 150000: + node.safe_psql( + 'postgres', + 'REVOKE EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) FROM backup') + else: + node.safe_psql( + 'postgres', + 'REVOKE EXECUTE ON FUNCTION pg_catalog.pg_backup_stop(boolean) FROM backup') + + + # Full backup in streaming mode + try: + self.backup_node( + backup_dir, 'node', node, + options=['--stream', '-U', 'backup']) + # we should die here because exception is what we expect to happen + if self.get_version(node) < 150000: + self.assertEqual( + 1, 0, + "Expecting Error because of missing permissions on pg_stop_backup " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + else: + self.assertEqual( + 1, 0, + "Expecting Error because of missing permissions on pg_backup_stop " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + if self.get_version(node) < 150000: + self.assertIn( + "ERROR: permission denied for function pg_stop_backup", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + else: + self.assertIn( + "ERROR: permission denied for function pg_backup_stop", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertIn( + "query was: SELECT pg_catalog.txid_snapshot_xmax", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_start_time(self): + """Test, that option --start-time allows to set backup_id and restore""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + startTime = int(time()) + self.backup_node( + backup_dir, 'node', node, backup_type='full', + options=['--stream', '--start-time={0}'.format(str(startTime))]) + # restore FULL backup by backup_id calculated from start-time + self.restore_node( + backup_dir, 'node', + data_dir=os.path.join(self.tmp_path, self.module_name, self.fname, 'node_restored_full'), + backup_id=base36enc(startTime)) + + #FULL backup with incorrect start time + try: + startTime = str(int(time()-100000)) + self.backup_node( + backup_dir, 'node', node, backup_type='full', + options=['--stream', '--start-time={0}'.format(startTime)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + 'Expecting Error because start time for new backup must be newer ' + '\n Output: {0} \n CMD: {1}'.format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertRegex( + e.message, + r"ERROR: Can't assign backup_id from requested start_time \(\w*\), this time must be later that backup \w*\n", + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # DELTA backup + startTime = int(time()) + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--stream', '--start-time={0}'.format(str(startTime))]) + # restore DELTA backup by backup_id calculated from start-time + self.restore_node( + backup_dir, 'node', + data_dir=os.path.join(self.tmp_path, self.module_name, self.fname, 'node_restored_delta'), + backup_id=base36enc(startTime)) + + # PAGE backup + startTime = int(time()) + self.backup_node( + backup_dir, 'node', node, backup_type='page', + options=['--stream', '--start-time={0}'.format(str(startTime))]) + # restore PAGE backup by backup_id calculated from start-time + self.restore_node( + backup_dir, 'node', + data_dir=os.path.join(self.tmp_path, self.module_name, self.fname, 'node_restored_page'), + backup_id=base36enc(startTime)) + + # PTRACK backup + if self.ptrack: + node.safe_psql( + 'postgres', + 'create extension ptrack') + + startTime = int(time()) + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + options=['--stream', '--start-time={0}'.format(str(startTime))]) + # restore PTRACK backup by backup_id calculated from start-time + self.restore_node( + backup_dir, 'node', + data_dir=os.path.join(self.tmp_path, self.module_name, self.fname, 'node_restored_ptrack'), + backup_id=base36enc(startTime)) + + # @unittest.skip("skip") + def test_start_time_few_nodes(self): + """Test, that we can synchronize backup_id's for different DBs""" + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + backup_dir1 = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup1') + self.init_pb(backup_dir1) + self.add_instance(backup_dir1, 'node1', node1) + self.set_archiving(backup_dir1, 'node1', node1) + node1.slow_start() + + node2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node2'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + backup_dir2 = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup2') + self.init_pb(backup_dir2) + self.add_instance(backup_dir2, 'node2', node2) + self.set_archiving(backup_dir2, 'node2', node2) + node2.slow_start() + + # FULL backup + startTime = str(int(time())) + self.backup_node( + backup_dir1, 'node1', node1, backup_type='full', + options=['--stream', '--start-time={0}'.format(startTime)]) + self.backup_node( + backup_dir2, 'node2', node2, backup_type='full', + options=['--stream', '--start-time={0}'.format(startTime)]) + show_backup1 = self.show_pb(backup_dir1, 'node1')[0] + show_backup2 = self.show_pb(backup_dir2, 'node2')[0] + self.assertEqual(show_backup1['id'], show_backup2['id']) + + # DELTA backup + startTime = str(int(time())) + self.backup_node( + backup_dir1, 'node1', node1, backup_type='delta', + options=['--stream', '--start-time={0}'.format(startTime)]) + self.backup_node( + backup_dir2, 'node2', node2, backup_type='delta', + options=['--stream', '--start-time={0}'.format(startTime)]) + show_backup1 = self.show_pb(backup_dir1, 'node1')[1] + show_backup2 = self.show_pb(backup_dir2, 'node2')[1] + self.assertEqual(show_backup1['id'], show_backup2['id']) + + # PAGE backup + startTime = str(int(time())) + self.backup_node( + backup_dir1, 'node1', node1, backup_type='page', + options=['--stream', '--start-time={0}'.format(startTime)]) + self.backup_node( + backup_dir2, 'node2', node2, backup_type='page', + options=['--stream', '--start-time={0}'.format(startTime)]) + show_backup1 = self.show_pb(backup_dir1, 'node1')[2] + show_backup2 = self.show_pb(backup_dir2, 'node2')[2] + self.assertEqual(show_backup1['id'], show_backup2['id']) + + # PTRACK backup + if self.ptrack: + node1.safe_psql( + 'postgres', + 'create extension ptrack') + node2.safe_psql( + 'postgres', + 'create extension ptrack') + + startTime = str(int(time())) + self.backup_node( + backup_dir1, 'node1', node1, backup_type='ptrack', + options=['--stream', '--start-time={0}'.format(startTime)]) + self.backup_node( + backup_dir2, 'node2', node2, backup_type='ptrack', + options=['--stream', '--start-time={0}'.format(startTime)]) + show_backup1 = self.show_pb(backup_dir1, 'node1')[3] + show_backup2 = self.show_pb(backup_dir2, 'node2')[3] + self.assertEqual(show_backup1['id'], show_backup2['id']) + + def test_regress_issue_585(self): + """https://github.com/postgrespro/pg_probackup/issues/585""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # create couple of files that looks like db files + with open(os.path.join(node.data_dir, 'pg_multixact/offsets/1000'),'wb') as f: + pass + with open(os.path.join(node.data_dir, 'pg_multixact/members/1000'),'wb') as f: + pass + + self.backup_node( + backup_dir, 'node', node, backup_type='full', + options=['--stream']) + + output = self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--stream'], + return_id=False, + ) + self.assertNotRegex(output, r'WARNING: [^\n]* was stored as .* but looks like') + + node.cleanup() + + output = self.restore_node(backup_dir, 'node', node) + self.assertNotRegex(output, r'WARNING: [^\n]* was stored as .* but looks like') + + def test_2_delta_backups(self): + """https://github.com/postgrespro/pg_probackup/issues/596""" + node = self.make_simple_node('node', + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + # self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL + full_backup_id = self.backup_node(backup_dir, 'node', node, options=["--stream"]) + + # delta backup mode + delta_backup_id1 = self.backup_node( + backup_dir, 'node', node, backup_type="delta", options=["--stream"]) + + delta_backup_id2 = self.backup_node( + backup_dir, 'node', node, backup_type="delta", options=["--stream"]) + + # postgresql.conf and pg_hba.conf shouldn't be copied + conf_file = os.path.join(backup_dir, 'backups', 'node', delta_backup_id1, 'database', 'postgresql.conf') + self.assertFalse( + os.path.exists(conf_file), + "File should not exist: {0}".format(conf_file)) + conf_file = os.path.join(backup_dir, 'backups', 'node', delta_backup_id2, 'database', 'postgresql.conf') + print(conf_file) + self.assertFalse( + os.path.exists(conf_file), + "File should not exist: {0}".format(conf_file)) diff --git a/tests/catchup_test.py b/tests/catchup_test.py new file mode 100644 index 000000000..cf8388dd2 --- /dev/null +++ b/tests/catchup_test.py @@ -0,0 +1,1626 @@ +import os +from pathlib import Path +import signal +import unittest +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException + +class CatchupTest(ProbackupTest, unittest.TestCase): + +######################################### +# Basic tests +######################################### + def test_basic_full_catchup(self): + """ + Test 'multithreaded basebackup' mode (aka FULL catchup) + """ + # preparation + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True + ) + src_pg.slow_start() + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question AS SELECT 42 AS answer") + src_query_result = src_pg.table_checksum("ultimate_question") + + # do full catchup + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # 1st check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # run&recover catchup'ed instance + src_pg.stop() + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + + # 2nd check: run verification query + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + # Cleanup + dst_pg.stop() + #self.assertEqual(1, 0, 'Stop test') + + def test_full_catchup_with_tablespace(self): + """ + Test tablespace transfers + """ + # preparation + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True + ) + src_pg.slow_start() + tblspace1_old_path = self.get_tblspace_path(src_pg, 'tblspace1_old') + self.create_tblspace_in_node(src_pg, 'tblspace1', tblspc_path = tblspace1_old_path) + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question TABLESPACE tblspace1 AS SELECT 42 AS answer") + src_query_result = src_pg.table_checksum("ultimate_question") + + # do full catchup with tablespace mapping + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + tblspace1_new_path = self.get_tblspace_path(dst_pg, 'tblspace1_new') + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', + '-p', str(src_pg.port), + '--stream', + '-T', '{0}={1}'.format(tblspace1_old_path, tblspace1_new_path) + ] + ) + + # 1st check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # make changes in master tablespace + src_pg.safe_psql( + "postgres", + "UPDATE ultimate_question SET answer = -1") + src_pg.stop() + + # run&recover catchup'ed instance + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + + # 2nd check: run verification query + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + # Cleanup + dst_pg.stop() + + def test_basic_delta_catchup(self): + """ + Test delta catchup + """ + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question(answer int)") + + # preparation 2: make clean shutdowned lagging behind replica + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + self.set_replica(src_pg, dst_pg) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start(replica = True) + dst_pg.stop() + + # preparation 3: make changes on master (source) + src_pg.pgbench_init(scale = 10) + pgbench = src_pg.pgbench(options=['-T', '10', '--no-vacuum']) + pgbench.wait() + src_pg.safe_psql("postgres", "INSERT INTO ultimate_question VALUES(42)") + src_query_result = src_pg.table_checksum("ultimate_question") + + # do delta catchup + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # 1st check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # run&recover catchup'ed instance + src_pg.stop() + self.set_replica(master = src_pg, replica = dst_pg) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start(replica = True) + + # 2nd check: run verification query + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + # Cleanup + dst_pg.stop() + #self.assertEqual(1, 0, 'Stop test') + + def test_basic_ptrack_catchup(self): + """ + Test ptrack catchup + """ + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + ptrack_enable = True, + initdb_params = ['--data-checksums'] + ) + src_pg.slow_start() + src_pg.safe_psql("postgres", "CREATE EXTENSION ptrack") + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question(answer int)") + + # preparation 2: make clean shutdowned lagging behind replica + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + self.set_replica(src_pg, dst_pg) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start(replica = True) + dst_pg.stop() + + # preparation 3: make changes on master (source) + src_pg.pgbench_init(scale = 10) + pgbench = src_pg.pgbench(options=['-T', '10', '--no-vacuum']) + pgbench.wait() + src_pg.safe_psql("postgres", "INSERT INTO ultimate_question VALUES(42)") + src_query_result = src_pg.table_checksum("ultimate_question") + + # do ptrack catchup + self.catchup_node( + backup_mode = 'PTRACK', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # 1st check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # run&recover catchup'ed instance + src_pg.stop() + self.set_replica(master = src_pg, replica = dst_pg) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start(replica = True) + + # 2nd check: run verification query + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + # Cleanup + dst_pg.stop() + #self.assertEqual(1, 0, 'Stop test') + + def test_tli_delta_catchup(self): + """ + Test that we correctly follow timeline change with delta catchup + """ + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + + # preparation 2: destination + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_pg.stop() + + # preparation 3: promote source + src_pg.stop() + self.set_replica(dst_pg, src_pg) # fake replication + src_pg.slow_start(replica = True) + src_pg.promote() + src_pg.safe_psql("postgres", "CREATE TABLE ultimate_question AS SELECT 42 AS answer") + src_query_result = src_pg.table_checksum("ultimate_question") + + # do catchup (src_tli = 2, dst_tli = 1) + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # 1st check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # run&recover catchup'ed instance + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + self.set_replica(master = src_pg, replica = dst_pg) + dst_pg.slow_start(replica = True) + + # 2nd check: run verification query + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + dst_pg.stop() + + # do catchup (src_tli = 2, dst_tli = 2) + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # Cleanup + src_pg.stop() + + def test_tli_ptrack_catchup(self): + """ + Test that we correctly follow timeline change with ptrack catchup + """ + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + ptrack_enable = True, + initdb_params = ['--data-checksums'] + ) + src_pg.slow_start() + src_pg.safe_psql("postgres", "CREATE EXTENSION ptrack") + + # preparation 2: destination + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_pg.stop() + + # preparation 3: promote source + src_pg.stop() + self.set_replica(dst_pg, src_pg) # fake replication + src_pg.slow_start(replica = True) + src_pg.promote() + + src_pg.safe_psql("postgres", "CHECKPOINT") # force postgres to update tli in 'SELECT timeline_id FROM pg_catalog.pg_control_checkpoint()' + src_tli = src_pg.safe_psql("postgres", "SELECT timeline_id FROM pg_catalog.pg_control_checkpoint()").decode('utf-8').rstrip() + self.assertEqual(src_tli, "2", "Postgres didn't update TLI after promote") + + src_pg.safe_psql("postgres", "CREATE TABLE ultimate_question AS SELECT 42 AS answer") + src_query_result = src_pg.table_checksum("ultimate_question") + + # do catchup (src_tli = 2, dst_tli = 1) + self.catchup_node( + backup_mode = 'PTRACK', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # 1st check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # run&recover catchup'ed instance + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + self.set_replica(master = src_pg, replica = dst_pg) + dst_pg.slow_start(replica = True) + + # 2nd check: run verification query + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + dst_pg.stop() + + # do catchup (src_tli = 2, dst_tli = 2) + self.catchup_node( + backup_mode = 'PTRACK', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # Cleanup + src_pg.stop() + +######################################### +# Test various corner conditions +######################################### + def test_table_drop_with_delta(self): + """ + Test that dropped table in source will be dropped in delta catchup'ed instance too + """ + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question AS SELECT 42 AS answer") + + # preparation 2: make clean shutdowned lagging behind replica + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_pg.stop() + + # preparation 3: make changes on master (source) + # perform checkpoint twice to ensure, that datafile is actually deleted on filesystem + src_pg.safe_psql("postgres", "DROP TABLE ultimate_question") + src_pg.safe_psql("postgres", "CHECKPOINT") + src_pg.safe_psql("postgres", "CHECKPOINT") + + # do delta catchup + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # Check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # Cleanup + src_pg.stop() + + def test_table_drop_with_ptrack(self): + """ + Test that dropped table in source will be dropped in ptrack catchup'ed instance too + """ + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + ptrack_enable = True, + initdb_params = ['--data-checksums'] + ) + src_pg.slow_start() + src_pg.safe_psql("postgres", "CREATE EXTENSION ptrack") + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question AS SELECT 42 AS answer") + + # preparation 2: make clean shutdowned lagging behind replica + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_pg.stop() + + # preparation 3: make changes on master (source) + # perform checkpoint twice to ensure, that datafile is actually deleted on filesystem + src_pg.safe_psql("postgres", "DROP TABLE ultimate_question") + src_pg.safe_psql("postgres", "CHECKPOINT") + src_pg.safe_psql("postgres", "CHECKPOINT") + + # do ptrack catchup + self.catchup_node( + backup_mode = 'PTRACK', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # Check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # Cleanup + src_pg.stop() + + def test_tablefile_truncation_with_delta(self): + """ + Test that truncated table in source will be truncated in delta catchup'ed instance too + """ + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + src_pg.safe_psql( + "postgres", + "CREATE SEQUENCE t_seq; " + "CREATE TABLE t_heap AS SELECT i AS id, " + "md5(i::text) AS text, " + "md5(repeat(i::text, 10))::tsvector AS tsvector " + "FROM generate_series(0, 1024) i") + src_pg.safe_psql("postgres", "VACUUM t_heap") + + # preparation 2: make clean shutdowned lagging behind replica + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dest_options = {} + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_pg.stop() + + # preparation 3: make changes on master (source) + src_pg.safe_psql("postgres", "DELETE FROM t_heap WHERE ctid >= '(11,0)'") + src_pg.safe_psql("postgres", "VACUUM t_heap") + + # do delta catchup + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # Check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # Cleanup + src_pg.stop() + + def test_tablefile_truncation_with_ptrack(self): + """ + Test that truncated table in source will be truncated in ptrack catchup'ed instance too + """ + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + ptrack_enable = True, + initdb_params = ['--data-checksums'] + ) + src_pg.slow_start() + src_pg.safe_psql("postgres", "CREATE EXTENSION ptrack") + src_pg.safe_psql( + "postgres", + "CREATE SEQUENCE t_seq; " + "CREATE TABLE t_heap AS SELECT i AS id, " + "md5(i::text) AS text, " + "md5(repeat(i::text, 10))::tsvector AS tsvector " + "FROM generate_series(0, 1024) i") + src_pg.safe_psql("postgres", "VACUUM t_heap") + + # preparation 2: make clean shutdowned lagging behind replica + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dest_options = {} + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_pg.stop() + + # preparation 3: make changes on master (source) + src_pg.safe_psql("postgres", "DELETE FROM t_heap WHERE ctid >= '(11,0)'") + src_pg.safe_psql("postgres", "VACUUM t_heap") + + # do ptrack catchup + self.catchup_node( + backup_mode = 'PTRACK', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # Check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # Cleanup + src_pg.stop() + +######################################### +# Test reaction on user errors +######################################### + def test_local_tablespace_without_mapping(self): + """ + Test that we detect absence of needed --tablespace-mapping option + """ + if self.remote: + self.skipTest('Skipped because this test tests local catchup error handling') + + src_pg = self.make_simple_node(base_dir = os.path.join(self.module_name, self.fname, 'src')) + src_pg.slow_start() + + tblspace_path = self.get_tblspace_path(src_pg, 'tblspace') + self.create_tblspace_in_node( + src_pg, 'tblspace', + tblspc_path = tblspace_path) + + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question TABLESPACE tblspace AS SELECT 42 AS answer") + + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + try: + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', + '-p', str(src_pg.port), + '--stream', + ] + ) + self.assertEqual(1, 0, "Expecting Error because '-T' parameter is not specified.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Local catchup executed, but source database contains tablespace', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + # Cleanup + src_pg.stop() + + def test_running_dest_postmaster(self): + """ + Test that we detect running postmaster in destination + """ + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + + # preparation 2: destination + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + # leave running destination postmaster + # so don't call dst_pg.stop() + + # try delta catchup + try: + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + self.assertEqual(1, 0, "Expecting Error because postmaster in destination is running.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Postmaster with pid ', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + # Cleanup + src_pg.stop() + + def test_same_db_id(self): + """ + Test that we detect different id's of source and destination + """ + # preparation: + # source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True + ) + src_pg.slow_start() + # destination + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_pg.stop() + # fake destination + fake_dst_pg = self.make_simple_node(base_dir = os.path.join(self.module_name, self.fname, 'fake_dst')) + # fake source + fake_src_pg = self.make_simple_node(base_dir = os.path.join(self.module_name, self.fname, 'fake_src')) + + # try delta catchup (src (with correct src conn), fake_dst) + try: + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = fake_dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + self.assertEqual(1, 0, "Expecting Error because database identifiers mismatch.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Database identifiers mismatch: ', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + # try delta catchup (fake_src (with wrong src conn), dst) + try: + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = fake_src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + self.assertEqual(1, 0, "Expecting Error because database identifiers mismatch.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Database identifiers mismatch: ', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + # Cleanup + src_pg.stop() + + def test_tli_destination_mismatch(self): + """ + Test that we detect TLI mismatch in destination + """ + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + + # preparation 2: destination + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + self.set_replica(src_pg, dst_pg) + dst_pg.slow_start(replica = True) + dst_pg.promote() + dst_pg.stop() + + # preparation 3: "useful" changes + src_pg.safe_psql("postgres", "CREATE TABLE ultimate_question AS SELECT 42 AS answer") + src_query_result = src_pg.table_checksum("ultimate_question") + + # try catchup + try: + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_query_result = dst_pg.table_checksum("ultimate_question") + dst_pg.stop() + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + except ProbackupException as e: + self.assertIn( + 'ERROR: Source is behind destination in timeline history', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + # Cleanup + src_pg.stop() + + def test_tli_source_mismatch(self): + """ + Test that we detect TLI mismatch in source history + """ + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + + # preparation 2: fake source (promouted copy) + fake_src_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'fake_src')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = fake_src_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + fake_src_options = {} + fake_src_options['port'] = str(fake_src_pg.port) + self.set_auto_conf(fake_src_pg, fake_src_options) + self.set_replica(src_pg, fake_src_pg) + fake_src_pg.slow_start(replica = True) + fake_src_pg.promote() + self.switch_wal_segment(fake_src_pg) + fake_src_pg.safe_psql( + "postgres", + "CREATE TABLE t_heap AS SELECT i AS id, " + "md5(i::text) AS text, " + "md5(repeat(i::text, 10))::tsvector AS tsvector " + "FROM generate_series(0, 256) i") + self.switch_wal_segment(fake_src_pg) + fake_src_pg.safe_psql("postgres", "CREATE TABLE ultimate_question AS SELECT 'trash' AS garbage") + + # preparation 3: destination + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_pg.stop() + + # preparation 4: "useful" changes + src_pg.safe_psql("postgres", "CREATE TABLE ultimate_question AS SELECT 42 AS answer") + src_query_result = src_pg.table_checksum("ultimate_question") + + # try catchup + try: + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = fake_src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(fake_src_pg.port), '--stream'] + ) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_query_result = dst_pg.table_checksum("ultimate_question") + dst_pg.stop() + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + except ProbackupException as e: + self.assertIn( + 'ERROR: Destination is not in source timeline history', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + # Cleanup + src_pg.stop() + fake_src_pg.stop() + +######################################### +# Test unclean destination +######################################### + def test_unclean_delta_catchup(self): + """ + Test that we correctly recover uncleanly shutdowned destination + """ + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question(answer int)") + + # preparation 2: destination + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # try #1 + try: + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + self.assertEqual(1, 0, "Expecting Error because destination pg is not cleanly shutdowned.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Destination directory contains "backup_label" file', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + # try #2 + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + self.assertNotEqual(dst_pg.pid, 0, "Cannot detect pid of running postgres") + dst_pg.kill() + + # preparation 3: make changes on master (source) + src_pg.pgbench_init(scale = 10) + pgbench = src_pg.pgbench(options=['-T', '10', '--no-vacuum']) + pgbench.wait() + src_pg.safe_psql("postgres", "INSERT INTO ultimate_question VALUES(42)") + src_query_result = src_pg.table_checksum("ultimate_question") + + # do delta catchup + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # 1st check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # run&recover catchup'ed instance + src_pg.stop() + self.set_replica(master = src_pg, replica = dst_pg) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start(replica = True) + + # 2nd check: run verification query + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + # Cleanup + dst_pg.stop() + + def test_unclean_ptrack_catchup(self): + """ + Test that we correctly recover uncleanly shutdowned destination + """ + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + ptrack_enable = True, + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + src_pg.safe_psql("postgres", "CREATE EXTENSION ptrack") + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question(answer int)") + + # preparation 2: destination + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # try #1 + try: + self.catchup_node( + backup_mode = 'PTRACK', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + self.assertEqual(1, 0, "Expecting Error because destination pg is not cleanly shutdowned.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Destination directory contains "backup_label" file', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + # try #2 + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + self.assertNotEqual(dst_pg.pid, 0, "Cannot detect pid of running postgres") + dst_pg.kill() + + # preparation 3: make changes on master (source) + src_pg.pgbench_init(scale = 10) + pgbench = src_pg.pgbench(options=['-T', '10', '--no-vacuum']) + pgbench.wait() + src_pg.safe_psql("postgres", "INSERT INTO ultimate_question VALUES(42)") + src_query_result = src_pg.table_checksum("ultimate_question") + + # do delta catchup + self.catchup_node( + backup_mode = 'PTRACK', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + # 1st check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # run&recover catchup'ed instance + src_pg.stop() + self.set_replica(master = src_pg, replica = dst_pg) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start(replica = True) + + # 2nd check: run verification query + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + # Cleanup + dst_pg.stop() + +######################################### +# Test replication slot logic +# +# -S, --slot=SLOTNAME replication slot to use +# --temp-slot use temporary replication slot +# -P --perm-slot create permanent replication slot +# --primary-slot-name=SLOTNAME value for primary_slot_name parameter +# +# 1. if "--slot" is used - try to use already existing slot with given name +# 2. if "--slot" and "--perm-slot" used - try to create permanent slot and use it. +# 3. If "--perm-slot " flag is used without "--slot" option - use generic slot name like "pg_probackup_perm_slot" +# 4. If "--perm-slot " flag is used and permanent slot already exists - fail with error. +# 5. "--perm-slot" and "--temp-slot" flags cannot be used together. +######################################### + def test_catchup_with_replication_slot(self): + """ + """ + # preparation + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True + ) + src_pg.slow_start() + + # 1a. --slot option + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst_1a')) + try: + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--slot=nonexistentslot_1a' + ] + ) + self.assertEqual(1, 0, "Expecting Error because replication slot does not exist.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: replication slot "nonexistentslot_1a" does not exist', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + # 1b. --slot option + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst_1b')) + src_pg.safe_psql("postgres", "SELECT pg_catalog.pg_create_physical_replication_slot('existentslot_1b')") + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--slot=existentslot_1b' + ] + ) + + # 2a. --slot --perm-slot + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst_2a')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--slot=nonexistentslot_2a', + '--perm-slot' + ] + ) + + # 2b. and 4. --slot --perm-slot + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst_2b')) + src_pg.safe_psql("postgres", "SELECT pg_catalog.pg_create_physical_replication_slot('existentslot_2b')") + try: + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--slot=existentslot_2b', + '--perm-slot' + ] + ) + self.assertEqual(1, 0, "Expecting Error because replication slot already exist.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: replication slot "existentslot_2b" already exists', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + # 3. --perm-slot --slot + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst_3')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--perm-slot' + ] + ) + slot_name = src_pg.safe_psql( + "postgres", + "SELECT slot_name FROM pg_catalog.pg_replication_slots " + "WHERE slot_name NOT LIKE '%existentslot%' " + "AND slot_type = 'physical'" + ).decode('utf-8').rstrip() + self.assertEqual(slot_name, 'pg_probackup_perm_slot', 'Slot name mismatch') + + # 5. --perm-slot --temp-slot (PG>=10) + if self.get_version(src_pg) >= self.version_to_num('10.0'): + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst_5')) + try: + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--perm-slot', + '--temp-slot' + ] + ) + self.assertEqual(1, 0, "Expecting Error because conflicting options --perm-slot and --temp-slot used together\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: You cannot specify "--perm-slot" option with the "--temp-slot" option', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) + + #self.assertEqual(1, 0, 'Stop test') + +######################################### +# --exclude-path +######################################### + def test_catchup_with_exclude_path(self): + """ + various syntetic tests for --exclude-path option + """ + # preparation + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True + ) + src_pg.slow_start() + + # test 1 + os.mkdir(os.path.join(src_pg.data_dir, 'src_usefull_dir')) + with open(os.path.join(os.path.join(src_pg.data_dir, 'src_usefull_dir', 'src_garbage_file')), 'w') as f: + f.write('garbage') + f.flush() + f.close + os.mkdir(os.path.join(src_pg.data_dir, 'src_garbage_dir')) + with open(os.path.join(os.path.join(src_pg.data_dir, 'src_garbage_dir', 'src_garbage_file')), 'w') as f: + f.write('garbage') + f.flush() + f.close + + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--exclude-path={0}'.format(os.path.join(src_pg.data_dir, 'src_usefull_dir', 'src_garbage_file')), + '-x', '{0}'.format(os.path.join(src_pg.data_dir, 'src_garbage_dir')), + ] + ) + + self.assertTrue(Path(os.path.join(dst_pg.data_dir, 'src_usefull_dir')).exists()) + self.assertFalse(Path(os.path.join(dst_pg.data_dir, 'src_usefull_dir', 'src_garbage_file')).exists()) + self.assertFalse(Path(os.path.join(dst_pg.data_dir, 'src_garbage_dir')).exists()) + + self.set_replica(src_pg, dst_pg) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start(replica = True) + dst_pg.stop() + + # test 2 + os.mkdir(os.path.join(dst_pg.data_dir, 'dst_garbage_dir')) + os.mkdir(os.path.join(dst_pg.data_dir, 'dst_usefull_dir')) + with open(os.path.join(os.path.join(dst_pg.data_dir, 'dst_usefull_dir', 'dst_usefull_file')), 'w') as f: + f.write('gems') + f.flush() + f.close + + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--exclude-path=src_usefull_dir/src_garbage_file', + '--exclude-path=src_garbage_dir', + '--exclude-path={0}'.format(os.path.join(dst_pg.data_dir, 'dst_usefull_dir')), + ] + ) + + self.assertTrue(Path(os.path.join(dst_pg.data_dir, 'src_usefull_dir')).exists()) + self.assertFalse(Path(os.path.join(dst_pg.data_dir, 'src_usefull_dir', 'src_garbage_file')).exists()) + self.assertFalse(Path(os.path.join(dst_pg.data_dir, 'src_garbage_dir')).exists()) + self.assertFalse(Path(os.path.join(dst_pg.data_dir, 'dst_garbage_dir')).exists()) + self.assertTrue(Path(os.path.join(dst_pg.data_dir, 'dst_usefull_dir', 'dst_usefull_file')).exists()) + + #self.assertEqual(1, 0, 'Stop test') + src_pg.stop() + + def test_config_exclusion(self): + """ + Test that catchup can preserve dest replication config + """ + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question(answer int)") + + # preparation 2: make lagging behind replica + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + self.set_replica(src_pg, dst_pg) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg._assign_master(src_pg) + dst_pg.slow_start(replica = True) + dst_pg.stop() + + # preparation 3: make changes on master (source) + src_pg.pgbench_init(scale = 10) + pgbench = src_pg.pgbench(options=['-T', '2', '--no-vacuum']) + pgbench.wait() + + # test 1: do delta catchup with relative exclusion paths + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--exclude-path=postgresql.conf', + '--exclude-path=postgresql.auto.conf', + '--exclude-path=recovery.conf', + ] + ) + + # run&recover catchup'ed instance + # don't set destination db port and recover options + dst_pg.slow_start(replica = True) + + # check: run verification query + src_pg.safe_psql("postgres", "INSERT INTO ultimate_question VALUES(42)") + src_query_result = src_pg.table_checksum("ultimate_question") + dst_pg.catchup() # wait for replication + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + # preparation 4: make changes on master (source) + dst_pg.stop() + #src_pg.pgbench_init(scale = 10) + pgbench = src_pg.pgbench(options=['-T', '2', '--no-vacuum']) + pgbench.wait() + + # test 2: do delta catchup with absolute source exclusion paths + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--exclude-path={0}/postgresql.conf'.format(src_pg.data_dir), + '--exclude-path={0}/postgresql.auto.conf'.format(src_pg.data_dir), + '--exclude-path={0}/recovery.conf'.format(src_pg.data_dir), + ] + ) + + # run&recover catchup'ed instance + # don't set destination db port and recover options + dst_pg.slow_start(replica = True) + + # check: run verification query + src_pg.safe_psql("postgres", "INSERT INTO ultimate_question VALUES(2*42)") + src_query_result = src_pg.table_checksum("ultimate_question") + dst_pg.catchup() # wait for replication + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + # preparation 5: make changes on master (source) + dst_pg.stop() + pgbench = src_pg.pgbench(options=['-T', '2', '--no-vacuum']) + pgbench.wait() + + # test 3: do delta catchup with absolute destination exclusion paths + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', '-p', str(src_pg.port), '--stream', + '--exclude-path={0}/postgresql.conf'.format(dst_pg.data_dir), + '--exclude-path={0}/postgresql.auto.conf'.format(dst_pg.data_dir), + '--exclude-path={0}/recovery.conf'.format(dst_pg.data_dir), + ] + ) + + # run&recover catchup'ed instance + # don't set destination db port and recover options + dst_pg.slow_start(replica = True) + + # check: run verification query + src_pg.safe_psql("postgres", "INSERT INTO ultimate_question VALUES(3*42)") + src_query_result = src_pg.table_checksum("ultimate_question") + dst_pg.catchup() # wait for replication + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + # Cleanup + src_pg.stop() + dst_pg.stop() + #self.assertEqual(1, 0, 'Stop test') + +######################################### +# --dry-run +######################################### + def test_dry_run_catchup_full(self): + """ + Test dry-run option for full catchup + """ + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True + ) + src_pg.slow_start() + + # preparation 2: make clean shutdowned lagging behind replica + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + + src_pg.pgbench_init(scale = 10) + pgbench = src_pg.pgbench(options=['-T', '10', '--no-vacuum']) + pgbench.wait() + + # save the condition before dry-run + content_before = self.pgdata_content(dst_pg.data_dir) + + # do full catchup + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream', '--dry-run'] + ) + + # compare data dirs before and after catchup + self.compare_pgdata( + content_before, + self.pgdata_content(dst_pg.data_dir) + ) + + # Cleanup + src_pg.stop() + + def test_dry_run_catchup_ptrack(self): + """ + Test dry-run option for catchup in incremental ptrack mode + """ + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + ptrack_enable = True, + initdb_params = ['--data-checksums'] + ) + src_pg.slow_start() + src_pg.safe_psql("postgres", "CREATE EXTENSION ptrack") + + src_pg.pgbench_init(scale = 10) + pgbench = src_pg.pgbench(options=['-T', '10', '--no-vacuum']) + pgbench.wait() + + # preparation 2: make clean shutdowned lagging behind replica + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + self.set_replica(src_pg, dst_pg) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start(replica = True) + dst_pg.stop() + + # save the condition before dry-run + content_before = self.pgdata_content(dst_pg.data_dir) + + # do incremental catchup + self.catchup_node( + backup_mode = 'PTRACK', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream', '--dry-run'] + ) + + # compare data dirs before and after cathup + self.compare_pgdata( + content_before, + self.pgdata_content(dst_pg.data_dir) + ) + + # Cleanup + src_pg.stop() + + def test_dry_run_catchup_delta(self): + """ + Test dry-run option for catchup in incremental delta mode + """ + + # preparation 1: source + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True, + initdb_params = ['--data-checksums'], + pg_options = { 'wal_log_hints': 'on' } + ) + src_pg.slow_start() + + src_pg.pgbench_init(scale = 10) + pgbench = src_pg.pgbench(options=['-T', '10', '--no-vacuum']) + pgbench.wait() + + # preparation 2: make clean shutdowned lagging behind replica + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + self.set_replica(src_pg, dst_pg) + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start(replica = True) + dst_pg.stop() + + # save the condition before dry-run + content_before = self.pgdata_content(dst_pg.data_dir) + + # do delta catchup + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = ['-d', 'postgres', '-p', str(src_pg.port), '--stream', "--dry-run"] + ) + + # compare data dirs before and after cathup + self.compare_pgdata( + content_before, + self.pgdata_content(dst_pg.data_dir) + ) + + # Cleanup + src_pg.stop() + + def test_pgdata_is_ignored(self): + """ In catchup we still allow PGDATA to be set either from command line + or from the env var. This test that PGDATA is actually ignored and + --source-pgadta is used instead + """ + node = self.make_simple_node('node', + set_replication = True + ) + node.slow_start() + + # do full catchup + dest = self.make_empty_node('dst') + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = node.data_dir, + destination_node = dest, + options = ['-d', 'postgres', '-p', str(node.port), '--stream', '--pgdata=xxx'] + ) + + self.compare_pgdata( + self.pgdata_content(node.data_dir), + self.pgdata_content(dest.data_dir) + ) + + os.environ['PGDATA']='xxx' + + dest2 = self.make_empty_node('dst') + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = node.data_dir, + destination_node = dest2, + options = ['-d', 'postgres', '-p', str(node.port), '--stream'] + ) + + self.compare_pgdata( + self.pgdata_content(node.data_dir), + self.pgdata_content(dest2.data_dir) + ) diff --git a/tests/cfs_backup.py b/tests/cfs_backup_test.py similarity index 89% rename from tests/cfs_backup.py rename to tests/cfs_backup_test.py index 5a3665518..fb4a6c6b8 100644 --- a/tests/cfs_backup.py +++ b/tests/cfs_backup_test.py @@ -6,7 +6,6 @@ from .helpers.cfs_helpers import find_by_extensions, find_by_name, find_by_pattern, corrupt_file from .helpers.ptrack_helpers import ProbackupTest, ProbackupException -module_name = 'cfs_backup' tblspace_name = 'cfs_tblspace' @@ -14,15 +13,14 @@ class CfsBackupNoEncTest(ProbackupTest, unittest.TestCase): # --- Begin --- # @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def setUp(self): - self.fname = self.id().split('.')[3] self.backup_dir = os.path.join( - self.tmp_path, module_name, self.fname, 'backup') + self.tmp_path, self.module_name, self.fname, 'backup') self.node = self.make_simple_node( - base_dir="{0}/{1}/node".format(module_name, self.fname), + base_dir="{0}/{1}/node".format(self.module_name, self.fname), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'ptrack_enable': 'on', 'cfs_encryption': 'off', 'max_wal_senders': '2', 'shared_buffers': '200MB' @@ -35,18 +33,26 @@ def setUp(self): self.node.slow_start() + self.node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.create_tblspace_in_node(self.node, tblspace_name, cfs=True) tblspace = self.node.safe_psql( "postgres", "SELECT * FROM pg_tablespace WHERE spcname='{0}'".format( - tblspace_name) - ) - self.assertTrue( - tblspace_name in tblspace and "compression=true" in tblspace, + tblspace_name)) + + self.assertIn( + tblspace_name, str(tblspace), "ERROR: The tablespace not created " - "or it create without compressions" - ) + "or it create without compressions") + + self.assertIn( + "compression=true", str(tblspace), + "ERROR: The tablespace not created " + "or it create without compressions") self.assertTrue( find_by_name( @@ -163,12 +169,18 @@ def test_fullbackup_after_create_table(self): "ERROR: File pg_compression not found in {0}".format( os.path.join(self.backup_dir, 'node', backup_id)) ) - self.assertTrue( - find_by_extensions( - [os.path.join(self.backup_dir, 'backups', 'node', backup_id)], - ['.cfm']), - "ERROR: .cfm files not found in backup dir" - ) + + # check cfm size + cfms = find_by_extensions( + [os.path.join(self.backup_dir, 'backups', 'node', backup_id)], + ['.cfm']) + self.assertTrue(cfms, "ERROR: .cfm files not found in backup dir") + for cfm in cfms: + size = os.stat(cfm).st_size + self.assertLessEqual(size, 4096, + "ERROR: {0} is not truncated (has size {1} > 4096)".format( + cfm, size + )) # @unittest.expectedFailure # @unittest.skip("skip") @@ -405,6 +417,55 @@ def test_fullbackup_empty_tablespace_page_after_create_table(self): "ERROR: .cfm files not found in backup dir" ) + @unittest.skipUnless(ProbackupTest.enterprise, 'skip') + def test_page_doesnt_store_unchanged_cfm(self): + """ + Case: Test page backup doesn't store cfm file if table were not modified + """ + + self.node.safe_psql( + "postgres", + "CREATE TABLE {0} TABLESPACE {1} " + "AS SELECT i AS id, MD5(i::text) AS text, " + "MD5(repeat(i::text,10))::tsvector AS tsvector " + "FROM generate_series(0,256) i".format('t1', tblspace_name) + ) + + self.node.safe_psql("postgres", "checkpoint") + + backup_id_full = self.backup_node( + self.backup_dir, 'node', self.node, backup_type='full') + + self.assertTrue( + find_by_extensions( + [os.path.join(self.backup_dir, 'backups', 'node', backup_id_full)], + ['.cfm']), + "ERROR: .cfm files not found in backup dir" + ) + + backup_id = self.backup_node( + self.backup_dir, 'node', self.node, backup_type='page') + + show_backup = self.show_pb(self.backup_dir, 'node', backup_id) + self.assertEqual( + "OK", + show_backup["status"], + "ERROR: Incremental backup status is not valid. \n " + "Current backup status={0}".format(show_backup["status"]) + ) + self.assertTrue( + find_by_name( + [self.get_tblspace_path(self.node, tblspace_name)], + ['pg_compression']), + "ERROR: File pg_compression not found" + ) + self.assertFalse( + find_by_extensions( + [os.path.join(self.backup_dir, 'backups', 'node', backup_id)], + ['.cfm']), + "ERROR: .cfm files is found in backup dir" + ) + # @unittest.expectedFailure # @unittest.skip("skip") @unittest.skipUnless(ProbackupTest.enterprise, 'skip') @@ -473,7 +534,7 @@ def test_fullbackup_empty_tablespace_page_after_create_table_stream(self): ) # --- Section: Incremental from fill tablespace --- # - @unittest.expectedFailure + # @unittest.expectedFailure # @unittest.skip("skip") @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_fullbackup_after_create_table_ptrack_after_create_table(self): @@ -537,7 +598,7 @@ def test_fullbackup_after_create_table_ptrack_after_create_table(self): ) ) - @unittest.expectedFailure + # @unittest.expectedFailure # @unittest.skip("skip") @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_fullbackup_after_create_table_ptrack_after_create_table_stream(self): @@ -603,7 +664,7 @@ def test_fullbackup_after_create_table_ptrack_after_create_table_stream(self): ) ) - @unittest.expectedFailure + # @unittest.expectedFailure # @unittest.skip("skip") @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_fullbackup_after_create_table_page_after_create_table(self): @@ -686,7 +747,7 @@ def test_multiple_segments(self): 't_heap', tblspace_name) ) - full_result = self.node.safe_psql("postgres", "SELECT * FROM t_heap") + full_result = self.node.table_checksum("t_heap") try: backup_id_full = self.backup_node( @@ -708,7 +769,7 @@ def test_multiple_segments(self): 't_heap') ) - page_result = self.node.safe_psql("postgres", "SELECT * FROM t_heap") + page_result = self.node.table_checksum("t_heap") try: backup_id_page = self.backup_node( @@ -738,16 +799,18 @@ def test_multiple_segments(self): # CHECK FULL BACKUP self.node.stop() self.node.cleanup() - shutil.rmtree( - self.get_tblspace_path(self.node, tblspace_name), - ignore_errors=True) + shutil.rmtree(self.get_tblspace_path(self.node, tblspace_name)) self.restore_node( - self.backup_dir, 'node', self.node, - backup_id=backup_id_full, options=["-j", "4"]) + self.backup_dir, 'node', self.node, backup_id=backup_id_full, + options=[ + "-j", "4", + "--recovery-target=immediate", + "--recovery-target-action=promote"]) + self.node.slow_start() self.assertEqual( full_result, - self.node.safe_psql("postgres", "SELECT * FROM t_heap"), + self.node.table_checksum("t_heap"), 'Lost data after restore') # CHECK PAGE BACKUP @@ -757,12 +820,16 @@ def test_multiple_segments(self): self.get_tblspace_path(self.node, tblspace_name), ignore_errors=True) self.restore_node( - self.backup_dir, 'node', self.node, - backup_id=backup_id_page, options=["-j", "4"]) + self.backup_dir, 'node', self.node, backup_id=backup_id_page, + options=[ + "-j", "4", + "--recovery-target=immediate", + "--recovery-target-action=promote"]) + self.node.slow_start() self.assertEqual( page_result, - self.node.safe_psql("postgres", "SELECT * FROM t_heap"), + self.node.table_checksum("t_heap"), 'Lost data after restore') # @unittest.expectedFailure @@ -786,8 +853,7 @@ def test_multiple_segments_in_multiple_tablespaces(self): "AS SELECT i AS id, MD5(i::text) AS text, " "MD5(repeat(i::text,10))::tsvector AS tsvector " "FROM generate_series(0,1005000) i".format( - 't_heap_1', tblspace_name_1) - ) + 't_heap_1', tblspace_name_1)) self.node.safe_psql( "postgres", @@ -795,13 +861,10 @@ def test_multiple_segments_in_multiple_tablespaces(self): "AS SELECT i AS id, MD5(i::text) AS text, " "MD5(repeat(i::text,10))::tsvector AS tsvector " "FROM generate_series(0,1005000) i".format( - 't_heap_2', tblspace_name_2) - ) + 't_heap_2', tblspace_name_2)) - full_result_1 = self.node.safe_psql( - "postgres", "SELECT * FROM t_heap_1") - full_result_2 = self.node.safe_psql( - "postgres", "SELECT * FROM t_heap_2") + full_result_1 = self.node.table_checksum("t_heap_1") + full_result_2 = self.node.table_checksum("t_heap_2") try: backup_id_full = self.backup_node( @@ -832,10 +895,8 @@ def test_multiple_segments_in_multiple_tablespaces(self): 't_heap_2') ) - page_result_1 = self.node.safe_psql( - "postgres", "SELECT * FROM t_heap_1") - page_result_2 = self.node.safe_psql( - "postgres", "SELECT * FROM t_heap_2") + page_result_1 = self.node.table_checksum("t_heap_1") + page_result_2 = self.node.table_checksum("t_heap_2") try: backup_id_page = self.backup_node( @@ -864,57 +925,47 @@ def test_multiple_segments_in_multiple_tablespaces(self): # CHECK FULL BACKUP self.node.stop() - self.node.cleanup() - shutil.rmtree( - self.get_tblspace_path(self.node, tblspace_name), - ignore_errors=True) - shutil.rmtree( - self.get_tblspace_path(self.node, tblspace_name_1), - ignore_errors=True) - shutil.rmtree( - self.get_tblspace_path(self.node, tblspace_name_2), - ignore_errors=True) self.restore_node( self.backup_dir, 'node', self.node, - backup_id=backup_id_full, options=["-j", "4"]) + backup_id=backup_id_full, + options=[ + "-j", "4", "--incremental-mode=checksum", + "--recovery-target=immediate", + "--recovery-target-action=promote"]) self.node.slow_start() + self.assertEqual( full_result_1, - self.node.safe_psql("postgres", "SELECT * FROM t_heap_1"), + self.node.table_checksum("t_heap_1"), 'Lost data after restore') self.assertEqual( full_result_2, - self.node.safe_psql("postgres", "SELECT * FROM t_heap_2"), + self.node.table_checksum("t_heap_2"), 'Lost data after restore') # CHECK PAGE BACKUP self.node.stop() - self.node.cleanup() - shutil.rmtree( - self.get_tblspace_path(self.node, tblspace_name), - ignore_errors=True) - shutil.rmtree( - self.get_tblspace_path(self.node, tblspace_name_1), - ignore_errors=True) - shutil.rmtree( - self.get_tblspace_path(self.node, tblspace_name_2), - ignore_errors=True) self.restore_node( self.backup_dir, 'node', self.node, - backup_id=backup_id_page, options=["-j", "4"]) + backup_id=backup_id_page, + options=[ + "-j", "4", "--incremental-mode=checksum", + "--recovery-target=immediate", + "--recovery-target-action=promote"]) self.node.slow_start() + self.assertEqual( page_result_1, - self.node.safe_psql("postgres", "SELECT * FROM t_heap_1"), + self.node.table_checksum("t_heap_1"), 'Lost data after restore') self.assertEqual( page_result_2, - self.node.safe_psql("postgres", "SELECT * FROM t_heap_2"), + self.node.table_checksum("t_heap_2"), 'Lost data after restore') - @unittest.expectedFailure + # @unittest.expectedFailure # @unittest.skip("skip") @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_fullbackup_after_create_table_page_after_create_table_stream(self): @@ -981,7 +1032,6 @@ def test_fullbackup_after_create_table_page_after_create_table_stream(self): ) # --- Make backup with not valid data(broken .cfm) --- # - @unittest.expectedFailure # @unittest.skip("skip") @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_delete_random_cfm_file_from_tablespace_dir(self): @@ -993,6 +1043,11 @@ def test_delete_random_cfm_file_from_tablespace_dir(self): "FROM generate_series(0,256) i".format('t1', tblspace_name) ) + self.node.safe_psql( + "postgres", + "CHECKPOINT" + ) + list_cmf = find_by_extensions( [self.get_tblspace_path(self.node, tblspace_name)], ['.cfm']) @@ -1042,6 +1097,11 @@ def test_delete_random_data_file_from_tablespace_dir(self): "FROM generate_series(0,256) i".format('t1', tblspace_name) ) + self.node.safe_psql( + "postgres", + "CHECKPOINT" + ) + list_data_files = find_by_pattern( [self.get_tblspace_path(self.node, tblspace_name)], '^.*/\d+$') @@ -1147,10 +1207,6 @@ def test_broken_file_pg_compression_into_tablespace_dir(self): ) # # --- End ---# -# @unittest.skipUnless(ProbackupTest.enterprise, 'skip') -# def tearDown(self): -# self.node.cleanup() -# self.del_test_dir(module_name, self.fname) #class CfsBackupEncTest(CfsBackupNoEncTest): diff --git a/tests/cfs_catchup_test.py b/tests/cfs_catchup_test.py new file mode 100644 index 000000000..f6760b72c --- /dev/null +++ b/tests/cfs_catchup_test.py @@ -0,0 +1,117 @@ +import os +import unittest +import random +import shutil + +from .helpers.cfs_helpers import find_by_extensions, find_by_name, find_by_pattern, corrupt_file +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException + + +class CfsCatchupNoEncTest(ProbackupTest, unittest.TestCase): + + @unittest.skipUnless(ProbackupTest.enterprise, 'skip') + def test_full_catchup_with_tablespace(self): + """ + Test tablespace transfers + """ + # preparation + src_pg = self.make_simple_node( + base_dir = os.path.join(self.module_name, self.fname, 'src'), + set_replication = True + ) + src_pg.slow_start() + tblspace1_old_path = self.get_tblspace_path(src_pg, 'tblspace1_old') + self.create_tblspace_in_node(src_pg, 'tblspace1', tblspc_path = tblspace1_old_path, cfs=True) + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question TABLESPACE tblspace1 AS SELECT 42 AS answer") + src_query_result = src_pg.table_checksum("ultimate_question") + src_pg.safe_psql( + "postgres", + "CHECKPOINT") + + # do full catchup with tablespace mapping + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + tblspace1_new_path = self.get_tblspace_path(dst_pg, 'tblspace1_new') + self.catchup_node( + backup_mode = 'FULL', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', + '-p', str(src_pg.port), + '--stream', + '-T', '{0}={1}'.format(tblspace1_old_path, tblspace1_new_path) + ] + ) + + # 1st check: compare data directories + self.compare_pgdata( + self.pgdata_content(src_pg.data_dir), + self.pgdata_content(dst_pg.data_dir) + ) + + # check cfm size + cfms = find_by_extensions([os.path.join(dst_pg.data_dir)], ['.cfm']) + self.assertTrue(cfms, "ERROR: .cfm files not found in backup dir") + for cfm in cfms: + size = os.stat(cfm).st_size + self.assertLessEqual(size, 4096, + "ERROR: {0} is not truncated (has size {1} > 4096)".format( + cfm, size + )) + + # make changes in master tablespace + src_pg.safe_psql( + "postgres", + "UPDATE ultimate_question SET answer = -1") + src_pg.safe_psql( + "postgres", + "CHECKPOINT") + + # run&recover catchup'ed instance + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + + # 2nd check: run verification query + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') + + # and now delta backup + dst_pg.stop() + + self.catchup_node( + backup_mode = 'DELTA', + source_pgdata = src_pg.data_dir, + destination_node = dst_pg, + options = [ + '-d', 'postgres', + '-p', str(src_pg.port), + '--stream', + '-T', '{0}={1}'.format(tblspace1_old_path, tblspace1_new_path) + ] + ) + + # check cfm size again + cfms = find_by_extensions([os.path.join(dst_pg.data_dir)], ['.cfm']) + self.assertTrue(cfms, "ERROR: .cfm files not found in backup dir") + for cfm in cfms: + size = os.stat(cfm).st_size + self.assertLessEqual(size, 4096, + "ERROR: {0} is not truncated (has size {1} > 4096)".format( + cfm, size + )) + + # run&recover catchup'ed instance + dst_options = {} + dst_options['port'] = str(dst_pg.port) + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + + + # 3rd check: run verification query + src_query_result = src_pg.table_checksum("ultimate_question") + dst_query_result = dst_pg.table_checksum("ultimate_question") + self.assertEqual(src_query_result, dst_query_result, 'Different answer from copy') diff --git a/tests/cfs_restore.py b/tests/cfs_restore_test.py similarity index 90% rename from tests/cfs_restore.py rename to tests/cfs_restore_test.py index f59c092af..2fa35e71a 100644 --- a/tests/cfs_restore.py +++ b/tests/cfs_restore_test.py @@ -15,20 +15,17 @@ from .helpers.cfs_helpers import find_by_name from .helpers.ptrack_helpers import ProbackupTest, ProbackupException - -module_name = 'cfs_restore' - tblspace_name = 'cfs_tblspace' tblspace_name_new = 'cfs_tblspace_new' class CfsRestoreBase(ProbackupTest, unittest.TestCase): + @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def setUp(self): - self.fname = self.id().split('.')[3] - self.backup_dir = os.path.join(self.tmp_path, module_name, self.fname, 'backup') + self.backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.node = self.make_simple_node( - base_dir="{0}/{1}/node".format(module_name, self.fname), + base_dir="{0}/{1}/node".format(self.module_name, self.fname), set_replication=True, initdb_params=['--data-checksums'], pg_options={ @@ -60,14 +57,11 @@ def setUp(self): def add_data_in_cluster(self): pass - def tearDown(self): - self.node.cleanup() - self.del_test_dir(module_name, self.fname) - class CfsRestoreNoencEmptyTablespaceTest(CfsRestoreBase): # @unittest.expectedFailure # @unittest.skip("skip") + @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_restore_empty_tablespace_from_fullbackup(self): """ Case: Restore empty tablespace from valid full backup. @@ -102,7 +96,7 @@ def test_restore_empty_tablespace_from_fullbackup(self): tblspace = self.node.safe_psql( "postgres", "SELECT * FROM pg_tablespace WHERE spcname='{0}'".format(tblspace_name) - ) + ).decode("UTF-8") self.assertTrue( tblspace_name in tblspace and "compression=true" in tblspace, "ERROR: The tablespace not restored or it restored without compressions" @@ -118,14 +112,12 @@ def add_data_in_cluster(self): MD5(repeat(i::text,10))::tsvector AS tsvector \ FROM generate_series(0,1e5) i'.format('t1', tblspace_name) ) - self.table_t1 = self.node.safe_psql( - "postgres", - "SELECT * FROM t1" - ) + self.table_t1 = self.node.table_checksum("t1") # --- Restore from full backup ---# # @unittest.expectedFailure # @unittest.skip("skip") + @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_restore_from_fullbackup_to_old_location(self): """ Case: Restore instance from valid full backup to old location. @@ -159,12 +151,13 @@ def test_restore_from_fullbackup_to_old_location(self): ) self.assertEqual( - repr(self.node.safe_psql("postgres", "SELECT * FROM %s" % 't1')), - repr(self.table_t1) + self.node.table_checksum("t1"), + self.table_t1 ) # @unittest.expectedFailure # @unittest.skip("skip") + @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_restore_from_fullbackup_to_old_location_3_jobs(self): """ Case: Restore instance from valid full backup to old location. @@ -197,12 +190,13 @@ def test_restore_from_fullbackup_to_old_location_3_jobs(self): ) self.assertEqual( - repr(self.node.safe_psql("postgres", "SELECT * FROM %s" % 't1')), - repr(self.table_t1) + self.node.table_checksum("t1"), + self.table_t1 ) # @unittest.expectedFailure # @unittest.skip("skip") + @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_restore_from_fullbackup_to_new_location(self): """ Case: Restore instance from valid full backup to new location. @@ -211,13 +205,12 @@ def test_restore_from_fullbackup_to_new_location(self): self.node.cleanup() shutil.rmtree(self.get_tblspace_path(self.node, tblspace_name)) - node_new = self.make_simple_node(base_dir="{0}/{1}/node_new_location".format(module_name, self.fname)) + node_new = self.make_simple_node(base_dir="{0}/{1}/node_new_location".format(self.module_name, self.fname)) node_new.cleanup() try: self.restore_node(self.backup_dir, 'node', node_new, backup_id=self.backup_id) - node_new.append_conf("postgresql.auto.conf", - "port = {0}".format(node_new.port)) + self.set_auto_conf(node_new, {'port': node_new.port}) except ProbackupException as e: self.fail( "ERROR: Restore from full backup failed. \n {0} \n {1}".format( @@ -240,13 +233,14 @@ def test_restore_from_fullbackup_to_new_location(self): ) self.assertEqual( - repr(node_new.safe_psql("postgres", "SELECT * FROM %s" % 't1')), - repr(self.table_t1) + node_new.table_checksum("t1"), + self.table_t1 ) node_new.cleanup() # @unittest.expectedFailure # @unittest.skip("skip") + @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_restore_from_fullbackup_to_new_location_5_jobs(self): """ Case: Restore instance from valid full backup to new location. @@ -255,13 +249,12 @@ def test_restore_from_fullbackup_to_new_location_5_jobs(self): self.node.cleanup() shutil.rmtree(self.get_tblspace_path(self.node, tblspace_name)) - node_new = self.make_simple_node(base_dir="{0}/{1}/node_new_location".format(module_name, self.fname)) + node_new = self.make_simple_node(base_dir="{0}/{1}/node_new_location".format(self.module_name, self.fname)) node_new.cleanup() try: self.restore_node(self.backup_dir, 'node', node_new, backup_id=self.backup_id, options=['-j', '5']) - node_new.append_conf("postgresql.auto.conf", - "port = {0}".format(node_new.port)) + self.set_auto_conf(node_new, {'port': node_new.port}) except ProbackupException as e: self.fail( "ERROR: Restore from full backup failed. \n {0} \n {1}".format( @@ -284,13 +277,14 @@ def test_restore_from_fullbackup_to_new_location_5_jobs(self): ) self.assertEqual( - repr(node_new.safe_psql("postgres", "SELECT * FROM %s" % 't1')), - repr(self.table_t1) + node_new.table_checksum("t1"), + self.table_t1 ) node_new.cleanup() # @unittest.expectedFailure # @unittest.skip("skip") + @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_restore_from_fullbackup_to_old_location_tablespace_new_location(self): self.node.stop() self.node.cleanup() @@ -331,12 +325,13 @@ def test_restore_from_fullbackup_to_old_location_tablespace_new_location(self): ) self.assertEqual( - repr(self.node.safe_psql("postgres", "SELECT * FROM %s" % 't1')), - repr(self.table_t1) + self.node.table_checksum("t1"), + self.table_t1 ) # @unittest.expectedFailure # @unittest.skip("skip") + @unittest.skipUnless(ProbackupTest.enterprise, 'skip') def test_restore_from_fullbackup_to_old_location_tablespace_new_location_3_jobs(self): self.node.stop() self.node.cleanup() @@ -377,8 +372,8 @@ def test_restore_from_fullbackup_to_old_location_tablespace_new_location_3_jobs( ) self.assertEqual( - repr(self.node.safe_psql("postgres", "SELECT * FROM %s" % 't1')), - repr(self.table_t1) + self.node.table_checksum("t1"), + self.table_t1 ) # @unittest.expectedFailure diff --git a/tests/cfs_validate_backup.py b/tests/cfs_validate_backup_test.py similarity index 94% rename from tests/cfs_validate_backup.py rename to tests/cfs_validate_backup_test.py index eea6f0e21..343020dfc 100644 --- a/tests/cfs_validate_backup.py +++ b/tests/cfs_validate_backup_test.py @@ -5,7 +5,6 @@ from .helpers.cfs_helpers import find_by_extensions, find_by_name, find_by_pattern, corrupt_file from .helpers.ptrack_helpers import ProbackupTest, ProbackupException -module_name = 'cfs_validate_backup' tblspace_name = 'cfs_tblspace' diff --git a/tests/checkdb.py b/tests/checkdb.py deleted file mode 100644 index 64a733ac7..000000000 --- a/tests/checkdb.py +++ /dev/null @@ -1,497 +0,0 @@ -import os -import unittest -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException -from datetime import datetime, timedelta -import subprocess -from testgres import QueryException -import shutil -import sys -import time - - -module_name = 'checkdb' - - -class CheckdbTest(ProbackupTest, unittest.TestCase): - - # @unittest.skip("skip") - def test_checkdb_amcheck_only_sanity(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir="{0}/{1}/node".format(module_name, fname), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "create table t_heap as select i" - " as id from generate_series(0,100) i") - - node.safe_psql( - "postgres", - "create index on t_heap(id)") - - try: - node.safe_psql( - "postgres", - "create extension amcheck") - except QueryException as e: - node.safe_psql( - "postgres", - "create extension amcheck_next") - - log_file_path = os.path.join( - backup_dir, 'log', 'pg_probackup.log') - - # simple sanity - try: - self.checkdb_node( - options=['--skip-block-validation']) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because --amcheck option is missing\n" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "ERROR: Option '--skip-block-validation' must be " - "used with '--amcheck' option", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - # simple sanity - output = self.checkdb_node( - options=[ - '--amcheck', - '--skip-block-validation', - '-d', 'postgres', '-p', str(node.port)]) - - self.assertIn( - 'INFO: checkdb --amcheck finished successfully', - output) - self.assertIn( - 'All checked indexes are valid', - output) - - # logging to file sanity - try: - self.checkdb_node( - options=[ - '--amcheck', - '--skip-block-validation', - '--log-level-file=verbose', - '-d', 'postgres', '-p', str(node.port)]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because log_directory missing\n" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "ERROR: Cannot save checkdb logs to a file. " - "You must specify --log-directory option when " - "running checkdb with --log-level-file option enabled", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - # If backup_dir provided, then instance name must be - # provided too - try: - self.checkdb_node( - backup_dir, - options=[ - '--amcheck', - '--skip-block-validation', - '--log-level-file=verbose', - '-d', 'postgres', '-p', str(node.port)]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because log_directory missing\n" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "ERROR: required parameter not specified: --instance", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - # checkdb can use default or set in config values, - # if backup_dir and instance name are provided - self.checkdb_node( - backup_dir, - 'node', - options=[ - '--amcheck', - '--skip-block-validation', - '--log-level-file=verbose', - '-d', 'postgres', '-p', str(node.port)]) - - # check that file present and full of messages - os.path.isfile(log_file_path) - with open(log_file_path) as f: - log_file_content = f.read() - self.assertIn( - 'INFO: checkdb --amcheck finished successfully', - log_file_content) - self.assertIn( - 'VERBOSE: (query)', - log_file_content) - os.unlink(log_file_path) - - # log-level-file and log-directory are provided - self.checkdb_node( - backup_dir, - 'node', - options=[ - '--amcheck', - '--skip-block-validation', - '--log-level-file=verbose', - '--log-directory={0}'.format( - os.path.join(backup_dir, 'log')), - '-d', 'postgres', '-p', str(node.port)]) - - # check that file present and full of messages - os.path.isfile(log_file_path) - with open(log_file_path) as f: - log_file_content = f.read() - self.assertIn( - 'INFO: checkdb --amcheck finished successfully', - log_file_content) - self.assertIn( - 'VERBOSE: (query)', - log_file_content) - os.unlink(log_file_path) - - gdb = self.checkdb_node( - gdb=True, - options=[ - '--amcheck', - '--skip-block-validation', - '--log-level-file=verbose', - '--log-directory={0}'.format( - os.path.join(backup_dir, 'log')), - '-d', 'postgres', '-p', str(node.port)]) - - gdb.set_breakpoint('amcheck_one_index') - gdb.run_until_break() - - node.safe_psql( - "postgres", - "drop table t_heap") - - gdb.remove_all_breakpoints() - - gdb.continue_execution_until_exit() - - # check that message about missing index is present - with open(log_file_path) as f: - log_file_content = f.read() - self.assertIn( - 'ERROR: checkdb --amcheck finished with failure', - log_file_content) - self.assertIn( - "WARNING: Thread [1]. Amcheck failed in database 'postgres' " - "for index: 'public.t_heap_id_idx':", - log_file_content) - self.assertIn( - 'ERROR: could not open relation with OID', - log_file_content) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_basic_checkdb_amcheck_only_sanity(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir="{0}/{1}/node".format(module_name, fname), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - # create two databases - node.safe_psql("postgres", "create database db1") - try: - node.safe_psql( - "db1", - "create extension amcheck") - except QueryException as e: - node.safe_psql( - "db1", - "create extension amcheck_next") - - node.safe_psql("postgres", "create database db2") - try: - node.safe_psql( - "db2", - "create extension amcheck") - except QueryException as e: - node.safe_psql( - "db2", - "create extension amcheck_next") - - # init pgbench in two databases and corrupt both indexes - node.pgbench_init(scale=5, dbname='db1') - node.pgbench_init(scale=5, dbname='db2') - - node.safe_psql( - "db2", - "alter index pgbench_accounts_pkey rename to some_index") - - index_path_1 = os.path.join( - node.data_dir, - node.safe_psql( - "db1", - "select pg_relation_filepath('pgbench_accounts_pkey')").rstrip()) - - index_path_2 = os.path.join( - node.data_dir, - node.safe_psql( - "db2", - "select pg_relation_filepath('some_index')").rstrip()) - - try: - self.checkdb_node( - options=[ - '--amcheck', - '--skip-block-validation', - '-d', 'postgres', '-p', str(node.port)]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because some db was not amchecked" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "ERROR: Some databases were not amchecked", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - node.stop() - - # Let`s do index corruption - with open(index_path_1, "rb+", 0) as f: - f.seek(42000) - f.write(b"blablahblahs") - f.flush() - f.close - - with open(index_path_2, "rb+", 0) as f: - f.seek(42000) - f.write(b"blablahblahs") - f.flush() - f.close - - node.slow_start() - - log_file_path = os.path.join( - backup_dir, 'log', 'pg_probackup.log') - - try: - self.checkdb_node( - options=[ - '--amcheck', - '--skip-block-validation', - '--log-level-file=verbose', - '--log-directory={0}'.format( - os.path.join(backup_dir, 'log')), - '-d', 'postgres', '-p', str(node.port)]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because some db was not amchecked" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "ERROR: checkdb --amcheck finished with failure", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - # corruption of both indexes in db1 and db2 must be detected - # also the that amcheck is not installed in 'postgres' - # should be logged - with open(log_file_path) as f: - log_file_content = f.read() - self.assertIn( - "WARNING: Thread [1]. Amcheck failed in database 'db1' " - "for index: 'public.pgbench_accounts_pkey':", - log_file_content) - - self.assertIn( - "WARNING: Thread [1]. Amcheck failed in database 'db2' " - "for index: 'public.some_index':", - log_file_content) - - self.assertIn( - "ERROR: checkdb --amcheck finished with failure", - log_file_content) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_checkdb_block_validation_sanity(self): - """make node, corrupt some pages, check that checkdb failed""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "create table t_heap as select 1 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i") - node.safe_psql( - "postgres", - "CHECKPOINT;") - - heap_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('t_heap')").rstrip() - - # sanity - try: - self.checkdb_node() - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because pgdata must be specified\n" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "ERROR: required parameter not specified: PGDATA (-D, --pgdata)", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - self.checkdb_node( - data_dir=node.data_dir, - options=['-d', 'postgres', '-p', str(node.port)]) - - self.checkdb_node( - backup_dir, 'node', - options=['-d', 'postgres', '-p', str(node.port)]) - - heap_full_path = os.path.join(node.data_dir, heap_path) - - with open(heap_full_path, "rb+", 0) as f: - f.seek(9000) - f.write(b"bla") - f.flush() - f.close - - with open(heap_full_path, "rb+", 0) as f: - f.seek(42000) - f.write(b"bla") - f.flush() - f.close - - try: - self.checkdb_node( - backup_dir, 'node', - options=['-d', 'postgres', '-p', str(node.port)]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because of data corruption\n" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "ERROR: Checkdb failed", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - self.assertIn( - 'WARNING: Corruption detected in file "{0}", block 1'.format( - os.path.normpath(heap_full_path)), - e.message) - - self.assertIn( - 'WARNING: Corruption detected in file "{0}", block 5'.format( - os.path.normpath(heap_full_path)), - e.message) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_checkdb_sigint_handling(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - try: - node.safe_psql( - "postgres", - "create extension amcheck") - except QueryException as e: - node.safe_psql( - "postgres", - "create extension amcheck_next") - - # FULL backup - gdb = self.checkdb_node( - backup_dir, 'node', gdb=True, - options=[ - '-d', 'postgres', '-j', '4', - '--skip-block-validation', - '--amcheck', '-p', str(node.port)]) - - gdb.set_breakpoint('amcheck_one_index') - gdb.run_until_break() - - gdb.continue_execution_until_break(10) - gdb.remove_all_breakpoints() - - gdb._execute('signal SIGINT') - gdb.continue_execution_until_error() - - with open(node.pg_log_file, 'r') as f: - output = f.read() - - self.assertNotIn('could not receive data from client', output) - self.assertNotIn('could not send data to client', output) - self.assertNotIn('connection to client lost', output) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/checkdb_test.py b/tests/checkdb_test.py new file mode 100644 index 000000000..eb46aea19 --- /dev/null +++ b/tests/checkdb_test.py @@ -0,0 +1,851 @@ +import os +import unittest +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +from datetime import datetime, timedelta +import subprocess +from testgres import QueryException +import shutil +import sys +import time + + +class CheckdbTest(ProbackupTest, unittest.TestCase): + + # @unittest.skip("skip") + def test_checkdb_amcheck_only_sanity(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir="{0}/{1}/node".format(self.module_name, self.fname), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "create table t_heap as select i" + " as id from generate_series(0,100) i") + + node.safe_psql( + "postgres", + "create index on t_heap(id)") + + node.safe_psql( + "postgres", + "create table idxpart (a int) " + "partition by range (a)") + + # there aren't partitioned indexes on 10 and lesser versions + if self.get_version(node) >= 110000: + node.safe_psql( + "postgres", + "create index on idxpart(a)") + + try: + node.safe_psql( + "postgres", + "create extension amcheck") + except QueryException as e: + node.safe_psql( + "postgres", + "create extension amcheck_next") + + log_file_path = os.path.join( + backup_dir, 'log', 'pg_probackup.log') + + # simple sanity + try: + self.checkdb_node( + options=['--skip-block-validation']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because --amcheck option is missing\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Option '--skip-block-validation' must be " + "used with '--amcheck' option", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # simple sanity + output = self.checkdb_node( + options=[ + '--amcheck', + '--skip-block-validation', + '-d', 'postgres', '-p', str(node.port)]) + + self.assertIn( + 'INFO: checkdb --amcheck finished successfully', + output) + self.assertIn( + 'All checked indexes are valid', + output) + + # logging to file sanity + try: + self.checkdb_node( + options=[ + '--amcheck', + '--skip-block-validation', + '--log-level-file=verbose', + '-d', 'postgres', '-p', str(node.port)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because log_directory missing\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Cannot save checkdb logs to a file. " + "You must specify --log-directory option when " + "running checkdb with --log-level-file option enabled", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # If backup_dir provided, then instance name must be + # provided too + try: + self.checkdb_node( + backup_dir, + options=[ + '--amcheck', + '--skip-block-validation', + '--log-level-file=verbose', + '-d', 'postgres', '-p', str(node.port)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because log_directory missing\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Required parameter not specified: --instance", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # checkdb can use default or set in config values, + # if backup_dir and instance name are provided + self.checkdb_node( + backup_dir, + 'node', + options=[ + '--amcheck', + '--skip-block-validation', + '--log-level-file=verbose', + '-d', 'postgres', '-p', str(node.port)]) + + # check that file present and full of messages + os.path.isfile(log_file_path) + with open(log_file_path) as f: + log_file_content = f.read() + self.assertIn( + 'INFO: checkdb --amcheck finished successfully', + log_file_content) + self.assertIn( + 'VERBOSE: (query)', + log_file_content) + os.unlink(log_file_path) + + # log-level-file and log-directory are provided + self.checkdb_node( + backup_dir, + 'node', + options=[ + '--amcheck', + '--skip-block-validation', + '--log-level-file=verbose', + '--log-directory={0}'.format( + os.path.join(backup_dir, 'log')), + '-d', 'postgres', '-p', str(node.port)]) + + # check that file present and full of messages + os.path.isfile(log_file_path) + with open(log_file_path) as f: + log_file_content = f.read() + self.assertIn( + 'INFO: checkdb --amcheck finished successfully', + log_file_content) + self.assertIn( + 'VERBOSE: (query)', + log_file_content) + os.unlink(log_file_path) + + gdb = self.checkdb_node( + gdb=True, + options=[ + '--amcheck', + '--skip-block-validation', + '--log-level-file=verbose', + '--log-directory={0}'.format( + os.path.join(backup_dir, 'log')), + '-d', 'postgres', '-p', str(node.port)]) + + gdb.set_breakpoint('amcheck_one_index') + gdb.run_until_break() + + node.safe_psql( + "postgres", + "drop table t_heap") + + gdb.remove_all_breakpoints() + + gdb.continue_execution_until_exit() + + # check that message about missing index is present + with open(log_file_path) as f: + log_file_content = f.read() + self.assertIn( + 'ERROR: checkdb --amcheck finished with failure', + log_file_content) + self.assertIn( + "WARNING: Thread [1]. Amcheck failed in database 'postgres' " + "for index: 'public.t_heap_id_idx':", + log_file_content) + self.assertIn( + 'ERROR: could not open relation with OID', + log_file_content) + + # Clean after yourself + gdb.kill() + node.stop() + + # @unittest.skip("skip") + def test_basic_checkdb_amcheck_only_sanity(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir="{0}/{1}/node".format(self.module_name, self.fname), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # create two databases + node.safe_psql("postgres", "create database db1") + try: + node.safe_psql( + "db1", + "create extension amcheck") + except QueryException as e: + node.safe_psql( + "db1", + "create extension amcheck_next") + + node.safe_psql("postgres", "create database db2") + try: + node.safe_psql( + "db2", + "create extension amcheck") + except QueryException as e: + node.safe_psql( + "db2", + "create extension amcheck_next") + + # init pgbench in two databases and corrupt both indexes + node.pgbench_init(scale=5, dbname='db1') + node.pgbench_init(scale=5, dbname='db2') + + node.safe_psql( + "db2", + "alter index pgbench_accounts_pkey rename to some_index") + + index_path_1 = os.path.join( + node.data_dir, + node.safe_psql( + "db1", + "select pg_relation_filepath('pgbench_accounts_pkey')").decode('utf-8').rstrip()) + + index_path_2 = os.path.join( + node.data_dir, + node.safe_psql( + "db2", + "select pg_relation_filepath('some_index')").decode('utf-8').rstrip()) + + try: + self.checkdb_node( + options=[ + '--amcheck', + '--skip-block-validation', + '-d', 'postgres', '-p', str(node.port)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because some db was not amchecked" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Some databases were not amchecked", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + node.stop() + + # Let`s do index corruption + with open(index_path_1, "rb+", 0) as f: + f.seek(42000) + f.write(b"blablahblahs") + f.flush() + f.close + + with open(index_path_2, "rb+", 0) as f: + f.seek(42000) + f.write(b"blablahblahs") + f.flush() + f.close + + node.slow_start() + + log_file_path = os.path.join( + backup_dir, 'log', 'pg_probackup.log') + + try: + self.checkdb_node( + options=[ + '--amcheck', + '--skip-block-validation', + '--log-level-file=verbose', + '--log-directory={0}'.format( + os.path.join(backup_dir, 'log')), + '-d', 'postgres', '-p', str(node.port)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because some db was not amchecked" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: checkdb --amcheck finished with failure", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # corruption of both indexes in db1 and db2 must be detected + # also the that amcheck is not installed in 'postgres' + # should be logged + with open(log_file_path) as f: + log_file_content = f.read() + self.assertIn( + "WARNING: Thread [1]. Amcheck failed in database 'db1' " + "for index: 'public.pgbench_accounts_pkey':", + log_file_content) + + self.assertIn( + "WARNING: Thread [1]. Amcheck failed in database 'db2' " + "for index: 'public.some_index':", + log_file_content) + + self.assertIn( + "ERROR: checkdb --amcheck finished with failure", + log_file_content) + + # Clean after yourself + node.stop() + + # @unittest.skip("skip") + def test_checkdb_block_validation_sanity(self): + """make node, corrupt some pages, check that checkdb failed""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1000) i") + node.safe_psql( + "postgres", + "CHECKPOINT;") + + heap_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + # sanity + try: + self.checkdb_node() + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because pgdata must be specified\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Required parameter not specified: PGDATA (-D, --pgdata)", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.checkdb_node( + data_dir=node.data_dir, + options=['-d', 'postgres', '-p', str(node.port)]) + + self.checkdb_node( + backup_dir, 'node', + options=['-d', 'postgres', '-p', str(node.port)]) + + heap_full_path = os.path.join(node.data_dir, heap_path) + + with open(heap_full_path, "rb+", 0) as f: + f.seek(9000) + f.write(b"bla") + f.flush() + f.close + + with open(heap_full_path, "rb+", 0) as f: + f.seek(42000) + f.write(b"bla") + f.flush() + f.close + + try: + self.checkdb_node( + backup_dir, 'node', + options=['-d', 'postgres', '-p', str(node.port)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of data corruption\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Checkdb failed", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertIn( + 'WARNING: Corruption detected in file "{0}", block 1'.format( + os.path.normpath(heap_full_path)), + e.message) + + self.assertIn( + 'WARNING: Corruption detected in file "{0}", block 5'.format( + os.path.normpath(heap_full_path)), + e.message) + + # Clean after yourself + node.stop() + + def test_checkdb_checkunique(self): + """Test checkunique parameter of amcheck.bt_index_check function""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + node.slow_start() + + try: + node.safe_psql( + "postgres", + "create extension amcheck") + except QueryException as e: + node.safe_psql( + "postgres", + "create extension amcheck_next") + + # Part of https://commitfest.postgresql.org/32/2976/ patch test + node.safe_psql( + "postgres", + "CREATE TABLE bttest_unique(a varchar(50), b varchar(1500), c bytea, d varchar(50)); " + "ALTER TABLE bttest_unique SET (autovacuum_enabled = false); " + "CREATE UNIQUE INDEX bttest_unique_idx ON bttest_unique(a,b); " + "UPDATE pg_catalog.pg_index SET indisunique = false " + "WHERE indrelid = (SELECT oid FROM pg_catalog.pg_class WHERE relname = 'bttest_unique'); " + "INSERT INTO bttest_unique " + " SELECT i::text::varchar, " + " array_to_string(array( " + " SELECT substr('ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789', ((random()*(36-1)+1)::integer), 1) " + " FROM generate_series(1,1300)),'')::varchar, " + " i::text::bytea, i::text::varchar " + " FROM generate_series(0,1) AS i, generate_series(0,30) AS x; " + "UPDATE pg_catalog.pg_index SET indisunique = true " + "WHERE indrelid = (SELECT oid FROM pg_catalog.pg_class WHERE relname = 'bttest_unique'); " + "DELETE FROM bttest_unique WHERE ctid::text='(0,2)'; " + "DELETE FROM bttest_unique WHERE ctid::text='(4,2)'; " + "DELETE FROM bttest_unique WHERE ctid::text='(4,3)'; " + "DELETE FROM bttest_unique WHERE ctid::text='(9,3)';") + + # run without checkunique option (error will not detected) + output = self.checkdb_node( + options=[ + '--amcheck', + '--skip-block-validation', + '-d', 'postgres', '-p', str(node.port)]) + + self.assertIn( + 'INFO: checkdb --amcheck finished successfully', + output) + self.assertIn( + 'All checked indexes are valid', + output) + + # run with checkunique option + try: + self.checkdb_node( + options=[ + '--amcheck', + '--skip-block-validation', + '--checkunique', + '-d', 'postgres', '-p', str(node.port)]) + if (ProbackupTest.enterprise and + (self.get_version(node) >= 111300 and self.get_version(node) < 120000 + or self.get_version(node) >= 120800 and self.get_version(node) < 130000 + or self.get_version(node) >= 130400)): + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of index corruption\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + else: + self.assertRegex( + self.output, + r"WARNING: Extension 'amcheck(|_next)' version [\d.]* in schema 'public' do not support 'checkunique' parameter") + except ProbackupException as e: + self.assertIn( + "ERROR: checkdb --amcheck finished with failure. Not all checked indexes are valid. All databases were amchecked.", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertIn( + "Amcheck failed in database 'postgres' for index: 'public.bttest_unique_idx': ERROR: index \"bttest_unique_idx\" is corrupted. There are tuples violating UNIQUE constraint", + e.message) + + # Clean after yourself + node.stop() + + # @unittest.skip("skip") + def test_checkdb_sigint_handling(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + try: + node.safe_psql( + "postgres", + "create extension amcheck") + except QueryException as e: + node.safe_psql( + "postgres", + "create extension amcheck_next") + + # FULL backup + gdb = self.checkdb_node( + backup_dir, 'node', gdb=True, + options=[ + '-d', 'postgres', '-j', '2', + '--skip-block-validation', + '--progress', + '--amcheck', '-p', str(node.port)]) + + gdb.set_breakpoint('amcheck_one_index') + gdb.run_until_break() + + gdb.continue_execution_until_break(20) + gdb.remove_all_breakpoints() + + gdb._execute('signal SIGINT') + gdb.continue_execution_until_error() + + with open(node.pg_log_file, 'r') as f: + output = f.read() + + self.assertNotIn('could not receive data from client', output) + self.assertNotIn('could not send data to client', output) + self.assertNotIn('connection to client lost', output) + + # Clean after yourself + gdb.kill() + node.stop() + + # @unittest.skip("skip") + def test_checkdb_with_least_privileges(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + 'CREATE DATABASE backupdb') + + try: + node.safe_psql( + "backupdb", + "create extension amcheck") + except QueryException as e: + node.safe_psql( + "backupdb", + "create extension amcheck_next") + + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC;") + + # PG 9.5 + if self.get_version(node) < 90600: + node.safe_psql( + 'backupdb', + 'CREATE ROLE backup WITH LOGIN; ' + 'GRANT CONNECT ON DATABASE backupdb to backup; ' + 'GRANT USAGE ON SCHEMA pg_catalog TO backup; ' + 'GRANT USAGE ON SCHEMA public TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_am TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_class TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_index TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_namespace TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.texteq(text, text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.namene(name, name) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.int8(integer) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.charne("char", "char") TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.string_to_array(text, text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.array_position(anyarray, anyelement) TO backup; ' + 'GRANT EXECUTE ON FUNCTION bt_index_check(regclass, bool) TO backup;') # amcheck-next function + + # PG 9.6 + elif self.get_version(node) > 90600 and self.get_version(node) < 100000: + node.safe_psql( + 'backupdb', + 'CREATE ROLE backup WITH LOGIN; ' + 'GRANT CONNECT ON DATABASE backupdb to backup; ' + 'GRANT USAGE ON SCHEMA pg_catalog TO backup; ' + 'GRANT USAGE ON SCHEMA public TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_am TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_class TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_index TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_namespace TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.texteq(text, text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.namene(name, name) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.int8(integer) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.charne("char", "char") TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.string_to_array(text, text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.array_position(anyarray, anyelement) TO backup; ' +# 'GRANT EXECUTE ON FUNCTION bt_index_check(regclass) TO backup; ' + 'GRANT EXECUTE ON FUNCTION bt_index_check(regclass, bool) TO backup;') + + # PG 10 + elif self.get_version(node) > 100000 and self.get_version(node) < 110000: + node.safe_psql( + 'backupdb', + 'CREATE ROLE backup WITH LOGIN; ' + 'GRANT CONNECT ON DATABASE backupdb to backup; ' + 'GRANT USAGE ON SCHEMA pg_catalog TO backup; ' + 'GRANT USAGE ON SCHEMA public TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_am TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_class TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_index TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_namespace TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.texteq(text, text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.namene(name, name) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.int8(integer) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.charne("char", "char") TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.string_to_array(text, text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.array_position(anyarray, anyelement) TO backup;' + 'GRANT EXECUTE ON FUNCTION bt_index_check(regclass) TO backup;') + + if ProbackupTest.enterprise: + # amcheck-1.1 + node.safe_psql( + 'backupdb', + 'GRANT EXECUTE ON FUNCTION bt_index_check(regclass, bool) TO backup') + else: + # amcheck-1.0 + node.safe_psql( + 'backupdb', + 'GRANT EXECUTE ON FUNCTION bt_index_check(regclass) TO backup') + # >= 11 < 14 + elif self.get_version(node) > 110000 and self.get_version(node) < 140000: + node.safe_psql( + 'backupdb', + 'CREATE ROLE backup WITH LOGIN; ' + 'GRANT CONNECT ON DATABASE backupdb to backup; ' + 'GRANT USAGE ON SCHEMA pg_catalog TO backup; ' + 'GRANT USAGE ON SCHEMA public TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_am TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_class TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_index TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_namespace TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.texteq(text, text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.namene(name, name) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.int8(integer) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.charne("char", "char") TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.string_to_array(text, text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.array_position(anyarray, anyelement) TO backup; ' + 'GRANT EXECUTE ON FUNCTION bt_index_check(regclass) TO backup; ' + 'GRANT EXECUTE ON FUNCTION bt_index_check(regclass, bool) TO backup;') + + # checkunique parameter + if ProbackupTest.enterprise: + if (self.get_version(node) >= 111300 and self.get_version(node) < 120000 + or self.get_version(node) >= 120800 and self.get_version(node) < 130000 + or self.get_version(node) >= 130400): + node.safe_psql( + "backupdb", + "GRANT EXECUTE ON FUNCTION bt_index_check(regclass, bool, bool) TO backup") + # >= 14 + else: + node.safe_psql( + 'backupdb', + 'CREATE ROLE backup WITH LOGIN; ' + 'GRANT CONNECT ON DATABASE backupdb to backup; ' + 'GRANT USAGE ON SCHEMA pg_catalog TO backup; ' + 'GRANT USAGE ON SCHEMA public TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_am TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_class TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_index TO backup; ' + 'GRANT SELECT ON TABLE pg_catalog.pg_namespace TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.texteq(text, text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.namene(name, name) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.int8(integer) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.charne("char", "char") TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.string_to_array(text, text) TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.array_position(anycompatiblearray, anycompatible) TO backup; ' + 'GRANT EXECUTE ON FUNCTION bt_index_check(regclass) TO backup; ' + 'GRANT EXECUTE ON FUNCTION bt_index_check(regclass, bool) TO backup;') + + # checkunique parameter + if ProbackupTest.enterprise: + node.safe_psql( + "backupdb", + "GRANT EXECUTE ON FUNCTION bt_index_check(regclass, bool, bool) TO backup") + + if ProbackupTest.pgpro: + node.safe_psql( + 'backupdb', + 'GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_version() TO backup; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_edition() TO backup;') + + # checkdb + try: + self.checkdb_node( + backup_dir, 'node', + options=[ + '--amcheck', '-U', 'backup', + '-d', 'backupdb', '-p', str(node.port)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because permissions are missing\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "INFO: Amcheck succeeded for database 'backupdb'", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertIn( + "WARNING: Extension 'amcheck' or 'amcheck_next' are " + "not installed in database postgres", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertIn( + "ERROR: Some databases were not amchecked", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # Clean after yourself + node.stop() diff --git a/tests/compatibility.py b/tests/compatibility.py deleted file mode 100644 index 7c7038828..000000000 --- a/tests/compatibility.py +++ /dev/null @@ -1,539 +0,0 @@ -import unittest -import subprocess -import os -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException -from sys import exit - -module_name = 'compatibility' - - -class CompatibilityTest(ProbackupTest, unittest.TestCase): - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_backward_compatibility_page(self): - """Description in jira issue PGPRO-434""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'} - ) - self.init_pb(backup_dir, old_binary=True) - self.show_pb(backup_dir) - - self.add_instance(backup_dir, 'node', node, old_binary=True) - self.show_pb(backup_dir) - - self.set_archiving(backup_dir, 'node', node, old_binary=True) - node.slow_start() - - node.pgbench_init(scale=10) - - # FULL backup with old binary - self.backup_node( - backup_dir, 'node', node, old_binary=True) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - self.show_pb(backup_dir) - - self.validate_pb(backup_dir) - - # RESTORE old FULL with new binary - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "--recovery-target-action=promote"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Page BACKUP with old binary - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "20"] - ) - pgbench.wait() - pgbench.stdout.close() - - self.backup_node( - backup_dir, 'node', node, backup_type='page', - old_binary=True) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "--recovery-target-action=promote"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Page BACKUP with new binary - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "20"] - ) - pgbench.wait() - pgbench.stdout.close() - - self.backup_node( - backup_dir, 'node', node, backup_type='page') - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "--recovery-target-action=promote"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_backward_compatibility_delta(self): - """Description in jira issue PGPRO-434""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'} - ) - self.init_pb(backup_dir, old_binary=True) - self.show_pb(backup_dir) - - self.add_instance(backup_dir, 'node', node, old_binary=True) - self.show_pb(backup_dir) - - self.set_archiving(backup_dir, 'node', node, old_binary=True) - node.slow_start() - - node.pgbench_init(scale=10) - - # FULL backup with old binary - self.backup_node( - backup_dir, 'node', node, old_binary=True) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - self.show_pb(backup_dir) - - self.validate_pb(backup_dir) - - # RESTORE old FULL with new binary - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "--recovery-target-action=promote"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Delta BACKUP with old binary - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "20"] - ) - pgbench.wait() - pgbench.stdout.close() - - self.backup_node( - backup_dir, 'node', node, backup_type='delta', - old_binary=True) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "--recovery-target-action=promote"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Delta BACKUP with new binary - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "20"] - ) - pgbench.wait() - pgbench.stdout.close() - - self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "--recovery-target-action=promote"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_backward_compatibility_ptrack(self): - """Description in jira issue PGPRO-434""" - - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off', - 'ptrack_enable': 'on'}) - - self.init_pb(backup_dir, old_binary=True) - self.show_pb(backup_dir) - - self.add_instance(backup_dir, 'node', node, old_binary=True) - self.show_pb(backup_dir) - - self.set_archiving(backup_dir, 'node', node, old_binary=True) - node.slow_start() - - node.pgbench_init(scale=10) - - # FULL backup with old binary - self.backup_node( - backup_dir, 'node', node, old_binary=True) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - self.show_pb(backup_dir) - - self.validate_pb(backup_dir) - - # RESTORE old FULL with new binary - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "--recovery-target-action=promote"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # ptrack BACKUP with old binary - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "20"] - ) - pgbench.wait() - pgbench.stdout.close() - - self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - old_binary=True) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "--recovery-target-action=promote"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Ptrack BACKUP with new binary - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "20"] - ) - pgbench.wait() - pgbench.stdout.close() - - self.backup_node( - backup_dir, 'node', node, backup_type='ptrack') - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "--recovery-target-action=promote"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_backward_compatibility_compression(self): - """Description in jira issue PGPRO-434""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) - - self.init_pb(backup_dir, old_binary=True) - self.add_instance(backup_dir, 'node', node, old_binary=True) - - self.set_archiving(backup_dir, 'node', node, old_binary=True) - node.slow_start() - - node.pgbench_init(scale=10) - - # FULL backup with OLD binary - backup_id = self.backup_node( - backup_dir, 'node', node, - old_binary=True, - options=['--compress']) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - # restore OLD FULL with new binary - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # PAGE backup with OLD binary - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "10"]) - pgbench.wait() - pgbench.stdout.close() - - self.backup_node( - backup_dir, 'node', node, - backup_type='page', - old_binary=True, - options=['--compress']) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # PAGE backup with new binary - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "10"]) - pgbench.wait() - pgbench.stdout.close() - - self.backup_node( - backup_dir, 'node', node, - backup_type='page', - options=['--compress']) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Delta backup with old binary - self.delete_pb(backup_dir, 'node', backup_id) - - self.backup_node( - backup_dir, 'node', node, - old_binary=True, - options=['--compress']) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "10"]) - - pgbench.wait() - pgbench.stdout.close() - - self.backup_node( - backup_dir, 'node', node, - backup_type='delta', - options=['--compress'], - old_binary=True) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Delta backup with new binary - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "10"]) - - pgbench.wait() - pgbench.stdout.close() - - self.backup_node( - backup_dir, 'node', node, - backup_type='delta', - options=['--compress']) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_backward_compatibility_merge(self): - """ - Create node, take FULL and PAGE backups with old binary, - merge them with new binary - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) - - self.init_pb(backup_dir, old_binary=True) - self.add_instance(backup_dir, 'node', node, old_binary=True) - - self.set_archiving(backup_dir, 'node', node, old_binary=True) - node.slow_start() - - # FULL backup with OLD binary - self.backup_node( - backup_dir, 'node', node, - old_binary=True) - - node.pgbench_init(scale=1) - - # PAGE backup with OLD binary - backup_id = self.backup_node( - backup_dir, 'node', node, - backup_type='page', old_binary=True) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - self.merge_backup(backup_dir, "node", backup_id) - - print(self.show_pb(backup_dir, as_text=True, as_json=False)) - - # restore OLD FULL with new binary - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, options=["-j", "4"]) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/compatibility_test.py b/tests/compatibility_test.py new file mode 100644 index 000000000..7ae8baf9f --- /dev/null +++ b/tests/compatibility_test.py @@ -0,0 +1,1503 @@ +import unittest +import subprocess +import os +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +from sys import exit +import shutil + + +def check_manual_tests_enabled(): + return 'PGPROBACKUP_MANUAL' in os.environ and os.environ['PGPROBACKUP_MANUAL'] == 'ON' + + +def check_ssh_agent_path_exists(): + return 'PGPROBACKUP_SSH_AGENT_PATH' in os.environ + + +class CrossCompatibilityTest(ProbackupTest, unittest.TestCase): + @unittest.skipUnless(check_manual_tests_enabled(), 'skip manual test') + @unittest.skipUnless(check_ssh_agent_path_exists(), 'skip no ssh agent path exist') + # @unittest.skip("skip") + def test_catchup_with_different_remote_major_pg(self): + """ + Decription in jira issue PBCKP-236 + This test exposures ticket error using pg_probackup builds for both PGPROEE11 and PGPROEE9_6 + + Prerequisites: + - pg_probackup git tag for PBCKP 2.5.1 + - master pg_probackup build should be made for PGPROEE11 + - agent pg_probackup build should be made for PGPROEE9_6 + + Calling probackup PGPROEE9_6 pg_probackup agent from PGPROEE11 pg_probackup master for DELTA backup causes + the PBCKP-236 problem + + Please give env variables PROBACKUP_MANUAL=ON;PGPROBACKUP_SSH_AGENT_PATH= + for the test + + Please make path for agent's pgprobackup_ssh_agent_path = '/home/avaness/postgres/postgres.build.ee.9.6/bin/' + without pg_probackup executable + """ + + self.verbose = True + self.remote = True + # please use your own local path like + # pgprobackup_ssh_agent_path = '/home/avaness/postgres/postgres.build.clean/bin/' + pgprobackup_ssh_agent_path = os.environ['PGPROBACKUP_SSH_AGENT_PATH'] + + src_pg = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'src'), + set_replication=True, + ) + src_pg.slow_start() + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question AS SELECT 42 AS answer") + + # do full catchup + dst_pg = self.make_empty_node(os.path.join(self.module_name, self.fname, 'dst')) + self.catchup_node( + backup_mode='FULL', + source_pgdata=src_pg.data_dir, + destination_node=dst_pg, + options=['-d', 'postgres', '-p', str(src_pg.port), '--stream'] + ) + + dst_options = {'port': str(dst_pg.port)} + self.set_auto_conf(dst_pg, dst_options) + dst_pg.slow_start() + dst_pg.stop() + + src_pg.safe_psql( + "postgres", + "CREATE TABLE ultimate_question2 AS SELECT 42 AS answer") + + # do delta catchup with remote pg_probackup agent with another postgres major version + # this DELTA backup should fail without PBCKP-236 patch. + self.catchup_node( + backup_mode='DELTA', + source_pgdata=src_pg.data_dir, + destination_node=dst_pg, + # here's substitution of --remoge-path pg_probackup agent compiled with another postgres version + options=['-d', 'postgres', '-p', str(src_pg.port), '--stream', '--remote-path=' + pgprobackup_ssh_agent_path] + ) + + +class CompatibilityTest(ProbackupTest, unittest.TestCase): + + def setUp(self): + super().setUp() + if not self.probackup_old_path: + self.skipTest('PGPROBACKUPBIN_OLD is not set') + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_backward_compatibility_page(self): + """Description in jira issue PGPRO-434""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.show_pb(backup_dir) + + self.add_instance(backup_dir, 'node', node, old_binary=True) + self.show_pb(backup_dir) + + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.pgbench_init(scale=10) + + # FULL backup with old binary + self.backup_node( + backup_dir, 'node', node, old_binary=True) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + self.show_pb(backup_dir) + + self.validate_pb(backup_dir) + + # RESTORE old FULL with new binary + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # Page BACKUP with old binary + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "20"] + ) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, backup_type='page', + old_binary=True) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # Page BACKUP with new binary + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "20"]) + + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + node.safe_psql( + 'postgres', + 'create table tmp as select * from pgbench_accounts where aid < 1000') + + node.safe_psql( + 'postgres', + 'delete from pgbench_accounts') + + node.safe_psql( + 'postgres', + 'VACUUM') + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + node.safe_psql( + 'postgres', + 'insert into pgbench_accounts select * from pgbench_accounts') + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_backward_compatibility_delta(self): + """Description in jira issue PGPRO-434""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.show_pb(backup_dir) + + self.add_instance(backup_dir, 'node', node, old_binary=True) + self.show_pb(backup_dir) + + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.pgbench_init(scale=10) + + # FULL backup with old binary + self.backup_node( + backup_dir, 'node', node, old_binary=True) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + self.show_pb(backup_dir) + + self.validate_pb(backup_dir) + + # RESTORE old FULL with new binary + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # Delta BACKUP with old binary + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "20"] + ) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + old_binary=True) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # Delta BACKUP with new binary + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "20"] + ) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node(backup_dir, 'node', node, backup_type='delta') + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + node.safe_psql( + 'postgres', + 'create table tmp as select * from pgbench_accounts where aid < 1000') + + node.safe_psql( + 'postgres', + 'delete from pgbench_accounts') + + node.safe_psql( + 'postgres', + 'VACUUM') + + self.backup_node(backup_dir, 'node', node, backup_type='delta') + + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + node.safe_psql( + 'postgres', + 'insert into pgbench_accounts select * from pgbench_accounts') + + self.backup_node(backup_dir, 'node', node, backup_type='delta') + + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_backward_compatibility_ptrack(self): + """Description in jira issue PGPRO-434""" + + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.show_pb(backup_dir) + + self.add_instance(backup_dir, 'node', node, old_binary=True) + self.show_pb(backup_dir) + + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + node.pgbench_init(scale=10) + + # FULL backup with old binary + self.backup_node( + backup_dir, 'node', node, old_binary=True) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + self.show_pb(backup_dir) + + self.validate_pb(backup_dir) + + # RESTORE old FULL with new binary + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # ptrack BACKUP with old binary + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "20"] + ) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + old_binary=True) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + self.restore_node( + backup_dir, 'node', node_restored, + options=[ + "-j", "4", + "--recovery-target=latest", + "--recovery-target-action=promote"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # Ptrack BACKUP with new binary + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "20"] + ) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack') + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + options=[ + "-j", "4", + "--recovery-target=latest", + "--recovery-target-action=promote"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_backward_compatibility_compression(self): + """Description in jira issue PGPRO-434""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.pgbench_init(scale=10) + + # FULL backup with OLD binary + backup_id = self.backup_node( + backup_dir, 'node', node, + old_binary=True, + options=['--compress']) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + # restore OLD FULL with new binary + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # PAGE backup with OLD binary + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "10"]) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, + backup_type='page', + old_binary=True, + options=['--compress']) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + self.restore_node( + backup_dir, 'node', node_restored, + options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # PAGE backup with new binary + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "10"]) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, + backup_type='page', + options=['--compress']) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # Delta backup with old binary + self.delete_pb(backup_dir, 'node', backup_id) + + self.backup_node( + backup_dir, 'node', node, + old_binary=True, + options=['--compress']) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "10"]) + + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, + backup_type='delta', + options=['--compress'], + old_binary=True) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # Delta backup with new binary + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "10"]) + + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, + backup_type='delta', + options=['--compress']) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_backward_compatibility_merge(self): + """ + Create node, take FULL and PAGE backups with old binary, + merge them with new binary + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + # FULL backup with OLD binary + self.backup_node( + backup_dir, 'node', node, + old_binary=True) + + node.pgbench_init(scale=1) + + # PAGE backup with OLD binary + backup_id = self.backup_node( + backup_dir, 'node', node, + backup_type='page', old_binary=True) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + self.merge_backup(backup_dir, "node", backup_id) + + self.show_pb(backup_dir, as_text=True, as_json=False) + + # restore OLD FULL with new binary + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_backward_compatibility_merge_1(self): + """ + Create node, take FULL and PAGE backups with old binary, + merge them with new binary. + old binary version =< 2.2.7 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.pgbench_init(scale=20) + + # FULL backup with OLD binary + self.backup_node(backup_dir, 'node', node, old_binary=True) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "1", "-T", "10", "--no-vacuum"]) + pgbench.wait() + pgbench.stdout.close() + + # PAGE1 backup with OLD binary + self.backup_node( + backup_dir, 'node', node, backup_type='page', old_binary=True) + + node.safe_psql( + 'postgres', + 'DELETE from pgbench_accounts') + + node.safe_psql( + 'postgres', + 'VACUUM pgbench_accounts') + + # PAGE2 backup with OLD binary + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type='page', old_binary=True) + + pgdata = self.pgdata_content(node.data_dir) + + # merge chain created by old binary with new binary + output = self.merge_backup(backup_dir, "node", backup_id) + + # check that in-place is disabled + self.assertIn( + "WARNING: In-place merge is disabled " + "because of storage format incompatibility", output) + + # restore merged backup + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node(backup_dir, 'node', node_restored) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_backward_compatibility_merge_2(self): + """ + Create node, take FULL and PAGE backups with old binary, + merge them with new binary. + old binary version =< 2.2.7 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.pgbench_init(scale=50) + + node.safe_psql( + 'postgres', + 'VACUUM pgbench_accounts') + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + # FULL backup with OLD binary + self.backup_node(backup_dir, 'node', node, old_binary=True) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "1", "-T", "10", "--no-vacuum"]) + pgbench.wait() + pgbench.stdout.close() + + # PAGE1 backup with OLD binary + page1 = self.backup_node( + backup_dir, 'node', node, + backup_type='page', old_binary=True) + + pgdata1 = self.pgdata_content(node.data_dir) + + node.safe_psql( + 'postgres', + "DELETE from pgbench_accounts where ctid > '(10,1)'") + + # PAGE2 backup with OLD binary + page2 = self.backup_node( + backup_dir, 'node', node, + backup_type='page', old_binary=True) + + pgdata2 = self.pgdata_content(node.data_dir) + + # PAGE3 backup with OLD binary + page3 = self.backup_node( + backup_dir, 'node', node, + backup_type='page', old_binary=True) + + pgdata3 = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "1", "-T", "10", "--no-vacuum"]) + pgbench.wait() + pgbench.stdout.close() + + # PAGE4 backup with NEW binary + page4 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + pgdata4 = self.pgdata_content(node.data_dir) + + # merge backups one by one and check data correctness + # merge PAGE1 + self.merge_backup( + backup_dir, "node", page1, options=['--log-level-file=VERBOSE']) + + # check data correctness for PAGE1 + node_restored.cleanup() + self.restore_node( + backup_dir, 'node', node_restored, backup_id=page1, + options=['--log-level-file=VERBOSE']) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata1, pgdata_restored) + + # merge PAGE2 + self.merge_backup(backup_dir, "node", page2) + + # check data correctness for PAGE2 + node_restored.cleanup() + self.restore_node(backup_dir, 'node', node_restored, backup_id=page2) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata2, pgdata_restored) + + # merge PAGE3 + self.show_pb(backup_dir, 'node', page3) + self.merge_backup(backup_dir, "node", page3) + self.show_pb(backup_dir, 'node', page3) + + # check data correctness for PAGE3 + node_restored.cleanup() + self.restore_node(backup_dir, 'node', node_restored, backup_id=page3) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata3, pgdata_restored) + + # merge PAGE4 + self.merge_backup(backup_dir, "node", page4) + + # check data correctness for PAGE4 + node_restored.cleanup() + self.restore_node(backup_dir, 'node', node_restored, backup_id=page4) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata4, pgdata_restored) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_backward_compatibility_merge_3(self): + """ + Create node, take FULL and PAGE backups with old binary, + merge them with new binary. + old binary version =< 2.2.7 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.pgbench_init(scale=50) + + node.safe_psql( + 'postgres', + 'VACUUM pgbench_accounts') + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + # FULL backup with OLD binary + self.backup_node( + backup_dir, 'node', node, old_binary=True, options=['--compress']) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "1", "-T", "10", "--no-vacuum"]) + pgbench.wait() + pgbench.stdout.close() + + # PAGE1 backup with OLD binary + page1 = self.backup_node( + backup_dir, 'node', node, + backup_type='page', old_binary=True, options=['--compress']) + + pgdata1 = self.pgdata_content(node.data_dir) + + node.safe_psql( + 'postgres', + "DELETE from pgbench_accounts where ctid > '(10,1)'") + + # PAGE2 backup with OLD binary + page2 = self.backup_node( + backup_dir, 'node', node, + backup_type='page', old_binary=True, options=['--compress']) + + pgdata2 = self.pgdata_content(node.data_dir) + + # PAGE3 backup with OLD binary + page3 = self.backup_node( + backup_dir, 'node', node, + backup_type='page', old_binary=True, options=['--compress']) + + pgdata3 = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "1", "-T", "10", "--no-vacuum"]) + pgbench.wait() + pgbench.stdout.close() + + # PAGE4 backup with NEW binary + page4 = self.backup_node( + backup_dir, 'node', node, backup_type='page', options=['--compress']) + pgdata4 = self.pgdata_content(node.data_dir) + + # merge backups one by one and check data correctness + # merge PAGE1 + self.merge_backup( + backup_dir, "node", page1, options=['--log-level-file=VERBOSE']) + + # check data correctness for PAGE1 + node_restored.cleanup() + self.restore_node( + backup_dir, 'node', node_restored, backup_id=page1, + options=['--log-level-file=VERBOSE']) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata1, pgdata_restored) + + # merge PAGE2 + self.merge_backup(backup_dir, "node", page2) + + # check data correctness for PAGE2 + node_restored.cleanup() + self.restore_node(backup_dir, 'node', node_restored, backup_id=page2) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata2, pgdata_restored) + + # merge PAGE3 + self.show_pb(backup_dir, 'node', page3) + self.merge_backup(backup_dir, "node", page3) + self.show_pb(backup_dir, 'node', page3) + + # check data correctness for PAGE3 + node_restored.cleanup() + self.restore_node(backup_dir, 'node', node_restored, backup_id=page3) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata3, pgdata_restored) + + # merge PAGE4 + self.merge_backup(backup_dir, "node", page4) + + # check data correctness for PAGE4 + node_restored.cleanup() + self.restore_node(backup_dir, 'node', node_restored, backup_id=page4) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata4, pgdata_restored) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_backward_compatibility_merge_4(self): + """ + Start merge between minor version, crash and retry it. + old binary version =< 2.4.0 + """ + if self.version_to_num(self.old_probackup_version) > self.version_to_num('2.4.0'): + self.assertTrue( + False, 'You need pg_probackup old_binary =< 2.4.0 for this test') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.pgbench_init(scale=20) + + node.safe_psql( + 'postgres', + 'VACUUM pgbench_accounts') + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + # FULL backup with OLD binary + self.backup_node( + backup_dir, 'node', node, old_binary=True, options=['--compress']) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "1", "-T", "20", "--no-vacuum"]) + pgbench.wait() + pgbench.stdout.close() + + # PAGE backup with NEW binary + page_id = self.backup_node( + backup_dir, 'node', node, backup_type='page', options=['--compress']) + pgdata = self.pgdata_content(node.data_dir) + + # merge PAGE4 + gdb = self.merge_backup(backup_dir, "node", page_id, gdb=True) + + gdb.set_breakpoint('rename') + gdb.run_until_break() + gdb.continue_execution_until_break(500) + gdb._execute('signal SIGKILL') + + try: + self.merge_backup(backup_dir, "node", page_id) + self.assertEqual( + 1, 0, + "Expecting Error because of format changes.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Retry of failed merge for backups with different " + "between minor versions is forbidden to avoid data corruption " + "because of storage format changes introduced in 2.4.0 version, " + "please take a new full backup", + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_backward_compatibility_merge_5(self): + """ + Create node, take FULL and PAGE backups with old binary, + merge them with new binary. + old binary version >= STORAGE_FORMAT_VERSION (2.4.4) + """ + if self.version_to_num(self.old_probackup_version) < self.version_to_num('2.4.4'): + self.assertTrue( + False, 'OLD pg_probackup binary must be >= 2.4.4 for this test') + + self.assertNotEqual( + self.version_to_num(self.old_probackup_version), + self.version_to_num(self.probackup_version)) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.pgbench_init(scale=20) + + # FULL backup with OLD binary + self.backup_node(backup_dir, 'node', node, old_binary=True) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "1", "-T", "10", "--no-vacuum"]) + pgbench.wait() + pgbench.stdout.close() + + # PAGE1 backup with OLD binary + self.backup_node( + backup_dir, 'node', node, backup_type='page', old_binary=True) + + node.safe_psql( + 'postgres', + 'DELETE from pgbench_accounts') + + node.safe_psql( + 'postgres', + 'VACUUM pgbench_accounts') + + # PAGE2 backup with OLD binary + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type='page', old_binary=True) + + pgdata = self.pgdata_content(node.data_dir) + + # merge chain created by old binary with new binary + output = self.merge_backup(backup_dir, "node", backup_id) + + # check that in-place is disabled + self.assertNotIn( + "WARNING: In-place merge is disabled " + "because of storage format incompatibility", output) + + # restore merged backup + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node(backup_dir, 'node', node_restored) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_page_vacuum_truncate(self): + """ + make node, create table, take full backup, + delete all data, vacuum relation, + take page backup, insert some data, + take second page backup, + restore latest page backup using new binary + and check data correctness + old binary should be 2.2.x version + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.safe_psql( + "postgres", + "create sequence t_seq; " + "create table t_heap as select i as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1024) i") + + node.safe_psql( + "postgres", + "vacuum t_heap") + + id1 = self.backup_node(backup_dir, 'node', node, old_binary=True) + pgdata1 = self.pgdata_content(node.data_dir) + + node.safe_psql( + "postgres", + "delete from t_heap") + + node.safe_psql( + "postgres", + "vacuum t_heap") + + id2 = self.backup_node( + backup_dir, 'node', node, backup_type='page', old_binary=True) + pgdata2 = self.pgdata_content(node.data_dir) + + node.safe_psql( + "postgres", + "insert into t_heap select i as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1) i") + + id3 = self.backup_node( + backup_dir, 'node', node, backup_type='page', old_binary=True) + pgdata3 = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + data_dir=node_restored.data_dir, backup_id=id1) + + # Physical comparison + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata1, pgdata_restored) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + data_dir=node_restored.data_dir, backup_id=id2) + + # Physical comparison + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata2, pgdata_restored) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + data_dir=node_restored.data_dir, backup_id=id3) + + # Physical comparison + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata3, pgdata_restored) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() + node_restored.cleanup() + + # @unittest.skip("skip") + def test_page_vacuum_truncate_compression(self): + """ + make node, create table, take full backup, + delete all data, vacuum relation, + take page backup, insert some data, + take second page backup, + restore latest page backup using new binary + and check data correctness + old binary should be 2.2.x version + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.safe_psql( + "postgres", + "create sequence t_seq; " + "create table t_heap as select i as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1024) i") + + node.safe_psql( + "postgres", + "vacuum t_heap") + + self.backup_node( + backup_dir, 'node',node, old_binary=True, options=['--compress']) + + node.safe_psql( + "postgres", + "delete from t_heap") + + node.safe_psql( + "postgres", + "vacuum t_heap") + + self.backup_node( + backup_dir, 'node', node, backup_type='page', + old_binary=True, options=['--compress']) + + node.safe_psql( + "postgres", + "insert into t_heap select i as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1) i") + + self.backup_node( + backup_dir, 'node', node, backup_type='page', + old_binary=True, options=['--compress']) + + pgdata = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node(backup_dir, 'node', node_restored) + + # Physical comparison + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() + + # @unittest.skip("skip") + def test_page_vacuum_truncate_compressed_1(self): + """ + make node, create table, take full backup, + delete all data, vacuum relation, + take page backup, insert some data, + take second page backup, + restore latest page backup using new binary + and check data correctness + old binary should be 2.2.x version + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + self.set_archiving(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + node.safe_psql( + "postgres", + "create sequence t_seq; " + "create table t_heap as select i as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1024) i") + + node.safe_psql( + "postgres", + "vacuum t_heap") + + id1 = self.backup_node( + backup_dir, 'node', node, + old_binary=True, options=['--compress']) + pgdata1 = self.pgdata_content(node.data_dir) + + node.safe_psql( + "postgres", + "delete from t_heap") + + node.safe_psql( + "postgres", + "vacuum t_heap") + + id2 = self.backup_node( + backup_dir, 'node', node, backup_type='page', + old_binary=True, options=['--compress']) + pgdata2 = self.pgdata_content(node.data_dir) + + node.safe_psql( + "postgres", + "insert into t_heap select i as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1) i") + + id3 = self.backup_node( + backup_dir, 'node', node, backup_type='page', + old_binary=True, options=['--compress']) + pgdata3 = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + data_dir=node_restored.data_dir, backup_id=id1) + + # Physical comparison + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata1, pgdata_restored) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + data_dir=node_restored.data_dir, backup_id=id2) + + # Physical comparison + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata2, pgdata_restored) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + data_dir=node_restored.data_dir, backup_id=id3) + + # Physical comparison + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata3, pgdata_restored) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() + node_restored.cleanup() + + # @unittest.skip("skip") + def test_hidden_files(self): + """ + old_version should be < 2.3.0 + Create hidden file in pgdata, take backup + with old binary, then try to delete backup + with new binary + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + open(os.path.join(node.data_dir, ".hidden_stuff"), 'a').close() + + backup_id = self.backup_node( + backup_dir, 'node',node, old_binary=True, options=['--stream']) + + self.delete_pb(backup_dir, 'node', backup_id) + + # @unittest.skip("skip") + def test_compatibility_tablespace(self): + """ + https://github.com/postgrespro/pg_probackup/issues/348 + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node, old_binary=True) + node.slow_start() + + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type="full", + options=["-j", "4", "--stream"], old_binary=True) + + tblspace_old_path = self.get_tblspace_path(node, 'tblspace_old') + + self.create_tblspace_in_node( + node, 'tblspace', + tblspc_path=tblspace_old_path) + + node.safe_psql( + "postgres", + "create table t_heap_lame tablespace tblspace " + "as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1000) i") + + tblspace_new_path = self.get_tblspace_path(node, 'tblspace_new') + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + try: + self.restore_node( + backup_dir, 'node', node_restored, + options=[ + "-j", "4", + "-T", "{0}={1}".format( + tblspace_old_path, tblspace_new_path)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because tablespace mapping is incorrect" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Backup {0} has no tablespaceses, ' + 'nothing to remap'.format(backup_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.backup_node( + backup_dir, 'node', node, backup_type="delta", + options=["-j", "4", "--stream"], old_binary=True) + + self.restore_node( + backup_dir, 'node', node_restored, + options=[ + "-j", "4", + "-T", "{0}={1}".format( + tblspace_old_path, tblspace_new_path)]) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) diff --git a/tests/compression.py b/tests/compression_test.py similarity index 82% rename from tests/compression.py rename to tests/compression_test.py index b8788a46c..55924b9d2 100644 --- a/tests/compression.py +++ b/tests/compression_test.py @@ -5,9 +5,6 @@ import subprocess -module_name = 'compression' - - class CompressionTest(ProbackupTest, unittest.TestCase): # @unittest.skip("skip") @@ -18,10 +15,9 @@ def test_basic_compression_stream_zlib(self): check data correctness in restored instance """ self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -36,7 +32,7 @@ def test_basic_compression_stream_zlib(self): "create table t_heap as select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,256) i") - full_result = node.execute("postgres", "SELECT * FROM t_heap") + full_result = node.table_checksum("t_heap") full_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='full', options=[ @@ -49,7 +45,7 @@ def test_basic_compression_stream_zlib(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(256,512) i") - page_result = node.execute("postgres", "SELECT * FROM t_heap") + page_result = node.table_checksum("t_heap") page_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='page', options=[ @@ -61,7 +57,7 @@ def test_basic_compression_stream_zlib(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(512,768) i") - delta_result = node.execute("postgres", "SELECT * FROM t_heap") + delta_result = node.table_checksum("t_heap") delta_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='delta', options=['--stream', '--compress-algorithm=zlib']) @@ -81,7 +77,7 @@ def test_basic_compression_stream_zlib(self): repr(self.output), self.cmd)) node.slow_start() - full_result_new = node.execute("postgres", "SELECT * FROM t_heap") + full_result_new = node.table_checksum("t_heap") self.assertEqual(full_result, full_result_new) node.cleanup() @@ -97,7 +93,7 @@ def test_basic_compression_stream_zlib(self): repr(self.output), self.cmd)) node.slow_start() - page_result_new = node.execute("postgres", "SELECT * FROM t_heap") + page_result_new = node.table_checksum("t_heap") self.assertEqual(page_result, page_result_new) node.cleanup() @@ -113,12 +109,8 @@ def test_basic_compression_stream_zlib(self): repr(self.output), self.cmd)) node.slow_start() - delta_result_new = node.execute("postgres", "SELECT * FROM t_heap") + delta_result_new = node.table_checksum("t_heap") self.assertEqual(delta_result, delta_result_new) - node.cleanup() - - # Clean after yourself - self.del_test_dir(module_name, fname) def test_compression_archive_zlib(self): """ @@ -126,10 +118,9 @@ def test_compression_archive_zlib(self): check data correctness in restored instance """ self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -143,7 +134,7 @@ def test_compression_archive_zlib(self): "postgres", "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,1) i") - full_result = node.execute("postgres", "SELECT * FROM t_heap") + full_result = node.table_checksum("t_heap") full_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='full', options=["--compress-algorithm=zlib"]) @@ -154,7 +145,7 @@ def test_compression_archive_zlib(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector " "from generate_series(0,2) i") - page_result = node.execute("postgres", "SELECT * FROM t_heap") + page_result = node.table_checksum("t_heap") page_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='page', options=["--compress-algorithm=zlib"]) @@ -164,7 +155,7 @@ def test_compression_archive_zlib(self): "postgres", "insert into t_heap select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,3) i") - delta_result = node.execute("postgres", "SELECT * FROM t_heap") + delta_result = node.table_checksum("t_heap") delta_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='delta', options=['--compress-algorithm=zlib']) @@ -184,7 +175,7 @@ def test_compression_archive_zlib(self): repr(self.output), self.cmd)) node.slow_start() - full_result_new = node.execute("postgres", "SELECT * FROM t_heap") + full_result_new = node.table_checksum("t_heap") self.assertEqual(full_result, full_result_new) node.cleanup() @@ -200,7 +191,7 @@ def test_compression_archive_zlib(self): repr(self.output), self.cmd)) node.slow_start() - page_result_new = node.execute("postgres", "SELECT * FROM t_heap") + page_result_new = node.table_checksum("t_heap") self.assertEqual(page_result, page_result_new) node.cleanup() @@ -216,23 +207,19 @@ def test_compression_archive_zlib(self): repr(self.output), self.cmd)) node.slow_start() - delta_result_new = node.execute("postgres", "SELECT * FROM t_heap") + delta_result_new = node.table_checksum("t_heap") self.assertEqual(delta_result, delta_result_new) node.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_compression_stream_pglz(self): """ make archive node, make full and page stream backups, check data correctness in restored instance """ self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -247,7 +234,7 @@ def test_compression_stream_pglz(self): "create table t_heap as select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,256) i") - full_result = node.execute("postgres", "SELECT * FROM t_heap") + full_result = node.table_checksum("t_heap") full_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='full', options=['--stream', '--compress-algorithm=pglz']) @@ -258,7 +245,7 @@ def test_compression_stream_pglz(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(256,512) i") - page_result = node.execute("postgres", "SELECT * FROM t_heap") + page_result = node.table_checksum("t_heap") page_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='page', options=['--stream', '--compress-algorithm=pglz']) @@ -269,7 +256,7 @@ def test_compression_stream_pglz(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(512,768) i") - delta_result = node.execute("postgres", "SELECT * FROM t_heap") + delta_result = node.table_checksum("t_heap") delta_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='delta', options=['--stream', '--compress-algorithm=pglz']) @@ -289,7 +276,7 @@ def test_compression_stream_pglz(self): repr(self.output), self.cmd)) node.slow_start() - full_result_new = node.execute("postgres", "SELECT * FROM t_heap") + full_result_new = node.table_checksum("t_heap") self.assertEqual(full_result, full_result_new) node.cleanup() @@ -305,7 +292,7 @@ def test_compression_stream_pglz(self): repr(self.output), self.cmd)) node.slow_start() - page_result_new = node.execute("postgres", "SELECT * FROM t_heap") + page_result_new = node.table_checksum("t_heap") self.assertEqual(page_result, page_result_new) node.cleanup() @@ -321,23 +308,19 @@ def test_compression_stream_pglz(self): repr(self.output), self.cmd)) node.slow_start() - delta_result_new = node.execute("postgres", "SELECT * FROM t_heap") + delta_result_new = node.table_checksum("t_heap") self.assertEqual(delta_result, delta_result_new) node.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_compression_archive_pglz(self): """ make archive node, make full and page backups, check data correctness in restored instance """ self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -352,7 +335,7 @@ def test_compression_archive_pglz(self): "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector " "from generate_series(0,100) i") - full_result = node.execute("postgres", "SELECT * FROM t_heap") + full_result = node.table_checksum("t_heap") full_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='full', options=['--compress-algorithm=pglz']) @@ -363,7 +346,7 @@ def test_compression_archive_pglz(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector " "from generate_series(100,200) i") - page_result = node.execute("postgres", "SELECT * FROM t_heap") + page_result = node.table_checksum("t_heap") page_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='page', options=['--compress-algorithm=pglz']) @@ -374,7 +357,7 @@ def test_compression_archive_pglz(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector " "from generate_series(200,300) i") - delta_result = node.execute("postgres", "SELECT * FROM t_heap") + delta_result = node.table_checksum("t_heap") delta_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='delta', options=['--compress-algorithm=pglz']) @@ -394,7 +377,7 @@ def test_compression_archive_pglz(self): repr(self.output), self.cmd)) node.slow_start() - full_result_new = node.execute("postgres", "SELECT * FROM t_heap") + full_result_new = node.table_checksum("t_heap") self.assertEqual(full_result, full_result_new) node.cleanup() @@ -410,7 +393,7 @@ def test_compression_archive_pglz(self): repr(self.output), self.cmd)) node.slow_start() - page_result_new = node.execute("postgres", "SELECT * FROM t_heap") + page_result_new = node.table_checksum("t_heap") self.assertEqual(page_result, page_result_new) node.cleanup() @@ -426,23 +409,19 @@ def test_compression_archive_pglz(self): repr(self.output), self.cmd)) node.slow_start() - delta_result_new = node.execute("postgres", "SELECT * FROM t_heap") + delta_result_new = node.table_checksum("t_heap") self.assertEqual(delta_result, delta_result_new) node.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_compression_wrong_algorithm(self): """ make archive node, make full and page backups, check data correctness in restored instance """ self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -464,13 +443,10 @@ def test_compression_wrong_algorithm(self): except ProbackupException as e: self.assertEqual( e.message, - 'ERROR: invalid compress algorithm value "bla-blah"\n', + 'ERROR: Invalid compress algorithm value "bla-blah"\n', '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_incompressible_pages(self): """ @@ -478,10 +454,9 @@ def test_incompressible_pages(self): take backup with compression, make sure that page was not compressed, restore backup and check data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -518,6 +493,3 @@ def test_incompressible_pages(self): self.compare_pgdata(pgdata, pgdata_restored) node.slow_start() - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/config.py b/tests/config.py deleted file mode 100644 index 4a382e13a..000000000 --- a/tests/config.py +++ /dev/null @@ -1,53 +0,0 @@ -import unittest -import subprocess -import os -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException -from sys import exit - -module_name = 'config' - - -class ConfigTest(ProbackupTest, unittest.TestCase): - - # @unittest.expectedFailure - # @unittest.skip("skip") - def test_remove_instance_config(self): - """remove pg_probackup.conf""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.show_pb(backup_dir) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.backup_node(backup_dir, 'node', node) - - self.backup_node( - backup_dir, 'node', node, backup_type='page') - - conf_file = os.path.join( - backup_dir, 'backups','node', 'pg_probackup.conf') - - os.unlink(os.path.join(backup_dir, 'backups','node', 'pg_probackup.conf')) - - try: - self.backup_node( - backup_dir, 'node', node, backup_type='page') - self.assertEqual( - 1, 0, - "Expecting Error because pg_probackup.conf is missing. " - ".\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - 'ERROR: could not open file "{0}": ' - 'No such file or directory'.format(conf_file), - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) diff --git a/tests/config_test.py b/tests/config_test.py new file mode 100644 index 000000000..b1a0f9295 --- /dev/null +++ b/tests/config_test.py @@ -0,0 +1,113 @@ +import unittest +import subprocess +import os +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +from sys import exit +from shutil import copyfile + + +class ConfigTest(ProbackupTest, unittest.TestCase): + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_remove_instance_config(self): + """remove pg_probackup.conself.f""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.show_pb(backup_dir) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + conf_file = os.path.join( + backup_dir, 'backups','node', 'pg_probackup.conf') + + os.unlink(os.path.join(backup_dir, 'backups','node', 'pg_probackup.conf')) + + try: + self.backup_node( + backup_dir, 'node', node, backup_type='page') + self.assertEqual( + 1, 0, + "Expecting Error because pg_probackup.conf is missing. " + ".\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: could not open file "{0}": ' + 'No such file or directory'.format(conf_file), + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_corrupt_backup_content(self): + """corrupt backup_content.control""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + full1_id = self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + 'postgres', + 'create table t1()') + + fulle2_id = self.backup_node(backup_dir, 'node', node) + + fulle1_conf_file = os.path.join( + backup_dir, 'backups','node', full1_id, 'backup_content.control') + + fulle2_conf_file = os.path.join( + backup_dir, 'backups','node', fulle2_id, 'backup_content.control') + + copyfile(fulle2_conf_file, fulle1_conf_file) + + try: + self.validate_pb(backup_dir, 'node') + self.assertEqual( + 1, 0, + "Expecting Error because pg_probackup.conf is missing. " + ".\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "WARNING: Invalid CRC of backup control file '{0}':".format(fulle1_conf_file), + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertIn( + "WARNING: Failed to get file list for backup {0}".format(full1_id), + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertIn( + "WARNING: Backup {0} file list is corrupted".format(full1_id), + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.show_pb(backup_dir, 'node', full1_id)['status'] + + self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], 'CORRUPT') + self.assertEqual(self.show_pb(backup_dir, 'node')[1]['status'], 'OK') diff --git a/tests/delete.py b/tests/delete_test.py similarity index 70% rename from tests/delete.py rename to tests/delete_test.py index ce6bbb98d..10100887d 100644 --- a/tests/delete.py +++ b/tests/delete_test.py @@ -2,10 +2,6 @@ import os from .helpers.ptrack_helpers import ProbackupTest, ProbackupException import subprocess -from sys import exit - - -module_name = 'delete' class DeleteTest(ProbackupTest, unittest.TestCase): @@ -14,12 +10,11 @@ class DeleteTest(ProbackupTest, unittest.TestCase): # @unittest.expectedFailure def test_delete_full_backups(self): """delete full backups""" - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -51,19 +46,15 @@ def test_delete_full_backups(self): self.assertEqual(show_backups[0]['id'], id_1) self.assertEqual(show_backups[1]['id'], id_3) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") # @unittest.expectedFailure def test_del_instance_archive(self): """delete full backups""" - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -83,19 +74,15 @@ def test_del_instance_archive(self): # Delete instance self.del_instance(backup_dir, 'node') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") # @unittest.expectedFailure def test_delete_archive_mix_compress_and_non_compressed_segments(self): """delete full backups""" - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir="{0}/{1}/node".format(module_name, fname), + base_dir="{0}/{1}/node".format(self.module_name, self.fname), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving( @@ -142,18 +129,14 @@ def test_delete_archive_mix_compress_and_non_compressed_segments(self): '--retention-redundancy=3', '--delete-expired']) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delete_increment_page(self): """delete increment and all after him""" - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -182,27 +165,27 @@ def test_delete_increment_page(self): self.assertEqual(show_backups[1]['backup-mode'], "FULL") self.assertEqual(show_backups[1]['status'], "OK") - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delete_increment_ptrack(self): """delete increment and all after him""" if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') + self.skipTest('Skipped because ptrack support is disabled') - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums'], - pg_options={'ptrack_enable': 'on'}) + base_dir=os.path.join(self.module_name, self.fname, 'node'), + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + 'postgres', + 'CREATE EXTENSION ptrack') + # full backup mode self.backup_node(backup_dir, 'node', node) # ptrack backup mode @@ -226,9 +209,6 @@ def test_delete_increment_ptrack(self): self.assertEqual(show_backups[1]['backup-mode'], "FULL") self.assertEqual(show_backups[1]['status'], "OK") - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delete_orphaned_wal_segments(self): """ @@ -236,12 +216,11 @@ def test_delete_orphaned_wal_segments(self): delete second backup without --wal option, then delete orphaned wals via --wal option """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -260,7 +239,7 @@ def test_delete_orphaned_wal_segments(self): # Check wals wals_dir = os.path.join(backup_dir, 'wal', 'node') - wals = [f for f in os.listdir(wals_dir) if os.path.isfile(os.path.join(wals_dir, f)) and not f.endswith('.backup')] + wals = [f for f in os.listdir(wals_dir) if os.path.isfile(os.path.join(wals_dir, f))] original_wal_quantity = len(wals) # delete second full backup @@ -283,23 +262,21 @@ def test_delete_orphaned_wal_segments(self): result = self.delete_pb(backup_dir, 'node', options=['--wal']) # delete useless wals - self.assertTrue('INFO: removed min WAL segment' in result - and 'INFO: removed max WAL segment' in result) + self.assertTrue('On timeline 1 WAL segments between ' in result + and 'will be removed' in result) + self.validate_pb(backup_dir) self.assertEqual(self.show_pb(backup_dir, 'node', backup_3_id)['status'], "OK") # Check quantity, it should be lower than original - wals = [f for f in os.listdir(wals_dir) if os.path.isfile(os.path.join(wals_dir, f)) and not f.endswith('.backup')] + wals = [f for f in os.listdir(wals_dir) if os.path.isfile(os.path.join(wals_dir, f))] self.assertTrue(original_wal_quantity > len(wals), "Number of wals not changed after 'delete --wal' which is illegal") # Delete last backup self.delete_pb(backup_dir, 'node', backup_3_id, options=['--wal']) - wals = [f for f in os.listdir(wals_dir) if os.path.isfile(os.path.join(wals_dir, f)) and not f.endswith('.backup')] + wals = [f for f in os.listdir(wals_dir) if os.path.isfile(os.path.join(wals_dir, f))] self.assertEqual (0, len(wals), "Number of wals should be equal to 0") - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delete_wal_between_multiple_timelines(self): """ @@ -310,12 +287,11 @@ def test_delete_wal_between_multiple_timelines(self): [A1, B1) are deleted and backups B1 and A2 keep their WAL """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -327,12 +303,11 @@ def test_delete_wal_between_multiple_timelines(self): node.pgbench_init(scale=3) node2 = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node2')) + base_dir=os.path.join(self.module_name, self.fname, 'node2')) node2.cleanup() self.restore_node(backup_dir, 'node', node2) - node2.append_conf( - 'postgresql.auto.conf', "port = {0}".format(node2.port)) + self.set_auto_conf(node2, {'port': node2.port}) node2.slow_start() # load some more data to node @@ -352,22 +327,18 @@ def test_delete_wal_between_multiple_timelines(self): self.validate_pb(backup_dir) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delete_backup_with_empty_control_file(self): """ take backup, truncate its control file, try to delete it via 'delete' command """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums'], set_replication=True) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -393,25 +364,20 @@ def test_delete_backup_with_empty_control_file(self): self.delete_pb(backup_dir, 'node', backup_id=backup_id) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delete_interleaved_incremental_chains(self): """complicated case of interleaved backup chains""" - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() # Take FULL BACKUPs - backup_id_a = self.backup_node(backup_dir, 'node', node) backup_id_b = self.backup_node(backup_dir, 'node', node) @@ -518,9 +484,6 @@ def test_delete_interleaved_incremental_chains(self): print(self.show_pb( backup_dir, 'node', as_json=False, as_text=True)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delete_multiple_descendants(self): """ @@ -533,12 +496,11 @@ def test_delete_multiple_descendants(self): FULLb | FULLa should be deleted """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -690,5 +652,171 @@ def test_delete_multiple_descendants(self): self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) - # Clean after yourself - self.del_test_dir(module_name, fname) + # @unittest.skip("skip") + def test_delete_multiple_descendants_dry_run(self): + """ + PAGEa3 + PAGEa2 / + \ / + PAGEa1 (delete target) + | + FULLa + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL BACKUP + node.pgbench_init(scale=1) + backup_id_a = self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + page_id_a1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + page_id_a2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + + # Change PAGEa2 to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a2, 'ERROR') + + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + page_id_a3 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Change PAGEa2 to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a2, 'OK') + + # Delete PAGEa1 + output = self.delete_pb( + backup_dir, 'node', page_id_a1, + options=['--dry-run', '--log-level-console=LOG', '--delete-wal']) + + print(output) + self.assertIn( + 'LOG: Backup {0} can be deleted'.format(page_id_a3), + output) + self.assertIn( + 'LOG: Backup {0} can be deleted'.format(page_id_a2), + output) + self.assertIn( + 'LOG: Backup {0} can be deleted'.format(page_id_a1), + output) + + self.assertIn( + 'INFO: Resident data size to free by ' + 'delete of backup {0} :'.format(page_id_a1), + output) + + self.assertIn( + 'On timeline 1 WAL segments between 000000010000000000000001 ' + 'and 000000010000000000000003 can be removed', + output) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) + + output = self.delete_pb( + backup_dir, 'node', page_id_a1, + options=['--log-level-console=LOG', '--delete-wal']) + + self.assertIn( + 'LOG: Backup {0} will be deleted'.format(page_id_a3), + output) + self.assertIn( + 'LOG: Backup {0} will be deleted'.format(page_id_a2), + output) + self.assertIn( + 'LOG: Backup {0} will be deleted'.format(page_id_a1), + output) + self.assertIn( + 'INFO: Resident data size to free by ' + 'delete of backup {0} :'.format(page_id_a1), + output) + + self.assertIn( + 'On timeline 1 WAL segments between 000000010000000000000001 ' + 'and 000000010000000000000003 will be removed', + output) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 1) + + self.validate_pb(backup_dir, 'node') + + def test_delete_error_backups(self): + """delete increment and all after him""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # full backup mode + self.backup_node(backup_dir, 'node', node) + # page backup mode + self.backup_node(backup_dir, 'node', node, backup_type="page") + + # Take FULL BACKUP + backup_id_a = self.backup_node(backup_dir, 'node', node) + # Take PAGE BACKUP + backup_id_b = self.backup_node(backup_dir, 'node', node, backup_type="page") + + backup_id_c = self.backup_node(backup_dir, 'node', node, backup_type="page") + + backup_id_d = self.backup_node(backup_dir, 'node', node, backup_type="page") + + # full backup mode + self.backup_node(backup_dir, 'node', node) + self.backup_node(backup_dir, 'node', node, backup_type="page") + backup_id_e = self.backup_node(backup_dir, 'node', node, backup_type="page") + self.backup_node(backup_dir, 'node', node, backup_type="page") + + # Change status to ERROR + self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_c, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_e, 'ERROR') + + print(self.show_pb(backup_dir, as_text=True, as_json=False)) + + show_backups = self.show_pb(backup_dir, 'node') + self.assertEqual(len(show_backups), 10) + + # delete error backups + output = self.delete_pb(backup_dir, 'node', options=['--status=ERROR', '--dry-run']) + show_backups = self.show_pb(backup_dir, 'node') + self.assertEqual(len(show_backups), 10) + + self.assertIn( + "Deleting all backups with status 'ERROR' in dry run mode", + output) + + self.assertIn( + "INFO: Backup {0} with status OK can be deleted".format(backup_id_d), + output) + + print(self.show_pb(backup_dir, as_text=True, as_json=False)) + + show_backups = self.show_pb(backup_dir, 'node') + output = self.delete_pb(backup_dir, 'node', options=['--status=ERROR']) + print(output) + show_backups = self.show_pb(backup_dir, 'node') + self.assertEqual(len(show_backups), 4) + + self.assertEqual(show_backups[0]['status'], "OK") + self.assertEqual(show_backups[1]['status'], "OK") + self.assertEqual(show_backups[2]['status'], "OK") + self.assertEqual(show_backups[3]['status'], "OK") diff --git a/tests/delta.py b/tests/delta_test.py similarity index 69% rename from tests/delta.py rename to tests/delta_test.py index 6a5a9d736..8736a079c 100644 --- a/tests/delta.py +++ b/tests/delta_test.py @@ -8,9 +8,6 @@ from threading import Thread -module_name = 'delta' - - class DeltaTest(ProbackupTest, unittest.TestCase): # @unittest.skip("skip") @@ -21,18 +18,14 @@ def test_basic_delta_vacuum_truncate(self): take delta backup, take second delta backup, restore latest delta backup and check data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -46,51 +39,40 @@ def test_basic_delta_vacuum_truncate(self): "create table t_heap as select i as id, " "md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1024) i;" - ) + "from generate_series(0,1024) i;") node.safe_psql( "postgres", - "vacuum t_heap" - ) + "vacuum t_heap") self.backup_node(backup_dir, 'node', node, options=['--stream']) node.safe_psql( "postgres", - "delete from t_heap where ctid >= '(11,0)'" - ) + "delete from t_heap where ctid >= '(11,0)'") node.safe_psql( "postgres", - "vacuum t_heap" - ) + "vacuum t_heap") self.backup_node( - backup_dir, 'node', node, backup_type='delta' - ) + backup_dir, 'node', node, backup_type='delta') self.backup_node( - backup_dir, 'node', node, backup_type='delta' - ) + backup_dir, 'node', node, backup_type='delta') pgdata = self.pgdata_content(node.data_dir) self.restore_node( - backup_dir, 'node', node_restored - ) + backup_dir, 'node', node_restored) # Physical comparison pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delta_vacuum_truncate_1(self): """ @@ -99,19 +81,14 @@ def test_delta_vacuum_truncate_1(self): take delta backup, take second delta backup, restore latest delta backup and check data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'autovacuum': 'off' - } ) node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored'), + base_dir=os.path.join(self.module_name, self.fname, 'node_restored'), ) self.init_pb(backup_dir) @@ -173,13 +150,9 @@ def test_delta_vacuum_truncate_1(self): pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delta_vacuum_truncate_2(self): """ @@ -188,19 +161,14 @@ def test_delta_vacuum_truncate_2(self): take delta backup, take second delta backup, restore latest delta backup and check data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'autovacuum': 'off' - } ) node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored'), + base_dir=os.path.join(self.module_name, self.fname, 'node_restored'), ) self.init_pb(backup_dir) @@ -218,7 +186,7 @@ def test_delta_vacuum_truncate_2(self): filepath = node.safe_psql( "postgres", "select pg_relation_filepath('t_heap')" - ).rstrip() + ).decode('utf-8').rstrip() self.backup_node(backup_dir, 'node', node) @@ -226,40 +194,32 @@ def test_delta_vacuum_truncate_2(self): os.unlink(os.path.join(node.data_dir, filepath + '.1')) self.backup_node( - backup_dir, 'node', node, backup_type='delta' - ) + backup_dir, 'node', node, backup_type='delta') self.backup_node( - backup_dir, 'node', node, backup_type='delta' - ) + backup_dir, 'node', node, backup_type='delta') pgdata = self.pgdata_content(node.data_dir) self.restore_node( - backup_dir, 'node', node_restored - ) + backup_dir, 'node', node_restored) # Physical comparison pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delta_stream(self): """ make archive node, take full and delta stream backups, restore them and check data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ @@ -279,7 +239,7 @@ def test_delta_stream(self): "md5(i::text)::tsvector as tsvector " "from generate_series(0,100) i") - full_result = node.execute("postgres", "SELECT * FROM t_heap") + full_result = node.table_checksum("t_heap") full_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='full', options=['--stream']) @@ -290,7 +250,7 @@ def test_delta_stream(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector " "from generate_series(100,200) i") - delta_result = node.execute("postgres", "SELECT * FROM t_heap") + delta_result = node.table_checksum("t_heap") delta_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='delta', options=['--stream']) @@ -310,7 +270,7 @@ def test_delta_stream(self): '\n Unexpected Error Message: {0}\n' ' CMD: {1}'.format(repr(self.output), self.cmd)) node.slow_start() - full_result_new = node.execute("postgres", "SELECT * FROM t_heap") + full_result_new = node.table_checksum("t_heap") self.assertEqual(full_result, full_result_new) node.cleanup() @@ -326,13 +286,10 @@ def test_delta_stream(self): '\n Unexpected Error Message: {0}\n' ' CMD: {1}'.format(repr(self.output), self.cmd)) node.slow_start() - delta_result_new = node.execute("postgres", "SELECT * FROM t_heap") + delta_result_new = node.table_checksum("t_heap") self.assertEqual(delta_result, delta_result_new) node.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delta_archive(self): """ @@ -340,10 +297,9 @@ def test_delta_archive(self): restore them and check data correctness """ self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -357,7 +313,7 @@ def test_delta_archive(self): "postgres", "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,1) i") - full_result = node.execute("postgres", "SELECT * FROM t_heap") + full_result = node.table_checksum("t_heap") full_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='full') @@ -366,7 +322,7 @@ def test_delta_archive(self): "postgres", "insert into t_heap select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,2) i") - delta_result = node.execute("postgres", "SELECT * FROM t_heap") + delta_result = node.table_checksum("t_heap") delta_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='delta') @@ -385,7 +341,7 @@ def test_delta_archive(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(self.output), self.cmd)) node.slow_start() - full_result_new = node.execute("postgres", "SELECT * FROM t_heap") + full_result_new = node.table_checksum("t_heap") self.assertEqual(full_result, full_result_new) node.cleanup() @@ -401,30 +357,25 @@ def test_delta_archive(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(self.output), self.cmd)) node.slow_start() - delta_result_new = node.execute("postgres", "SELECT * FROM t_heap") + delta_result_new = node.table_checksum("t_heap") self.assertEqual(delta_result, delta_result_new) node.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delta_multiple_segments(self): """ Make node, create table with multiple segments, write some data to it, check delta and data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'fsync': 'off', 'shared_buffers': '1GB', 'maintenance_work_mem': '1GB', - 'autovacuum': 'off', 'full_page_writes': 'off' } ) @@ -449,36 +400,36 @@ def test_delta_multiple_segments(self): node.safe_psql("postgres", "checkpoint") # GET LOGICAL CONTENT FROM NODE - result = node.safe_psql("postgres", "select * from pgbench_accounts") + result = node.table_checksum("pgbench_accounts") # delta BACKUP self.backup_node( - backup_dir, 'node', node, backup_type='delta', - options=['--stream']) + backup_dir, 'node', node, + backup_type='delta', options=['--stream']) # GET PHYSICAL CONTENT FROM NODE pgdata = self.pgdata_content(node.data_dir) # RESTORE NODE restored_node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'restored_node')) + base_dir=os.path.join(self.module_name, self.fname, 'restored_node')) restored_node.cleanup() tblspc_path = self.get_tblspace_path(node, 'somedata') tblspc_path_new = self.get_tblspace_path( restored_node, 'somedata_restored') - self.restore_node(backup_dir, 'node', restored_node, options=[ - "-j", "4", "-T", "{0}={1}".format(tblspc_path, tblspc_path_new), - "--recovery-target-action=promote"]) + self.restore_node( + backup_dir, 'node', restored_node, + options=[ + "-j", "4", "-T", "{0}={1}".format( + tblspc_path, tblspc_path_new)]) # GET PHYSICAL CONTENT FROM NODE_RESTORED pgdata_restored = self.pgdata_content(restored_node.data_dir) # START RESTORED NODE - restored_node.append_conf( - "postgresql.auto.conf", "port = {0}".format(restored_node.port)) + self.set_auto_conf(restored_node, {'port': restored_node.port}) restored_node.slow_start() - result_new = restored_node.safe_psql( - "postgres", "select * from pgbench_accounts") + result_new = restored_node.table_checksum("pgbench_accounts") # COMPARE RESTORED FILES self.assertEqual(result, result_new, 'data is lost') @@ -486,32 +437,26 @@ def test_delta_multiple_segments(self): if self.paranoia: self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delta_vacuum_full(self): """ make node, make full and delta stream backups, restore them and check data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s' - } - ) + initdb_params=['--data-checksums']) + node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored'), - ) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) - node_restored.cleanup() node.slow_start() self.create_tblspace_in_node(node, 'somedata') @@ -535,7 +480,7 @@ def test_delta_vacuum_full(self): process.start() while not gdb.stopped_in_breakpoint: - sleep(1) + time.sleep(1) gdb.continue_execution_until_break(20) @@ -563,30 +508,23 @@ def test_delta_vacuum_full(self): pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_create_db(self): """ Make node, take full backup, create database db1, take delta backup, restore database and check it presense """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'max_wal_size': '10GB', - 'checkpoint_timeout': '5min', - 'autovacuum': 'off' } ) @@ -600,7 +538,7 @@ def test_create_db(self): "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,100) i") - node.safe_psql("postgres", "SELECT * FROM t_heap") + node.table_checksum("t_heap") self.backup_node( backup_dir, 'node', node, options=["--stream"]) @@ -624,7 +562,7 @@ def test_create_db(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') + base_dir=os.path.join(self.module_name, self.fname, 'node_restored') ) node_restored.cleanup() @@ -644,8 +582,7 @@ def test_create_db(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() # DROP DATABASE DB1 @@ -679,8 +616,7 @@ def test_create_db(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() try: @@ -697,25 +633,20 @@ def test_create_db(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_exists_in_previous_backup(self): """ Make node, take full backup, create table, take page backup, take delta backup, check that file is no fully copied to delta backup """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'max_wal_size': '10GB', 'checkpoint_timeout': '5min', - 'autovacuum': 'off' } ) @@ -730,10 +661,10 @@ def test_exists_in_previous_backup(self): "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,100) i") - node.safe_psql("postgres", "SELECT * FROM t_heap") + node.table_checksum("t_heap") filepath = node.safe_psql( "postgres", - "SELECT pg_relation_filepath('t_heap')").rstrip() + "SELECT pg_relation_filepath('t_heap')").decode('utf-8').rstrip() self.backup_node( backup_dir, 'node', @@ -781,7 +712,7 @@ def test_exists_in_previous_backup(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') + base_dir=os.path.join(self.module_name, self.fname, 'node_restored') ) node_restored.cleanup() @@ -801,27 +732,21 @@ def test_exists_in_previous_backup(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_alter_table_set_tablespace_delta(self): """ Make node, create tablespace with table, take full backup, alter tablespace location, take delta backup, restore database. """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'checkpoint_timeout': '30s', - 'autovacuum': 'off' } ) @@ -835,8 +760,8 @@ def test_alter_table_set_tablespace_delta(self): "postgres", "create table t_heap tablespace somedata as select i as id," " md5(i::text) as text, md5(i::text)::tsvector as tsvector" - " from generate_series(0,100) i" - ) + " from generate_series(0,100) i") + # FULL backup self.backup_node(backup_dir, 'node', node, options=["--stream"]) @@ -844,24 +769,21 @@ def test_alter_table_set_tablespace_delta(self): self.create_tblspace_in_node(node, 'somedata_new') node.safe_psql( "postgres", - "alter table t_heap set tablespace somedata_new" - ) + "alter table t_heap set tablespace somedata_new") # DELTA BACKUP - result = node.safe_psql( - "postgres", "select * from t_heap") + result = node.table_checksum("t_heap") self.backup_node( backup_dir, 'node', node, backup_type='delta', - options=["--stream"] - ) + options=["--stream"]) + if self.paranoia: pgdata = self.pgdata_content(node.data_dir) # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') - ) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() self.restore_node( @@ -875,8 +797,7 @@ def test_alter_table_set_tablespace_delta(self): "-T", "{0}={1}".format( self.get_tblspace_path(node, 'somedata_new'), self.get_tblspace_path(node_restored, 'somedata_new') - ), - "--recovery-target-action=promote" + ) ] ) @@ -886,18 +807,13 @@ def test_alter_table_set_tablespace_delta(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - result_new = node_restored.safe_psql( - "postgres", "select * from t_heap") + result_new = node_restored.table_checksum("t_heap") self.assertEqual(result, result_new, 'lost some data after restore') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_alter_database_set_tablespace_delta(self): """ @@ -905,15 +821,11 @@ def test_alter_database_set_tablespace_delta(self): take delta backup, alter database tablespace location, take delta backup restore last delta backup. """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off' - } ) self.init_pb(backup_dir) @@ -959,7 +871,7 @@ def test_alter_database_set_tablespace_delta(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') + base_dir=os.path.join(self.module_name, self.fname, 'node_restored') ) node_restored.cleanup() @@ -984,27 +896,21 @@ def test_alter_database_set_tablespace_delta(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_delta_delete(self): """ Make node, create tablespace with table, take full backup, alter tablespace location, take delta backup, restore database. """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'checkpoint_timeout': '30s', - 'autovacuum': 'off' } ) @@ -1047,7 +953,7 @@ def test_delta_delete(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') + base_dir=os.path.join(self.module_name, self.fname, 'node_restored') ) node_restored.cleanup() @@ -1068,171 +974,19 @@ def test_delta_delete(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_delta_corruption_heal_via_ptrack_1(self): - """make node, corrupt some page, check that backup failed""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - self.backup_node( - backup_dir, 'node', node, - backup_type="full", options=["-j", "4", "--stream"]) - - node.safe_psql( - "postgres", - "create table t_heap as select 1 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i") - node.safe_psql( - "postgres", - "CHECKPOINT;") - - heap_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('t_heap')").rstrip() - - with open(os.path.join(node.data_dir, heap_path), "rb+", 0) as f: - f.seek(9000) - f.write(b"bla") - f.flush() - f.close - - self.backup_node( - backup_dir, 'node', node, - backup_type="delta", - options=["-j", "4", "--stream", '--log-level-file=verbose']) - - # open log file and check - with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: - log_content = f.read() - self.assertIn('block 1, try to fetch via SQL', log_content) - self.assertIn('SELECT pg_catalog.pg_ptrack_get_block', log_content) - f.close - - self.assertTrue( - self.show_pb(backup_dir, 'node')[1]['status'] == 'OK', - "Backup Status should be OK") - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_page_corruption_heal_via_ptrack_2(self): - """make node, corrupt some page, check that backup failed""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - self.backup_node( - backup_dir, 'node', node, backup_type="full", - options=["-j", "4", "--stream"]) - - node.safe_psql( - "postgres", - "create table t_heap as select 1 as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i") - node.safe_psql( - "postgres", - "CHECKPOINT;") - - heap_path = node.safe_psql( - "postgres", - "select pg_relation_filepath('t_heap')").rstrip() - node.stop() - - with open(os.path.join(node.data_dir, heap_path), "rb+", 0) as f: - f.seek(9000) - f.write(b"bla") - f.flush() - f.close - node.slow_start() - - try: - self.backup_node( - backup_dir, 'node', node, backup_type="delta", - options=["-j", "4", "--stream", "--log-level-console=LOG"]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because of page " - "corruption in PostgreSQL instance.\n" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - if self.remote: - self.assertTrue( - "LOG: File" in e.message and - "try to fetch via SQL" in e.message and - "WARNING: page verification failed, " - "calculated checksum" in e.message and - "ERROR: query failed: " - "ERROR: invalid page in block" in e.message and - "query was: SELECT pg_catalog.pg_ptrack_get_block" in e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - else: - self.assertTrue( - "WARNING: File" in e.message and - "blknum" in e.message and - "have wrong checksum" in e.message and - "try to fetch via SQL" in e.message and - "WARNING: page verification failed, " - "calculated checksum" in e.message and - "ERROR: query failed: " - "ERROR: invalid page in block" in e.message and - "query was: SELECT pg_catalog.pg_ptrack_get_block" in e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - self.assertTrue( - self.show_pb(backup_dir, 'node')[1]['status'] == 'ERROR', - "Backup Status should be ERROR") - - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_delta_nullified_heap_page_backup(self): """ make node, take full backup, nullify some heap block, take delta backup, restore, physically compare pgdata`s """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1242,7 +996,7 @@ def test_delta_nullified_heap_page_backup(self): file_path = node.safe_psql( "postgres", - "select pg_relation_filepath('pgbench_accounts')").rstrip() + "select pg_relation_filepath('pgbench_accounts')").decode('utf-8').rstrip() node.safe_psql( "postgres", @@ -1275,15 +1029,15 @@ def test_delta_nullified_heap_page_backup(self): content = f.read() self.assertIn( - "LOG: File: {0} blknum 1, empty page".format(file), + 'VERBOSE: File: "{0}" blknum 1, empty page'.format(file), content) self.assertNotIn( - "Skipping blknum: 1 in file: {0}".format(file), + "Skipping blknum 1 in file: {0}".format(file), content) # Restore DELTA backup node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() self.restore_node( @@ -1293,21 +1047,17 @@ def test_delta_nullified_heap_page_backup(self): pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_delta_backup_from_past(self): """ make node, take FULL stream backup, take DELTA stream backup, restore FULL backup, try to take second DELTA stream backup """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -1348,5 +1098,100 @@ def test_delta_backup_from_past(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) + @unittest.skip("skip") + # @unittest.expectedFailure + def test_delta_pg_resetxlog(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'shared_buffers': '512MB', + 'max_wal_size': '3GB'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # Create table + node.safe_psql( + "postgres", + "create extension bloom; create sequence t_seq; " + "create table t_heap " + "as select nextval('t_seq')::int as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " +# "from generate_series(0,25600) i") + "from generate_series(0,2560) i") + + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + node.safe_psql( + 'postgres', + "update t_heap set id = nextval('t_seq'), text = md5(text), " + "tsvector = md5(repeat(tsvector::text, 10))::tsvector") + + # kill the bastard + if self.verbose: + print('Killing postmaster. Losing Ptrack changes') + node.stop(['-m', 'immediate', '-D', node.data_dir]) + + # now smack it with sledgehammer + if node.major_version >= 10: + pg_resetxlog_path = self.get_bin_path('pg_resetwal') + wal_dir = 'pg_wal' + else: + pg_resetxlog_path = self.get_bin_path('pg_resetxlog') + wal_dir = 'pg_xlog' + + self.run_binary( + [ + pg_resetxlog_path, + '-D', + node.data_dir, + '-o 42', + '-f' + ], + asynchronous=False) + + if not node.status(): + node.slow_start() + else: + print("Die! Die! Why won't you die?... Why won't you die?") + exit(1) + + # take ptrack backup +# self.backup_node( +# backup_dir, 'node', node, +# backup_type='delta', options=['--stream']) + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['--stream']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because instance was brutalized by pg_resetxlog" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd) + ) + except ProbackupException as e: + self.assertIn( + 'Insert error message', + e.message, + '\n Unexpected Error Message: {0}\n' + ' CMD: {1}'.format(repr(e.message), self.cmd)) + +# pgdata = self.pgdata_content(node.data_dir) +# +# node_restored = self.make_simple_node( +# base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) +# node_restored.cleanup() +# +# self.restore_node( +# backup_dir, 'node', node_restored) +# +# pgdata_restored = self.pgdata_content(node_restored.data_dir) +# self.compare_pgdata(pgdata, pgdata_restored) diff --git a/tests/exclude.py b/tests/exclude.py deleted file mode 100644 index 2e644d7e9..000000000 --- a/tests/exclude.py +++ /dev/null @@ -1,156 +0,0 @@ -import os -import unittest -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException - - -module_name = 'exclude' - - -class ExcludeTest(ProbackupTest, unittest.TestCase): - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_exclude_temp_tables(self): - """ - make node without archiving, create temp table, take full backup, - check that temp table not present in backup catalogue - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - conn = node.connect() - with node.connect("postgres") as conn: - - conn.execute( - "create temp table test as " - "select generate_series(0,50050000)::text") - conn.commit() - - temp_schema_name = conn.execute( - "SELECT nspname FROM pg_namespace " - "WHERE oid = pg_my_temp_schema()")[0][0] - conn.commit() - - temp_toast_schema_name = "pg_toast_" + temp_schema_name.replace( - "pg_", "") - conn.commit() - - conn.execute("create index test_idx on test (generate_series)") - conn.commit() - - heap_path = conn.execute( - "select pg_relation_filepath('test')")[0][0] - conn.commit() - - index_path = conn.execute( - "select pg_relation_filepath('test_idx')")[0][0] - conn.commit() - - heap_oid = conn.execute("select 'test'::regclass::oid")[0][0] - conn.commit() - - toast_path = conn.execute( - "select pg_relation_filepath('{0}.{1}')".format( - temp_toast_schema_name, "pg_toast_" + str(heap_oid)))[0][0] - conn.commit() - - toast_idx_path = conn.execute( - "select pg_relation_filepath('{0}.{1}')".format( - temp_toast_schema_name, - "pg_toast_" + str(heap_oid) + "_index"))[0][0] - conn.commit() - - temp_table_filename = os.path.basename(heap_path) - temp_idx_filename = os.path.basename(index_path) - temp_toast_filename = os.path.basename(toast_path) - temp_idx_toast_filename = os.path.basename(toast_idx_path) - - self.backup_node( - backup_dir, 'node', node, backup_type='full', options=['--stream']) - - for root, dirs, files in os.walk(backup_dir): - for file in files: - if file in [ - temp_table_filename, temp_table_filename + ".1", - temp_idx_filename, - temp_idx_filename + ".1", - temp_toast_filename, - temp_toast_filename + ".1", - temp_idx_toast_filename, - temp_idx_toast_filename + ".1" - ]: - self.assertEqual( - 1, 0, - "Found temp table file in backup catalogue.\n " - "Filepath: {0}".format(file)) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_exclude_unlogged_tables_1(self): - """ - make node without archiving, create unlogged table, take full backup, - alter table to unlogged, take delta backup, restore delta backup, - check that PGDATA`s are physically the same - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off', - "shared_buffers": "10MB"}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - conn = node.connect() - with node.connect("postgres") as conn: - - conn.execute( - "create unlogged table test as " - "select generate_series(0,5005000)::text") - conn.commit() - - conn.execute("create index test_idx on test (generate_series)") - conn.commit() - - self.backup_node( - backup_dir, 'node', node, - backup_type='full', options=['--stream']) - - node.safe_psql('postgres', "alter table test set logged") - - self.backup_node( - backup_dir, 'node', node, backup_type='delta', - options=['--stream'] - ) - - pgdata = self.pgdata_content(node.data_dir) - - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored, options=["-j", "4"]) - - # Physical comparison - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/exclude_test.py b/tests/exclude_test.py new file mode 100644 index 000000000..cb3530cd5 --- /dev/null +++ b/tests/exclude_test.py @@ -0,0 +1,338 @@ +import os +import unittest +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException + + +class ExcludeTest(ProbackupTest, unittest.TestCase): + + # @unittest.skip("skip") + def test_exclude_temp_files(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'logging_collector': 'on', + 'log_filename': 'postgresql.log'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + oid = node.safe_psql( + 'postgres', + "select oid from pg_database where datname = 'postgres'").rstrip() + + pgsql_tmp_dir = os.path.join(node.data_dir, 'base', 'pgsql_tmp') + + os.mkdir(pgsql_tmp_dir) + + file = os.path.join(pgsql_tmp_dir, 'pgsql_tmp7351.16') + with open(file, 'w') as f: + f.write("HELLO") + f.flush() + f.close + + full_id = self.backup_node( + backup_dir, 'node', node, backup_type='full', options=['--stream']) + + file = os.path.join( + backup_dir, 'backups', 'node', full_id, + 'database', 'base', 'pgsql_tmp', 'pgsql_tmp7351.16') + + self.assertFalse( + os.path.exists(file), + "File must be excluded: {0}".format(file)) + + # TODO check temporary tablespaces + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_exclude_temp_tables(self): + """ + make node without archiving, create temp table, take full backup, + check that temp table not present in backup catalogue + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + with node.connect("postgres") as conn: + + conn.execute( + "create temp table test as " + "select generate_series(0,50050000)::text") + conn.commit() + + temp_schema_name = conn.execute( + "SELECT nspname FROM pg_namespace " + "WHERE oid = pg_my_temp_schema()")[0][0] + conn.commit() + + temp_toast_schema_name = "pg_toast_" + temp_schema_name.replace( + "pg_", "") + conn.commit() + + conn.execute("create index test_idx on test (generate_series)") + conn.commit() + + heap_path = conn.execute( + "select pg_relation_filepath('test')")[0][0] + conn.commit() + + index_path = conn.execute( + "select pg_relation_filepath('test_idx')")[0][0] + conn.commit() + + heap_oid = conn.execute("select 'test'::regclass::oid")[0][0] + conn.commit() + + toast_path = conn.execute( + "select pg_relation_filepath('{0}.{1}')".format( + temp_toast_schema_name, "pg_toast_" + str(heap_oid)))[0][0] + conn.commit() + + toast_idx_path = conn.execute( + "select pg_relation_filepath('{0}.{1}')".format( + temp_toast_schema_name, + "pg_toast_" + str(heap_oid) + "_index"))[0][0] + conn.commit() + + temp_table_filename = os.path.basename(heap_path) + temp_idx_filename = os.path.basename(index_path) + temp_toast_filename = os.path.basename(toast_path) + temp_idx_toast_filename = os.path.basename(toast_idx_path) + + self.backup_node( + backup_dir, 'node', node, backup_type='full', options=['--stream']) + + for root, dirs, files in os.walk(backup_dir): + for file in files: + if file in [ + temp_table_filename, temp_table_filename + ".1", + temp_idx_filename, + temp_idx_filename + ".1", + temp_toast_filename, + temp_toast_filename + ".1", + temp_idx_toast_filename, + temp_idx_toast_filename + ".1" + ]: + self.assertEqual( + 1, 0, + "Found temp table file in backup catalogue.\n " + "Filepath: {0}".format(file)) + + # @unittest.skip("skip") + def test_exclude_unlogged_tables_1(self): + """ + make node without archiving, create unlogged table, take full backup, + alter table to unlogged, take delta backup, restore delta backup, + check that PGDATA`s are physically the same + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + "shared_buffers": "10MB"}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + conn = node.connect() + with node.connect("postgres") as conn: + + conn.execute( + "create unlogged table test as " + "select generate_series(0,5005000)::text") + conn.commit() + + conn.execute("create index test_idx on test (generate_series)") + conn.commit() + + self.backup_node( + backup_dir, 'node', node, + backup_type='full', options=['--stream']) + + node.safe_psql('postgres', "alter table test set logged") + + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) + + # Physical comparison + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_exclude_unlogged_tables_2(self): + """ + 1. make node, create unlogged, take FULL, DELTA, PAGE, + check that unlogged table files was not backed up + 2. restore FULL, DELTA, PAGE to empty db, + ensure unlogged table exist and is epmty + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + "shared_buffers": "10MB"}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_ids = [] + + for backup_type in ['full', 'delta', 'page']: + + if backup_type == 'full': + node.safe_psql( + 'postgres', + 'create unlogged table test as select generate_series(0,20050000)::text') + else: + node.safe_psql( + 'postgres', + 'insert into test select generate_series(0,20050000)::text') + + rel_path = node.execute( + 'postgres', + "select pg_relation_filepath('test')")[0][0] + + backup_id = self.backup_node( + backup_dir, 'node', node, + backup_type=backup_type, options=['--stream']) + + backup_ids.append(backup_id) + + filelist = self.get_backup_filelist( + backup_dir, 'node', backup_id) + + self.assertNotIn( + rel_path, filelist, + "Unlogged table was not excluded") + + self.assertNotIn( + rel_path + '.1', filelist, + "Unlogged table was not excluded") + + self.assertNotIn( + rel_path + '.2', filelist, + "Unlogged table was not excluded") + + self.assertNotIn( + rel_path + '.3', filelist, + "Unlogged table was not excluded") + + # ensure restoring retrieves back only empty unlogged table + for backup_id in backup_ids: + node.stop() + node.cleanup() + + self.restore_node(backup_dir, 'node', node, backup_id=backup_id) + + node.slow_start() + + self.assertEqual( + node.execute( + 'postgres', + 'select count(*) from test')[0][0], + 0) + + # @unittest.skip("skip") + def test_exclude_log_dir(self): + """ + check that by default 'log' and 'pg_log' directories are not backed up + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'logging_collector': 'on', + 'log_filename': 'postgresql.log'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.backup_node( + backup_dir, 'node', node, + backup_type='full', options=['--stream']) + + log_dir = node.safe_psql( + 'postgres', + 'show log_directory').decode('utf-8').rstrip() + + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]) + + # check that PGDATA/log or PGDATA/pg_log do not exists + path = os.path.join(node.data_dir, log_dir) + log_file = os.path.join(path, 'postgresql.log') + self.assertTrue(os.path.exists(path)) + self.assertFalse(os.path.exists(log_file)) + + # @unittest.skip("skip") + def test_exclude_log_dir_1(self): + """ + check that "--backup-pg-log" works correctly + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'logging_collector': 'on', + 'log_filename': 'postgresql.log'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + log_dir = node.safe_psql( + 'postgres', + 'show log_directory').decode('utf-8').rstrip() + + self.backup_node( + backup_dir, 'node', node, + backup_type='full', options=['--stream', '--backup-pg-log']) + + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]) + + # check that PGDATA/log or PGDATA/pg_log do not exists + path = os.path.join(node.data_dir, log_dir) + log_file = os.path.join(path, 'postgresql.log') + self.assertTrue(os.path.exists(path)) + self.assertTrue(os.path.exists(log_file)) diff --git a/tests/expected/option_help.out b/tests/expected/option_help.out index 2d7ae3b76..f0c77ae16 100644 --- a/tests/expected/option_help.out +++ b/tests/expected/option_help.out @@ -5,13 +5,14 @@ pg_probackup - utility to manage backup/recovery of PostgreSQL database. pg_probackup version - pg_probackup init -B backup-path + pg_probackup init -B backup-dir - pg_probackup set-config -B backup-path --instance=instance_name + pg_probackup set-config -B backup-dir --instance=instance-name [-D pgdata-path] [--external-dirs=external-directories-paths] [--log-level-console=log-level-console] [--log-level-file=log-level-file] + [--log-format-file=log-format-file] [--log-filename=log-filename] [--error-log-filename=error-log-filename] [--log-directory=log-directory] @@ -19,6 +20,7 @@ pg_probackup - utility to manage backup/recovery of PostgreSQL database. [--log-rotation-age=log-rotation-age] [--retention-redundancy=retention-redundancy] [--retention-window=retention-window] + [--wal-depth=wal-depth] [--compress-algorithm=compress-algorithm] [--compress-level=compress-level] [--archive-timeout=timeout] @@ -26,28 +28,40 @@ pg_probackup - utility to manage backup/recovery of PostgreSQL database. [--remote-proto] [--remote-host] [--remote-port] [--remote-path] [--remote-user] [--ssh-options] + [--restore-command=cmdline] [--archive-host=destination] + [--archive-port=port] [--archive-user=username] [--help] - pg_probackup show-config -B backup-path --instance=instance_name + pg_probackup set-backup -B backup-dir --instance=instance-name + -i backup-id [--ttl=interval] [--expire-time=timestamp] + [--note=text] + [--help] + + pg_probackup show-config -B backup-dir --instance=instance-name [--format=format] + [--no-scale-units] [--help] - pg_probackup backup -B backup-path -b backup-mode --instance=instance_name + pg_probackup backup -B backup-dir -b backup-mode --instance=instance-name [-D pgdata-path] [-C] - [--stream [-S slot-name]] [--temp-slot] + [--stream [-S slot-name] [--temp-slot]] [--backup-pg-log] [-j num-threads] [--progress] [--no-validate] [--skip-block-validation] [--external-dirs=external-directories-paths] + [--no-sync] [--log-level-console=log-level-console] [--log-level-file=log-level-file] + [--log-format-console=log-format-console] + [--log-format-file=log-format-file] [--log-filename=log-filename] [--error-log-filename=error-log-filename] [--log-directory=log-directory] [--log-rotation-size=log-rotation-size] - [--log-rotation-age=log-rotation-age] + [--log-rotation-age=log-rotation-age] [--no-color] [--delete-expired] [--delete-wal] [--merge-expired] [--retention-redundancy=retention-redundancy] [--retention-window=retention-window] + [--wal-depth=wal-depth] [--compress] [--compress-algorithm=compress-algorithm] [--compress-level=compress-level] @@ -57,9 +71,10 @@ pg_probackup - utility to manage backup/recovery of PostgreSQL database. [--remote-proto] [--remote-host] [--remote-port] [--remote-path] [--remote-user] [--ssh-options] + [--ttl=interval] [--expire-time=timestamp] [--note=text] [--help] - pg_probackup restore -B backup-path --instance=instance_name + pg_probackup restore -B backup-dir --instance=instance-name [-D pgdata-path] [-i backup-id] [-j num-threads] [--recovery-target-time=time|--recovery-target-xid=xid |--recovery-target-lsn=lsn [--recovery-target-inclusive=boolean]] @@ -67,17 +82,26 @@ pg_probackup - utility to manage backup/recovery of PostgreSQL database. [--recovery-target=immediate|latest] [--recovery-target-name=target-name] [--recovery-target-action=pause|promote|shutdown] - [--restore-as-replica] + [--restore-command=cmdline] + [-R | --restore-as-replica] [--force] + [--primary-conninfo=primary_conninfo] + [-S | --primary-slot-name=slotname] [--no-validate] [--skip-block-validation] [-T OLDDIR=NEWDIR] [--progress] [--external-mapping=OLDDIR=NEWDIR] - [--skip-external-dirs] + [--skip-external-dirs] [--no-sync] + [-X WALDIR | --waldir=WALDIR] + [-I | --incremental-mode=none|checksum|lsn] + [--db-include | --db-exclude] + [--destroy-all-other-dbs] [--remote-proto] [--remote-host] [--remote-port] [--remote-path] [--remote-user] [--ssh-options] + [--archive-host=hostname] + [--archive-port=port] [--archive-user=username] [--help] - pg_probackup validate -B backup-path [--instance=instance_name] + pg_probackup validate -B backup-dir [--instance=instance-name] [-i backup-id] [--progress] [-j num-threads] [--recovery-target-time=time|--recovery-target-xid=xid |--recovery-target-lsn=lsn [--recovery-target-inclusive=boolean]] @@ -86,43 +110,51 @@ pg_probackup - utility to manage backup/recovery of PostgreSQL database. [--skip-block-validation] [--help] - pg_probackup checkdb [-B backup-path] [--instance=instance_name] + pg_probackup checkdb [-B backup-dir] [--instance=instance-name] [-D pgdata-path] [--progress] [-j num-threads] [--amcheck] [--skip-block-validation] - [--heapallindexed] + [--heapallindexed] [--checkunique] [--help] - pg_probackup show -B backup-path - [--instance=instance_name [-i backup-id]] - [--format=format] - [--help] + pg_probackup show -B backup-dir + [--instance=instance-name [-i backup-id]] + [--format=format] [--archive] + [--no-color] [--help] - pg_probackup delete -B backup-path --instance=instance_name - [--wal] [-i backup-id | --expired | --merge-expired] - [--dry-run] + pg_probackup delete -B backup-dir --instance=instance-name + [-j num-threads] [--progress] + [--retention-redundancy=retention-redundancy] + [--retention-window=retention-window] + [--wal-depth=wal-depth] + [-i backup-id | --delete-expired | --merge-expired | --status=backup_status] + [--delete-wal] + [--dry-run] [--no-validate] [--no-sync] [--help] - pg_probackup merge -B backup-path --instance=instance_name + pg_probackup merge -B backup-dir --instance=instance-name -i backup-id [--progress] [-j num-threads] + [--no-validate] [--no-sync] [--help] - pg_probackup add-instance -B backup-path -D pgdata-path - --instance=instance_name + pg_probackup add-instance -B backup-dir -D pgdata-path + --instance=instance-name [--external-dirs=external-directories-paths] [--remote-proto] [--remote-host] [--remote-port] [--remote-path] [--remote-user] [--ssh-options] [--help] - pg_probackup del-instance -B backup-path - --instance=instance_name + pg_probackup del-instance -B backup-dir + --instance=instance-name [--help] - pg_probackup archive-push -B backup-path --instance=instance_name - --wal-file-path=wal-file-path + pg_probackup archive-push -B backup-dir --instance=instance-name --wal-file-name=wal-file-name - [--overwrite] - [--compress] + [--wal-file-path=wal-file-path] + [-j num-threads] [--batch-size=batch_size] + [--archive-timeout=timeout] + [--no-ready-rename] [--no-sync] + [--overwrite] [--compress] [--compress-algorithm=compress-algorithm] [--compress-level=compress-level] [--remote-proto] [--remote-host] @@ -130,13 +162,30 @@ pg_probackup - utility to manage backup/recovery of PostgreSQL database. [--ssh-options] [--help] - pg_probackup archive-get -B backup-path --instance=instance_name + pg_probackup archive-get -B backup-dir --instance=instance-name --wal-file-path=wal-file-path --wal-file-name=wal-file-name + [-j num-threads] [--batch-size=batch_size] + [--no-validate-wal] [--remote-proto] [--remote-host] [--remote-port] [--remote-path] [--remote-user] [--ssh-options] [--help] -Read the website for details. + pg_probackup catchup -b catchup-mode + --source-pgdata=path_to_pgdata_on_remote_server + --destination-pgdata=path_to_local_dir + [--stream [-S slot-name] [--temp-slot | --perm-slot]] + [-j num-threads] + [-T OLDDIR=NEWDIR] + [--exclude-path=path_prefix] + [-d dbname] [-h host] [-p port] [-U username] + [-w --no-password] [-W --password] + [--remote-proto] [--remote-host] + [--remote-port] [--remote-path] [--remote-user] + [--ssh-options] + [--dry-run] + [--help] + +Read the website for details . Report bugs to . diff --git a/tests/expected/option_help_ru.out b/tests/expected/option_help_ru.out new file mode 100644 index 000000000..bd6d76970 --- /dev/null +++ b/tests/expected/option_help_ru.out @@ -0,0 +1,191 @@ + +pg_probackup - утилита для управления резервным копированием/восстановлением базы данных PostgreSQL. + + pg_probackup help [COMMAND] + + pg_probackup version + + pg_probackup init -B backup-dir + + pg_probackup set-config -B backup-dir --instance=instance-name + [-D pgdata-path] + [--external-dirs=external-directories-paths] + [--log-level-console=log-level-console] + [--log-level-file=log-level-file] + [--log-format-file=log-format-file] + [--log-filename=log-filename] + [--error-log-filename=error-log-filename] + [--log-directory=log-directory] + [--log-rotation-size=log-rotation-size] + [--log-rotation-age=log-rotation-age] + [--retention-redundancy=retention-redundancy] + [--retention-window=retention-window] + [--wal-depth=wal-depth] + [--compress-algorithm=compress-algorithm] + [--compress-level=compress-level] + [--archive-timeout=timeout] + [-d dbname] [-h host] [-p port] [-U username] + [--remote-proto] [--remote-host] + [--remote-port] [--remote-path] [--remote-user] + [--ssh-options] + [--restore-command=cmdline] [--archive-host=destination] + [--archive-port=port] [--archive-user=username] + [--help] + + pg_probackup set-backup -B backup-dir --instance=instance-name + -i backup-id [--ttl=interval] [--expire-time=timestamp] + [--note=text] + [--help] + + pg_probackup show-config -B backup-dir --instance=instance-name + [--format=format] + [--no-scale-units] + [--help] + + pg_probackup backup -B backup-dir -b backup-mode --instance=instance-name + [-D pgdata-path] [-C] + [--stream [-S slot-name] [--temp-slot]] + [--backup-pg-log] [-j num-threads] [--progress] + [--no-validate] [--skip-block-validation] + [--external-dirs=external-directories-paths] + [--no-sync] + [--log-level-console=log-level-console] + [--log-level-file=log-level-file] + [--log-format-console=log-format-console] + [--log-format-file=log-format-file] + [--log-filename=log-filename] + [--error-log-filename=error-log-filename] + [--log-directory=log-directory] + [--log-rotation-size=log-rotation-size] + [--log-rotation-age=log-rotation-age] [--no-color] + [--delete-expired] [--delete-wal] [--merge-expired] + [--retention-redundancy=retention-redundancy] + [--retention-window=retention-window] + [--wal-depth=wal-depth] + [--compress] + [--compress-algorithm=compress-algorithm] + [--compress-level=compress-level] + [--archive-timeout=archive-timeout] + [-d dbname] [-h host] [-p port] [-U username] + [-w --no-password] [-W --password] + [--remote-proto] [--remote-host] + [--remote-port] [--remote-path] [--remote-user] + [--ssh-options] + [--ttl=interval] [--expire-time=timestamp] [--note=text] + [--help] + + pg_probackup restore -B backup-dir --instance=instance-name + [-D pgdata-path] [-i backup-id] [-j num-threads] + [--recovery-target-time=time|--recovery-target-xid=xid + |--recovery-target-lsn=lsn [--recovery-target-inclusive=boolean]] + [--recovery-target-timeline=timeline] + [--recovery-target=immediate|latest] + [--recovery-target-name=target-name] + [--recovery-target-action=pause|promote|shutdown] + [--restore-command=cmdline] + [-R | --restore-as-replica] [--force] + [--primary-conninfo=primary_conninfo] + [-S | --primary-slot-name=slotname] + [--no-validate] [--skip-block-validation] + [-T OLDDIR=NEWDIR] [--progress] + [--external-mapping=OLDDIR=NEWDIR] + [--skip-external-dirs] [--no-sync] + [-X WALDIR | --waldir=WALDIR] + [-I | --incremental-mode=none|checksum|lsn] + [--db-include | --db-exclude] + [--destroy-all-other-dbs] + [--remote-proto] [--remote-host] + [--remote-port] [--remote-path] [--remote-user] + [--ssh-options] + [--archive-host=hostname] + [--archive-port=port] [--archive-user=username] + [--help] + + pg_probackup validate -B backup-dir [--instance=instance-name] + [-i backup-id] [--progress] [-j num-threads] + [--recovery-target-time=time|--recovery-target-xid=xid + |--recovery-target-lsn=lsn [--recovery-target-inclusive=boolean]] + [--recovery-target-timeline=timeline] + [--recovery-target-name=target-name] + [--skip-block-validation] + [--help] + + pg_probackup checkdb [-B backup-dir] [--instance=instance-name] + [-D pgdata-path] [--progress] [-j num-threads] + [--amcheck] [--skip-block-validation] + [--heapallindexed] [--checkunique] + [--help] + + pg_probackup show -B backup-dir + [--instance=instance-name [-i backup-id]] + [--format=format] [--archive] + [--no-color] [--help] + + pg_probackup delete -B backup-dir --instance=instance-name + [-j num-threads] [--progress] + [--retention-redundancy=retention-redundancy] + [--retention-window=retention-window] + [--wal-depth=wal-depth] + [-i backup-id | --delete-expired | --merge-expired | --status=backup_status] + [--delete-wal] + [--dry-run] [--no-validate] [--no-sync] + [--help] + + pg_probackup merge -B backup-dir --instance=instance-name + -i backup-id [--progress] [-j num-threads] + [--no-validate] [--no-sync] + [--help] + + pg_probackup add-instance -B backup-dir -D pgdata-path + --instance=instance-name + [--external-dirs=external-directories-paths] + [--remote-proto] [--remote-host] + [--remote-port] [--remote-path] [--remote-user] + [--ssh-options] + [--help] + + pg_probackup del-instance -B backup-dir + --instance=instance-name + [--help] + + pg_probackup archive-push -B backup-dir --instance=instance-name + --wal-file-name=wal-file-name + [--wal-file-path=wal-file-path] + [-j num-threads] [--batch-size=batch_size] + [--archive-timeout=timeout] + [--no-ready-rename] [--no-sync] + [--overwrite] [--compress] + [--compress-algorithm=compress-algorithm] + [--compress-level=compress-level] + [--remote-proto] [--remote-host] + [--remote-port] [--remote-path] [--remote-user] + [--ssh-options] + [--help] + + pg_probackup archive-get -B backup-dir --instance=instance-name + --wal-file-path=wal-file-path + --wal-file-name=wal-file-name + [-j num-threads] [--batch-size=batch_size] + [--no-validate-wal] + [--remote-proto] [--remote-host] + [--remote-port] [--remote-path] [--remote-user] + [--ssh-options] + [--help] + + pg_probackup catchup -b catchup-mode + --source-pgdata=path_to_pgdata_on_remote_server + --destination-pgdata=path_to_local_dir + [--stream [-S slot-name] [--temp-slot | --perm-slot]] + [-j num-threads] + [-T OLDDIR=NEWDIR] + [--exclude-path=path_prefix] + [-d dbname] [-h host] [-p port] [-U username] + [-w --no-password] [-W --password] + [--remote-proto] [--remote-host] + [--remote-port] [--remote-path] [--remote-user] + [--ssh-options] + [--dry-run] + [--help] + +Подробнее читайте на сайте . +Сообщайте об ошибках в . diff --git a/tests/expected/option_version.out b/tests/expected/option_version.out deleted file mode 100644 index 9c96acc51..000000000 --- a/tests/expected/option_version.out +++ /dev/null @@ -1 +0,0 @@ -pg_probackup 2.1.5 \ No newline at end of file diff --git a/tests/external.py b/tests/external_test.py similarity index 86% rename from tests/external.py rename to tests/external_test.py index 1e383c709..53f3c5449 100644 --- a/tests/external.py +++ b/tests/external_test.py @@ -6,9 +6,7 @@ import shutil -module_name = 'external' - - +# TODO: add some ptrack tests class ExternalTest(ProbackupTest, unittest.TestCase): # @unittest.skip("skip") @@ -19,15 +17,14 @@ def test_basic_external(self): with external directory, restore backup, check that external directory was successfully copied """ - fname = self.id().split('.')[3] - core_dir = os.path.join(self.tmp_path, module_name, fname) + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums'], set_replication=True) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') external_dir = self.get_tblspace_path(node, 'somedirectory') # create directory in external_directory @@ -38,7 +35,8 @@ def test_basic_external(self): # take FULL backup with external directory pointing to a file file_path = os.path.join(core_dir, 'file') - open(file_path, "w+") + with open(file_path, "w+") as f: + pass try: self.backup_node( @@ -90,9 +88,6 @@ def test_basic_external(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") # @unittest.expectedFailure def test_external_none(self): @@ -102,13 +97,12 @@ def test_external_none(self): restore delta backup, check that external directory was not copied """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums'], set_replication=True) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') external_dir = self.get_tblspace_path(node, 'somedirectory') # create directory in external_directory @@ -152,9 +146,6 @@ def test_external_none(self): node.base_dir, exclude_dirs=['logs']) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") # @unittest.expectedFailure def test_external_dirs_overlapping(self): @@ -163,13 +154,12 @@ def test_external_dirs_overlapping(self): take backup with two external directories pointing to the same directory, backup should fail """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums'], set_replication=True) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') external_dir1 = self.get_tblspace_path(node, 'external_dir1') external_dir2 = self.get_tblspace_path(node, 'external_dir2') @@ -206,9 +196,6 @@ def test_external_dirs_overlapping(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_external_dir_mapping(self): """ @@ -217,13 +204,12 @@ def test_external_dir_mapping(self): check that restore with external-dir mapping will end with success """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -246,7 +232,7 @@ def test_external_dir_mapping(self): data_dir=external_dir2, options=["-j", "4"]) node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() external_dir1_new = self.get_tblspace_path(node_restored, 'external_dir1') @@ -299,20 +285,16 @@ def test_external_dir_mapping(self): node_restored.base_dir, exclude_dirs=['logs']) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") # @unittest.expectedFailure def test_backup_multiple_external(self): """check that cmdline has priority over config""" - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -360,9 +342,6 @@ def test_backup_multiple_external(self): node.base_dir, exclude_dirs=['logs']) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_backward_compatibility(self): @@ -372,13 +351,14 @@ def test_external_backward_compatibility(self): restore delta backup, check that incremental chain restored correctly """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + if not self.probackup_old_path: + self.skipTest("You must specify PGPROBACKUPBIN_OLD" + " for run this test") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir, old_binary=True) self.show_pb(backup_dir) @@ -444,7 +424,7 @@ def test_external_backward_compatibility(self): # RESTORE chain with new binary node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() @@ -463,9 +443,6 @@ def test_external_backward_compatibility(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_backward_compatibility_merge_1(self): @@ -474,13 +451,14 @@ def test_external_backward_compatibility_merge_1(self): take delta backup with new binary and 2 external directories merge delta backup ajd restore it """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + if not self.probackup_old_path: + self.skipTest("You must specify PGPROBACKUPBIN_OLD" + " for run this test") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir, old_binary=True) self.show_pb(backup_dir) @@ -537,7 +515,7 @@ def test_external_backward_compatibility_merge_1(self): # Restore merged backup node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() @@ -556,9 +534,6 @@ def test_external_backward_compatibility_merge_1(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_backward_compatibility_merge_2(self): @@ -567,13 +542,14 @@ def test_external_backward_compatibility_merge_2(self): take delta backup with new binary and 2 external directories merge delta backup and restore it """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + if not self.probackup_old_path: + self.skipTest("You must specify PGPROBACKUPBIN_OLD" + " for run this test") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir, old_binary=True) self.show_pb(backup_dir) @@ -659,7 +635,7 @@ def test_external_backward_compatibility_merge_2(self): # Restore merged backup node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() @@ -682,20 +658,18 @@ def test_external_backward_compatibility_merge_2(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_merge(self): """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + if not self.probackup_old_path: + self.skipTest("You must specify PGPROBACKUPBIN_OLD" + " for run this test") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node, old_binary=True) @@ -743,8 +717,11 @@ def test_external_merge(self): pgdata = self.pgdata_content( node.base_dir, exclude_dirs=['logs']) + print(self.show_pb(backup_dir, 'node', as_json=False, as_text=True)) + # Merge - self.merge_backup(backup_dir, 'node', backup_id=backup_id) + print(self.merge_backup(backup_dir, 'node', backup_id=backup_id, + options=['--log-level-file=VERBOSE'])) # RESTORE node.cleanup() @@ -767,21 +744,15 @@ def test_external_merge(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_merge_skip_external_dirs(self): """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -867,21 +838,15 @@ def test_external_merge_skip_external_dirs(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_merge_1(self): """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -949,20 +914,15 @@ def test_external_merge_1(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_merge_3(self): """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1043,21 +1003,15 @@ def test_external_merge_3(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_merge_2(self): """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1139,21 +1093,15 @@ def test_external_merge_2(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_restore_external_changed_data(self): """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1239,21 +1187,16 @@ def test_restore_external_changed_data(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_restore_external_changed_data_1(self): """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ - 'autovacuum': 'off', 'max_wal_size': '32MB'}) self.init_pb(backup_dir) @@ -1347,21 +1290,16 @@ def test_restore_external_changed_data_1(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_merge_external_changed_data(self): """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ - 'autovacuum': 'off', 'max_wal_size': '32MB'}) self.init_pb(backup_dir) @@ -1451,23 +1389,17 @@ def test_merge_external_changed_data(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_restore_skip_external(self): """ Check that --skip-external-dirs works correctly """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1524,9 +1456,6 @@ def test_restore_skip_external(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_dir_is_symlink(self): @@ -1536,18 +1465,15 @@ def test_external_dir_is_symlink(self): but restored as directory """ if os.name == 'nt': - return unittest.skip('Skipped for Windows') + self.skipTest('Skipped for Windows') - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1560,7 +1486,7 @@ def test_external_dir_is_symlink(self): backup_dir, 'node', node, options=["-j", "4", "--stream"]) # fill some directory with data - core_dir = os.path.join(self.tmp_path, module_name, fname) + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) symlinked_dir = os.path.join(core_dir, 'symlinked') self.restore_node( @@ -1584,7 +1510,7 @@ def test_external_dir_is_symlink(self): node.base_dir, exclude_dirs=['logs']) node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) # RESTORE node_restored.cleanup() @@ -1609,9 +1535,6 @@ def test_external_dir_is_symlink(self): backup_dir, 'node', backup_id=backup_id)['external-dirs']) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_dir_contain_symlink_on_dir(self): @@ -1621,14 +1544,13 @@ def test_external_dir_contain_symlink_on_dir(self): but restored as directory """ if os.name == 'nt': - return unittest.skip('Skipped for Windows') + self.skipTest('Skipped for Windows') - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -1644,7 +1566,7 @@ def test_external_dir_contain_symlink_on_dir(self): backup_dir, 'node', node, options=["-j", "4", "--stream"]) # fill some directory with data - core_dir = os.path.join(self.tmp_path, module_name, fname) + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) symlinked_dir = os.path.join(core_dir, 'symlinked') self.restore_node( @@ -1669,7 +1591,7 @@ def test_external_dir_contain_symlink_on_dir(self): node.base_dir, exclude_dirs=['logs']) node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) # RESTORE node_restored.cleanup() @@ -1694,9 +1616,6 @@ def test_external_dir_contain_symlink_on_dir(self): backup_dir, 'node', backup_id=backup_id)['external-dirs']) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_dir_contain_symlink_on_file(self): @@ -1706,14 +1625,13 @@ def test_external_dir_contain_symlink_on_file(self): but restored as directory """ if os.name == 'nt': - return unittest.skip('Skipped for Windows') + self.skipTest('Skipped for Windows') - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -1729,7 +1647,7 @@ def test_external_dir_contain_symlink_on_file(self): backup_dir, 'node', node, options=["-j", "4", "--stream"]) # fill some directory with data - core_dir = os.path.join(self.tmp_path, module_name, fname) + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) symlinked_dir = os.path.join(core_dir, 'symlinked') self.restore_node( @@ -1742,7 +1660,7 @@ def test_external_dir_contain_symlink_on_file(self): # create symlink to directory in external directory src_file = os.path.join(symlinked_dir, 'postgresql.conf') os.mkdir(external_dir) - os.chmod(external_dir, 0700) + os.chmod(external_dir, 0o0700) os.symlink(src_file, file_in_external_dir) # FULL backup with external directories @@ -1756,7 +1674,7 @@ def test_external_dir_contain_symlink_on_file(self): node.base_dir, exclude_dirs=['logs']) node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) # RESTORE node_restored.cleanup() @@ -1781,9 +1699,6 @@ def test_external_dir_contain_symlink_on_file(self): backup_dir, 'node', backup_id=backup_id)['external-dirs']) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_external_dir_is_tablespace(self): @@ -1791,16 +1706,13 @@ def test_external_dir_is_tablespace(self): Check that backup fails with error if external directory points to tablespace """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1833,25 +1745,19 @@ def test_external_dir_is_tablespace(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_restore_external_dir_not_empty(self): """ Check that backup fails with error if external directory point to not empty tablespace and if remapped directory also isn`t empty """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1915,9 +1821,6 @@ def test_restore_external_dir_not_empty(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_restore_external_dir_is_missing(self): """ take FULL backup with not empty external directory @@ -1925,16 +1828,13 @@ def test_restore_external_dir_is_missing(self): take DELTA backup with external directory, which should fail """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1999,9 +1899,6 @@ def test_restore_external_dir_is_missing(self): node.base_dir, exclude_dirs=['logs']) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_merge_external_dir_is_missing(self): """ take FULL backup with not empty external directory @@ -2012,16 +1909,13 @@ def test_merge_external_dir_is_missing(self): merge it into FULL, restore and check data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -2089,9 +1983,6 @@ def test_merge_external_dir_is_missing(self): node.base_dir, exclude_dirs=['logs']) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_restore_external_dir_is_empty(self): """ take FULL backup with not empty external directory @@ -2100,16 +1991,13 @@ def test_restore_external_dir_is_empty(self): restore DELRA backup, check that restored external directory is empty """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -2120,7 +2008,7 @@ def test_restore_external_dir_is_empty(self): # create empty file in external directory # open(os.path.join(external_dir, 'file'), 'a').close() os.mkdir(external_dir) - os.chmod(external_dir, 0700) + os.chmod(external_dir, 0o0700) with open(os.path.join(external_dir, 'file'), 'w+') as f: f.close() @@ -2155,9 +2043,6 @@ def test_restore_external_dir_is_empty(self): node.base_dir, exclude_dirs=['logs']) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_merge_external_dir_is_empty(self): """ take FULL backup with not empty external directory @@ -2166,16 +2051,13 @@ def test_merge_external_dir_is_empty(self): merge backups and restore FULL, check that restored external directory is empty """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -2186,7 +2068,7 @@ def test_merge_external_dir_is_empty(self): # create empty file in external directory # open(os.path.join(external_dir, 'file'), 'a').close() os.mkdir(external_dir) - os.chmod(external_dir, 0700) + os.chmod(external_dir, 0o0700) with open(os.path.join(external_dir, 'file'), 'w+') as f: f.close() @@ -2224,9 +2106,6 @@ def test_merge_external_dir_is_empty(self): node.base_dir, exclude_dirs=['logs']) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_restore_external_dir_string_order(self): """ take FULL backup with not empty external directory @@ -2235,16 +2114,13 @@ def test_restore_external_dir_string_order(self): restore DELRA backup, check that restored external directory is empty """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -2255,12 +2131,12 @@ def test_restore_external_dir_string_order(self): # create empty file in external directory os.mkdir(external_dir_1) - os.chmod(external_dir_1, 0700) + os.chmod(external_dir_1, 0o0700) with open(os.path.join(external_dir_1, 'fileA'), 'w+') as f: f.close() os.mkdir(external_dir_2) - os.chmod(external_dir_2, 0700) + os.chmod(external_dir_2, 0o0700) with open(os.path.join(external_dir_2, 'fileZ'), 'w+') as f: f.close() @@ -2306,9 +2182,6 @@ def test_restore_external_dir_string_order(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_merge_external_dir_string_order(self): """ take FULL backup with not empty external directory @@ -2317,16 +2190,13 @@ def test_merge_external_dir_string_order(self): restore DELRA backup, check that restored external directory is empty """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - core_dir = os.path.join(self.tmp_path, module_name, fname) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + core_dir = os.path.join(self.tmp_path, self.module_name, self.fname) shutil.rmtree(core_dir, ignore_errors=True) node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -2337,12 +2207,12 @@ def test_merge_external_dir_string_order(self): # create empty file in external directory os.mkdir(external_dir_1) - os.chmod(external_dir_1, 0700) + os.chmod(external_dir_1, 0o0700) with open(os.path.join(external_dir_1, 'fileA'), 'w+') as f: f.close() os.mkdir(external_dir_2) - os.chmod(external_dir_2, 0700) + os.chmod(external_dir_2, 0o0700) with open(os.path.join(external_dir_2, 'fileZ'), 'w+') as f: f.close() @@ -2391,9 +2261,6 @@ def test_merge_external_dir_string_order(self): self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_smart_restore_externals(self): """ @@ -2402,13 +2269,12 @@ def test_smart_restore_externals(self): make sure that files from externals are not copied during restore https://github.com/postgrespro/pg_probackup/issues/63 """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -2455,7 +2321,7 @@ def test_smart_restore_externals(self): logfile = os.path.join(backup_dir, 'log', 'pg_probackup.log') with open(logfile, 'r') as f: - logfile_content = f.read() + logfile_content = f.read() # get delta between FULL and PAGE filelists filelist_full = self.get_backup_filelist( @@ -2470,9 +2336,6 @@ def test_smart_restore_externals(self): for file in filelist_diff: self.assertNotIn(file, logfile_content) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_external_validation(self): """ @@ -2481,13 +2344,12 @@ def test_external_validation(self): corrupt external file in backup, run validate which should fail """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -2541,6 +2403,3 @@ def test_external_validation(self): 'CORRUPT', self.show_pb(backup_dir, 'node', full_id)['status'], 'Backup STATUS should be "CORRUPT"') - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/false_positive.py b/tests/false_positive.py deleted file mode 100644 index d3b27f877..000000000 --- a/tests/false_positive.py +++ /dev/null @@ -1,296 +0,0 @@ -import unittest -import os -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException -from datetime import datetime, timedelta -import subprocess - - -module_name = 'false_positive' - - -class FalsePositive(ProbackupTest, unittest.TestCase): - - # @unittest.skip("skip") - @unittest.expectedFailure - def test_validate_wal_lost_segment(self): - """ - Loose segment located between backups. ExpectedFailure. This is BUG - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.backup_node(backup_dir, 'node', node) - - # make some wals - node.pgbench_init(scale=5) - - # delete last wal segment - wals_dir = os.path.join(backup_dir, "wal", 'node') - wals = [f for f in os.listdir(wals_dir) if os.path.isfile( - os.path.join(wals_dir, f)) and not f.endswith('.backup')] - wals = map(int, wals) - os.remove(os.path.join(wals_dir, '0000000' + str(max(wals)))) - - # We just lost a wal segment and know nothing about it - self.backup_node(backup_dir, 'node', node) - self.assertTrue( - 'validation completed successfully' in self.validate_pb( - backup_dir, 'node')) - ######## - - # Clean after yourself - self.del_test_dir(module_name, fname) - - @unittest.expectedFailure - # Need to force validation of ancestor-chain - def test_incremental_backup_corrupt_full_1(self): - """page-level backup with corrupted full backup""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node(backup_dir, 'node', node) - file = os.path.join( - backup_dir, "backups", "node", - backup_id.decode("utf-8"), "database", "postgresql.conf") - os.remove(file) - - try: - self.backup_node(backup_dir, 'node', node, backup_type="page") - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because page backup should not be " - "possible without valid full backup.\n " - "Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertEqual( - e.message, - 'ERROR: Valid backup on current timeline is not found. ' - 'Create new FULL backup before an incremental one.\n', - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - sleep(1) - self.assertFalse( - True, - "Expecting Error because page backup should not be " - "possible without valid full backup.\n " - "Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertEqual( - e.message, - 'ERROR: Valid backup on current timeline is not found. ' - 'Create new FULL backup before an incremental one.\n', - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - self.assertEqual( - self.show_pb(backup_dir, 'node')[0]['Status'], "ERROR") - - # Clean after yourself - self.del_test_dir(module_name, fname) - - @unittest.expectedFailure - def test_ptrack_concurrent_get_and_clear_1(self): - """make node, make full and ptrack stream backups," - " restore them and check data correctness""" - - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'ptrack_enable': 'on' - } - ) - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "create table t_heap as select i" - " as id from generate_series(0,1) i" - ) - - self.backup_node(backup_dir, 'node', node, options=['--stream']) - gdb = self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - options=['--stream'], - gdb=True - ) - - gdb.set_breakpoint('make_pagemap_from_ptrack') - gdb.run_until_break() - - node.safe_psql( - "postgres", - "update t_heap set id = 100500") - - tablespace_oid = node.safe_psql( - "postgres", - "select oid from pg_tablespace where spcname = 'pg_default'").rstrip() - - relfilenode = node.safe_psql( - "postgres", - "select 't_heap'::regclass::oid").rstrip() - - node.safe_psql( - "postgres", - "SELECT pg_ptrack_get_and_clear({0}, {1})".format( - tablespace_oid, relfilenode)) - - gdb.continue_execution_until_exit() - - self.backup_node( - backup_dir, 'node', node, - backup_type='ptrack', options=['--stream'] - ) - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - result = node.safe_psql("postgres", "SELECT * FROM t_heap") - node.cleanup() - self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) - - # Physical comparison - if self.paranoia: - pgdata_restored = self.pgdata_content( - node.data_dir, ignore_ptrack=False) - self.compare_pgdata(pgdata, pgdata_restored) - - node.slow_start() - # Logical comparison - self.assertEqual( - result, - node.safe_psql("postgres", "SELECT * FROM t_heap")) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - @unittest.expectedFailure - def test_ptrack_concurrent_get_and_clear_2(self): - """make node, make full and ptrack stream backups," - " restore them and check data correctness""" - - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'ptrack_enable': 'on' - } - ) - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.safe_psql( - "postgres", - "create table t_heap as select i" - " as id from generate_series(0,1) i" - ) - - self.backup_node(backup_dir, 'node', node, options=['--stream']) - gdb = self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - options=['--stream'], - gdb=True - ) - - gdb.set_breakpoint('pthread_create') - gdb.run_until_break() - - node.safe_psql( - "postgres", - "update t_heap set id = 100500") - - tablespace_oid = node.safe_psql( - "postgres", - "select oid from pg_tablespace " - "where spcname = 'pg_default'").rstrip() - - relfilenode = node.safe_psql( - "postgres", - "select 't_heap'::regclass::oid").rstrip() - - node.safe_psql( - "postgres", - "SELECT pg_ptrack_get_and_clear({0}, {1})".format( - tablespace_oid, relfilenode)) - - gdb._execute("delete breakpoints") - gdb.continue_execution_until_exit() - - try: - self.backup_node( - backup_dir, 'node', node, - backup_type='ptrack', options=['--stream'] - ) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because of LSN mismatch from ptrack_control " - "and previous backup ptrack_lsn.\n" - " Output: {0} \n CMD: {1}".format(repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertTrue( - 'ERROR: LSN from ptrack_control' in e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - result = node.safe_psql("postgres", "SELECT * FROM t_heap") - node.cleanup() - self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) - - # Physical comparison - if self.paranoia: - pgdata_restored = self.pgdata_content( - node.data_dir, ignore_ptrack=False) - self.compare_pgdata(pgdata, pgdata_restored) - - node.slow_start() - # Logical comparison - self.assertEqual( - result, - node.safe_psql("postgres", "SELECT * FROM t_heap") - ) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/false_positive_test.py b/tests/false_positive_test.py new file mode 100644 index 000000000..ea82cb18f --- /dev/null +++ b/tests/false_positive_test.py @@ -0,0 +1,340 @@ +import unittest +import os +from time import sleep + +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +from datetime import datetime, timedelta +import subprocess + + +class FalsePositive(ProbackupTest, unittest.TestCase): + + # @unittest.skip("skip") + @unittest.expectedFailure + def test_validate_wal_lost_segment(self): + """ + Loose segment located between backups. ExpectedFailure. This is BUG + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + + # make some wals + node.pgbench_init(scale=5) + + # delete last wal segment + wals_dir = os.path.join(backup_dir, "wal", 'node') + wals = [f for f in os.listdir(wals_dir) if os.path.isfile( + os.path.join(wals_dir, f)) and not f.endswith('.backup')] + wals = map(int, wals) + os.remove(os.path.join(wals_dir, '0000000' + str(max(wals)))) + + # We just lost a wal segment and know nothing about it + self.backup_node(backup_dir, 'node', node) + self.assertTrue( + 'validation completed successfully' in self.validate_pb( + backup_dir, 'node')) + ######## + + @unittest.expectedFailure + # Need to force validation of ancestor-chain + def test_incremental_backup_corrupt_full_1(self): + """page-level backup with corrupted full backup""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node(backup_dir, 'node', node) + file = os.path.join( + backup_dir, "backups", "node", + backup_id.decode("utf-8"), "database", "postgresql.conf") + os.remove(file) + + try: + self.backup_node(backup_dir, 'node', node, backup_type="page") + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because page backup should not be " + "possible without valid full backup.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertEqual( + e.message, + 'ERROR: Valid full backup on current timeline is not found. ' + 'Create new FULL backup before an incremental one.\n', + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + self.assertFalse( + True, + "Expecting Error because page backup should not be " + "possible without valid full backup.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertEqual( + e.message, + 'ERROR: Valid full backup on current timeline is not found. ' + 'Create new FULL backup before an incremental one.\n', + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['Status'], "ERROR") + + # @unittest.skip("skip") + @unittest.expectedFailure + def test_pg_10_waldir(self): + """ + test group access for PG >= 11 + """ + if self.pg_config_version < self.version_to_num('10.0'): + self.skipTest('You need PostgreSQL >= 10 for this test') + + wal_dir = os.path.join( + os.path.join(self.tmp_path, self.module_name, self.fname), 'wal_dir') + import shutil + shutil.rmtree(wal_dir, ignore_errors=True) + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=[ + '--data-checksums', + '--waldir={0}'.format(wal_dir)]) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # take FULL backup + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + # restore backup + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored) + + # compare pgdata permissions + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + self.assertTrue( + os.path.islink(os.path.join(node_restored.data_dir, 'pg_wal')), + 'pg_wal should be symlink') + + @unittest.expectedFailure + # @unittest.skip("skip") + def test_recovery_target_time_backup_victim(self): + """ + Check that for validation to recovery target + probackup chooses valid backup + https://github.com/postgrespro/pg_probackup/issues/104 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + target_time = node.safe_psql( + "postgres", + "select now()").rstrip() + + node.safe_psql( + "postgres", + "create table t_heap1 as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,100) i") + + gdb = self.backup_node(backup_dir, 'node', node, gdb=True) + + # Attention! This breakpoint is set to a probackup internal fuction, not a postgres core one + gdb.set_breakpoint('pg_stop_backup') + gdb.run_until_break() + gdb.remove_all_breakpoints() + gdb._execute('signal SIGINT') + gdb.continue_execution_until_error() + + backup_id = self.show_pb(backup_dir, 'node')[1]['id'] + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node', backup_id)['status'], + 'Backup STATUS should be "ERROR"') + + self.validate_pb( + backup_dir, 'node', + options=['--recovery-target-time={0}'.format(target_time)]) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_recovery_target_lsn_backup_victim(self): + """ + Check that for validation to recovery target + probackup chooses valid backup + https://github.com/postgrespro/pg_probackup/issues/104 + + @y.sokolov: looks like this test should pass. + So I commented 'expectedFailure' + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + node.safe_psql( + "postgres", + "create table t_heap1 as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,100) i") + + gdb = self.backup_node( + backup_dir, 'node', node, + options=['--log-level-console=LOG'], gdb=True) + + # Attention! This breakpoint is set to a probackup internal fuction, not a postgres core one + gdb.set_breakpoint('pg_stop_backup') + gdb.run_until_break() + gdb.remove_all_breakpoints() + gdb._execute('signal SIGINT') + gdb.continue_execution_until_error() + + backup_id = self.show_pb(backup_dir, 'node')[1]['id'] + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node', backup_id)['status'], + 'Backup STATUS should be "ERROR"') + + self.switch_wal_segment(node) + + target_lsn = self.show_pb(backup_dir, 'node', backup_id)['start-lsn'] + + self.validate_pb( + backup_dir, 'node', + options=['--recovery-target-lsn={0}'.format(target_lsn)]) + + # @unittest.skip("skip") + @unittest.expectedFailure + def test_streaming_timeout(self): + """ + Illustrate the problem of loosing exact error + message because our WAL streaming engine is "borrowed" + from pg_receivexlog + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '1h', + 'wal_sender_timeout': '5s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + gdb = self.backup_node( + backup_dir, 'node', node, gdb=True, + options=['--stream', '--log-level-file=LOG']) + + # Attention! This breakpoint is set to a probackup internal fuction, not a postgres core one + gdb.set_breakpoint('pg_stop_backup') + gdb.run_until_break() + + sleep(10) + gdb.continue_execution_until_error() + gdb._execute('detach') + sleep(2) + + log_file_path = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(log_file_path) as f: + log_content = f.read() + + self.assertIn( + 'could not receive data from WAL stream', + log_content) + + self.assertIn( + 'ERROR: Problem in receivexlog', + log_content) + + # @unittest.skip("skip") + @unittest.expectedFailure + def test_validate_all_empty_catalog(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + + try: + self.validate_pb(backup_dir) + self.assertEqual( + 1, 0, + "Expecting Error because backup_dir is empty.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: This backup catalog contains no backup instances', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) diff --git a/tests/helpers/__init__.py b/tests/helpers/__init__.py index ac64c4230..2e5ed40e8 100644 --- a/tests/helpers/__init__.py +++ b/tests/helpers/__init__.py @@ -1,2 +1,9 @@ -__all__ = ['ptrack_helpers', 'cfs_helpers', 'expected_errors'] -#from . import * \ No newline at end of file +__all__ = ['ptrack_helpers', 'cfs_helpers', 'data_helpers'] + +import unittest + +# python 2.7 compatibility +if not hasattr(unittest.TestCase, "skipTest"): + def skipTest(self, reason): + raise unittest.SkipTest(reason) + unittest.TestCase.skipTest = skipTest \ No newline at end of file diff --git a/tests/helpers/cfs_helpers.py b/tests/helpers/cfs_helpers.py index 67e2b331b..31af76f2e 100644 --- a/tests/helpers/cfs_helpers.py +++ b/tests/helpers/cfs_helpers.py @@ -88,4 +88,6 @@ def corrupt_file(filename): def random_string(n): a = string.ascii_letters + string.digits - return ''.join([random.choice(a) for i in range(int(n)+1)]) \ No newline at end of file + random_str = ''.join([random.choice(a) for i in range(int(n)+1)]) + return str.encode(random_str) +# return ''.join([random.choice(a) for i in range(int(n)+1)]) diff --git a/tests/helpers/data_helpers.py b/tests/helpers/data_helpers.py new file mode 100644 index 000000000..27cb66c3d --- /dev/null +++ b/tests/helpers/data_helpers.py @@ -0,0 +1,78 @@ +import re +import unittest +import functools +import time + +def _tail_file(file, linetimeout, totaltimeout): + start = time.time() + with open(file, 'r') as f: + waits = 0 + while waits < linetimeout: + line = f.readline() + if line == '': + waits += 1 + time.sleep(1) + continue + waits = 0 + yield line + if time.time() - start > totaltimeout: + raise TimeoutError("total timeout tailing %s" % (file,)) + else: + raise TimeoutError("line timeout tailing %s" % (file,)) + + +class tail_file(object): # snake case to immitate function + def __init__(self, filename, *, linetimeout=10, totaltimeout=60, collect=False): + self.filename = filename + self.tailer = _tail_file(filename, linetimeout, totaltimeout) + self.collect = collect + self.lines = [] + self._content = None + + def __iter__(self): + return self + + def __next__(self): + line = next(self.tailer) + if self.collect: + self.lines.append(line) + self._content = None + return line + + @property + def content(self): + if not self.collect: + raise AttributeError("content collection is not enabled", + name="content", obj=self) + if not self._content: + self._content = "".join(self.lines) + return self._content + + def drop_content(self): + self.lines.clear() + self._content = None + + def stop_collect(self): + self.drop_content() + self.collect = False + + def wait(self, *, contains:str = None, regex:str = None): + assert contains != None or regex != None + assert contains == None or regex == None + try: + for line in self: + if contains is not None and contains in line: + break + if regex is not None and re.search(regex, line): + break + except TimeoutError: + msg = "Didn't found expected " + if contains is not None: + msg += repr(contains) + elif regex is not None: + msg += f"/{regex}/" + msg += f" in {self.filename}" + raise unittest.TestCase.failureException(msg) + + def wait_shutdown(self): + self.wait(contains='database system is shut down') diff --git a/tests/helpers/ptrack_helpers.py b/tests/helpers/ptrack_helpers.py index a6e151ae8..27d982856 100644 --- a/tests/helpers/ptrack_helpers.py +++ b/tests/helpers/ptrack_helpers.py @@ -1,6 +1,9 @@ # you need os for unittest to work import os +import gc +import unittest from sys import exit, argv, version_info +import signal import subprocess import shutil import six @@ -12,6 +15,7 @@ from time import sleep import re import json +import random idx_ptrack = { 't_heap': { @@ -86,25 +90,56 @@ def dir_files(base_dir): return out_list +def is_pgpro(): + # pg_config --help + cmd = [os.environ['PG_CONFIG'], '--help'] + + result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, check=True) + return b'postgrespro' in result.stdout + + def is_enterprise(): # pg_config --help - if os.name == 'posix': - cmd = [os.environ['PG_CONFIG'], '--help'] - - elif os.name == 'nt': - cmd = [[os.environ['PG_CONFIG']], ['--help']] - - p = subprocess.Popen( - cmd, - stdout=subprocess.PIPE, - stderr=subprocess.PIPE - ) - if b'postgrespro.ru' in p.communicate()[0]: - return True - else: + cmd = [os.environ['PG_CONFIG'], '--help'] + + p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, check=True) + # PostgresPro std or ent + if b'postgrespro' in p.stdout: + cmd = [os.environ['PG_CONFIG'], '--pgpro-edition'] + p = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, check=True) + + return b'enterprise' in p.stdout + else: # PostgreSQL return False +def is_nls_enabled(): + cmd = [os.environ['PG_CONFIG'], '--configure'] + + result = subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, check=True) + return b'enable-nls' in result.stdout + + +def base36enc(number): + """Converts an integer to a base36 string.""" + alphabet = '0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ' + base36 = '' + sign = '' + + if number < 0: + sign = '-' + number = -number + + if 0 <= number < len(alphabet): + return sign + alphabet[number] + + while number != 0: + number, i = divmod(number, len(alphabet)) + base36 = alphabet[i] + base36 + + return sign + base36 + + class ProbackupException(Exception): def __init__(self, message, cmd): self.message = message @@ -113,38 +148,105 @@ def __init__(self, message, cmd): def __str__(self): return '\n ERROR: {0}\n CMD: {1}'.format(repr(self.message), self.cmd) +class PostgresNodeExtended(testgres.PostgresNode): -def slow_start(self, replica=False): + def __init__(self, base_dir=None, *args, **kwargs): + super(PostgresNodeExtended, self).__init__(name='test', base_dir=base_dir, *args, **kwargs) + self.is_started = False - # wait for https://github.com/postgrespro/testgres/pull/50 -# self.start() -# self.poll_query_until( -# "postgres", -# "SELECT not pg_is_in_recovery()", -# suppress={testgres.NodeConnection}) - if replica: - query = 'SELECT pg_is_in_recovery()' - else: - query = 'SELECT not pg_is_in_recovery()' + def slow_start(self, replica=False): - self.start() - while True: - try: - if self.safe_psql('postgres', query) == 't\n': - break - except testgres.QueryException as e: - if 'database system is starting up' in e[0]: - continue + # wait for https://github.com/postgrespro/testgres/pull/50 + # self.start() + # self.poll_query_until( + # "postgres", + # "SELECT not pg_is_in_recovery()", + # suppress={testgres.NodeConnection}) + if replica: + query = 'SELECT pg_is_in_recovery()' + else: + query = 'SELECT not pg_is_in_recovery()' + + self.start() + while True: + try: + output = self.safe_psql('template1', query).decode("utf-8").rstrip() + + if output == 't': + break + + except testgres.QueryException as e: + if 'database system is starting up' in e.message: + pass + elif 'FATAL: the database system is not accepting connections' in e.message: + pass + elif replica and 'Hot standby mode is disabled' in e.message: + raise e + else: + raise e + + sleep(0.5) + + def start(self, *args, **kwargs): + if not self.is_started: + super(PostgresNodeExtended, self).start(*args, **kwargs) + self.is_started = True + return self + + def stop(self, *args, **kwargs): + if self.is_started: + result = super(PostgresNodeExtended, self).stop(*args, **kwargs) + self.is_started = False + return result + + def kill(self, someone = None): + if self.is_started: + sig = signal.SIGKILL if os.name != 'nt' else signal.SIGBREAK + if someone == None: + os.kill(self.pid, sig) else: - raise e + os.kill(self.auxiliary_pids[someone][0], sig) + self.is_started = False + + def table_checksum(self, table, dbname="postgres"): + con = self.connect(dbname=dbname) + + curname = "cur_"+str(random.randint(0,2**48)) + + con.execute(""" + DECLARE %s NO SCROLL CURSOR FOR + SELECT t::text FROM %s as t + """ % (curname, table)) + sum = hashlib.md5() + while True: + rows = con.execute("FETCH FORWARD 5000 FROM %s" % curname) + if not rows: + break + for row in rows: + # hash uses SipHash since Python3.4, therefore it is good enough + sum.update(row[0].encode('utf8')) + + con.execute("CLOSE %s; ROLLBACK;" % curname) + + con.close() + return sum.hexdigest() class ProbackupTest(object): # Class attributes enterprise = is_enterprise() + enable_nls = is_nls_enabled() + pgpro = is_pgpro() def __init__(self, *args, **kwargs): super(ProbackupTest, self).__init__(*args, **kwargs) + + self.nodes_to_cleanup = [] + + if isinstance(self, unittest.TestCase): + self.module_name = self.id().split('.')[1] + self.fname = self.id().split('.')[3] + if '-v' in argv or '--verbose' in argv: self.verbose = True else: @@ -175,15 +277,15 @@ def __init__(self, *args, **kwargs): self.test_env['LC_MESSAGES'] = 'C' self.test_env['LC_TIME'] = 'C' - self.paranoia = False - if 'PG_PROBACKUP_PARANOIA' in self.test_env: - if self.test_env['PG_PROBACKUP_PARANOIA'] == 'ON': - self.paranoia = True + self.gdb = 'PGPROBACKUP_GDB' in self.test_env and \ + self.test_env['PGPROBACKUP_GDB'] == 'ON' + + self.paranoia = 'PG_PROBACKUP_PARANOIA' in self.test_env and \ + self.test_env['PG_PROBACKUP_PARANOIA'] == 'ON' + + self.archive_compress = 'ARCHIVE_COMPRESSION' in self.test_env and \ + self.test_env['ARCHIVE_COMPRESSION'] == 'ON' - self.archive_compress = False - if 'ARCHIVE_COMPRESSION' in self.test_env: - if self.test_env['ARCHIVE_COMPRESSION'] == 'ON': - self.archive_compress = True try: testgres.configure_testgres( cache_initdb=False, @@ -208,10 +310,7 @@ def __init__(self, *args, **kwargs): self.user = self.get_username() self.probackup_path = None if 'PGPROBACKUPBIN' in self.test_env: - if ( - os.path.isfile(self.test_env["PGPROBACKUPBIN"]) and - os.access(self.test_env["PGPROBACKUPBIN"], os.X_OK) - ): + if shutil.which(self.test_env["PGPROBACKUPBIN"]): self.probackup_path = self.test_env["PGPROBACKUPBIN"] else: if self.verbose: @@ -265,6 +364,32 @@ def __init__(self, *args, **kwargs): if self.verbose: print('PGPROBACKUPBIN_OLD is not an executable file') + self.probackup_version = None + self.old_probackup_version = None + + try: + self.probackup_version_output = subprocess.check_output( + [self.probackup_path, "--version"], + stderr=subprocess.STDOUT, + ).decode('utf-8') + except subprocess.CalledProcessError as e: + raise ProbackupException(e.output.decode('utf-8')) + + if self.probackup_old_path: + old_probackup_version_output = subprocess.check_output( + [self.probackup_old_path, "--version"], + stderr=subprocess.STDOUT, + ).decode('utf-8') + self.old_probackup_version = re.search( + r"\d+\.\d+\.\d+", + subprocess.check_output( + [self.probackup_old_path, "--version"], + stderr=subprocess.STDOUT, + ).decode('utf-8') + ).group(0) + + self.probackup_version = re.search(r"\d+\.\d+\.\d+", self.probackup_version_output).group(0) + self.remote = False self.remote_host = None self.remote_port = None @@ -277,10 +402,54 @@ def __init__(self, *args, **kwargs): self.ptrack = False if 'PG_PROBACKUP_PTRACK' in self.test_env: if self.test_env['PG_PROBACKUP_PTRACK'] == 'ON': - self.ptrack = True + if self.pg_config_version >= self.version_to_num('11.0'): + self.ptrack = True os.environ["PGAPPNAME"] = "pg_probackup" + def is_test_result_ok(test_case): + # sources of solution: + # 1. python versions 2.7 - 3.10, verified on 3.10, 3.7, 2.7, taken from: + # https://tousu.in/qa/?qa=555402/unit-testing-getting-pythons-unittest-results-in-a-teardown-method&show=555403#a555403 + # + # 2. python versions 3.11+ mixin, verified on 3.11, taken from: https://stackoverflow.com/a/39606065 + + if not isinstance(test_case, unittest.TestCase): + raise AssertionError("test_case is not instance of unittest.TestCase") + + if hasattr(test_case, '_outcome'): # Python 3.4+ + if hasattr(test_case._outcome, 'errors'): + # Python 3.4 - 3.10 (These two methods have no side effects) + result = test_case.defaultTestResult() # These two methods have no side effects + test_case._feedErrorsToResult(result, test_case._outcome.errors) + else: + # Python 3.11+ and pytest 5.3.5+ + result = test_case._outcome.result + if not hasattr(result, 'errors'): + result.errors = [] + if not hasattr(result, 'failures'): + result.failures = [] + else: # Python 2.7, 3.0-3.3 + result = getattr(test_case, '_outcomeForDoCleanups', test_case._resultForDoCleanups) + + ok = all(test != test_case for test, text in result.errors + result.failures) + + return ok + + def tearDown(self): + if self.is_test_result_ok(): + for node in self.nodes_to_cleanup: + node.cleanup() + self.del_test_dir(self.module_name, self.fname) + + else: + for node in self.nodes_to_cleanup: + # TODO make decorator with proper stop() vs cleanup() + node._try_shutdown(max_attempts=1) + # node.cleanup() + + self.nodes_to_cleanup.clear() + @property def pg_config_version(self): return self.version_to_num( @@ -304,53 +473,150 @@ def pg_config_version(self): # print('PGPROBACKUP_SSH_USER is not set') # exit(1) + def make_empty_node( + self, + base_dir=None): + real_base_dir = os.path.join(self.tmp_path, base_dir) + shutil.rmtree(real_base_dir, ignore_errors=True) + os.makedirs(real_base_dir) + node = PostgresNodeExtended(base_dir=real_base_dir) + node.should_rm_dirs = True + self.nodes_to_cleanup.append(node) + + return node def make_simple_node( self, base_dir=None, set_replication=False, + ptrack_enable=False, initdb_params=[], pg_options={}): - real_base_dir = os.path.join(self.tmp_path, base_dir) - shutil.rmtree(real_base_dir, ignore_errors=True) - os.makedirs(real_base_dir) - - node = testgres.get_new_node('test', base_dir=real_base_dir) - # bound method slow_start() to 'node' class instance - node.slow_start = slow_start.__get__(node) - node.should_rm_dirs = True + node = self.make_empty_node(base_dir) node.init( initdb_params=initdb_params, allow_streaming=set_replication) - # Sane default parameters - node.append_conf('postgresql.auto.conf', 'max_connections = 100') - node.append_conf('postgresql.auto.conf', 'shared_buffers = 10MB') - node.append_conf('postgresql.auto.conf', 'fsync = off') - node.append_conf('postgresql.auto.conf', 'wal_level = logical') - node.append_conf('postgresql.auto.conf', 'hot_standby = off') - - node.append_conf( - 'postgresql.auto.conf', "log_line_prefix = '%t [%p]: [%l-1] '") - node.append_conf('postgresql.auto.conf', 'log_statement = none') - node.append_conf('postgresql.auto.conf', 'log_duration = on') - node.append_conf( - 'postgresql.auto.conf', 'log_min_duration_statement = 0') - node.append_conf('postgresql.auto.conf', 'log_connections = on') - node.append_conf('postgresql.auto.conf', 'log_disconnections = on') + # set major version + with open(os.path.join(node.data_dir, 'PG_VERSION')) as f: + node.major_version_str = str(f.read().rstrip()) + node.major_version = float(node.major_version_str) - # Apply given parameters - for key, value in six.iteritems(pg_options): - node.append_conf('postgresql.auto.conf', '%s = %s' % (key, value)) + # Sane default parameters + options = {} + options['max_connections'] = 100 + options['shared_buffers'] = '10MB' + options['fsync'] = 'off' + + options['wal_level'] = 'logical' + options['hot_standby'] = 'off' + + options['log_line_prefix'] = '%t [%p]: [%l-1] ' + options['log_statement'] = 'none' + options['log_duration'] = 'on' + options['log_min_duration_statement'] = 0 + options['log_connections'] = 'on' + options['log_disconnections'] = 'on' + options['restart_after_crash'] = 'off' + options['autovacuum'] = 'off' # Allow replication in pg_hba.conf if set_replication: - node.append_conf( - 'postgresql.auto.conf', - 'max_wal_senders = 10') + options['max_wal_senders'] = 10 + + if ptrack_enable: + options['ptrack.map_size'] = '128' + options['shared_preload_libraries'] = 'ptrack' + + if node.major_version >= 13: + options['wal_keep_size'] = '200MB' + else: + options['wal_keep_segments'] = '100' + + # set default values + self.set_auto_conf(node, options) + + # Apply given parameters + self.set_auto_conf(node, pg_options) + + # kludge for testgres + # https://github.com/postgrespro/testgres/issues/54 + # for PG >= 13 remove 'wal_keep_segments' parameter + if node.major_version >= 13: + self.set_auto_conf( + node, {}, 'postgresql.conf', ['wal_keep_segments']) return node + + def simple_bootstrap(self, node, role) -> None: + + node.safe_psql( + 'postgres', + 'CREATE ROLE {0} WITH LOGIN REPLICATION'.format(role)) + + # PG 9.5 + if self.get_version(node) < 90600: + node.safe_psql( + 'postgres', + 'GRANT USAGE ON SCHEMA pg_catalog TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO {0};'.format(role)) + # PG 9.6 + elif self.get_version(node) > 90600 and self.get_version(node) < 100000: + node.safe_psql( + 'postgres', + 'GRANT USAGE ON SCHEMA pg_catalog TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_xlog_replay_location() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_checkpoint() TO {0};'.format(role)) + # >= 10 && < 15 + elif self.get_version(node) >= 100000 and self.get_version(node) < 150000: + node.safe_psql( + 'postgres', + 'GRANT USAGE ON SCHEMA pg_catalog TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_checkpoint() TO {0};'.format(role)) + # >= 15 + else: + node.safe_psql( + 'postgres', + 'GRANT USAGE ON SCHEMA pg_catalog TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_start(text, boolean) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_stop(boolean) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_current() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO {0}; ' + 'GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_checkpoint() TO {0};'.format(role)) def create_tblspace_in_node(self, node, tblspc_name, tblspc_path=None, cfs=False): res = node.execute( @@ -381,6 +647,35 @@ def create_tblspace_in_node(self, node, tblspc_name, tblspc_path=None, cfs=False # res[0], 0, # 'Failed to create tablespace with cmd: {0}'.format(cmd)) + def drop_tblspace(self, node, tblspc_name): + res = node.execute( + 'postgres', + 'select exists' + " (select 1 from pg_tablespace where spcname = '{0}')".format( + tblspc_name) + ) + # Check that tablespace with name 'tblspc_name' do not exists already + self.assertTrue( + res[0][0], + 'Tablespace "{0}" do not exists'.format(tblspc_name) + ) + + rels = node.execute( + "postgres", + "SELECT relname FROM pg_class c " + "LEFT JOIN pg_tablespace t ON c.reltablespace = t.oid " + "where c.relkind = 'r' and t.spcname = '{0}'".format(tblspc_name)) + + for rel in rels: + node.safe_psql( + 'postgres', + "DROP TABLE {0}".format(rel[0])) + + node.safe_psql( + 'postgres', + 'DROP TABLESPACE {0}'.format(tblspc_name)) + + def get_tblspace_path(self, node, tblspc_name): return os.path.join(node.base_dir, tblspc_name) @@ -400,7 +695,8 @@ def get_fork_path(self, node, fork_name): def get_md5_per_page_for_fork(self, file, size_in_pages): pages_per_segment = {} md5_per_page = {} - nsegments = size_in_pages/131072 + size_in_pages = int(size_in_pages) + nsegments = int(size_in_pages/131072) if size_in_pages % 131072 != 0: nsegments = nsegments + 1 @@ -452,10 +748,10 @@ def get_ptrack_bits_per_page_for_fork(self, node, file, size=[]): if os.path.getsize(file) == 0: return ptrack_bits_for_fork byte_size = os.path.getsize(file + '_ptrack') - npages = byte_size/8192 + npages = int(byte_size/8192) if byte_size % 8192 != 0: print('Ptrack page is not 8k aligned') - sys.exit(1) + exit(1) file = os.open(file + '_ptrack', os.O_RDONLY) @@ -477,6 +773,28 @@ def get_ptrack_bits_per_page_for_fork(self, node, file, size=[]): os.close(file) return ptrack_bits_for_fork + def check_ptrack_map_sanity(self, node, idx_ptrack): + success = True + for i in idx_ptrack: + # get new size of heap and indexes. size calculated in pages + idx_ptrack[i]['new_size'] = self.get_fork_size(node, i) + # update path to heap and index files in case they`ve changed + idx_ptrack[i]['path'] = self.get_fork_path(node, i) + # calculate new md5sums for pages + idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) + # get ptrack for every idx + idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( + node, idx_ptrack[i]['path'], + [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + + # compare pages and check ptrack sanity + if not self.check_ptrack_sanity(idx_ptrack[i]): + success = False + + self.assertTrue( + success, 'Ptrack has failed to register changes in data files') + def check_ptrack_sanity(self, idx_dict): success = True if idx_dict['new_size'] > idx_dict['old_size']: @@ -628,7 +946,7 @@ def check_ptrack_clean(self, idx_dict, size): ) ) - def run_pb(self, command, asynchronous=False, gdb=False, old_binary=False, return_id=True): + def run_pb(self, command, asynchronous=False, gdb=False, old_binary=False, return_id=True, env=None): if not self.probackup_old_path and old_binary: print('PGPROBACKUPBIN_OLD is not set') exit(1) @@ -638,24 +956,27 @@ def run_pb(self, command, asynchronous=False, gdb=False, old_binary=False, retur else: binary_path = self.probackup_path + if not env: + env=self.test_env + try: self.cmd = [' '.join(map(str, [binary_path] + command))] if self.verbose: print(self.cmd) if gdb: - return GDBobj([binary_path] + command, self.verbose) + return GDBobj([binary_path] + command, self) if asynchronous: return subprocess.Popen( - self.cmd, + [binary_path] + command, stdout=subprocess.PIPE, stderr=subprocess.PIPE, - env=self.test_env + env=env ) else: self.output = subprocess.check_output( [binary_path] + command, stderr=subprocess.STDOUT, - env=self.test_env + env=env ).decode('utf-8') if command[0] == 'backup' and return_id: # return backup ID @@ -665,9 +986,14 @@ def run_pb(self, command, asynchronous=False, gdb=False, old_binary=False, retur else: return self.output except subprocess.CalledProcessError as e: - raise ProbackupException(e.output.decode('utf-8'), self.cmd) + raise ProbackupException(e.output.decode('utf-8').replace("\r",""), + self.cmd) + + def run_binary(self, command, asynchronous=False, env=None): + + if not env: + env = self.test_env - def run_binary(self, command, asynchronous=False): if self.verbose: print([' '.join(map(str, command))]) try: @@ -677,13 +1003,13 @@ def run_binary(self, command, asynchronous=False): stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE, - env=self.test_env + env=env ) else: self.output = subprocess.check_output( command, stderr=subprocess.STDOUT, - env=self.test_env + env=env ).decode('utf-8') return self.output except subprocess.CalledProcessError as e: @@ -733,6 +1059,22 @@ def set_config(self, backup_dir, instance, old_binary=False, options=[]): return self.run_pb(cmd + options, old_binary=old_binary) + def set_backup(self, backup_dir, instance, backup_id=False, + old_binary=False, options=[]): + + cmd = [ + 'set-backup', + '-B', backup_dir + ] + + if instance: + cmd = cmd + ['--instance={0}'.format(instance)] + + if backup_id: + cmd = cmd + ['-i', backup_id] + + return self.run_pb(cmd + options, old_binary=old_binary) + def del_instance(self, backup_dir, instance, old_binary=False): return self.run_pb([ @@ -750,7 +1092,8 @@ def backup_node( self, backup_dir, instance, node, data_dir=False, backup_type='full', datname=False, options=[], asynchronous=False, gdb=False, - old_binary=False, return_id=True + old_binary=False, return_id=True, no_remote=False, + env=None ): if not node and not data_dir: print('You must provide ether node or data_dir for backup') @@ -772,7 +1115,7 @@ def backup_node( cmd_list += ['-D', data_dir] # don`t forget to kill old_binary after remote ssh release - if self.remote and not old_binary: + if self.remote and not old_binary and not no_remote: options = options + [ '--remote-proto=ssh', '--remote-host=localhost'] @@ -780,11 +1123,14 @@ def backup_node( if backup_type: cmd_list += ['-b', backup_type] - return self.run_pb(cmd_list + options, asynchronous, gdb, old_binary, return_id) + if not old_binary: + cmd_list += ['--no-sync'] + + return self.run_pb(cmd_list + options, asynchronous, gdb, old_binary, return_id, env=env) def checkdb_node( self, backup_dir=False, instance=False, data_dir=False, - options=[], async=False, gdb=False, old_binary=False + options=[], asynchronous=False, gdb=False, old_binary=False ): cmd_list = ["checkdb"] @@ -798,7 +1144,7 @@ def checkdb_node( if data_dir: cmd_list += ["-D", data_dir] - return self.run_pb(cmd_list + options, async, gdb, old_binary) + return self.run_pb(cmd_list + options, asynchronous, gdb, old_binary) def merge_backup( self, backup_dir, instance, backup_id, asynchronous=False, @@ -814,7 +1160,8 @@ def merge_backup( def restore_node( self, backup_dir, instance, node=False, - data_dir=None, backup_id=None, old_binary=False, options=[] + data_dir=None, backup_id=None, old_binary=False, options=[], + gdb=False ): if data_dir is None: @@ -836,11 +1183,37 @@ def restore_node( if backup_id: cmd_list += ['-i', backup_id] - return self.run_pb(cmd_list + options, old_binary=old_binary) + if not old_binary: + cmd_list += ['--no-sync'] + + return self.run_pb(cmd_list + options, gdb=gdb, old_binary=old_binary) + + def catchup_node( + self, + backup_mode, source_pgdata, destination_node, + options = [] + ): + + cmd_list = [ + 'catchup', + '--backup-mode={0}'.format(backup_mode), + '--source-pgdata={0}'.format(source_pgdata), + '--destination-pgdata={0}'.format(destination_node.data_dir) + ] + if self.remote: + cmd_list += ['--remote-proto=ssh', '--remote-host=localhost'] + if self.verbose: + cmd_list += [ + '--log-level-file=VERBOSE', + '--log-directory={0}'.format(destination_node.logs_dir) + ] + + return self.run_pb(cmd_list + options) def show_pb( self, backup_dir, instance=None, backup_id=None, - options=[], as_text=False, as_json=True, old_binary=False + options=[], as_text=False, as_json=True, old_binary=False, + env=None ): backup_list = [] @@ -861,7 +1234,7 @@ def show_pb( if as_text: # You should print it when calling as_text=true - return self.run_pb(cmd_list + options, old_binary=old_binary) + return self.run_pb(cmd_list + options, old_binary=old_binary, env=env) # get show result as list of lines if as_json: @@ -879,10 +1252,14 @@ def show_pb( return backup else: backup_list.append(backup) + + if backup_id is not None: + self.assertTrue(False, "Failed to find backup with ID: {0}".format(backup_id)) + return backup_list else: show_splitted = self.run_pb( - cmd_list + options, old_binary=old_binary).splitlines() + cmd_list + options, old_binary=old_binary, env=env).splitlines() if instance is not None and backup_id is None: # cut header(ID, Mode, etc) from show as single string header = show_splitted[1:2][0] @@ -933,11 +1310,68 @@ def show_pb( var = var.strip('"') var = var.strip("'") specific_record[name.strip()] = var + + if not specific_record: + self.assertTrue(False, "Failed to find backup with ID: {0}".format(backup_id)) + return specific_record + def show_archive( + self, backup_dir, instance=None, options=[], + as_text=False, as_json=True, old_binary=False, + tli=0 + ): + + cmd_list = [ + 'show', + '--archive', + '-B', backup_dir, + ] + if instance: + cmd_list += ['--instance={0}'.format(instance)] + + # AHTUNG, WARNING will break json parsing + if as_json: + cmd_list += ['--format=json', '--log-level-console=error'] + + if as_text: + # You should print it when calling as_text=true + return self.run_pb(cmd_list + options, old_binary=old_binary) + + if as_json: + if as_text: + data = self.run_pb(cmd_list + options, old_binary=old_binary) + else: + data = json.loads(self.run_pb(cmd_list + options, old_binary=old_binary)) + + if instance: + instance_timelines = None + for instance_name in data: + if instance_name['instance'] == instance: + instance_timelines = instance_name['timelines'] + break + + if tli > 0: + timeline_data = None + for timeline in instance_timelines: + if timeline['tli'] == tli: + return timeline + + return {} + + if instance_timelines: + return instance_timelines + + return data + else: + show_splitted = self.run_pb( + cmd_list + options, old_binary=old_binary).splitlines() + print(show_splitted) + exit(1) + def validate_pb( - self, backup_dir, instance=None, - backup_id=None, options=[], old_binary=False, gdb=False + self, backup_dir, instance=None, backup_id=None, + options=[], old_binary=False, gdb=False, asynchronous=False ): cmd_list = [ @@ -949,11 +1383,11 @@ def validate_pb( if backup_id: cmd_list += ['-i', backup_id] - return self.run_pb(cmd_list + options, old_binary=old_binary, gdb=gdb) + return self.run_pb(cmd_list + options, old_binary=old_binary, gdb=gdb, asynchronous=asynchronous) def delete_pb( - self, backup_dir, instance, - backup_id=None, options=[], old_binary=False): + self, backup_dir, instance, backup_id=None, + options=[], old_binary=False, gdb=False, asynchronous=False): cmd_list = [ 'delete', '-B', backup_dir @@ -963,7 +1397,7 @@ def delete_pb( if backup_id: cmd_list += ['-i', backup_id] - return self.run_pb(cmd_list + options, old_binary=old_binary) + return self.run_pb(cmd_list + options, old_binary=old_binary, gdb=gdb, asynchronous=asynchronous) def delete_expired( self, backup_dir, instance, options=[], old_binary=False): @@ -991,8 +1425,16 @@ def show_config(self, backup_dir, instance, old_binary=False): def get_recovery_conf(self, node): out_dict = {} + + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_conf_path = os.path.join(node.data_dir, 'postgresql.auto.conf') + with open(recovery_conf_path, 'r') as f: + print(f.read()) + else: + recovery_conf_path = os.path.join(node.data_dir, 'recovery.conf') + with open( - os.path.join(node.data_dir, 'recovery.conf'), 'r' + recovery_conf_path, 'r' ) as recovery_conf: for line in recovery_conf: try: @@ -1004,73 +1446,185 @@ def get_recovery_conf(self, node): def set_archiving( self, backup_dir, instance, node, replica=False, - overwrite=False, compress=False, old_binary=False): + overwrite=False, compress=True, old_binary=False, + log_level=False, archive_timeout=False, + custom_archive_command=None): + # parse postgresql.auto.conf + options = {} if replica: - archive_mode = 'always' - node.append_conf('postgresql.auto.conf', 'hot_standby = on') + options['archive_mode'] = 'always' + options['hot_standby'] = 'on' else: - archive_mode = 'on' + options['archive_mode'] = 'on' - node.append_conf( - 'postgresql.auto.conf', - 'archive_mode = {0}'.format(archive_mode) - ) + if custom_archive_command is None: + if os.name == 'posix': + options['archive_command'] = '"{0}" archive-push -B {1} --instance={2} '.format( + self.probackup_path, backup_dir, instance) + + elif os.name == 'nt': + options['archive_command'] = '"{0}" archive-push -B {1} --instance={2} '.format( + self.probackup_path.replace("\\","\\\\"), + backup_dir.replace("\\","\\\\"), instance) + + # don`t forget to kill old_binary after remote ssh release + if self.remote and not old_binary: + options['archive_command'] += '--remote-proto=ssh ' + options['archive_command'] += '--remote-host=localhost ' + + if self.archive_compress and compress: + options['archive_command'] += '--compress ' + + if overwrite: + options['archive_command'] += '--overwrite ' + + options['archive_command'] += '--log-level-console=VERBOSE ' + options['archive_command'] += '-j 5 ' + options['archive_command'] += '--batch-size 10 ' + options['archive_command'] += '--no-sync ' + + if archive_timeout: + options['archive_command'] += '--archive-timeout={0} '.format( + archive_timeout) + + if os.name == 'posix': + options['archive_command'] += '--wal-file-path=%p --wal-file-name=%f' + + elif os.name == 'nt': + options['archive_command'] += '--wal-file-path="%p" --wal-file-name="%f"' + + if log_level: + options['archive_command'] += ' --log-level-console={0}'.format(log_level) + options['archive_command'] += ' --log-level-file={0} '.format(log_level) + else: # custom_archive_command is not None + options['archive_command'] = custom_archive_command + + self.set_auto_conf(node, options) + + def get_restore_command(self, backup_dir, instance, node): + + # parse postgresql.auto.conf + restore_command = '' if os.name == 'posix': - archive_command = '"{0}" archive-push -B {1} --instance={2} '.format( + restore_command += '{0} archive-get -B {1} --instance={2} '.format( self.probackup_path, backup_dir, instance) elif os.name == 'nt': - archive_command = '"{0}" archive-push -B {1} --instance={2} '.format( + restore_command += '"{0}" archive-get -B {1} --instance={2} '.format( self.probackup_path.replace("\\","\\\\"), - backup_dir.replace("\\","\\\\"), - instance) + backup_dir.replace("\\","\\\\"), instance) # don`t forget to kill old_binary after remote ssh release - if self.remote and not old_binary: - archive_command = archive_command + '--remote-proto=ssh --remote-host=localhost ' - - if self.archive_compress or compress: - archive_command = archive_command + '--compress ' - - if overwrite: - archive_command = archive_command + '--overwrite ' + if self.remote: + restore_command += '--remote-proto=ssh ' + restore_command += '--remote-host=localhost ' if os.name == 'posix': - archive_command = archive_command + '--wal-file-path %p --wal-file-name %f' + restore_command += '--wal-file-path=%p --wal-file-name=%f' elif os.name == 'nt': - archive_command = archive_command + '--wal-file-path "%p" --wal-file-name "%f"' + restore_command += '--wal-file-path="%p" --wal-file-name="%f"' + + return restore_command + + # rm_options - list of parameter name that should be deleted from current config, + # example: ['wal_keep_segments', 'max_wal_size'] + def set_auto_conf(self, node, options, config='postgresql.auto.conf', rm_options={}): + + # parse postgresql.auto.conf + path = os.path.join(node.data_dir, config) + + with open(path, 'r') as f: + raw_content = f.read() - node.append_conf( - 'postgresql.auto.conf', - "archive_command = '{0}'".format( - archive_command)) + current_options = {} + current_directives = [] + for line in raw_content.splitlines(): + + # ignore comments + if line.startswith('#'): + continue + + if line == '': + continue + + if line.startswith('include'): + current_directives.append(line) + continue + + name, var = line.partition('=')[::2] + name = name.strip() + var = var.strip() + var = var.strip('"') + var = var.strip("'") + + # remove options specified in rm_options list + if name in rm_options: + continue + + current_options[name] = var + + for option in options: + current_options[option] = options[option] + + auto_conf = '' + for option in current_options: + auto_conf += "{0} = '{1}'\n".format( + option, current_options[option]) + + for directive in current_directives: + auto_conf += directive + "\n" + + with open(path, 'wt') as f: + f.write(auto_conf) + f.flush() + f.close() def set_replica( self, master, replica, replica_name='replica', - synchronous=False + synchronous=False, + log_shipping=False ): - replica.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(replica.port)) - replica.append_conf('postgresql.auto.conf', 'hot_standby = on') - replica.append_conf('recovery.conf', 'standby_mode = on') - replica.append_conf( - 'recovery.conf', - "primary_conninfo = 'user={0} port={1} application_name={2}" - " sslmode=prefer sslcompression=1'".format( - self.user, master.port, replica_name) - ) + + self.set_auto_conf( + replica, + options={ + 'port': replica.port, + 'hot_standby': 'on'}) + + if self.get_version(replica) >= self.version_to_num('12.0'): + with open(os.path.join(replica.data_dir, "standby.signal"), 'w') as f: + f.flush() + f.close() + + config = 'postgresql.auto.conf' + + if not log_shipping: + self.set_auto_conf( + replica, + {'primary_conninfo': 'user={0} port={1} application_name={2} ' + ' sslmode=prefer sslcompression=1'.format( + self.user, master.port, replica_name)}, + config) + else: + replica.append_conf('recovery.conf', 'standby_mode = on') + + if not log_shipping: + replica.append_conf( + 'recovery.conf', + "primary_conninfo = 'user={0} port={1} application_name={2}" + " sslmode=prefer sslcompression=1'".format( + self.user, master.port, replica_name)) + if synchronous: - master.append_conf( - 'postgresql.auto.conf', - "synchronous_standby_names='{0}'".format(replica_name) - ) - master.append_conf( - 'postgresql.auto.conf', - "synchronous_commit='remote_apply'" - ) + self.set_auto_conf( + master, + options={ + 'synchronous_standby_names': replica_name, + 'synchronous_commit': 'remote_apply'}) + master.reload() def change_backup_status(self, backup_dir, instance, backup_id, status): @@ -1144,7 +1698,7 @@ def version_to_num(self, version): parts.append('0') num = 0 for part in parts: - num = num * 100 + int(re.sub("[^\d]", "", part)) + num = num * 100 + int(re.sub(r"[^\d]", "", part)) return num def switch_wal_segment(self, node): @@ -1156,7 +1710,7 @@ def switch_wal_segment(self, node): """ if isinstance(node, testgres.PostgresNode): if self.version_to_num( - node.safe_psql('postgres', 'show server_version') + node.safe_psql('postgres', 'show server_version').decode('utf-8') ) >= self.version_to_num('10.0'): node.safe_psql('postgres', 'select pg_switch_wal()') else: @@ -1173,10 +1727,11 @@ def switch_wal_segment(self, node): def wait_until_replica_catch_with_master(self, master, replica): - if self.version_to_num( - master.safe_psql( - 'postgres', - 'show server_version')) >= self.version_to_num('10.0'): + version = master.safe_psql( + 'postgres', + 'show server_version').decode('utf-8').rstrip() + + if self.version_to_num(version) >= self.version_to_num('10.0'): master_function = 'pg_catalog.pg_current_wal_lsn()' replica_function = 'pg_catalog.pg_last_wal_replay_lsn()' else: @@ -1185,7 +1740,7 @@ def wait_until_replica_catch_with_master(self, master, replica): lsn = master.safe_psql( 'postgres', - 'SELECT {0}'.format(master_function)).rstrip() + 'SELECT {0}'.format(master_function)).decode('utf-8').rstrip() # Wait until replica catch up with master replica.poll_query_until( @@ -1196,15 +1751,18 @@ def get_version(self, node): return self.version_to_num( testgres.get_pg_config()['VERSION'].split(" ")[1]) + def get_ptrack_version(self, node): + version = node.safe_psql( + "postgres", + "SELECT extversion " + "FROM pg_catalog.pg_extension WHERE extname = 'ptrack'").decode('utf-8').rstrip() + return self.version_to_num(version) + def get_bin_path(self, binary): return testgres.get_bin_path(binary) def del_test_dir(self, module_name, fname): """ Del testdir and optimistically try to del module dir""" - try: - testgres.clean_all() - except: - pass shutil.rmtree( os.path.join( @@ -1214,10 +1772,6 @@ def del_test_dir(self, module_name, fname): ), ignore_errors=True ) - try: - os.rmdir(os.path.join(self.tmp_path, module_name)) - except: - pass def pgdata_content(self, pgdata, ignore_ptrack=True, exclude_dirs=None): """ return dict with directory content. " @@ -1230,7 +1784,10 @@ def pgdata_content(self, pgdata, ignore_ptrack=True, exclude_dirs=None): 'postmaster.pid', 'postmaster.opts', 'pg_internal.init', 'postgresql.auto.conf', 'backup_label', 'tablespace_map', 'recovery.conf', - 'ptrack_control', 'ptrack_init', 'pg_control' + 'ptrack_control', 'ptrack_init', 'pg_control', + 'probackup_recovery.conf', 'recovery.signal', + 'standby.signal', 'ptrack.map', 'ptrack.map.mmap', + 'ptrack.map.tmp', 'recovery.done','backup_label.old' ] if exclude_dirs: @@ -1253,194 +1810,225 @@ def pgdata_content(self, pgdata, ignore_ptrack=True, exclude_dirs=None): file_fullpath = os.path.join(root, file) file_relpath = os.path.relpath(file_fullpath, pgdata) - directory_dict['files'][file_relpath] = {'is_datafile': False} - directory_dict['files'][file_relpath]['md5'] = hashlib.md5( - open(file_fullpath, 'rb').read()).hexdigest() + cfile = ContentFile(file.isdigit()) + directory_dict['files'][file_relpath] = cfile + with open(file_fullpath, 'rb') as f: + # truncate cfm's content's zero tail + if file_relpath.endswith('.cfm'): + content = f.read() + zero64 = b"\x00"*64 + l = len(content) + while l > 64: + s = (l - 1) & ~63 + if content[s:l] != zero64[:l-s]: + break + l = s + content = content[:l] + digest = hashlib.md5(content) + else: + digest = hashlib.md5() + while True: + b = f.read(64*1024) + if not b: break + digest.update(b) + cfile.md5 = digest.hexdigest() # crappy algorithm - if file.isdigit(): - directory_dict['files'][file_relpath]['is_datafile'] = True + if cfile.is_datafile: size_in_pages = os.path.getsize(file_fullpath)/8192 - directory_dict['files'][file_relpath][ - 'md5_per_page'] = self.get_md5_per_page_for_fork( + cfile.md5_per_page = self.get_md5_per_page_for_fork( file_fullpath, size_in_pages ) - for root, dirs, files in os.walk(pgdata, topdown=False, followlinks=True): for directory in dirs: directory_path = os.path.join(root, directory) directory_relpath = os.path.relpath(directory_path, pgdata) - - found = False - for d in dirs_to_ignore: - if d in directory_relpath: - found = True - break - - # check if directory already here as part of larger directory - if not found: - for d in directory_dict['dirs']: - # print("OLD dir {0}".format(d)) - if directory_relpath in d: - found = True - break - - if not found: - directory_dict['dirs'][directory_relpath] = {} + parent = os.path.dirname(directory_relpath) + if parent in directory_dict['dirs']: + del directory_dict['dirs'][parent] + directory_dict['dirs'][directory_relpath] = ContentDir() # get permissions for every file and directory - for file in directory_dict['dirs']: - full_path = os.path.join(pgdata, file) - directory_dict['dirs'][file]['mode'] = os.stat( - full_path).st_mode + for dir, cdir in directory_dict['dirs'].items(): + full_path = os.path.join(pgdata, dir) + cdir.mode = os.stat(full_path).st_mode - for file in directory_dict['files']: + for file, cfile in directory_dict['files'].items(): full_path = os.path.join(pgdata, file) - directory_dict['files'][file]['mode'] = os.stat( - full_path).st_mode + cfile.mode = os.stat(full_path).st_mode return directory_dict - def compare_pgdata(self, original_pgdata, restored_pgdata): - """ return dict with directory content. DO IT BEFORE RECOVERY""" - fail = False - error_message = 'Restored PGDATA is not equal to original!\n' - - # Compare directories - for directory in restored_pgdata['dirs']: - if directory not in original_pgdata['dirs']: - fail = True - error_message += '\nDirectory was not present' - error_message += ' in original PGDATA: {0}\n'.format( - os.path.join(restored_pgdata['pgdata'], directory)) - else: - if ( - restored_pgdata['dirs'][directory]['mode'] != - original_pgdata['dirs'][directory]['mode'] - ): - fail = True - error_message += '\nDir permissions mismatch:\n' - error_message += ' Dir old: {0} Permissions: {1}\n'.format( - os.path.join(original_pgdata['pgdata'], directory), - original_pgdata['dirs'][directory]['mode']) - error_message += ' Dir new: {0} Permissions: {1}\n'.format( - os.path.join(restored_pgdata['pgdata'], directory), - restored_pgdata['dirs'][directory]['mode']) + def get_known_bugs_comparision_exclusion_dict(self, node): + """ get dict of known datafiles difference, that can be used in compare_pgdata() """ + comparision_exclusion_dict = dict() + # bug in spgist metapage update (PGPRO-5707) + spgist_filelist = node.safe_psql( + "postgres", + "SELECT pg_catalog.pg_relation_filepath(pg_class.oid) " + "FROM pg_am, pg_class " + "WHERE pg_am.amname = 'spgist' " + "AND pg_class.relam = pg_am.oid" + ).decode('utf-8').rstrip().splitlines() + for filename in spgist_filelist: + comparision_exclusion_dict[filename] = set([0]) + return comparision_exclusion_dict - for directory in original_pgdata['dirs']: - if directory not in restored_pgdata['dirs']: - fail = True - error_message += '\nDirectory dissappeared' - error_message += ' in restored PGDATA: {0}\n'.format( - os.path.join(restored_pgdata['pgdata'], directory)) + def compare_pgdata(self, original_pgdata, restored_pgdata, exclusion_dict = dict()): + """ + return dict with directory content. DO IT BEFORE RECOVERY + exclusion_dict is used for exclude files (and it block_no) from comparision + it is a dict with relative filenames as keys and set of block numbers as values + """ + fail = False + error_message = 'Restored PGDATA is not equal to original!\n' - for file in restored_pgdata['files']: + # Compare directories + restored_dirs = set(restored_pgdata['dirs']) + original_dirs = set(original_pgdata['dirs']) + + for directory in sorted(restored_dirs - original_dirs): + fail = True + error_message += '\nDirectory was not present' + error_message += ' in original PGDATA: {0}\n'.format( + os.path.join(restored_pgdata['pgdata'], directory)) + + for directory in sorted(original_dirs - restored_dirs): + fail = True + error_message += '\nDirectory dissappeared' + error_message += ' in restored PGDATA: {0}\n'.format( + os.path.join(restored_pgdata['pgdata'], directory)) + + for directory in sorted(original_dirs & restored_dirs): + original = original_pgdata['dirs'][directory] + restored = restored_pgdata['dirs'][directory] + if original.mode != restored.mode: + fail = True + error_message += '\nDir permissions mismatch:\n' + error_message += ' Dir old: {0} Permissions: {1}\n'.format( + os.path.join(original_pgdata['pgdata'], directory), + original.mode) + error_message += ' Dir new: {0} Permissions: {1}\n'.format( + os.path.join(restored_pgdata['pgdata'], directory), + restored.mode) + + restored_files = set(restored_pgdata['files']) + original_files = set(original_pgdata['files']) + + for file in sorted(restored_files - original_files): # File is present in RESTORED PGDATA # but not present in ORIGINAL # only backup_label is allowed - if file not in original_pgdata['files']: - fail = True - error_message += '\nFile is not present' - error_message += ' in original PGDATA: {0}\n'.format( - os.path.join(restored_pgdata['pgdata'], file)) - - for file in original_pgdata['files']: - if file in restored_pgdata['files']: - - if ( - restored_pgdata['files'][file]['mode'] != - original_pgdata['files'][file]['mode'] - ): - fail = True - error_message += '\nFile permissions mismatch:\n' - error_message += ' File_old: {0} Permissions: {1}\n'.format( - os.path.join(original_pgdata['pgdata'], file), - original_pgdata['files'][file]['mode']) - error_message += ' File_new: {0} Permissions: {1}\n'.format( - os.path.join(restored_pgdata['pgdata'], file), - restored_pgdata['files'][file]['mode']) + fail = True + error_message += '\nFile is not present' + error_message += ' in original PGDATA: {0}\n'.format( + os.path.join(restored_pgdata['pgdata'], file)) + + for file in sorted(original_files - restored_files): + error_message += ( + '\nFile disappearance.\n ' + 'File: {0}\n').format( + os.path.join(restored_pgdata['pgdata'], file) + ) + fail = True - if ( - original_pgdata['files'][file]['md5'] != - restored_pgdata['files'][file]['md5'] - ): + for file in sorted(original_files & restored_files): + original = original_pgdata['files'][file] + restored = restored_pgdata['files'][file] + if restored.mode != original.mode: + fail = True + error_message += '\nFile permissions mismatch:\n' + error_message += ' File_old: {0} Permissions: {1:o}\n'.format( + os.path.join(original_pgdata['pgdata'], file), + original.mode) + error_message += ' File_new: {0} Permissions: {1:o}\n'.format( + os.path.join(restored_pgdata['pgdata'], file), + restored.mode) + + if original.md5 != restored.md5: + if file not in exclusion_dict: fail = True error_message += ( - '\nFile Checksumm mismatch.\n' - 'File_old: {0}\nChecksumm_old: {1}\n' - 'File_new: {2}\nChecksumm_new: {3}\n').format( + '\nFile Checksum mismatch.\n' + 'File_old: {0}\nChecksum_old: {1}\n' + 'File_new: {2}\nChecksum_new: {3}\n').format( os.path.join(original_pgdata['pgdata'], file), - original_pgdata['files'][file]['md5'], + original.md5, os.path.join(restored_pgdata['pgdata'], file), - restored_pgdata['files'][file]['md5'] + restored.md5 ) - if original_pgdata['files'][file]['is_datafile']: - for page in original_pgdata['files'][file]['md5_per_page']: - if page not in restored_pgdata['files'][file]['md5_per_page']: - error_message += ( - '\n Page {0} dissappeared.\n ' - 'File: {1}\n').format( - page, - os.path.join( - restored_pgdata['pgdata'], - file - ) - ) - continue - - if original_pgdata['files'][file][ - 'md5_per_page'][page] != restored_pgdata[ - 'files'][file]['md5_per_page'][page]: - error_message += ( - '\n Page checksumm mismatch: {0}\n ' - ' PAGE Checksumm_old: {1}\n ' - ' PAGE Checksumm_new: {2}\n ' - ' File: {3}\n' - ).format( - page, - original_pgdata['files'][file][ - 'md5_per_page'][page], - restored_pgdata['files'][file][ - 'md5_per_page'][page], - os.path.join( - restored_pgdata['pgdata'], file) - ) - for page in restored_pgdata['files'][file]['md5_per_page']: - if page not in original_pgdata['files'][file]['md5_per_page']: - error_message += '\n Extra page {0}\n File: {1}\n'.format( - page, - os.path.join( - restored_pgdata['pgdata'], file)) + if not original.is_datafile: + continue + + original_pages = set(original.md5_per_page) + restored_pages = set(restored.md5_per_page) - else: - error_message += ( - '\nFile disappearance.\n ' - 'File: {0}\n').format( - os.path.join(restored_pgdata['pgdata'], file) + for page in sorted(original_pages - restored_pages): + error_message += '\n Page {0} dissappeared.\n File: {1}\n'.format( + page, + os.path.join(restored_pgdata['pgdata'], file) ) - fail = True + + + for page in sorted(restored_pages - original_pages): + error_message += '\n Extra page {0}\n File: {1}\n'.format( + page, + os.path.join(restored_pgdata['pgdata'], file)) + + for page in sorted(original_pages & restored_pages): + if file in exclusion_dict and page in exclusion_dict[file]: + continue + + if original.md5_per_page[page] != restored.md5_per_page[page]: + fail = True + error_message += ( + '\n Page checksum mismatch: {0}\n ' + ' PAGE Checksum_old: {1}\n ' + ' PAGE Checksum_new: {2}\n ' + ' File: {3}\n' + ).format( + page, + original.md5_per_page[page], + restored.md5_per_page[page], + os.path.join( + restored_pgdata['pgdata'], file) + ) + self.assertFalse(fail, error_message) def gdb_attach(self, pid): - return GDBobj([str(pid)], self.verbose, attach=True) + return GDBobj([str(pid)], self, attach=True) + + def _check_gdb_flag_or_skip_test(self): + if not self.gdb: + self.skipTest( + "Specify PGPROBACKUP_GDB and build without " + "optimizations for run this test" + ) class GdbException(Exception): - def __init__(self, message=False): + def __init__(self, message="False"): self.message = message def __str__(self): return '\n ERROR: {0}\n'.format(repr(self.message)) -class GDBobj(ProbackupTest): - def __init__(self, cmd, verbose, attach=False): - self.verbose = verbose +class GDBobj: + def __init__(self, cmd, env, attach=False): + self.verbose = env.verbose + self.output = '' + # Check gdb flag is set up + if not env.gdb: + raise GdbException("No `PGPROBACKUP_GDB=on` is set, " + "test should call ProbackupTest::check_gdb_flag_or_skip_test() on its start " + "and be skipped") # Check gdb presense try: gdb_version, _ = subprocess.Popen( @@ -1463,7 +2051,7 @@ def __init__(self, cmd, verbose, attach=False): # Get version gdb_version_number = re.search( - b"^GNU gdb [^\d]*(\d+)\.(\d)", + br"^GNU gdb [^\d]*(\d+)\.(\d)", gdb_version) self.major_version = int(gdb_version_number.group(1)) self.minor_version = int(gdb_version_number.group(2)) @@ -1477,14 +2065,13 @@ def __init__(self, cmd, verbose, attach=False): stdout=subprocess.PIPE, stderr=subprocess.STDOUT, bufsize=0, - universal_newlines=True + text=True, + errors='replace', ) self.gdb_pid = self.proc.pid - # discard data from pipe, - # is there a way to do it a less derpy way? while True: - line = self.proc.stdout.readline() + line = self.get_line() if 'No such process' in line: raise GdbException(line) @@ -1494,6 +2081,15 @@ def __init__(self, cmd, verbose, attach=False): else: break + def get_line(self): + line = self.proc.stdout.readline() + self.output += line + return line + + def kill(self): + self.proc.kill() + self.proc.wait() + def set_breakpoint(self, location): result = self._execute('break ' + location) @@ -1585,6 +2181,9 @@ def continue_execution_until_error(self): return if line.startswith('*stopped,reason="exited'): return + if line.startswith( + '*stopped,reason="signal-received",signal-name="SIGABRT"'): + return raise GdbException( 'Failed to continue execution until error.\n') @@ -1608,16 +2207,17 @@ def continue_execution_until_break(self, ignore_count=0): 'Failed to continue execution until break.\n') def stopped_in_breakpoint(self): - output = [] while True: - line = self.proc.stdout.readline() - output += [line] + line = self.get_line() if self.verbose: print(line) if line.startswith('*stopped,reason="breakpoint-hit"'): return True return False + def quit(self): + self.proc.terminate() + # use for breakpoint, run, continue def _execute(self, cmd, running=True): output = [] @@ -1628,7 +2228,7 @@ def _execute(self, cmd, running=True): # look for command we just send while True: - line = self.proc.stdout.readline() + line = self.get_line() if self.verbose: print(repr(line)) @@ -1638,7 +2238,7 @@ def _execute(self, cmd, running=True): break while True: - line = self.proc.stdout.readline() + line = self.get_line() output += [line] if self.verbose: print(repr(line)) @@ -1650,3 +2250,10 @@ def _execute(self, cmd, running=True): # if running and line.startswith('*running'): break return output +class ContentFile(object): + __slots__ = ('is_datafile', 'mode', 'md5', 'md5_per_page') + def __init__(self, is_datafile: bool): + self.is_datafile = is_datafile + +class ContentDir(object): + __slots__ = ('mode') diff --git a/tests/incr_restore_test.py b/tests/incr_restore_test.py new file mode 100644 index 000000000..6a2164098 --- /dev/null +++ b/tests/incr_restore_test.py @@ -0,0 +1,2632 @@ +import os +import unittest +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +import subprocess +from datetime import datetime +import sys +from time import sleep +from datetime import datetime, timedelta +import hashlib +import shutil +import json +from testgres import QueryException, StartNodeException +import stat +from stat import S_ISDIR + +class IncrRestoreTest(ProbackupTest, unittest.TestCase): + + # @unittest.skip("skip") + def test_basic_incr_restore(self): + """incremental restore in CHECKSUM mode""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=50) + + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1']) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + backup_id = self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgdata = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + node.stop() + + self.restore_node( + backup_dir, 'node', node, + options=["-j", "4", "--incremental-mode=checksum"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_basic_incr_restore_into_missing_directory(self): + """""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=10) + + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1']) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgdata = self.pgdata_content(node.data_dir) + + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=["-j", "4", "--incremental-mode=checksum"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_checksum_corruption_detection(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=10) + + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1']) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + backup_id = self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgdata = self.pgdata_content(node.data_dir) + + node.stop() + + self.restore_node( + backup_dir, 'node', node, + options=["-j", "4", "--incremental-mode=lsn"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_incr_restore_with_tablespace(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + tblspace = self.get_tblspace_path(node, 'tblspace') + some_directory = self.get_tblspace_path(node, 'some_directory') + + # stuff new destination with garbage + self.restore_node(backup_dir, 'node', node, data_dir=some_directory) + + self.create_tblspace_in_node(node, 'tblspace') + node.pgbench_init(scale=10, tablespace='tblspace') + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + pgdata = self.pgdata_content(node.data_dir) + + node.stop() + + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", "--incremental-mode=checksum", "--force", + "-T{0}={1}".format(tblspace, some_directory)]) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_incr_restore_with_tablespace_1(self): + """recovery to target timeline""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + set_replication=True) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + tblspace = self.get_tblspace_path(node, 'tblspace') + some_directory = self.get_tblspace_path(node, 'some_directory') + + self.restore_node(backup_dir, 'node', node, data_dir=some_directory) + + self.create_tblspace_in_node(node, 'tblspace') + node.pgbench_init(scale=10, tablespace='tblspace') + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, backup_type='delta', options=['--stream']) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node( + backup_dir, 'node', node, backup_type='delta', options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + node.stop() + + self.restore_node( + backup_dir, 'node', node, + options=["-j", "4", "--incremental-mode=checksum"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_incr_restore_with_tablespace_2(self): + """ + If "--tablespace-mapping" option is used with incremental restore, + then new directory must be empty. + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + set_replication=True) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + node_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_1')) + + # fill node1 with data + out = self.restore_node( + backup_dir, 'node', node, + data_dir=node_1.data_dir, + options=['--incremental-mode=checksum', '--force']) + + self.assertIn("WARNING: Backup catalog was initialized for system id", out) + + tblspace = self.get_tblspace_path(node, 'tblspace') + self.create_tblspace_in_node(node, 'tblspace') + node.pgbench_init(scale=5, tablespace='tblspace') + + node.safe_psql( + 'postgres', + 'vacuum') + + self.backup_node(backup_dir, 'node', node, backup_type='delta', options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + try: + self.restore_node( + backup_dir, 'node', node, + data_dir=node_1.data_dir, + options=['--incremental-mode=checksum', '-T{0}={1}'.format(tblspace, tblspace)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because remapped directory is not empty.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Remapped tablespace destination is not empty', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + out = self.restore_node( + backup_dir, 'node', node, + data_dir=node_1.data_dir, + options=[ + '--force', '--incremental-mode=checksum', + '-T{0}={1}'.format(tblspace, tblspace)]) + + pgdata_restored = self.pgdata_content(node_1.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_incr_restore_with_tablespace_3(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.create_tblspace_in_node(node, 'tblspace1') + node.pgbench_init(scale=10, tablespace='tblspace1') + + # take backup with tblspace1 + self.backup_node(backup_dir, 'node', node, options=['--stream']) + pgdata = self.pgdata_content(node.data_dir) + + self.drop_tblspace(node, 'tblspace1') + + self.create_tblspace_in_node(node, 'tblspace2') + node.pgbench_init(scale=10, tablespace='tblspace2') + + node.stop() + + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", + "--incremental-mode=checksum"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_incr_restore_with_tablespace_4(self): + """ + Check that system ID mismatch is detected, + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.create_tblspace_in_node(node, 'tblspace1') + node.pgbench_init(scale=10, tablespace='tblspace1') + + # take backup of node1 with tblspace1 + self.backup_node(backup_dir, 'node', node, options=['--stream']) + pgdata = self.pgdata_content(node.data_dir) + + self.drop_tblspace(node, 'tblspace1') + node.cleanup() + + # recreate node + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + node.slow_start() + + self.create_tblspace_in_node(node, 'tblspace1') + node.pgbench_init(scale=10, tablespace='tblspace1') + node.stop() + + try: + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", + "--incremental-mode=checksum"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because destination directory has wrong system id.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'WARNING: Backup catalog was initialized for system id', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + self.assertIn( + 'ERROR: Incremental restore is not allowed', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + out = self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", "--force", + "--incremental-mode=checksum"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.expectedFailure + @unittest.skip("skip") + def test_incr_restore_with_tablespace_5(self): + """ + More complicated case, we restore backup + with tablespace, which we remap into directory + with some old content, that belongs to an instance + with different system id. + """ + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node1) + node1.slow_start() + + self.create_tblspace_in_node(node1, 'tblspace') + node1.pgbench_init(scale=10, tablespace='tblspace') + + # take backup of node1 with tblspace + self.backup_node(backup_dir, 'node', node1, options=['--stream']) + pgdata = self.pgdata_content(node1.data_dir) + + node1.stop() + + # recreate node + node2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node2'), + set_replication=True, + initdb_params=['--data-checksums']) + node2.slow_start() + + self.create_tblspace_in_node(node2, 'tblspace') + node2.pgbench_init(scale=10, tablespace='tblspace') + node2.stop() + + tblspc1_path = self.get_tblspace_path(node1, 'tblspace') + tblspc2_path = self.get_tblspace_path(node2, 'tblspace') + + out = self.restore_node( + backup_dir, 'node', node1, + options=[ + "-j", "4", "--force", + "--incremental-mode=checksum", + "-T{0}={1}".format(tblspc1_path, tblspc2_path)]) + + # check that tblspc1_path is empty + self.assertFalse( + os.listdir(tblspc1_path), + "Dir is not empty: '{0}'".format(tblspc1_path)) + + pgdata_restored = self.pgdata_content(node1.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_incr_restore_with_tablespace_6(self): + """ + Empty pgdata, not empty tablespace + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.create_tblspace_in_node(node, 'tblspace') + node.pgbench_init(scale=10, tablespace='tblspace') + + # take backup of node with tblspace + self.backup_node(backup_dir, 'node', node, options=['--stream']) + pgdata = self.pgdata_content(node.data_dir) + + node.cleanup() + + try: + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", + "--incremental-mode=checksum"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because there is running postmaster " + "process in destination directory.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: PGDATA is empty, but tablespace destination is not', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + out = self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", "--force", + "--incremental-mode=checksum"]) + + self.assertIn( + "INFO: Destination directory and tablespace directories are empty, " + "disable incremental restore", out) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_incr_restore_with_tablespace_7(self): + """ + Restore backup without tablespace into + PGDATA with tablespace. + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # take backup of node with tblspace + self.backup_node(backup_dir, 'node', node, options=['--stream']) + pgdata = self.pgdata_content(node.data_dir) + + self.create_tblspace_in_node(node, 'tblspace') + node.pgbench_init(scale=5, tablespace='tblspace') + node.stop() + +# try: +# self.restore_node( +# backup_dir, 'node', node, +# options=[ +# "-j", "4", +# "--incremental-mode=checksum"]) +# # we should die here because exception is what we expect to happen +# self.assertEqual( +# 1, 0, +# "Expecting Error because there is running postmaster " +# "process in destination directory.\n " +# "Output: {0} \n CMD: {1}".format( +# repr(self.output), self.cmd)) +# except ProbackupException as e: +# self.assertIn( +# 'ERROR: PGDATA is empty, but tablespace destination is not', +# e.message, +# '\n Unexpected Error Message: {0}\n CMD: {1}'.format( +# repr(e.message), self.cmd)) + + out = self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", "--incremental-mode=checksum"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_basic_incr_restore_sanity(self): + """recovery to target timeline""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + set_replication=True) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + try: + self.restore_node( + backup_dir, 'node', node, + options=["-j", "4", "--incremental-mode=checksum"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because there is running postmaster " + "process in destination directory.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'WARNING: Postmaster with pid', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + self.assertIn( + 'ERROR: Incremental restore is not allowed', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + node_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_1')) + + try: + self.restore_node( + backup_dir, 'node', node_1, data_dir=node_1.data_dir, + options=["-j", "4", "--incremental-mode=checksum"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because destination directory has wrong system id.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'WARNING: Backup catalog was initialized for system id', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + self.assertIn( + 'ERROR: Incremental restore is not allowed', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_incr_checksum_restore(self): + """ + /----C-----D + ------A----B---*--------X + + X - is instance, we want to return it to C state. + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + pg_options={'wal_log_hints': 'on'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=50) + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + xid = node.safe_psql( + 'postgres', + 'select txid_current()').decode('utf-8').rstrip() + + # --A-----B--------X + pgbench = node.pgbench(options=['-T', '30', '-c', '1', '--no-vacuum']) + pgbench.wait() + node.stop(['-m', 'immediate', '-D', node.data_dir]) + + node_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_1')) + node_1.cleanup() + + self.restore_node( + backup_dir, 'node', node_1, data_dir=node_1.data_dir, + options=[ + '--recovery-target-action=promote', + '--recovery-target-xid={0}'.format(xid)]) + + self.set_auto_conf(node_1, {'port': node_1.port}) + node_1.slow_start() + + # /-- + # --A-----B----*----X + pgbench = node_1.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # /--C + # --A-----B----*----X + self.backup_node(backup_dir, 'node', node_1, + data_dir=node_1.data_dir, backup_type='page') + + # /--C------ + # --A-----B----*----X + pgbench = node_1.pgbench(options=['-T', '50', '-c', '1']) + pgbench.wait() + + # /--C------D + # --A-----B----*----X + self.backup_node(backup_dir, 'node', node_1, + data_dir=node_1.data_dir, backup_type='page') + + pgdata = self.pgdata_content(node_1.data_dir) + + self.restore_node( + backup_dir, 'node', node, + options=["-j", "4", "--incremental-mode=checksum"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + + self.set_auto_conf(node, {'port': node.port}) + node.slow_start() + + self.compare_pgdata(pgdata, pgdata_restored) + + + # @unittest.skip("skip") + def test_incr_lsn_restore(self): + """ + /----C-----D + ------A----B---*--------X + + X - is instance, we want to return it to C state. + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + pg_options={'wal_log_hints': 'on'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=50) + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + xid = node.safe_psql( + 'postgres', + 'select txid_current()').decode('utf-8').rstrip() + + # --A-----B--------X + pgbench = node.pgbench(options=['-T', '30', '-c', '1', '--no-vacuum']) + pgbench.wait() + node.stop(['-m', 'immediate', '-D', node.data_dir]) + + node_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_1')) + node_1.cleanup() + + self.restore_node( + backup_dir, 'node', node_1, data_dir=node_1.data_dir, + options=[ + '--recovery-target-action=promote', + '--recovery-target-xid={0}'.format(xid)]) + + self.set_auto_conf(node_1, {'port': node_1.port}) + node_1.slow_start() + + # /-- + # --A-----B----*----X + pgbench = node_1.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # /--C + # --A-----B----*----X + self.backup_node(backup_dir, 'node', node_1, + data_dir=node_1.data_dir, backup_type='page') + + # /--C------ + # --A-----B----*----X + pgbench = node_1.pgbench(options=['-T', '50', '-c', '1']) + pgbench.wait() + + # /--C------D + # --A-----B----*----X + self.backup_node(backup_dir, 'node', node_1, + data_dir=node_1.data_dir, backup_type='page') + + pgdata = self.pgdata_content(node_1.data_dir) + + self.restore_node( + backup_dir, 'node', node, options=["-j", "4", "--incremental-mode=lsn"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + + self.set_auto_conf(node, {'port': node.port}) + node.slow_start() + + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_incr_lsn_sanity(self): + """ + /----A-----B + F------*--------X + + X - is instance, we want to return it to state B. + fail is expected behaviour in case of lsn restore. + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + pg_options={'wal_log_hints': 'on'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=10) + + node_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_1')) + node_1.cleanup() + + self.restore_node( + backup_dir, 'node', node_1, data_dir=node_1.data_dir) + + self.set_auto_conf(node_1, {'port': node_1.port}) + node_1.slow_start() + + pgbench = node_1.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + self.backup_node(backup_dir, 'node', node_1, + data_dir=node_1.data_dir, backup_type='full') + + pgbench = node_1.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + page_id = self.backup_node(backup_dir, 'node', node_1, + data_dir=node_1.data_dir, backup_type='page') + + node.stop() + + try: + self.restore_node( + backup_dir, 'node', node, data_dir=node.data_dir, + options=["-j", "4", "--incremental-mode=lsn"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because incremental restore in lsn mode is impossible\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Cannot perform incremental restore of " + "backup chain {0} in 'lsn' mode".format(page_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_incr_checksum_sanity(self): + """ + /----A-----B + F------*--------X + + X - is instance, we want to return it to state B. + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=20) + + node_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_1')) + node_1.cleanup() + + self.restore_node( + backup_dir, 'node', node_1, data_dir=node_1.data_dir) + + self.set_auto_conf(node_1, {'port': node_1.port}) + node_1.slow_start() + + pgbench = node_1.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + self.backup_node(backup_dir, 'node', node_1, + data_dir=node_1.data_dir, backup_type='full') + + pgbench = node_1.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + page_id = self.backup_node(backup_dir, 'node', node_1, + data_dir=node_1.data_dir, backup_type='page') + pgdata = self.pgdata_content(node_1.data_dir) + + node.stop() + + self.restore_node( + backup_dir, 'node', node, data_dir=node.data_dir, + options=["-j", "4", "--incremental-mode=checksum"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_incr_checksum_corruption_detection(self): + """ + check that corrupted page got detected and replaced + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), +# initdb_params=['--data-checksums'], + pg_options={'wal_log_hints': 'on'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=20) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + self.backup_node(backup_dir, 'node', node, + data_dir=node.data_dir, backup_type='full') + + heap_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('pgbench_accounts')").decode('utf-8').rstrip() + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + page_id = self.backup_node(backup_dir, 'node', node, + data_dir=node.data_dir, backup_type='page') + + pgdata = self.pgdata_content(node.data_dir) + + node.stop() + + path = os.path.join(node.data_dir, heap_path) + with open(path, "rb+", 0) as f: + f.seek(22000) + f.write(b"bla") + f.flush() + f.close + + self.restore_node( + backup_dir, 'node', node, data_dir=node.data_dir, + options=["-j", "4", "--incremental-mode=checksum"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_incr_lsn_corruption_detection(self): + """ + check that corrupted page got detected and replaced + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + pg_options={'wal_log_hints': 'on'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=20) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + self.backup_node(backup_dir, 'node', node, + data_dir=node.data_dir, backup_type='full') + + heap_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('pgbench_accounts')").decode('utf-8').rstrip() + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + page_id = self.backup_node(backup_dir, 'node', node, + data_dir=node.data_dir, backup_type='page') + + pgdata = self.pgdata_content(node.data_dir) + + node.stop() + + path = os.path.join(node.data_dir, heap_path) + with open(path, "rb+", 0) as f: + f.seek(22000) + f.write(b"bla") + f.flush() + f.close + + self.restore_node( + backup_dir, 'node', node, data_dir=node.data_dir, + options=["-j", "4", "--incremental-mode=lsn"]) + + pgdata_restored = self.pgdata_content(node.data_dir) + + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_incr_restore_multiple_external(self): + """check that cmdline has priority over config""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + external_dir1 = self.get_tblspace_path(node, 'external_dir1') + external_dir2 = self.get_tblspace_path(node, 'external_dir2') + + # FULL backup + node.pgbench_init(scale=20) + self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4"]) + + # fill external directories with data + self.restore_node( + backup_dir, 'node', node, + data_dir=external_dir1, options=["-j", "4"]) + + self.restore_node( + backup_dir, 'node', node, + data_dir=external_dir2, options=["-j", "4"]) + + self.set_config( + backup_dir, 'node', + options=['-E{0}{1}{2}'.format( + external_dir1, self.EXTERNAL_DIRECTORY_DELIMITER, external_dir2)]) + + # cmdline option MUST override options in config + self.backup_node( + backup_dir, 'node', node, + backup_type='full', options=["-j", "4"]) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # cmdline option MUST override options in config + self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=["-j", "4"]) + + pgdata = self.pgdata_content( + node.base_dir, exclude_dirs=['logs']) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + node.stop() + + self.restore_node( + backup_dir, 'node', node, + options=["-j", "4", '--incremental-mode=checksum']) + + pgdata_restored = self.pgdata_content( + node.base_dir, exclude_dirs=['logs']) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_incr_lsn_restore_multiple_external(self): + """check that cmdline has priority over config""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + external_dir1 = self.get_tblspace_path(node, 'external_dir1') + external_dir2 = self.get_tblspace_path(node, 'external_dir2') + + # FULL backup + node.pgbench_init(scale=20) + self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4"]) + + # fill external directories with data + self.restore_node( + backup_dir, 'node', node, + data_dir=external_dir1, options=["-j", "4"]) + + self.restore_node( + backup_dir, 'node', node, + data_dir=external_dir2, options=["-j", "4"]) + + self.set_config( + backup_dir, 'node', + options=['-E{0}{1}{2}'.format( + external_dir1, self.EXTERNAL_DIRECTORY_DELIMITER, external_dir2)]) + + # cmdline option MUST override options in config + self.backup_node( + backup_dir, 'node', node, + backup_type='full', options=["-j", "4"]) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # cmdline option MUST override options in config + self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=["-j", "4"]) + + pgdata = self.pgdata_content( + node.base_dir, exclude_dirs=['logs']) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + node.stop() + + self.restore_node( + backup_dir, 'node', node, + options=["-j", "4", '--incremental-mode=lsn']) + + pgdata_restored = self.pgdata_content( + node.base_dir, exclude_dirs=['logs']) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_incr_lsn_restore_backward(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={'wal_log_hints': 'on', 'hot_standby': 'on'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + node.pgbench_init(scale=2) + full_id = self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4"]) + + full_pgdata = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + page_id = self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=["-j", "4"]) + + page_pgdata = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + delta_id = self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=["-j", "4"]) + + delta_pgdata = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=full_id, + options=[ + "-j", "4", + '--incremental-mode=lsn', + '--recovery-target=immediate', + '--recovery-target-action=pause']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(full_pgdata, pgdata_restored) + + node.slow_start(replica=True) + node.stop() + + try: + self.restore_node( + backup_dir, 'node', node, backup_id=page_id, + options=[ + "-j", "4", '--incremental-mode=lsn', + '--recovery-target=immediate', '--recovery-target-action=pause']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because incremental restore in lsn mode is impossible\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "Cannot perform incremental restore of backup chain", + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.restore_node( + backup_dir, 'node', node, backup_id=page_id, + options=[ + "-j", "4", '--incremental-mode=checksum', + '--recovery-target=immediate', '--recovery-target-action=pause']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(page_pgdata, pgdata_restored) + + node.slow_start(replica=True) + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=delta_id, + options=[ + "-j", "4", + '--incremental-mode=lsn', + '--recovery-target=immediate', + '--recovery-target-action=pause']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(delta_pgdata, pgdata_restored) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_incr_checksum_restore_backward(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'hot_standby': 'on'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + node.pgbench_init(scale=20) + full_id = self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4"]) + + full_pgdata = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + page_id = self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=["-j", "4"]) + + page_pgdata = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + delta_id = self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=["-j", "4"]) + + delta_pgdata = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=full_id, + options=[ + "-j", "4", + '--incremental-mode=checksum', + '--recovery-target=immediate', + '--recovery-target-action=pause']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(full_pgdata, pgdata_restored) + + node.slow_start(replica=True) + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=page_id, + options=[ + "-j", "4", + '--incremental-mode=checksum', + '--recovery-target=immediate', + '--recovery-target-action=pause']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(page_pgdata, pgdata_restored) + + node.slow_start(replica=True) + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=delta_id, + options=[ + "-j", "4", + '--incremental-mode=checksum', + '--recovery-target=immediate', + '--recovery-target-action=pause']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(delta_pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_make_replica_via_incr_checksum_restore(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums']) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', master) + self.set_archiving(backup_dir, 'node', master, replica=True) + master.slow_start() + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + master.pgbench_init(scale=20) + + self.backup_node(backup_dir, 'node', master) + + self.restore_node( + backup_dir, 'node', replica, options=['-R']) + + # Settings for Replica + self.set_replica(master, replica, synchronous=False) + + replica.slow_start(replica=True) + + pgbench = master.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # PROMOTIONS + replica.promote() + new_master = replica + + # old master is going a bit further + old_master = master + pgbench = old_master.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + old_master.stop() + + pgbench = new_master.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # take backup from new master + self.backup_node( + backup_dir, 'node', new_master, + data_dir=new_master.data_dir, backup_type='page') + + # restore old master as replica + self.restore_node( + backup_dir, 'node', old_master, data_dir=old_master.data_dir, + options=['-R', '--incremental-mode=checksum']) + + self.set_replica(new_master, old_master, synchronous=True) + + old_master.slow_start(replica=True) + + pgbench = new_master.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # @unittest.skip("skip") + def test_make_replica_via_incr_lsn_restore(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums']) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', master) + self.set_archiving(backup_dir, 'node', master, replica=True) + master.slow_start() + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + master.pgbench_init(scale=20) + + self.backup_node(backup_dir, 'node', master) + + self.restore_node( + backup_dir, 'node', replica, options=['-R']) + + # Settings for Replica + self.set_replica(master, replica, synchronous=False) + + replica.slow_start(replica=True) + + pgbench = master.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # PROMOTIONS + replica.promote() + new_master = replica + + # old master is going a bit further + old_master = master + pgbench = old_master.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + old_master.stop() + + pgbench = new_master.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # take backup from new master + self.backup_node( + backup_dir, 'node', new_master, + data_dir=new_master.data_dir, backup_type='page') + + # restore old master as replica + self.restore_node( + backup_dir, 'node', old_master, data_dir=old_master.data_dir, + options=['-R', '--incremental-mode=lsn']) + + self.set_replica(new_master, old_master, synchronous=True) + + old_master.slow_start(replica=True) + + pgbench = new_master.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_incr_checksum_long_xact(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + 'create extension pageinspect') + + # FULL backup + con = node.connect("postgres") + con.execute("CREATE TABLE t1 (a int)") + con.commit() + + + con.execute("INSERT INTO t1 values (1)") + con.commit() + + # leave uncommited + con2 = node.connect("postgres") + con.execute("INSERT INTO t1 values (2)") + con2.execute("INSERT INTO t1 values (3)") + + full_id = self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + + self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + + con.commit() + + node.safe_psql( + 'postgres', + 'select * from t1') + + con2.commit() + node.safe_psql( + 'postgres', + 'select * from t1') + + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=full_id, + options=["-j", "4", '--incremental-mode=checksum']) + + node.slow_start() + + self.assertEqual( + node.safe_psql( + 'postgres', + 'select count(*) from t1').decode('utf-8').rstrip(), + '1') + + # @unittest.skip("skip") + # @unittest.expectedFailure + # This test will pass with Enterprise + # because it has checksums enabled by default + @unittest.skipIf(ProbackupTest.enterprise, 'skip') + def test_incr_lsn_long_xact_1(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + 'create extension pageinspect') + + # FULL backup + con = node.connect("postgres") + con.execute("CREATE TABLE t1 (a int)") + con.commit() + + + con.execute("INSERT INTO t1 values (1)") + con.commit() + + # leave uncommited + con2 = node.connect("postgres") + con.execute("INSERT INTO t1 values (2)") + con2.execute("INSERT INTO t1 values (3)") + + full_id = self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + + self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + + con.commit() + + # when does LSN gets stamped when checksum gets updated ? + node.safe_psql( + 'postgres', + 'select * from t1') + + con2.commit() + node.safe_psql( + 'postgres', + 'select * from t1') + + node.stop() + + try: + self.restore_node( + backup_dir, 'node', node, backup_id=full_id, + options=["-j", "4", '--incremental-mode=lsn']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because incremental restore in lsn mode is impossible\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Incremental restore in 'lsn' mode require data_checksums to be " + "enabled in destination data directory", + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_incr_lsn_long_xact_2(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'full_page_writes': 'off', + 'wal_log_hints': 'off'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + 'create extension pageinspect') + + # FULL backup + con = node.connect("postgres") + con.execute("CREATE TABLE t1 (a int)") + con.commit() + + + con.execute("INSERT INTO t1 values (1)") + con.commit() + + # leave uncommited + con2 = node.connect("postgres") + con.execute("INSERT INTO t1 values (2)") + con2.execute("INSERT INTO t1 values (3)") + + full_id = self.backup_node( + backup_dir, 'node', node, + backup_type="full", options=["-j", "4", "--stream"]) + + self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + +# print(node.safe_psql( +# 'postgres', +# "select * from page_header(get_raw_page('t1', 0))")) + + con.commit() + + # when does LSN gets stamped when checksum gets updated ? + node.safe_psql( + 'postgres', + 'select * from t1') + +# print(node.safe_psql( +# 'postgres', +# "select * from page_header(get_raw_page('t1', 0))")) + + con2.commit() + node.safe_psql( + 'postgres', + 'select * from t1') + +# print(node.safe_psql( +# 'postgres', +# "select * from page_header(get_raw_page('t1', 0))")) + + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=full_id, + options=["-j", "4", '--incremental-mode=lsn']) + + node.slow_start() + + self.assertEqual( + node.safe_psql( + 'postgres', + 'select count(*) from t1').decode('utf-8').rstrip(), + '1') + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_incr_restore_zero_size_file_checksum(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + fullpath = os.path.join(node.data_dir, 'simple_file') + with open(fullpath, "w+b", 0) as f: + f.flush() + f.close + + # FULL backup + id1 = self.backup_node( + backup_dir, 'node', node, + options=["-j", "4", "--stream"]) + + pgdata1 = self.pgdata_content(node.data_dir) + + with open(fullpath, "rb+", 0) as f: + f.seek(9000) + f.write(b"bla") + f.flush() + f.close + + id2 = self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + pgdata2 = self.pgdata_content(node.data_dir) + + with open(fullpath, "w") as f: + f.close() + + id3 = self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + pgdata3 = self.pgdata_content(node.data_dir) + + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=id1, + options=["-j", "4", '-I', 'checksum']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata1, pgdata_restored) + + self.restore_node( + backup_dir, 'node', node, backup_id=id2, + options=["-j", "4", '-I', 'checksum']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata2, pgdata_restored) + + self.restore_node( + backup_dir, 'node', node, backup_id=id3, + options=["-j", "4", '-I', 'checksum']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata3, pgdata_restored) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_incr_restore_zero_size_file_lsn(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + fullpath = os.path.join(node.data_dir, 'simple_file') + with open(fullpath, "w+b", 0) as f: + f.flush() + f.close + + # FULL backup + id1 = self.backup_node( + backup_dir, 'node', node, + options=["-j", "4", "--stream"]) + + pgdata1 = self.pgdata_content(node.data_dir) + + with open(fullpath, "rb+", 0) as f: + f.seek(9000) + f.write(b"bla") + f.flush() + f.close + + id2 = self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + pgdata2 = self.pgdata_content(node.data_dir) + + with open(fullpath, "w") as f: + f.close() + + id3 = self.backup_node( + backup_dir, 'node', node, + backup_type="delta", options=["-j", "4", "--stream"]) + pgdata3 = self.pgdata_content(node.data_dir) + + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=id1, + options=["-j", "4", '-I', 'checksum']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata1, pgdata_restored) + + node.slow_start() + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=id2, + options=["-j", "4", '-I', 'checksum']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata2, pgdata_restored) + + node.slow_start() + node.stop() + + self.restore_node( + backup_dir, 'node', node, backup_id=id3, + options=["-j", "4", '-I', 'checksum']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata3, pgdata_restored) + + def test_incremental_partial_restore_exclude_checksum(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + db_list_raw = node.safe_psql( + 'postgres', + 'SELECT to_json(a) ' + 'FROM (SELECT oid, datname FROM pg_database) a').rstrip() + + db_list_splitted = db_list_raw.splitlines() + + db_list = {} + for line in db_list_splitted: + line = json.loads(line) + db_list[line['datname']] = line['oid'] + + node.pgbench_init(scale=20) + + # FULL backup + self.backup_node(backup_dir, 'node', node) + pgdata = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # PAGE backup + backup_id = self.backup_node(backup_dir, 'node', node, backup_type='page') + + # restore FULL backup into second node2 + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1')) + node1.cleanup() + + node2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node2')) + node2.cleanup() + + # restore some data into node2 + self.restore_node(backup_dir, 'node', node2) + + # partial restore backup into node1 + self.restore_node( + backup_dir, 'node', + node1, options=[ + "--db-exclude=db1", + "--db-exclude=db5"]) + + pgdata1 = self.pgdata_content(node1.data_dir) + + # partial incremental restore backup into node2 + self.restore_node( + backup_dir, 'node', + node2, options=[ + "--db-exclude=db1", + "--db-exclude=db5", + "-I", "checksum", + "--destroy-all-other-dbs", + ]) + + pgdata2 = self.pgdata_content(node2.data_dir) + + self.compare_pgdata(pgdata1, pgdata2) + + self.set_auto_conf(node2, {'port': node2.port}) + + node2.slow_start() + + node2.safe_psql( + 'postgres', + 'select 1') + + try: + node2.safe_psql( + 'db1', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + try: + node2.safe_psql( + 'db5', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + with open(node2.pg_log_file, 'r') as f: + output = f.read() + + self.assertNotIn('PANIC', output) + + def test_incremental_partial_restore_exclude_lsn(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + db_list_raw = node.safe_psql( + 'postgres', + 'SELECT to_json(a) ' + 'FROM (SELECT oid, datname FROM pg_database) a').rstrip() + + db_list_splitted = db_list_raw.splitlines() + + db_list = {} + for line in db_list_splitted: + line = json.loads(line) + db_list[line['datname']] = line['oid'] + + node.pgbench_init(scale=20) + + # FULL backup + self.backup_node(backup_dir, 'node', node) + pgdata = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1']) + pgbench.wait() + + # PAGE backup + backup_id = self.backup_node(backup_dir, 'node', node, backup_type='page') + + node.stop() + + # restore FULL backup into second node2 + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1')) + node1.cleanup() + + node2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node2')) + node2.cleanup() + + # restore some data into node2 + self.restore_node(backup_dir, 'node', node2) + + # partial restore backup into node1 + self.restore_node( + backup_dir, 'node', + node1, options=[ + "--db-exclude=db1", + "--db-exclude=db5"]) + + pgdata1 = self.pgdata_content(node1.data_dir) + + # partial incremental restore backup into node2 + node2.port = node.port + node2.slow_start() + node2.stop() + self.restore_node( + backup_dir, 'node', + node2, options=[ + "--db-exclude=db1", + "--db-exclude=db5", + "-I", "lsn", + "--destroy-all-other-dbs", + ]) + + pgdata2 = self.pgdata_content(node2.data_dir) + + self.compare_pgdata(pgdata1, pgdata2) + + self.set_auto_conf(node2, {'port': node2.port}) + + node2.slow_start() + + node2.safe_psql( + 'postgres', + 'select 1') + + try: + node2.safe_psql( + 'db1', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + try: + node2.safe_psql( + 'db5', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + with open(node2.pg_log_file, 'r') as f: + output = f.read() + + self.assertNotIn('PANIC', output) + + def test_incremental_partial_restore_exclude_tablespace_checksum(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # cat_version = node.get_control_data()["Catalog version number"] + # version_specific_dir = 'PG_' + node.major_version_str + '_' + cat_version + + # PG_10_201707211 + # pg_tblspc/33172/PG_9.5_201510051/16386/ + + self.create_tblspace_in_node(node, 'somedata') + + node_tablespace = self.get_tblspace_path(node, 'somedata') + + tbl_oid = node.safe_psql( + 'postgres', + "SELECT oid " + "FROM pg_tablespace " + "WHERE spcname = 'somedata'").rstrip() + + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0} tablespace somedata'.format(i)) + + db_list_raw = node.safe_psql( + 'postgres', + 'SELECT to_json(a) ' + 'FROM (SELECT oid, datname FROM pg_database) a').rstrip() + + db_list_splitted = db_list_raw.splitlines() + + db_list = {} + for line in db_list_splitted: + line = json.loads(line) + db_list[line['datname']] = line['oid'] + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + + # node1 + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1')) + node1.cleanup() + node1_tablespace = self.get_tblspace_path(node1, 'somedata') + + # node2 + node2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node2')) + node2.cleanup() + node2_tablespace = self.get_tblspace_path(node2, 'somedata') + + # in node2 restore full backup + self.restore_node( + backup_dir, 'node', + node2, options=[ + "-T", "{0}={1}".format( + node_tablespace, node2_tablespace)]) + + # partial restore into node1 + self.restore_node( + backup_dir, 'node', + node1, options=[ + "--db-exclude=db1", + "--db-exclude=db5", + "-T", "{0}={1}".format( + node_tablespace, node1_tablespace)]) + + pgdata1 = self.pgdata_content(node1.data_dir) + + # partial incremental restore into node2 + try: + self.restore_node( + backup_dir, 'node', + node2, options=[ + "-I", "checksum", + "--db-exclude=db1", + "--db-exclude=db5", + "-T", "{0}={1}".format( + node_tablespace, node2_tablespace), + "--destroy-all-other-dbs"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because remapped tablespace contain old data .\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Remapped tablespace destination is not empty:', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.restore_node( + backup_dir, 'node', + node2, options=[ + "-I", "checksum", "--force", + "--db-exclude=db1", + "--db-exclude=db5", + "-T", "{0}={1}".format( + node_tablespace, node2_tablespace), + "--destroy-all-other-dbs", + ]) + + pgdata2 = self.pgdata_content(node2.data_dir) + + self.compare_pgdata(pgdata1, pgdata2) + + self.set_auto_conf(node2, {'port': node2.port}) + node2.slow_start() + + node2.safe_psql( + 'postgres', + 'select 1') + + try: + node2.safe_psql( + 'db1', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + try: + node2.safe_psql( + 'db5', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + with open(node2.pg_log_file, 'r') as f: + output = f.read() + + self.assertNotIn('PANIC', output) + + def test_incremental_partial_restore_deny(self): + """ + Do now allow partial incremental restore into non-empty PGDATA + becase we can't limit WAL replay to a single database. + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + for i in range(1, 3): + node.safe_psql('postgres', f'CREATE database db{i}') + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + pgdata = self.pgdata_content(node.data_dir) + + try: + self.restore_node(backup_dir, 'node', node, options=["--db-include=db1", '-I', 'LSN']) + self.fail("incremental partial restore is not allowed") + except ProbackupException as e: + self.assertIn("Incremental restore is not allowed: Postmaster is running.", e.message) + + node.safe_psql('db2', 'create table x (id int)') + node.safe_psql('db2', 'insert into x values (42)') + + node.stop() + + try: + self.restore_node(backup_dir, 'node', node, options=["--db-include=db1", '-I', 'LSN']) + self.fail("because incremental partial restore is not allowed") + except ProbackupException as e: + self.assertIn("Incremental restore is not allowed: Partial incremental restore into non-empty PGDATA is forbidden", e.message) + + node.slow_start() + value = node.execute('db2', 'select * from x')[0][0] + self.assertEqual(42, value) + + def test_deny_incremental_partial_restore_exclude_tablespace_checksum(self): + """ + Do now allow partial incremental restore into non-empty PGDATA + becase we can't limit WAL replay to a single database. + (case of tablespaces) + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.create_tblspace_in_node(node, 'somedata') + + node_tablespace = self.get_tblspace_path(node, 'somedata') + + tbl_oid = node.safe_psql( + 'postgres', + "SELECT oid " + "FROM pg_tablespace " + "WHERE spcname = 'somedata'").rstrip() + + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0} tablespace somedata'.format(i)) + + db_list_raw = node.safe_psql( + 'postgres', + 'SELECT to_json(a) ' + 'FROM (SELECT oid, datname FROM pg_database) a').rstrip() + + db_list_splitted = db_list_raw.splitlines() + + db_list = {} + for line in db_list_splitted: + line = json.loads(line) + db_list[line['datname']] = line['oid'] + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + + # node2 + node2 = self.make_simple_node('node2') + node2.cleanup() + node2_tablespace = self.get_tblspace_path(node2, 'somedata') + + # in node2 restore full backup + self.restore_node( + backup_dir, 'node', + node2, options=[ + "-T", f"{node_tablespace}={node2_tablespace}"]) + + # partial incremental restore into node2 + try: + self.restore_node(backup_dir, 'node', node2, + options=["-I", "checksum", + "--db-exclude=db1", + "--db-exclude=db5", + "-T", f"{node_tablespace}={node2_tablespace}"]) + self.fail("remapped tablespace contain old data") + except ProbackupException as e: + pass + + try: + self.restore_node(backup_dir, 'node', node2, + options=[ + "-I", "checksum", "--force", + "--db-exclude=db1", "--db-exclude=db5", + "-T", f"{node_tablespace}={node2_tablespace}"]) + self.fail("incremental partial restore is not allowed") + except ProbackupException as e: + self.assertIn("Incremental restore is not allowed: Partial incremental restore into non-empty PGDATA is forbidden", e.message) + + def test_incremental_pg_filenode_map(self): + """ + https://github.com/postgrespro/pg_probackup/issues/320 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1'), + initdb_params=['--data-checksums']) + node1.cleanup() + + node.pgbench_init(scale=5) + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + + # in node1 restore full backup + self.restore_node(backup_dir, 'node', node1) + self.set_auto_conf(node1, {'port': node1.port}) + node1.slow_start() + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1']) + + pgbench = node1.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1']) + + node.safe_psql( + 'postgres', + 'reindex index pg_type_oid_index') + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + + node1.stop() + + # incremental restore into node1 + self.restore_node(backup_dir, 'node', node1, options=["-I", "checksum"]) + + self.set_auto_conf(node1, {'port': node1.port}) + node1.slow_start() + + node1.safe_psql( + 'postgres', + 'select 1') + +# check that MinRecPoint and BackupStartLsn are correctly used in case of --incrementa-lsn + + # @unittest.skip("skip") + def test_incr_restore_issue_313(self): + """ + Check that failed incremental restore can be restarted + """ + self._check_gdb_flag_or_skip_test + node = self.make_simple_node('node', + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale = 50) + + full_backup_id = self.backup_node(backup_dir, 'node', node, backup_type='full') + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + last_backup_id = self.backup_node(backup_dir, 'node', node, backup_type='delta') + + pgdata = self.pgdata_content(node.data_dir) + node.cleanup() + + self.restore_node(backup_dir, 'node', node, backup_id=full_backup_id) + + count = 0 + filelist = self.get_backup_filelist(backup_dir, 'node', last_backup_id) + for file in filelist: + # count only nondata files + if int(filelist[file]['is_datafile']) == 0 and \ + not stat.S_ISDIR(int(filelist[file]['mode'])) and \ + not filelist[file]['size'] == '0' and \ + file != 'database_map': + count += 1 + + gdb = self.restore_node(backup_dir, 'node', node, gdb=True, + backup_id=last_backup_id, options=['--progress', '--incremental-mode=checksum']) + gdb.verbose = False + gdb.set_breakpoint('restore_non_data_file') + gdb.run_until_break() + gdb.continue_execution_until_break(count - 1) + gdb.quit() + + bak_file = os.path.join(node.data_dir, 'global', 'pg_control.pbk.bak') + self.assertTrue( + os.path.exists(bak_file), + "pg_control bak File should not exist: {0}".format(bak_file)) + + try: + node.slow_start() + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because backup is not fully restored") + except StartNodeException as e: + self.assertIn( + 'Cannot start node', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + with open(os.path.join(node.logs_dir, 'postgresql.log'), 'r') as f: + if self.pg_config_version >= 120000: + self.assertIn( + "PANIC: could not read file \"global/pg_control\"", + f.read()) + else: + self.assertIn( + "PANIC: could not read from control file", + f.read()) + self.restore_node(backup_dir, 'node', node, + backup_id=last_backup_id, options=['--progress', '--incremental-mode=checksum']) + node.slow_start() + self.compare_pgdata(pgdata, self.pgdata_content(node.data_dir)) + + # @unittest.skip("skip") + def test_skip_pages_at_non_zero_segment_checksum(self): + if self.remote: + self.skipTest("Skipped because this test doesn't work properly in remote mode yet") + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + pg_options={'wal_log_hints': 'on'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # create table of size > 1 GB, so it will have several segments + node.safe_psql( + 'postgres', + "create table t as select i as a, i*2 as b, i*3 as c, i*4 as d, i*5 as e " + "from generate_series(1,20600000) i; " + "CHECKPOINT ") + + filepath = node.safe_psql( + 'postgres', + "SELECT pg_relation_filepath('t')" + ).decode('utf-8').rstrip() + + # segment .1 must exist in order to proceed this test + self.assertTrue(os.path.exists(f'{os.path.join(node.data_dir, filepath)}.1')) + + # do full backup + self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + 'postgres', + "DELETE FROM t WHERE a < 101; " + "CHECKPOINT") + + # do incremental backup + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgdata = self.pgdata_content(node.data_dir) + + node.safe_psql( + 'postgres', + "DELETE FROM t WHERE a < 201; " + "CHECKPOINT") + + node.stop() + + self.restore_node( + backup_dir, 'node', node, options=["-j", "4", "--incremental-mode=checksum", "--log-level-console=INFO"]) + + self.assertNotIn('WARNING: Corruption detected in file', self.output, + 'Incremental restore copied pages from .1 datafile segment that were not changed') + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_skip_pages_at_non_zero_segment_lsn(self): + if self.remote: + self.skipTest("Skipped because this test doesn't work properly in remote mode yet") + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + pg_options={'wal_log_hints': 'on'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # create table of size > 1 GB, so it will have several segments + node.safe_psql( + 'postgres', + "create table t as select i as a, i*2 as b, i*3 as c, i*4 as d, i*5 as e " + "from generate_series(1,20600000) i; " + "CHECKPOINT ") + + filepath = node.safe_psql( + 'postgres', + "SELECT pg_relation_filepath('t')" + ).decode('utf-8').rstrip() + + # segment .1 must exist in order to proceed this test + self.assertTrue(os.path.exists(f'{os.path.join(node.data_dir, filepath)}.1')) + + # do full backup + self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + 'postgres', + "DELETE FROM t WHERE a < 101; " + "CHECKPOINT") + + # do incremental backup + self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgdata = self.pgdata_content(node.data_dir) + + node.safe_psql( + 'postgres', + "DELETE FROM t WHERE a < 201; " + "CHECKPOINT") + + node.stop() + + self.restore_node( + backup_dir, 'node', node, options=["-j", "4", "--incremental-mode=lsn", "--log-level-console=INFO"]) + + self.assertNotIn('WARNING: Corruption detected in file', self.output, + 'Incremental restore copied pages from .1 datafile segment that were not changed') + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) diff --git a/tests/init.py b/tests/init_test.py similarity index 57% rename from tests/init.py rename to tests/init_test.py index 059e24d95..4e000c78f 100644 --- a/tests/init.py +++ b/tests/init_test.py @@ -1,9 +1,7 @@ import os import unittest from .helpers.ptrack_helpers import dir_files, ProbackupTest, ProbackupException - - -module_name = 'init' +import shutil class InitTest(ProbackupTest, unittest.TestCase): @@ -12,9 +10,8 @@ class InitTest(ProbackupTest, unittest.TestCase): # @unittest.expectedFailure def test_success(self): """Success normal init""" - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node(base_dir=os.path.join(module_name, fname, 'node')) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node(base_dir=os.path.join(self.module_name, self.fname, 'node')) self.init_pb(backup_dir) self.assertEqual( dir_files(backup_dir), @@ -59,19 +56,16 @@ def test_success(self): repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( - "ERROR: Required parameter not specified: PGDATA (-D, --pgdata)", + "ERROR: No postgres data directory specified.\n" + "Please specify it either using environment variable PGDATA or\ncommand line option --pgdata (-D)", e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format(e.message, self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_already_exist(self): """Failure with backup catalog already existed""" - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node(base_dir=os.path.join(module_name, fname, 'node')) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node(base_dir=os.path.join(self.module_name, self.fname, 'node')) self.init_pb(backup_dir) try: self.show_pb(backup_dir, 'node') @@ -83,15 +77,11 @@ def test_already_exist(self): e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_abs_path(self): """failure with backup catalog should be given as absolute path""" - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node(base_dir=os.path.join(module_name, fname, 'node')) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node(base_dir=os.path.join(self.module_name, self.fname, 'node')) try: self.run_pb(["init", "-B", os.path.relpath("%s/backup" % node.base_dir, self.dir_path)]) self.assertEqual(1, 0, 'Expecting Error due to initialization with non-absolute path in --backup-path. Output: {0} \n CMD: {1}'.format( @@ -102,5 +92,48 @@ def test_abs_path(self): e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_add_instance_idempotence(self): + """ + https://github.com/postgrespro/pg_probackup/issues/219 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node(base_dir=os.path.join(self.module_name, self.fname, 'node')) + self.init_pb(backup_dir) + + self.add_instance(backup_dir, 'node', node) + shutil.rmtree(os.path.join(backup_dir, 'backups', 'node')) + + dir_backups = os.path.join(backup_dir, 'backups', 'node') + dir_wal = os.path.join(backup_dir, 'wal', 'node') + + try: + self.add_instance(backup_dir, 'node', node) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because page backup should not be possible " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Instance 'node' WAL archive directory already exists: ", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + try: + self.add_instance(backup_dir, 'node', node) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because page backup should not be possible " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Instance 'node' WAL archive directory already exists: ", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) diff --git a/tests/locking.py b/tests/locking_test.py similarity index 60% rename from tests/locking.py rename to tests/locking_test.py index 54389bde6..5367c2610 100644 --- a/tests/locking.py +++ b/tests/locking_test.py @@ -4,9 +4,6 @@ from .helpers.ptrack_helpers import ProbackupTest, ProbackupException -module_name = 'locking' - - class LockingTest(ProbackupTest, unittest.TestCase): # @unittest.skip("skip") @@ -17,12 +14,13 @@ def test_locking_running_validate_1(self): run validate, expect it to successfully executed, concurrent RUNNING backup with pid file and active process is legal """ - fname = self.id().split('.')[3] + self._check_gdb_flag_or_skip_test() + node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -33,7 +31,7 @@ def test_locking_running_validate_1(self): gdb = self.backup_node( backup_dir, 'node', node, gdb=True) - gdb.set_breakpoint('copy_file') + gdb.set_breakpoint('backup_non_data_file') gdb.run_until_break() gdb.continue_execution_until_break(20) @@ -50,7 +48,7 @@ def test_locking_running_validate_1(self): backup_id = self.show_pb(backup_dir, 'node')[1]['id'] self.assertIn( - "is using backup {0} and still is running".format(backup_id), + "is using backup {0}, and is still running".format(backup_id), validate_output, '\n Unexpected Validate Output: {0}\n'.format(repr(validate_output))) @@ -61,7 +59,7 @@ def test_locking_running_validate_1(self): 'RUNNING', self.show_pb(backup_dir, 'node')[1]['status']) # Clean after yourself - # self.del_test_dir(module_name, fname) + gdb.kill() def test_locking_running_validate_2(self): """ @@ -71,12 +69,13 @@ def test_locking_running_validate_2(self): RUNNING backup with pid file AND without active pid is legal, but his status must be changed to ERROR and pid file is deleted """ - fname = self.id().split('.')[3] + self._check_gdb_flag_or_skip_test() + node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -87,7 +86,7 @@ def test_locking_running_validate_2(self): gdb = self.backup_node( backup_dir, 'node', node, gdb=True) - gdb.set_breakpoint('copy_file') + gdb.set_breakpoint('backup_non_data_file') gdb.run_until_break() gdb.continue_execution_until_break(20) @@ -129,7 +128,7 @@ def test_locking_running_validate_2(self): 'ERROR', self.show_pb(backup_dir, 'node')[1]['status']) # Clean after yourself - self.del_test_dir(module_name, fname) + gdb.kill() def test_locking_running_validate_2_specific_id(self): """ @@ -140,12 +139,13 @@ def test_locking_running_validate_2_specific_id(self): RUNNING backup with pid file AND without active pid is legal, but his status must be changed to ERROR and pid file is deleted """ - fname = self.id().split('.')[3] + self._check_gdb_flag_or_skip_test() + node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -156,7 +156,7 @@ def test_locking_running_validate_2_specific_id(self): gdb = self.backup_node( backup_dir, 'node', node, gdb=True) - gdb.set_breakpoint('copy_file') + gdb.set_breakpoint('backup_non_data_file') gdb.run_until_break() gdb.continue_execution_until_break(20) @@ -227,7 +227,7 @@ def test_locking_running_validate_2_specific_id(self): repr(e.message), self.cmd)) # Clean after yourself - self.del_test_dir(module_name, fname) + gdb.kill() def test_locking_running_3(self): """ @@ -237,12 +237,13 @@ def test_locking_running_3(self): RUNNING backup without pid file AND without active pid is legal, his status must be changed to ERROR """ - fname = self.id().split('.')[3] + self._check_gdb_flag_or_skip_test() + node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -253,7 +254,7 @@ def test_locking_running_3(self): gdb = self.backup_node( backup_dir, 'node', node, gdb=True) - gdb.set_breakpoint('copy_file') + gdb.set_breakpoint('backup_non_data_file') gdb.run_until_break() gdb.continue_execution_until_break(20) @@ -296,22 +297,23 @@ def test_locking_running_3(self): 'ERROR', self.show_pb(backup_dir, 'node')[1]['status']) # Clean after yourself - self.del_test_dir(module_name, fname) + gdb.kill() def test_locking_restore_locked(self): """ make node, take full backup, take two page backups, launch validate on PAGE1 and stop it in the middle, launch restore of PAGE2. - Expect restore to fail because validation of - intermediate backup is impossible + Expect restore to sucseed because read-only locks + do not conflict """ - fname = self.id().split('.')[3] + self._check_gdb_flag_or_skip_test() + node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -334,24 +336,12 @@ def test_locking_restore_locked(self): node.cleanup() - try: - self.restore_node(backup_dir, 'node', node) - self.assertEqual( - 1, 0, - "Expecting Error because restore without whole chain validation " - "is prohibited unless --no-validate provided.\n " - "Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertTrue( - "ERROR: Cannot lock backup {0} directory\n".format(full_id) in e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) + self.restore_node(backup_dir, 'node', node) # Clean after yourself - self.del_test_dir(module_name, fname) + gdb.kill() - def test_locking_restore_locked_without_validation(self): + def test_concurrent_delete_and_restore(self): """ make node, take full backup, take page backup, launch validate on FULL and stop it in the middle, @@ -359,12 +349,13 @@ def test_locking_restore_locked_without_validation(self): Expect restore to fail because validation of intermediate backup is impossible """ - fname = self.id().split('.')[3] + self._check_gdb_flag_or_skip_test() + node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -376,10 +367,11 @@ def test_locking_restore_locked_without_validation(self): # PAGE1 restore_id = self.backup_node(backup_dir, 'node', node, backup_type='page') - gdb = self.validate_pb( + gdb = self.delete_pb( backup_dir, 'node', backup_id=backup_id, gdb=True) - gdb.set_breakpoint('pgBackupValidate') + # gdb.set_breakpoint('pgFileDelete') + gdb.set_breakpoint('delete_backup_files') gdb.run_until_break() node.cleanup() @@ -397,14 +389,14 @@ def test_locking_restore_locked_without_validation(self): self.assertTrue( "Backup {0} is used without validation".format( restore_id) in e.message and - 'is using backup {0} and still is running'.format( + 'is using backup {0}, and is still running'.format( backup_id) in e.message and - 'ERROR: Cannot lock backup directory' in e.message, + 'ERROR: Cannot lock backup' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) # Clean after yourself - self.del_test_dir(module_name, fname) + gdb.kill() def test_locking_concurrent_validate_and_backup(self): """ @@ -412,12 +404,13 @@ def test_locking_concurrent_validate_and_backup(self): and stop it in the middle, take page backup. Expect PAGE backup to be successfully executed """ - fname = self.id().split('.')[3] + self._check_gdb_flag_or_skip_test() + node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -439,4 +432,198 @@ def test_locking_concurrent_validate_and_backup(self): self.backup_node(backup_dir, 'node', node, backup_type='page') # Clean after yourself - self.del_test_dir(module_name, fname) + gdb.kill() + + def test_locking_concurren_restore_and_delete(self): + """ + make node, take full backup, launch restore + and stop it in the middle, delete full backup. + Expect it to fail. + """ + self._check_gdb_flag_or_skip_test() + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL + full_id = self.backup_node(backup_dir, 'node', node) + + node.cleanup() + gdb = self.restore_node(backup_dir, 'node', node, gdb=True) + + gdb.set_breakpoint('create_data_directories') + gdb.run_until_break() + + try: + self.delete_pb(backup_dir, 'node', full_id) + self.assertEqual( + 1, 0, + "Expecting Error because backup is locked\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Cannot lock backup {0} directory".format(full_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # Clean after yourself + gdb.kill() + + def test_backup_directory_name(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL + full_id_1 = self.backup_node(backup_dir, 'node', node) + page_id_1 = self.backup_node(backup_dir, 'node', node, backup_type='page') + + full_id_2 = self.backup_node(backup_dir, 'node', node) + page_id_2 = self.backup_node(backup_dir, 'node', node, backup_type='page') + + node.cleanup() + + old_path = os.path.join(backup_dir, 'backups', 'node', full_id_1) + new_path = os.path.join(backup_dir, 'backups', 'node', 'hello_kitty') + + os.rename(old_path, new_path) + + # This PAGE backup is expected to be successfull + self.show_pb(backup_dir, 'node', full_id_1) + + self.validate_pb(backup_dir) + self.validate_pb(backup_dir, 'node') + self.validate_pb(backup_dir, 'node', full_id_1) + + self.restore_node(backup_dir, 'node', node, backup_id=full_id_1) + + self.delete_pb(backup_dir, 'node', full_id_1) + + old_path = os.path.join(backup_dir, 'backups', 'node', full_id_2) + new_path = os.path.join(backup_dir, 'backups', 'node', 'hello_kitty') + + self.set_backup( + backup_dir, 'node', full_id_2, options=['--note=hello']) + + self.merge_backup(backup_dir, 'node', page_id_2, options=["-j", "4"]) + + self.assertNotIn( + 'note', + self.show_pb(backup_dir, 'node', page_id_2)) + + # Clean after yourself + + def test_empty_lock_file(self): + """ + https://github.com/postgrespro/pg_probackup/issues/308 + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Fill with data + node.pgbench_init(scale=100) + + # FULL + backup_id = self.backup_node(backup_dir, 'node', node) + + lockfile = os.path.join(backup_dir, 'backups', 'node', backup_id, 'backup.pid') + with open(lockfile, "w+") as f: + f.truncate() + + out = self.validate_pb(backup_dir, 'node', backup_id) + + self.assertIn( + "Waiting 30 seconds on empty exclusive lock for backup", out) + +# lockfile = os.path.join(backup_dir, 'backups', 'node', backup_id, 'backup.pid') +# with open(lockfile, "w+") as f: +# f.truncate() +# +# p1 = self.validate_pb(backup_dir, 'node', backup_id, asynchronous=True, +# options=['--log-level-file=LOG', '--log-filename=validate.log']) +# sleep(3) +# p2 = self.delete_pb(backup_dir, 'node', backup_id, asynchronous=True, +# options=['--log-level-file=LOG', '--log-filename=delete.log']) +# +# p1.wait() +# p2.wait() + + def test_shared_lock(self): + """ + Make sure that shared lock leaves no files with pids + """ + self._check_gdb_flag_or_skip_test() + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Fill with data + node.pgbench_init(scale=1) + + # FULL + backup_id = self.backup_node(backup_dir, 'node', node) + + lockfile_excl = os.path.join(backup_dir, 'backups', 'node', backup_id, 'backup.pid') + lockfile_shr = os.path.join(backup_dir, 'backups', 'node', backup_id, 'backup_ro.pid') + + self.validate_pb(backup_dir, 'node', backup_id) + + self.assertFalse( + os.path.exists(lockfile_excl), + "File should not exist: {0}".format(lockfile_excl)) + + self.assertFalse( + os.path.exists(lockfile_shr), + "File should not exist: {0}".format(lockfile_shr)) + + gdb = self.validate_pb(backup_dir, 'node', backup_id, gdb=True) + + gdb.set_breakpoint('validate_one_page') + gdb.run_until_break() + gdb.kill() + + self.assertTrue( + os.path.exists(lockfile_shr), + "File should exist: {0}".format(lockfile_shr)) + + self.validate_pb(backup_dir, 'node', backup_id) + + self.assertFalse( + os.path.exists(lockfile_excl), + "File should not exist: {0}".format(lockfile_excl)) + + self.assertFalse( + os.path.exists(lockfile_shr), + "File should not exist: {0}".format(lockfile_shr)) + diff --git a/tests/logging.py b/tests/logging_test.py similarity index 68% rename from tests/logging.py rename to tests/logging_test.py index efde1d0b9..c5cdfa344 100644 --- a/tests/logging.py +++ b/tests/logging_test.py @@ -3,22 +3,22 @@ from .helpers.ptrack_helpers import ProbackupTest, ProbackupException import datetime -module_name = 'logging' - - class LogTest(ProbackupTest, unittest.TestCase): # @unittest.skip("skip") # @unittest.expectedFailure # PGPRO-2154 def test_log_rotation(self): - fname = self.id().split('.')[3] + """ + """ + self._check_gdb_flag_or_skip_test() + node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -39,17 +39,13 @@ def test_log_rotation(self): gdb.run_until_break() gdb.continue_execution_until_exit() - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_log_filename_strftime(self): - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -72,17 +68,13 @@ def test_log_filename_strftime(self): self.assertTrue(os.path.isfile(path)) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_truncate_rotation_file(self): - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -147,17 +139,13 @@ def test_truncate_rotation_file(self): self.assertTrue(os.path.isfile(rotation_file_path)) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_unlink_rotation_file(self): - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -219,17 +207,13 @@ def test_unlink_rotation_file(self): os.stat(log_file_path).st_size, log_file_size) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_garbage_in_rotation_file(self): - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -255,11 +239,8 @@ def test_garbage_in_rotation_file(self): self.assertTrue(os.path.isfile(rotation_file_path)) # mangle .rotation file - with open(rotation_file_path, "wt", 0) as f: + with open(rotation_file_path, "w+b", 0) as f: f.write(b"blah") - f.flush() - f.close - output = self.backup_node( backup_dir, 'node', node, options=[ @@ -298,5 +279,67 @@ def test_garbage_in_rotation_file(self): os.stat(log_file_path).st_size, log_file_size) - # Clean after yourself - self.del_test_dir(module_name, fname) + def test_issue_274(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + self.restore_node(backup_dir, 'node', replica) + + # Settings for Replica + self.set_replica(node, replica, synchronous=True) + self.set_archiving(backup_dir, 'node', replica, replica=True) + self.set_auto_conf(replica, {'port': replica.port}) + + replica.slow_start(replica=True) + + node.safe_psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,45600) i") + + log_dir = os.path.join(backup_dir, "somedir") + + try: + self.backup_node( + backup_dir, 'node', replica, backup_type='page', + options=[ + '--log-level-console=verbose', '--log-level-file=verbose', + '--log-directory={0}'.format(log_dir), '-j1', + '--log-filename=somelog.txt', '--archive-timeout=5s', + '--no-validate', '--log-rotation-size=100KB']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of archiving timeout" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: WAL segment', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + log_file_path = os.path.join( + log_dir, 'somelog.txt') + + self.assertTrue(os.path.isfile(log_file_path)) + + with open(log_file_path, "r+") as f: + log_content = f.read() + + self.assertIn('INFO: command:', log_content) diff --git a/tests/merge.py b/tests/merge_test.py similarity index 62% rename from tests/merge.py rename to tests/merge_test.py index d17d6b603..1d40af7f7 100644 --- a/tests/merge.py +++ b/tests/merge_test.py @@ -3,12 +3,11 @@ import unittest import os from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +from testgres import QueryException import shutil from datetime import datetime, timedelta import time - -module_name = "merge" - +import subprocess class MergeTest(ProbackupTest, unittest.TestCase): @@ -16,14 +15,12 @@ def test_basic_merge_full_page(self): """ Test MERGE command, it merges FULL backup with target PAGE backups """ - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, "backup") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, "backup") # Initialize instance and backup directory node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=["--data-checksums"] - ) + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=["--data-checksums"]) self.init_pb(backup_dir) self.add_instance(backup_dir, "node", node) @@ -31,7 +28,7 @@ def test_basic_merge_full_page(self): node.slow_start() # Do full backup - self.backup_node(backup_dir, "node", node) + self.backup_node(backup_dir, "node", node, options=['--compress']) show_backup = self.show_pb(backup_dir, "node")[0] self.assertEqual(show_backup["status"], "OK") @@ -45,7 +42,7 @@ def test_basic_merge_full_page(self): conn.commit() # Do first page backup - self.backup_node(backup_dir, "node", node, backup_type="page") + self.backup_node(backup_dir, "node", node, backup_type="page", options=['--compress']) show_backup = self.show_pb(backup_dir, "node")[1] # sanity check @@ -60,7 +57,9 @@ def test_basic_merge_full_page(self): conn.commit() # Do second page backup - self.backup_node(backup_dir, "node", node, backup_type="page") + self.backup_node( + backup_dir, "node", node, + backup_type="page", options=['--compress']) show_backup = self.show_pb(backup_dir, "node")[2] page_id = show_backup["id"] @@ -97,22 +96,16 @@ def test_basic_merge_full_page(self): count2 = node.execute("postgres", "select count(*) from test") self.assertEqual(count1, count2) - # Clean after yourself - node.cleanup() - self.del_test_dir(module_name, fname) - def test_merge_compressed_backups(self): """ Test MERGE command with compressed backups """ - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, "backup") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, "backup") # Initialize instance and backup directory node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=["--data-checksums"] - ) + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=["--data-checksums"]) self.init_pb(backup_dir) self.add_instance(backup_dir, "node", node) @@ -120,8 +113,7 @@ def test_merge_compressed_backups(self): node.slow_start() # Do full compressed backup - self.backup_node(backup_dir, "node", node, options=[ - '--compress-algorithm=zlib']) + self.backup_node(backup_dir, "node", node, options=['--compress']) show_backup = self.show_pb(backup_dir, "node")[0] self.assertEqual(show_backup["status"], "OK") @@ -137,8 +129,7 @@ def test_merge_compressed_backups(self): # Do compressed page backup self.backup_node( - backup_dir, "node", node, backup_type="page", - options=['--compress-algorithm=zlib']) + backup_dir, "node", node, backup_type="page", options=['--compress']) show_backup = self.show_pb(backup_dir, "node")[1] page_id = show_backup["id"] @@ -146,7 +137,7 @@ def test_merge_compressed_backups(self): self.assertEqual(show_backup["backup-mode"], "PAGE") # Merge all backups - self.merge_backup(backup_dir, "node", page_id) + self.merge_backup(backup_dir, "node", page_id, options=['-j2']) show_backups = self.show_pb(backup_dir, "node") self.assertEqual(len(show_backups), 1) @@ -164,23 +155,17 @@ def test_merge_compressed_backups(self): # Clean after yourself node.cleanup() - self.del_test_dir(module_name, fname) def test_merge_compressed_backups_1(self): """ Test MERGE command with compressed backups """ - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, "backup") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, "backup") # Initialize instance and backup directory node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, initdb_params=["--data-checksums"], - pg_options={ - 'autovacuum': 'off' - } - ) + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, initdb_params=["--data-checksums"]) self.init_pb(backup_dir) self.add_instance(backup_dir, "node", node) @@ -188,33 +173,31 @@ def test_merge_compressed_backups_1(self): node.slow_start() # Fill with data - node.pgbench_init(scale=5) + node.pgbench_init(scale=10) # Do compressed FULL backup - self.backup_node(backup_dir, "node", node, options=[ - '--compress-algorithm=zlib', '--stream']) + self.backup_node(backup_dir, "node", node, options=['--compress', '--stream']) show_backup = self.show_pb(backup_dir, "node")[0] self.assertEqual(show_backup["status"], "OK") self.assertEqual(show_backup["backup-mode"], "FULL") # Change data - pgbench = node.pgbench(options=['-T', '20', '-c', '2', '--no-vacuum']) + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) pgbench.wait() # Do compressed DELTA backup self.backup_node( - backup_dir, "node", node, backup_type="delta", - options=['--compress-algorithm=zlib', '--stream']) + backup_dir, "node", node, + backup_type="delta", options=['--compress', '--stream']) # Change data - pgbench = node.pgbench(options=['-T', '20', '-c', '2', '--no-vacuum']) + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) pgbench.wait() # Do compressed PAGE backup self.backup_node( - backup_dir, "node", node, backup_type="page", - options=['--compress-algorithm=zlib']) + backup_dir, "node", node, backup_type="page", options=['--compress']) pgdata = self.pgdata_content(node.data_dir) @@ -225,7 +208,7 @@ def test_merge_compressed_backups_1(self): self.assertEqual(show_backup["backup-mode"], "PAGE") # Merge all backups - self.merge_backup(backup_dir, "node", page_id) + self.merge_backup(backup_dir, "node", page_id, options=['-j2']) show_backups = self.show_pb(backup_dir, "node") self.assertEqual(len(show_backups), 1) @@ -241,22 +224,17 @@ def test_merge_compressed_backups_1(self): # Clean after yourself node.cleanup() - self.del_test_dir(module_name, fname) def test_merge_compressed_and_uncompressed_backups(self): """ Test MERGE command with compressed and uncompressed backups """ - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, "backup") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, "backup") # Initialize instance and backup directory node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=["--data-checksums"], - pg_options={ - 'autovacuum': 'off' - } ) self.init_pb(backup_dir) @@ -265,7 +243,7 @@ def test_merge_compressed_and_uncompressed_backups(self): node.slow_start() # Fill with data - node.pgbench_init(scale=5) + node.pgbench_init(scale=10) # Do compressed FULL backup self.backup_node(backup_dir, "node", node, options=[ @@ -276,21 +254,20 @@ def test_merge_compressed_and_uncompressed_backups(self): self.assertEqual(show_backup["backup-mode"], "FULL") # Change data - pgbench = node.pgbench(options=['-T', '20', '-c', '2', '--no-vacuum']) + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) pgbench.wait() # Do compressed DELTA backup self.backup_node( backup_dir, "node", node, backup_type="delta", - options=['--compress-algorithm=zlib', '--stream']) + options=['--compress', '--stream']) # Change data - pgbench = node.pgbench(options=['-T', '20', '-c', '2', '--no-vacuum']) + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) pgbench.wait() # Do uncompressed PAGE backup - self.backup_node( - backup_dir, "node", node, backup_type="page") + self.backup_node(backup_dir, "node", node, backup_type="page") pgdata = self.pgdata_content(node.data_dir) @@ -301,7 +278,7 @@ def test_merge_compressed_and_uncompressed_backups(self): self.assertEqual(show_backup["backup-mode"], "PAGE") # Merge all backups - self.merge_backup(backup_dir, "node", page_id) + self.merge_backup(backup_dir, "node", page_id, options=['-j2']) show_backups = self.show_pb(backup_dir, "node") self.assertEqual(len(show_backups), 1) @@ -317,22 +294,17 @@ def test_merge_compressed_and_uncompressed_backups(self): # Clean after yourself node.cleanup() - self.del_test_dir(module_name, fname) def test_merge_compressed_and_uncompressed_backups_1(self): """ Test MERGE command with compressed and uncompressed backups """ - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, "backup") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, "backup") # Initialize instance and backup directory node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=["--data-checksums"], - pg_options={ - 'autovacuum': 'off' - } ) self.init_pb(backup_dir) @@ -352,7 +324,7 @@ def test_merge_compressed_and_uncompressed_backups_1(self): self.assertEqual(show_backup["backup-mode"], "FULL") # Change data - pgbench = node.pgbench(options=['-T', '20', '-c', '2', '--no-vacuum']) + pgbench = node.pgbench(options=['-T', '20', '-c', '1', '--no-vacuum']) pgbench.wait() # Do uncompressed DELTA backup @@ -361,7 +333,7 @@ def test_merge_compressed_and_uncompressed_backups_1(self): options=['--stream']) # Change data - pgbench = node.pgbench(options=['-T', '20', '-c', '2', '--no-vacuum']) + pgbench = node.pgbench(options=['-T', '20', '-c', '1', '--no-vacuum']) pgbench.wait() # Do compressed PAGE backup @@ -394,22 +366,17 @@ def test_merge_compressed_and_uncompressed_backups_1(self): # Clean after yourself node.cleanup() - self.del_test_dir(module_name, fname) def test_merge_compressed_and_uncompressed_backups_2(self): """ Test MERGE command with compressed and uncompressed backups """ - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, "backup") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, "backup") # Initialize instance and backup directory node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=["--data-checksums"], - pg_options={ - 'autovacuum': 'off' - } ) self.init_pb(backup_dir) @@ -418,7 +385,7 @@ def test_merge_compressed_and_uncompressed_backups_2(self): node.slow_start() # Fill with data - node.pgbench_init(scale=5) + node.pgbench_init(scale=20) # Do uncompressed FULL backup self.backup_node(backup_dir, "node", node) @@ -428,7 +395,7 @@ def test_merge_compressed_and_uncompressed_backups_2(self): self.assertEqual(show_backup["backup-mode"], "FULL") # Change data - pgbench = node.pgbench(options=['-T', '20', '-c', '2', '--no-vacuum']) + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) pgbench.wait() # Do compressed DELTA backup @@ -437,7 +404,7 @@ def test_merge_compressed_and_uncompressed_backups_2(self): options=['--compress-algorithm=zlib', '--stream']) # Change data - pgbench = node.pgbench(options=['-T', '20', '-c', '2', '--no-vacuum']) + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) pgbench.wait() # Do uncompressed PAGE backup @@ -467,11 +434,6 @@ def test_merge_compressed_and_uncompressed_backups_2(self): pgdata_restored = self.pgdata_content(node.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - node.cleanup() - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") def test_merge_tablespaces(self): """ @@ -480,14 +442,10 @@ def test_merge_tablespaces(self): tablespace, take page backup, merge it and restore """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off' - } ) self.init_pb(backup_dir) @@ -558,14 +516,10 @@ def test_merge_tablespaces_1(self): drop first tablespace and take delta backup, merge it and restore """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off' - } ) self.init_pb(backup_dir) @@ -630,9 +584,6 @@ def test_merge_tablespaces_1(self): pgdata_restored = self.pgdata_content(node.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_merge_page_truncate(self): """ make node, create table, take full backup, @@ -640,19 +591,16 @@ def test_merge_page_truncate(self): take page backup, merge full and page, restore last page backup and check data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '300s', - 'autovacuum': 'off' - } - ) + 'checkpoint_timeout': '300s'}) + node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -700,32 +648,22 @@ def test_merge_page_truncate(self): backup_dir, 'node', node_restored, options=[ "-j", "4", - "-T", "{0}={1}".format(old_tablespace, new_tablespace), - "--recovery-target-action=promote"]) + "-T", "{0}={1}".format(old_tablespace, new_tablespace)]) # Physical comparison if self.paranoia: pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() # Logical comparison - result1 = node.safe_psql( - "postgres", - "select * from t_heap") - - result2 = node_restored.safe_psql( - "postgres", - "select * from t_heap") + result1 = node.table_checksum("t_heap") + result2 = node_restored.table_checksum("t_heap") self.assertEqual(result1, result2) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_merge_delta_truncate(self): """ make node, create table, take full backup, @@ -733,19 +671,16 @@ def test_merge_delta_truncate(self): take page backup, merge full and page, restore last page backup and check data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '300s', - 'autovacuum': 'off' - } - ) + 'checkpoint_timeout': '300s'}) + node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -793,32 +728,22 @@ def test_merge_delta_truncate(self): backup_dir, 'node', node_restored, options=[ "-j", "4", - "-T", "{0}={1}".format(old_tablespace, new_tablespace), - "--recovery-target-action=promote"]) + "-T", "{0}={1}".format(old_tablespace, new_tablespace)]) # Physical comparison if self.paranoia: pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() # Logical comparison - result1 = node.safe_psql( - "postgres", - "select * from t_heap") - - result2 = node_restored.safe_psql( - "postgres", - "select * from t_heap") + result1 = node.table_checksum("t_heap") + result2 = node_restored.table_checksum("t_heap") self.assertEqual(result1, result2) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_merge_ptrack_truncate(self): """ make node, create table, take full backup, @@ -827,28 +752,24 @@ def test_merge_ptrack_truncate(self): restore last page backup and check data correctness """ if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') + self.skipTest('Skipped because ptrack support is disabled') - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'autovacuum': 'off', - 'ptrack_enable': 'on' - } - ) - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + ptrack_enable=True) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) - node_restored.cleanup() node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.create_tblspace_in_node(node, 'somedata') node.safe_psql( @@ -883,6 +804,10 @@ def test_merge_ptrack_truncate(self): self.validate_pb(backup_dir) + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + old_tablespace = self.get_tblspace_path(node, 'somedata') new_tablespace = self.get_tblspace_path(node_restored, 'somedata_new') @@ -890,32 +815,22 @@ def test_merge_ptrack_truncate(self): backup_dir, 'node', node_restored, options=[ "-j", "4", - "-T", "{0}={1}".format(old_tablespace, new_tablespace), - "--recovery-target-action=promote"]) + "-T", "{0}={1}".format(old_tablespace, new_tablespace)]) # Physical comparison if self.paranoia: pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() # Logical comparison - result1 = node.safe_psql( - "postgres", - "select * from t_heap") - - result2 = node_restored.safe_psql( - "postgres", - "select * from t_heap") + result1 = node.table_checksum("t_heap") + result2 = node_restored.table_checksum("t_heap") self.assertEqual(result1, result2) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_merge_delta_delete(self): """ @@ -923,14 +838,12 @@ def test_merge_delta_delete(self): alter tablespace location, take delta backup, merge full and delta, restore database. """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'checkpoint_timeout': '30s', - 'autovacuum': 'off' } ) @@ -976,7 +889,7 @@ def test_merge_delta_delete(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') + base_dir=os.path.join(self.module_name, self.fname, 'node_restored') ) node_restored.cleanup() @@ -997,23 +910,20 @@ def test_merge_delta_delete(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_continue_failed_merge(self): """ Check that failed MERGE can be continued """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( base_dir=os.path.join( - module_name, fname, 'node'), + self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -1059,12 +969,14 @@ def test_continue_failed_merge(self): gdb = self.merge_backup(backup_dir, "node", backup_id, gdb=True) - gdb.set_breakpoint('copy_file') + gdb.set_breakpoint('backup_non_data_file_internal') gdb.run_until_break() gdb.continue_execution_until_break(5) gdb._execute('signal SIGKILL') + gdb._execute('detach') + time.sleep(1) print(self.show_pb(backup_dir, as_text=True, as_json=False)) @@ -1075,18 +987,16 @@ def test_continue_failed_merge(self): node.cleanup() self.restore_node(backup_dir, 'node', node) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_continue_failed_merge_with_corrupted_delta_backup(self): """ Fail merge via gdb, corrupt DELTA backup, try to continue merge """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) @@ -1101,42 +1011,37 @@ def test_continue_failed_merge_with_corrupted_delta_backup(self): "postgres", "create table t_heap as select i as id," " md5(i::text) as text, md5(i::text)::tsvector as tsvector" - " from generate_series(0,1000) i" - ) + " from generate_series(0,1000) i") old_path = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap')").rstrip() + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() # DELTA BACKUP self.backup_node( - backup_dir, 'node', node, backup_type='delta' - ) + backup_dir, 'node', node, backup_type='delta') node.safe_psql( "postgres", - "update t_heap set id = 100500" - ) + "update t_heap set id = 100500") node.safe_psql( "postgres", - "vacuum full t_heap" - ) + "vacuum full t_heap") new_path = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap')").rstrip() + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() # DELTA BACKUP backup_id_2 = self.backup_node( - backup_dir, 'node', node, backup_type='delta' - ) + backup_dir, 'node', node, backup_type='delta') backup_id = self.show_pb(backup_dir, "node")[1]["id"] # Failed MERGE gdb = self.merge_backup(backup_dir, "node", backup_id, gdb=True) - gdb.set_breakpoint('copy_file') + gdb.set_breakpoint('backup_non_data_file_internal') gdb.run_until_break() gdb.continue_execution_until_break(2) @@ -1165,7 +1070,7 @@ def test_continue_failed_merge_with_corrupted_delta_backup(self): # Try to continue failed MERGE try: - self.merge_backup(backup_dir, "node", backup_id) + print(self.merge_backup(backup_dir, "node", backup_id)) self.assertEqual( 1, 0, "Expecting Error because of incremental backup corruption.\n " @@ -1173,22 +1078,20 @@ def test_continue_failed_merge_with_corrupted_delta_backup(self): repr(self.output), self.cmd)) except ProbackupException as e: self.assertTrue( - "ERROR: Merging of backup {0} failed".format( + "ERROR: Backup {0} has status CORRUPT, merge is aborted".format( backup_id) in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_continue_failed_merge_2(self): """ Check that failed MERGE on delete can be continued """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) @@ -1232,8 +1135,12 @@ def test_continue_failed_merge_2(self): gdb.run_until_break() + gdb._execute('thread apply all bt') + gdb.continue_execution_until_break(20) + gdb._execute('thread apply all bt') + gdb._execute('signal SIGKILL') print(self.show_pb(backup_dir, as_text=True, as_json=False)) @@ -1244,18 +1151,17 @@ def test_continue_failed_merge_2(self): # Try to continue failed MERGE self.merge_backup(backup_dir, "node", backup_id) - # Clean after yourself - self.del_test_dir(module_name, fname) def test_continue_failed_merge_3(self): """ - Check that failed MERGE can`t be continued after target backup deleting - Create FULL and 2 PAGE backups + Check that failed MERGE cannot be continued if intermediate + backup is missing. """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) self.init_pb(backup_dir) @@ -1312,14 +1218,14 @@ def test_continue_failed_merge_3(self): gdb = self.merge_backup(backup_dir, "node", backup_id_merge, gdb=True) - gdb.set_breakpoint('copy_file') + gdb.set_breakpoint('backup_non_data_file_internal') gdb.run_until_break() gdb.continue_execution_until_break(2) gdb._execute('signal SIGKILL') print(self.show_pb(backup_dir, as_text=True, as_json=False)) - print(os.path.join(backup_dir, "backups", "node", backup_id_delete)) + # print(os.path.join(backup_dir, "backups", "node", backup_id_delete)) # DELETE PAGE1 shutil.rmtree( @@ -1335,22 +1241,18 @@ def test_continue_failed_merge_3(self): repr(self.output), self.cmd)) except ProbackupException as e: self.assertTrue( - "ERROR: Parent full backup for the given backup {0} was not found".format( - backup_id_merge) in e.message, + "ERROR: Incremental chain is broken, " + "merge is impossible to finish" in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_merge_different_compression_algo(self): """ Check that backups with different compression algorithms can be merged """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -1393,17 +1295,14 @@ def test_merge_different_compression_algo(self): self.merge_backup(backup_dir, "node", backup_id) - self.del_test_dir(module_name, fname) - def test_merge_different_wal_modes(self): """ Check that backups with different wal modes can be merged correctly """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -1435,20 +1334,18 @@ def test_merge_different_wal_modes(self): self.assertEqual( 'STREAM', self.show_pb(backup_dir, 'node', backup_id)['wal']) - self.del_test_dir(module_name, fname) - def test_crash_after_opening_backup_control_1(self): """ check that crashing after opening backup.control for writing will not result in losing backup metadata """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1486,20 +1383,20 @@ def test_crash_after_opening_backup_control_1(self): self.assertEqual( 'MERGING', self.show_pb(backup_dir, 'node')[1]['status']) - self.del_test_dir(module_name, fname) - + # @unittest.skip("skip") def test_crash_after_opening_backup_control_2(self): """ check that crashing after opening backup_content.control for writing will not result in losing metadata about backup files + TODO: rewrite """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1519,7 +1416,7 @@ def test_crash_after_opening_backup_control_2(self): path = node.safe_psql( 'postgres', - "select pg_relation_filepath('pgbench_accounts')").rstrip() + "select pg_relation_filepath('pgbench_accounts')").decode('utf-8').rstrip() fsm_path = path + '_fsm' @@ -1540,8 +1437,8 @@ def test_crash_after_opening_backup_control_2(self): gdb.set_breakpoint('write_backup_filelist') gdb.run_until_break() - gdb.set_breakpoint('fio_fwrite') - gdb.continue_execution_until_break(2) +# gdb.set_breakpoint('sprintf') +# gdb.continue_execution_until_break(1) gdb._execute('signal SIGKILL') @@ -1560,7 +1457,7 @@ def test_crash_after_opening_backup_control_2(self): backup_dir, 'backups', 'node', full_id, 'database', fsm_path) - print(file_to_remove) + # print(file_to_remove) os.remove(file_to_remove) @@ -1576,20 +1473,20 @@ def test_crash_after_opening_backup_control_2(self): self.compare_pgdata(pgdata, pgdata_restored) - self.del_test_dir(module_name, fname) - + # @unittest.skip("skip") def test_losing_file_after_failed_merge(self): """ check that crashing after opening backup_content.control for writing will not result in losing metadata about backup files + TODO: rewrite """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1610,7 +1507,7 @@ def test_losing_file_after_failed_merge(self): path = node.safe_psql( 'postgres', - "select pg_relation_filepath('pgbench_accounts')").rstrip() + "select pg_relation_filepath('pgbench_accounts')").decode('utf-8').rstrip() node.safe_psql( 'postgres', @@ -1631,8 +1528,8 @@ def test_losing_file_after_failed_merge(self): gdb.set_breakpoint('write_backup_filelist') gdb.run_until_break() - gdb.set_breakpoint('fio_fwrite') - gdb.continue_execution_until_break(2) +# gdb.set_breakpoint('sprintf') +# gdb.continue_execution_until_break(20) gdb._execute('signal SIGKILL') @@ -1666,18 +1563,16 @@ def test_losing_file_after_failed_merge(self): pgdata_restored = self.pgdata_content(node.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - self.del_test_dir(module_name, fname) - def test_failed_merge_after_delete(self): """ """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1691,7 +1586,7 @@ def test_failed_merge_after_delete(self): dboid = node.safe_psql( "postgres", - "select oid from pg_database where datname = 'testdb'").rstrip() + "select oid from pg_database where datname = 'testdb'").decode('utf-8').rstrip() # take FULL backup full_id = self.backup_node( @@ -1716,9 +1611,6 @@ def test_failed_merge_after_delete(self): gdb.set_breakpoint('delete_backup_files') gdb.run_until_break() - gdb.set_breakpoint('parray_bsearch') - gdb.continue_execution_until_break() - gdb.set_breakpoint('pgFileDelete') gdb.continue_execution_until_break(20) @@ -1726,7 +1618,7 @@ def test_failed_merge_after_delete(self): # backup half-merged self.assertEqual( - 'MERGING', self.show_pb(backup_dir, 'node')[0]['status']) + 'MERGED', self.show_pb(backup_dir, 'node')[0]['status']) self.assertEqual( full_id, self.show_pb(backup_dir, 'node')[0]['id']) @@ -1746,24 +1638,21 @@ def test_failed_merge_after_delete(self): repr(self.output), self.cmd)) except ProbackupException as e: self.assertTrue( - "ERROR: Parent full backup for the given " - "backup {0} was not found".format( - page_id_2) in e.message, + "ERROR: Full backup {0} has unfinished merge with backup {1}".format( + full_id, page_id) in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - self.del_test_dir(module_name, fname) - def test_failed_merge_after_delete_1(self): """ """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -1779,7 +1668,7 @@ def test_failed_merge_after_delete_1(self): page_1 = self.backup_node( backup_dir, 'node', node, backup_type='page') - # Change FULL B backup status to ERROR + # Change PAGE1 backup status to ERROR self.change_backup_status(backup_dir, 'node', page_1, 'ERROR') pgdata = self.pgdata_content(node.data_dir) @@ -1788,11 +1677,11 @@ def test_failed_merge_after_delete_1(self): pgbench = node.pgbench(options=['-T', '10', '-c', '2', '--no-vacuum']) pgbench.wait() - # take PAGE backup + # take PAGE2 backup page_id = self.backup_node( backup_dir, 'node', node, backup_type='page') - # Change FULL B backup status to ERROR + # Change PAGE1 backup status to OK self.change_backup_status(backup_dir, 'node', page_1, 'OK') gdb = self.merge_backup( @@ -1802,8 +1691,8 @@ def test_failed_merge_after_delete_1(self): gdb.set_breakpoint('delete_backup_files') gdb.run_until_break() - gdb.set_breakpoint('parray_bsearch') - gdb.continue_execution_until_break() +# gdb.set_breakpoint('parray_bsearch') +# gdb.continue_execution_until_break() gdb.set_breakpoint('pgFileDelete') gdb.continue_execution_until_break(30) @@ -1815,6 +1704,7 @@ def test_failed_merge_after_delete_1(self): # restore node.cleanup() try: + #self.restore_node(backup_dir, 'node', node, backup_id=page_1) self.restore_node(backup_dir, 'node', node) self.assertEqual( 1, 0, @@ -1828,123 +1718,277 @@ def test_failed_merge_after_delete_1(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_merge_backup_from_future(self): + def test_failed_merge_after_delete_2(self): """ - take FULL backup, table PAGE backup from future, - try to merge page with FULL """ - fname = self.id().split('.')[3] + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() - # Take FULL - self.backup_node(backup_dir, 'node', node) + # take FULL backup + full_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) - node.pgbench_init(scale=3) + node.pgbench_init(scale=1) - # Take PAGE from future - backup_id = self.backup_node( + page_1 = self.backup_node( backup_dir, 'node', node, backup_type='page') - with open( - os.path.join( - backup_dir, 'backups', 'node', - backup_id, "backup.control"), "a") as conf: - conf.write("start-time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() + timedelta(days=3))) - - # rename directory - new_id = self.show_pb(backup_dir, 'node')[1]['id'] - - os.rename( - os.path.join(backup_dir, 'backups', 'node', backup_id), - os.path.join(backup_dir, 'backups', 'node', new_id)) - - pgbench = node.pgbench(options=['-T', '3', '-c', '2', '--no-vacuum']) + # add data + pgbench = node.pgbench(options=['-T', '10', '-c', '2', '--no-vacuum']) pgbench.wait() - backup_id = self.backup_node( + # take PAGE2 backup + page_2 = self.backup_node( backup_dir, 'node', node, backup_type='page') - pgdata = self.pgdata_content(node.data_dir) - - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - node_restored.cleanup() - self.restore_node( - backup_dir, 'node', - node_restored, backup_id=backup_id) + gdb = self.merge_backup( + backup_dir, 'node', page_2, gdb=True, + options=['--log-level-console=VERBOSE']) - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) + gdb.set_breakpoint('pgFileDelete') + gdb.run_until_break() + gdb.continue_execution_until_break(2) + gdb._execute('signal SIGKILL') - # check that merged backup has the same state as - node_restored.cleanup() - self.merge_backup(backup_dir, 'node', backup_id=backup_id) - self.restore_node( - backup_dir, 'node', - node_restored, backup_id=backup_id) - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) + self.delete_pb(backup_dir, 'node', backup_id=page_2) - # Clean after yourself - self.del_test_dir(module_name, fname) + # rerun merge + try: + #self.restore_node(backup_dir, 'node', node, backup_id=page_1) + self.merge_backup(backup_dir, 'node', page_1) + self.assertEqual( + 1, 0, + "Expecting Error because of backup is missing.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Full backup {0} has unfinished merge " + "with backup {1}".format(full_id, page_2), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) - # @unittest.skip("skip") - def test_merge_multiple_descendants(self): + def test_failed_merge_after_delete_3(self): """ - PAGEb3 - | PAGEa3 - PAGEb2 / - | PAGEa2 / - PAGEb1 \ / - | PAGEa1 - FULLb | - FULLa """ - fname = self.id().split('.')[3] + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() - # Take FULL BACKUPs - backup_id_a = self.backup_node(backup_dir, 'node', node) + # add database + node.safe_psql( + 'postgres', + 'CREATE DATABASE testdb') - backup_id_b = self.backup_node(backup_dir, 'node', node) + dboid = node.safe_psql( + "postgres", + "select oid from pg_database where datname = 'testdb'").rstrip() - # Change FULLb backup status to ERROR - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + # take FULL backup + full_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) - page_id_a1 = self.backup_node( + # drop database + node.safe_psql( + 'postgres', + 'DROP DATABASE testdb') + + # take PAGE backup + page_id = self.backup_node( backup_dir, 'node', node, backup_type='page') - # Change FULLb backup status to OK - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + # create database + node.safe_psql( + 'postgres', + 'create DATABASE testdb') - # Change PAGEa1 backup status to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') + page_id_2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') - # PAGEa1 ERROR - # FULLb OK + gdb = self.merge_backup( + backup_dir, 'node', page_id, + gdb=True, options=['--log-level-console=verbose']) + + gdb.set_breakpoint('delete_backup_files') + gdb.run_until_break() + + gdb.set_breakpoint('pgFileDelete') + gdb.continue_execution_until_break(20) + + gdb._execute('signal SIGKILL') + + # backup half-merged + self.assertEqual( + 'MERGED', self.show_pb(backup_dir, 'node')[0]['status']) + + self.assertEqual( + full_id, self.show_pb(backup_dir, 'node')[0]['id']) + + db_path = os.path.join( + backup_dir, 'backups', 'node', full_id) + + # FULL backup is missing now + shutil.rmtree(db_path) + + try: + self.merge_backup( + backup_dir, 'node', page_id_2, + options=['--log-level-console=verbose']) + self.assertEqual( + 1, 0, + "Expecting Error because of missing parent.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertTrue( + "ERROR: Failed to find parent full backup for {0}".format( + page_id_2) in e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # Skipped, because backups from the future are invalid. + # This cause a "ERROR: Can't assign backup_id, there is already a backup in future" + # now (PBCKP-259). We can conduct such a test again when we + # untie 'backup_id' from 'start_time' + @unittest.skip("skip") + def test_merge_backup_from_future(self): + """ + take FULL backup, table PAGE backup from future, + try to merge page with FULL + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL + self.backup_node(backup_dir, 'node', node) + + node.pgbench_init(scale=5) + + # Take PAGE from future + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + with open( + os.path.join( + backup_dir, 'backups', 'node', + backup_id, "backup.control"), "a") as conf: + conf.write("start-time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() + timedelta(days=3))) + + # rename directory + new_id = self.show_pb(backup_dir, 'node')[1]['id'] + + os.rename( + os.path.join(backup_dir, 'backups', 'node', backup_id), + os.path.join(backup_dir, 'backups', 'node', new_id)) + + pgbench = node.pgbench(options=['-T', '5', '-c', '1', '--no-vacuum']) + pgbench.wait() + + backup_id = self.backup_node(backup_dir, 'node', node, backup_type='page') + pgdata = self.pgdata_content(node.data_dir) + + result = node.table_checksum("pgbench_accounts") + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', + node_restored, backup_id=backup_id) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # check that merged backup has the same state as + node_restored.cleanup() + self.merge_backup(backup_dir, 'node', backup_id=backup_id) + self.restore_node( + backup_dir, 'node', + node_restored, backup_id=backup_id) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + + self.set_auto_conf( + node_restored, + {'port': node_restored.port}) + node_restored.slow_start() + + result_new = node_restored.table_checksum("pgbench_accounts") + + self.assertEqual(result, result_new) + + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_merge_multiple_descendants(self): + """ + PAGEb3 + | PAGEa3 + PAGEb2 / + | PAGEa2 / + PAGEb1 \ / + | PAGEa1 + FULLb | + FULLa + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL BACKUPs + backup_id_a = self.backup_node(backup_dir, 'node', node) + + backup_id_b = self.backup_node(backup_dir, 'node', node) + + # Change FULLb backup status to ERROR + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + page_id_a1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Change FULLb backup status to OK + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # Change PAGEa1 backup status to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') + + # PAGEa1 ERROR + # FULLb OK # FULLa OK page_id_b1 = self.backup_node( @@ -2084,15 +2128,11 @@ def test_merge_multiple_descendants(self): repr(self.output), self.cmd)) except ProbackupException as e: self.assertTrue( - "ERROR: Parent full backup for the given " - "backup {0} was not found".format( + "ERROR: Failed to find parent full backup for {0}".format( page_id_a3) in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_smart_merge(self): """ @@ -2102,13 +2142,12 @@ def test_smart_merge(self): copied during restore https://github.com/postgrespro/pg_probackup/issues/63 """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -2150,20 +2189,593 @@ def test_smart_merge(self): with open(logfile, 'r') as f: logfile_content = f.read() - # Clean after yourself - self.del_test_dir(module_name, fname) + def test_idempotent_merge(self): + """ + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # add database + node.safe_psql( + 'postgres', + 'CREATE DATABASE testdb') + + # take FULL backup + full_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + # create database + node.safe_psql( + 'postgres', + 'create DATABASE testdb1') + + # take PAGE backup + page_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # create database + node.safe_psql( + 'postgres', + 'create DATABASE testdb2') + + page_id_2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + gdb = self.merge_backup( + backup_dir, 'node', page_id_2, + gdb=True, options=['--log-level-console=verbose']) + + gdb.set_breakpoint('delete_backup_files') + gdb.run_until_break() + gdb.remove_all_breakpoints() + + gdb.set_breakpoint('rename') + gdb.continue_execution_until_break() + gdb.continue_execution_until_break(2) + + gdb._execute('signal SIGKILL') + + show_backups = self.show_pb(backup_dir, "node") + self.assertEqual(len(show_backups), 1) + + self.assertEqual( + 'MERGED', self.show_pb(backup_dir, 'node')[0]['status']) + + self.assertEqual( + full_id, self.show_pb(backup_dir, 'node')[0]['id']) + + self.merge_backup(backup_dir, 'node', page_id_2) + + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'node')[0]['status']) + + self.assertEqual( + page_id_2, self.show_pb(backup_dir, 'node')[0]['id']) + + def test_merge_correct_inheritance(self): + """ + Make sure that backup metainformation fields + 'note' and 'expire-time' are correctly inherited + during merge + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # add database + node.safe_psql( + 'postgres', + 'CREATE DATABASE testdb') + + # take FULL backup + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + # create database + node.safe_psql( + 'postgres', + 'create DATABASE testdb1') + + # take PAGE backup + page_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + self.set_backup( + backup_dir, 'node', page_id, options=['--note=hello', '--ttl=20d']) + + page_meta = self.show_pb(backup_dir, 'node', page_id) + + self.merge_backup(backup_dir, 'node', page_id) + + print(self.show_pb(backup_dir, 'node', page_id)) + + self.assertEqual( + page_meta['note'], + self.show_pb(backup_dir, 'node', page_id)['note']) + + self.assertEqual( + page_meta['expire-time'], + self.show_pb(backup_dir, 'node', page_id)['expire-time']) + + def test_merge_correct_inheritance_1(self): + """ + Make sure that backup metainformation fields + 'note' and 'expire-time' are correctly inherited + during merge + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # add database + node.safe_psql( + 'postgres', + 'CREATE DATABASE testdb') + + # take FULL backup + self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--note=hello', '--ttl=20d']) + + # create database + node.safe_psql( + 'postgres', + 'create DATABASE testdb1') + + # take PAGE backup + page_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + self.merge_backup(backup_dir, 'node', page_id) + + self.assertNotIn( + 'note', + self.show_pb(backup_dir, 'node', page_id)) + + self.assertNotIn( + 'expire-time', + self.show_pb(backup_dir, 'node', page_id)) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_multi_timeline_merge(self): + """ + Check that backup in PAGE mode choose + parent backup correctly: + t12 /---P--> + ... + t3 /----> + t2 /----> + t1 -F-----D-> + + P must have F as parent + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql("postgres", "create extension pageinspect") + + try: + node.safe_psql( + "postgres", + "create extension amcheck") + except QueryException as e: + node.safe_psql( + "postgres", + "create extension amcheck_next") + + node.pgbench_init(scale=20) + full_id = self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + self.backup_node(backup_dir, 'node', node, backup_type='delta') + + node.cleanup() + self.restore_node( + backup_dir, 'node', node, backup_id=full_id, + options=[ + '--recovery-target=immediate', + '--recovery-target-action=promote']) + + node.slow_start() + + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + # create timelines + for i in range(2, 7): + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target=latest', + '--recovery-target-action=promote', + '--recovery-target-timeline={0}'.format(i)]) + node.slow_start() + + # at this point there is i+1 timeline + pgbench = node.pgbench(options=['-T', '20', '-c', '1', '--no-vacuum']) + pgbench.wait() + + # create backup at 2, 4 and 6 timeline + if i % 2 == 0: + self.backup_node(backup_dir, 'node', node, backup_type='page') + + page_id = self.backup_node(backup_dir, 'node', node, backup_type='page') + pgdata = self.pgdata_content(node.data_dir) + + self.merge_backup(backup_dir, 'node', page_id) + + result = node.table_checksum("pgbench_accounts") + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node(backup_dir, 'node', node_restored) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() + + result_new = node_restored.table_checksum("pgbench_accounts") + + self.assertEqual(result, result_new) + + self.compare_pgdata(pgdata, pgdata_restored) + + self.checkdb_node( + backup_dir, + 'node', + options=[ + '--amcheck', + '-d', 'postgres', '-p', str(node.port)]) + + self.checkdb_node( + backup_dir, + 'node', + options=[ + '--amcheck', + '-d', 'postgres', '-p', str(node_restored.port)]) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_merge_page_header_map_retry(self): + """ + page header map cannot be trusted when + running retry + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=20) + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + delta_id = self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + gdb = self.merge_backup(backup_dir, 'node', delta_id, gdb=True) + + # our goal here is to get full backup with merged data files, + # but with old page header map + gdb.set_breakpoint('cleanup_header_map') + gdb.run_until_break() + gdb._execute('signal SIGKILL') + + self.merge_backup(backup_dir, 'node', delta_id) + + node.cleanup() + + self.restore_node(backup_dir, 'node', node) + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_missing_data_file(self): + """ + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Add data + node.pgbench_init(scale=1) + + # FULL backup + self.backup_node(backup_dir, 'node', node) + + # Change data + pgbench = node.pgbench(options=['-T', '5', '-c', '1']) + pgbench.wait() + + # DELTA backup + delta_id = self.backup_node(backup_dir, 'node', node, backup_type='delta') + + path = node.safe_psql( + 'postgres', + "select pg_relation_filepath('pgbench_accounts')").decode('utf-8').rstrip() + + gdb = self.merge_backup( + backup_dir, "node", delta_id, + options=['--log-level-file=VERBOSE'], gdb=True) + gdb.set_breakpoint('merge_files') + gdb.run_until_break() + + # remove data file in incremental backup + file_to_remove = os.path.join( + backup_dir, 'backups', + 'node', delta_id, 'database', path) + + os.remove(file_to_remove) + + gdb.continue_execution_until_error() + + logfile = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(logfile, 'r') as f: + logfile_content = f.read() + + self.assertIn( + 'ERROR: Cannot open backup file "{0}": No such file or directory'.format(file_to_remove), + logfile_content) + + # @unittest.skip("skip") + def test_missing_non_data_file(self): + """ + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + self.backup_node(backup_dir, 'node', node) + + # DELTA backup + delta_id = self.backup_node(backup_dir, 'node', node, backup_type='delta') + + gdb = self.merge_backup( + backup_dir, "node", delta_id, + options=['--log-level-file=VERBOSE'], gdb=True) + gdb.set_breakpoint('merge_files') + gdb.run_until_break() + + # remove data file in incremental backup + file_to_remove = os.path.join( + backup_dir, 'backups', + 'node', delta_id, 'database', 'backup_label') + + os.remove(file_to_remove) + + gdb.continue_execution_until_error() + + logfile = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(logfile, 'r') as f: + logfile_content = f.read() + + self.assertIn( + 'ERROR: File "{0}" is not found'.format(file_to_remove), + logfile_content) + + self.assertIn( + 'ERROR: Backup files merging failed', + logfile_content) + + self.assertEqual( + 'MERGING', self.show_pb(backup_dir, 'node')[0]['status']) + + self.assertEqual( + 'MERGING', self.show_pb(backup_dir, 'node')[1]['status']) + + # @unittest.skip("skip") + def test_merge_remote_mode(self): + """ + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + full_id = self.backup_node(backup_dir, 'node', node) + + # DELTA backup + delta_id = self.backup_node(backup_dir, 'node', node, backup_type='delta') + + self.set_config(backup_dir, 'node', options=['--retention-window=1']) + + backups = os.path.join(backup_dir, 'backups', 'node') + with open( + os.path.join( + backups, full_id, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=5))) + + gdb = self.backup_node( + backup_dir, "node", node, + options=['--log-level-file=VERBOSE', '--merge-expired'], gdb=True) + gdb.set_breakpoint('merge_files') + gdb.run_until_break() + + logfile = os.path.join(backup_dir, 'log', 'pg_probackup.log') + + with open(logfile, "w+") as f: + f.truncate() + + gdb.continue_execution_until_exit() + + logfile = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(logfile, 'r') as f: + logfile_content = f.read() + + self.assertNotIn( + 'SSH', logfile_content) + + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'node')[0]['status']) + + def test_merge_pg_filenode_map(self): + """ + https://github.com/postgrespro/pg_probackup/issues/320 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1'), + initdb_params=['--data-checksums']) + node1.cleanup() + + node.pgbench_init(scale=5) + + # FULL backup + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '1']) + + self.backup_node(backup_dir, 'node', node, backup_type='delta') + + node.safe_psql( + 'postgres', + 'reindex index pg_type_oid_index') + + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + self.merge_backup(backup_dir, 'node', backup_id) + + node.cleanup() + + self.restore_node(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + 'select 1') + + def test_unfinished_merge(self): + """ Test when parent has unfinished merge with a different backup. """ + self._check_gdb_flag_or_skip_test() + cases = [('fail_merged', 'write_backup_filelist', ['MERGED', 'MERGING', 'OK']), + ('fail_merging', 'pgBackupWriteControl', ['MERGING', 'OK', 'OK'])] + + for name, terminate_at, states in cases: + node_name = 'node_' + name + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, name) + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, node_name), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, node_name, node) + self.set_archiving(backup_dir, node_name, node) + node.slow_start() + + full_id=self.backup_node(backup_dir, node_name, node, options=['--stream']) + + backup_id = self.backup_node(backup_dir, node_name, node, backup_type='delta') + second_backup_id = self.backup_node(backup_dir, node_name, node, backup_type='delta') + + gdb = self.merge_backup(backup_dir, node_name, backup_id, gdb=True) + gdb.set_breakpoint(terminate_at) + gdb.run_until_break() + gdb.remove_all_breakpoints() + gdb._execute('signal SIGINT') + gdb.continue_execution_until_error() -# 1. always use parent link when merging (intermediates may be from different chain) -# 2. page backup we are merging with may disappear after failed merge, -# it should not be possible to continue merge after that -# PAGE_A MERGING (disappear) -# FULL MERGING + print(self.show_pb(backup_dir, node_name, as_json=False, as_text=True)) -# FULL MERGING + backup_infos = self.show_pb(backup_dir, node_name) + self.assertEqual(len(backup_infos), len(states)) + for expected, real in zip(states, backup_infos): + self.assertEqual(expected, real['status']) -# PAGE_B OK (new backup) -# FULL MERGING + with self.assertRaisesRegex(ProbackupException, + f"Full backup {full_id} has unfinished merge with backup {backup_id}"): + self.merge_backup(backup_dir, node_name, second_backup_id, gdb=False) -# 3. Need new test with corrupted FULL backup -# 4. different compression levels +# 1. Need new test with corrupted FULL backup +# 2. different compression levels diff --git a/tests/option.py b/tests/option_test.py similarity index 71% rename from tests/option.py rename to tests/option_test.py index 9cf0b8f36..d1e8cb3a6 100644 --- a/tests/option.py +++ b/tests/option_test.py @@ -1,10 +1,7 @@ import unittest import os from .helpers.ptrack_helpers import ProbackupTest, ProbackupException - - -module_name = 'option' - +import locale class OptionTest(ProbackupTest, unittest.TestCase): @@ -12,49 +9,33 @@ class OptionTest(ProbackupTest, unittest.TestCase): # @unittest.expectedFailure def test_help_1(self): """help options""" - self.maxDiff = None - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') with open(os.path.join(self.dir_path, "expected/option_help.out"), "rb") as help_out: self.assertEqual( self.run_pb(["--help"]), help_out.read().decode("utf-8") ) - # @unittest.skip("skip") - def test_version_2(self): - """help options""" - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - with open(os.path.join(self.dir_path, "expected/option_version.out"), "rb") as version_out: - self.assertIn( - version_out.read().decode("utf-8"), - self.run_pb(["--version"]) - ) - # @unittest.skip("skip") def test_without_backup_path_3(self): """backup command failure without backup mode option""" - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') try: self.run_pb(["backup", "-b", "full"]) self.assertEqual(1, 0, "Expecting Error because '-B' parameter is not specified.\n Output: {0} \n CMD: {1}".format( repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( - 'ERROR: required parameter not specified: BACKUP_PATH (-B, --backup-path)', + 'ERROR: No backup catalog path specified.\n' + \ + 'Please specify it either using environment variable BACKUP_PATH or\n' + \ + 'command line option --backup-path (-B)', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) - # @unittest.skip("skip") def test_options_4(self): """check options test""" - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node')) + base_dir=os.path.join(self.module_name, self.fname, 'node')) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -66,7 +47,7 @@ def test_options_4(self): repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( - 'ERROR: required parameter not specified: --instance', + 'ERROR: Required parameter not specified: --instance', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) @@ -77,7 +58,7 @@ def test_options_4(self): repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( - 'ERROR: required parameter not specified: BACKUP_MODE (-b, --backup-mode)', + 'ERROR: No backup mode specified.\nPlease specify it either using environment variable BACKUP_MODE or\ncommand line option --backup-mode (-b)', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) @@ -88,7 +69,7 @@ def test_options_4(self): repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( - 'ERROR: invalid backup-mode "bad"', + 'ERROR: Invalid backup-mode "bad"', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) @@ -100,7 +81,8 @@ def test_options_4(self): repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( - 'ERROR: You must specify at least one of the delete options: --expired |--wal |--merge-expired |--delete-invalid |--backup_id', + 'ERROR: You must specify at least one of the delete options: ' + '--delete-expired |--delete-wal |--merge-expired |--status |(-i, --backup-id)', e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) @@ -113,29 +95,20 @@ def test_options_4(self): repr(self.output), self.cmd)) except ProbackupException as e: self.assertIn( - "option requires an argument -- 'i'", + "Option '-i' requires an argument", e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_options_5(self): """check options test""" - fname = self.id().split(".")[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node')) + base_dir=os.path.join(self.module_name, self.fname, 'node')) output = self.init_pb(backup_dir) - self.assertIn( - "INFO: Backup catalog", - output) + self.assertIn(f"INFO: Backup catalog '{backup_dir}' successfully initialized", output) - self.assertIn( - "successfully inited", - output) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -230,5 +203,51 @@ def test_options_5(self): e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format(repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) + # @unittest.skip("skip") + def test_help_6(self): + """help options""" + if ProbackupTest.enable_nls: + if check_locale('ru_RU.utf-8'): + self.test_env['LC_ALL'] = 'ru_RU.utf-8' + with open(os.path.join(self.dir_path, "expected/option_help_ru.out"), "rb") as help_out: + self.assertEqual( + self.run_pb(["--help"]), + help_out.read().decode("utf-8") + ) + else: + self.skipTest( + "Locale ru_RU.utf-8 doesn't work. You need install ru_RU.utf-8 locale for this test") + else: + self.skipTest( + 'You need configure PostgreSQL with --enabled-nls option for this test') + + # @unittest.skip("skip") + def test_options_no_scale_units(self): + """check --no-scale-units option""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node')) + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + # check that --no-scale-units option works correctly + output = self.run_pb(["show-config", "--backup-path", backup_dir, "--instance=node"]) + self.assertIn(container=output, member="archive-timeout = 5min") + output = self.run_pb(["show-config", "--backup-path", backup_dir, "--instance=node", "--no-scale-units"]) + self.assertIn(container=output, member="archive-timeout = 300") + self.assertNotIn(container=output, member="archive-timeout = 300s") + # check that we have now quotes ("") in json output + output = self.run_pb(["show-config", "--backup-path", backup_dir, "--instance=node", "--no-scale-units", "--format=json"]) + self.assertIn(container=output, member='"archive-timeout": 300,') + self.assertIn(container=output, member='"retention-redundancy": 0,') + self.assertNotIn(container=output, member='"archive-timeout": "300",') + +def check_locale(locale_name): + ret=True + old_locale = locale.setlocale(locale.LC_CTYPE,"") + try: + locale.setlocale(locale.LC_CTYPE, locale_name) + except locale.Error: + ret=False + finally: + locale.setlocale(locale.LC_CTYPE, old_locale) + return ret diff --git a/tests/page.py b/tests/page_test.py similarity index 62% rename from tests/page.py rename to tests/page_test.py index 9a717c234..a66d6d413 100644 --- a/tests/page.py +++ b/tests/page_test.py @@ -6,11 +6,9 @@ import subprocess import gzip import shutil +import time -module_name = 'page' - - -class PageBackupTest(ProbackupTest, unittest.TestCase): +class PageTest(ProbackupTest, unittest.TestCase): # @unittest.skip("skip") def test_basic_page_vacuum_truncate(self): @@ -20,19 +18,16 @@ def test_basic_page_vacuum_truncate(self): take page backup, take second page backup, restore last page backup and check data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '300s', - 'autovacuum': 'off' - } - ) + 'checkpoint_timeout': '300s'}) + node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -55,6 +50,7 @@ def test_basic_page_vacuum_truncate(self): self.backup_node(backup_dir, 'node', node) + # TODO: make it dynamic node.safe_psql( "postgres", "delete from t_heap where ctid >= '(11,0)'") @@ -78,31 +74,90 @@ def test_basic_page_vacuum_truncate(self): backup_dir, 'node', node_restored, options=[ "-j", "4", - "-T", "{0}={1}".format(old_tablespace, new_tablespace), - "--recovery-target-action=promote"]) + "-T", "{0}={1}".format(old_tablespace, new_tablespace)]) # Physical comparison if self.paranoia: pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() # Logical comparison - result1 = node.safe_psql( + result1 = node.table_checksum("t_heap") + result2 = node_restored.table_checksum("t_heap") + + self.assertEqual(result1, result2) + + # @unittest.skip("skip") + def test_page_vacuum_truncate_1(self): + """ + make node, create table, take full backup, + delete all data, vacuum relation, + take page backup, insert some data, + take second page backup and check data correctness + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( "postgres", - "select * from t_heap") + "create sequence t_seq; " + "create table t_heap as select i as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1024) i") - result2 = node_restored.safe_psql( + node.safe_psql( "postgres", - "select * from t_heap") + "vacuum t_heap") - self.assertEqual(result1, result2) + self.backup_node(backup_dir, 'node', node) - # Clean after yourself - self.del_test_dir(module_name, fname) + node.safe_psql( + "postgres", + "delete from t_heap") + + node.safe_psql( + "postgres", + "vacuum t_heap") + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + node.safe_psql( + "postgres", + "insert into t_heap select i as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1) i") + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + pgdata = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node(backup_dir, 'node', node_restored) + + # Physical comparison + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() # @unittest.skip("skip") def test_page_stream(self): @@ -111,10 +166,9 @@ def test_page_stream(self): restore them and check data correctness """ self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ @@ -133,7 +187,7 @@ def test_page_stream(self): "md5(i::text)::tsvector as tsvector " "from generate_series(0,100) i") - full_result = node.execute("postgres", "SELECT * FROM t_heap") + full_result = node.table_checksum("t_heap") full_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='full', options=['--stream']) @@ -144,7 +198,7 @@ def test_page_stream(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector " "from generate_series(100,200) i") - page_result = node.execute("postgres", "SELECT * FROM t_heap") + page_result = node.table_checksum("t_heap") page_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='page', options=['--stream', '-j', '4']) @@ -165,7 +219,7 @@ def test_page_stream(self): ' CMD: {1}'.format(repr(self.output), self.cmd)) node.slow_start() - full_result_new = node.execute("postgres", "SELECT * FROM t_heap") + full_result_new = node.table_checksum("t_heap") self.assertEqual(full_result, full_result_new) node.cleanup() @@ -184,13 +238,10 @@ def test_page_stream(self): self.compare_pgdata(pgdata, pgdata_restored) node.slow_start() - page_result_new = node.execute("postgres", "SELECT * FROM t_heap") + page_result_new = node.table_checksum("t_heap") self.assertEqual(page_result, page_result_new) node.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_page_archive(self): """ @@ -198,10 +249,9 @@ def test_page_archive(self): restore them and check data correctness """ self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ @@ -218,7 +268,7 @@ def test_page_archive(self): "postgres", "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,100) i") - full_result = node.execute("postgres", "SELECT * FROM t_heap") + full_result = node.table_checksum("t_heap") full_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='full') @@ -228,7 +278,7 @@ def test_page_archive(self): "insert into t_heap select i as id, " "md5(i::text) as text, md5(i::text)::tsvector as tsvector " "from generate_series(100, 200) i") - page_result = node.execute("postgres", "SELECT * FROM t_heap") + page_result = node.table_checksum("t_heap") page_backup_id = self.backup_node( backup_dir, 'node', node, backup_type='page', options=["-j", "4"]) @@ -254,7 +304,7 @@ def test_page_archive(self): node.slow_start() - full_result_new = node.execute("postgres", "SELECT * FROM t_heap") + full_result_new = node.table_checksum("t_heap") self.assertEqual(full_result, full_result_new) node.cleanup() @@ -278,33 +328,26 @@ def test_page_archive(self): node.slow_start() - page_result_new = node.execute("postgres", "SELECT * FROM t_heap") + page_result_new = node.table_checksum("t_heap") self.assertEqual(page_result, page_result_new) node.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_page_multiple_segments(self): """ Make node, create table with multiple segments, write some data to it, check page and data correctness """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'fsync': 'off', 'shared_buffers': '1GB', 'maintenance_work_mem': '1GB', - 'autovacuum': 'off', - 'full_page_writes': 'off' - } - ) + 'full_page_writes': 'off'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -321,19 +364,18 @@ def test_page_multiple_segments(self): # PGBENCH STUFF pgbench = node.pgbench(options=['-T', '50', '-c', '1', '--no-vacuum']) pgbench.wait() - node.safe_psql("postgres", "checkpoint") # GET LOGICAL CONTENT FROM NODE - result = node.safe_psql("postgres", "select * from pgbench_accounts") + result = node.table_checksum("pgbench_accounts") # PAGE BACKUP - self.backup_node( - backup_dir, 'node', node, backup_type='page') + self.backup_node(backup_dir, 'node', node, backup_type='page') + # GET PHYSICAL CONTENT FROM NODE pgdata = self.pgdata_content(node.data_dir) # RESTORE NODE restored_node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'restored_node')) + base_dir=os.path.join(self.module_name, self.fname, 'restored_node')) restored_node.cleanup() tblspc_path = self.get_tblspace_path(node, 'somedata') tblspc_path_new = self.get_tblspace_path( @@ -343,19 +385,16 @@ def test_page_multiple_segments(self): backup_dir, 'node', restored_node, options=[ "-j", "4", - "--recovery-target-action=promote", "-T", "{0}={1}".format(tblspc_path, tblspc_path_new)]) # GET PHYSICAL CONTENT FROM NODE_RESTORED pgdata_restored = self.pgdata_content(restored_node.data_dir) # START RESTORED NODE - restored_node.append_conf( - "postgresql.auto.conf", "port = {0}".format(restored_node.port)) + self.set_auto_conf(restored_node, {'port': restored_node.port}) restored_node.slow_start() - result_new = restored_node.safe_psql( - "postgres", "select * from pgbench_accounts") + result_new = restored_node.table_checksum("pgbench_accounts") # COMPARE RESTORED FILES self.assertEqual(result, result_new, 'data is lost') @@ -363,9 +402,6 @@ def test_page_multiple_segments(self): if self.paranoia: self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_page_delete(self): """ @@ -373,14 +409,12 @@ def test_page_delete(self): delete everything from table, vacuum table, take page backup, restore page backup, compare . """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'checkpoint_timeout': '30s', - 'autovacuum': 'off' } ) @@ -396,18 +430,15 @@ def test_page_delete(self): "postgres", "create table t_heap tablespace somedata as select i as id," " md5(i::text) as text, md5(i::text)::tsvector as tsvector" - " from generate_series(0,100) i" - ) + " from generate_series(0,100) i") node.safe_psql( "postgres", - "delete from t_heap" - ) + "delete from t_heap") node.safe_psql( "postgres", - "vacuum t_heap" - ) + "vacuum t_heap") # PAGE BACKUP self.backup_node( @@ -417,8 +448,7 @@ def test_page_delete(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') - ) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() self.restore_node( @@ -437,13 +467,9 @@ def test_page_delete(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_page_delete_1(self): """ @@ -451,14 +477,13 @@ def test_page_delete_1(self): delete everything from table, vacuum table, take page backup, restore page backup, compare . """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, initdb_params=['--data-checksums'], + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], pg_options={ 'checkpoint_timeout': '30s', - 'autovacuum': 'off' } ) @@ -496,7 +521,7 @@ def test_page_delete_1(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') + base_dir=os.path.join(self.module_name, self.fname, 'node_restored') ) node_restored.cleanup() @@ -516,30 +541,25 @@ def test_page_delete_1(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_parallel_pagemap(self): """ Test for parallel WAL segments reading, during which pagemap is built """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') # Initialize instance and backup directory node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums'], pg_options={ "hot_standby": "on" } ) node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored'), + base_dir=os.path.join(self.module_name, self.fname, 'node_restored'), ) self.init_pb(backup_dir) @@ -584,8 +604,7 @@ def test_parallel_pagemap(self): pgdata_restored = self.pgdata_content(node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() # Check restored node @@ -596,18 +615,16 @@ def test_parallel_pagemap(self): # Clean after yourself node.cleanup() node_restored.cleanup() - self.del_test_dir(module_name, fname) def test_parallel_pagemap_1(self): """ Test for parallel WAL segments reading, during which pagemap is built """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') # Initialize instance and backup directory node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums'], pg_options={} ) @@ -648,7 +665,6 @@ def test_parallel_pagemap_1(self): # Clean after yourself node.cleanup() - self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_page_backup_with_lost_wal_segment(self): @@ -659,12 +675,11 @@ def test_page_backup_with_lost_wal_segment(self): run page backup, expecting error because of missing wal segment make sure that backup status is 'ERROR' """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -678,7 +693,7 @@ def test_page_backup_with_lost_wal_segment(self): # delete last wal segment wals_dir = os.path.join(backup_dir, 'wal', 'node') wals = [f for f in os.listdir(wals_dir) if os.path.isfile(os.path.join( - wals_dir, f)) and not f.endswith('.backup') and not f.endswith('.partial')] + wals_dir, f)) and not f.endswith('.backup') and not f.endswith('.part')] wals = map(str, wals) file = os.path.join(wals_dir, max(wals)) os.remove(file) @@ -696,8 +711,6 @@ def test_page_backup_with_lost_wal_segment(self): self.output, self.cmd)) except ProbackupException as e: self.assertTrue( - 'INFO: Wait for LSN' in e.message and - 'in archived WAL segment' in e.message and 'Could not read WAL record at' in e.message and 'is absent' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( @@ -721,8 +734,6 @@ def test_page_backup_with_lost_wal_segment(self): self.output, self.cmd)) except ProbackupException as e: self.assertTrue( - 'INFO: Wait for LSN' in e.message and - 'in archived WAL segment' in e.message and 'Could not read WAL record at' in e.message and 'is absent' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( @@ -733,9 +744,6 @@ def test_page_backup_with_lost_wal_segment(self): self.show_pb(backup_dir, 'node')[2]['status'], 'Backup {0} should have STATUS "ERROR"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_page_backup_with_corrupted_wal_segment(self): """ @@ -745,12 +753,11 @@ def test_page_backup_with_corrupted_wal_segment(self): run page backup, expecting error because of missing wal segment make sure that backup status is 'ERROR' """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -759,7 +766,7 @@ def test_page_backup_with_corrupted_wal_segment(self): self.backup_node(backup_dir, 'node', node) # make some wals - node.pgbench_init(scale=4) + node.pgbench_init(scale=10) # delete last wal segment wals_dir = os.path.join(backup_dir, 'wal', 'node') @@ -811,10 +818,7 @@ def test_page_backup_with_corrupted_wal_segment(self): self.output, self.cmd)) except ProbackupException as e: self.assertTrue( - 'INFO: Wait for LSN' in e.message and - 'in archived WAL segment' in e.message and 'Could not read WAL record at' in e.message and - 'incorrect resource manager data checksum in record at' in e.message and 'Possible WAL corruption. Error has occured during reading WAL segment' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) @@ -836,10 +840,7 @@ def test_page_backup_with_corrupted_wal_segment(self): self.output, self.cmd)) except ProbackupException as e: self.assertTrue( - 'INFO: Wait for LSN' in e.message and - 'in archived WAL segment' in e.message and 'Could not read WAL record at' in e.message and - 'incorrect resource manager data checksum in record at' in e.message and 'Possible WAL corruption. Error has occured during reading WAL segment "{0}"'.format( file) in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( @@ -850,9 +851,6 @@ def test_page_backup_with_corrupted_wal_segment(self): self.show_pb(backup_dir, 'node')[2]['status'], 'Backup {0} should have STATUS "ERROR"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_page_backup_with_alien_wal_segment(self): """ @@ -864,15 +862,17 @@ def test_page_backup_with_alien_wal_segment(self): expecting error because of alien wal segment make sure that backup status is 'ERROR' """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, initdb_params=['--data-checksums']) alien_node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'alien_node')) + base_dir=os.path.join(self.module_name, self.fname, 'alien_node'), + set_replication=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -882,8 +882,10 @@ def test_page_backup_with_alien_wal_segment(self): self.set_archiving(backup_dir, 'alien_node', alien_node) alien_node.slow_start() - self.backup_node(backup_dir, 'node', node) - self.backup_node(backup_dir, 'alien_node', alien_node) + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + self.backup_node( + backup_dir, 'alien_node', alien_node, options=['--stream']) # make some wals node.safe_psql( @@ -892,7 +894,7 @@ def test_page_backup_with_alien_wal_segment(self): "create table t_heap as select i as id, " "md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1000) i;") + "from generate_series(0,10000) i;") alien_node.safe_psql( "postgres", @@ -904,7 +906,7 @@ def test_page_backup_with_alien_wal_segment(self): "create table t_heap_alien as select i as id, " "md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,100000) i;") + "from generate_series(0,10000) i;") # copy latest wal segment wals_dir = os.path.join(backup_dir, 'wal', 'alien_node') @@ -915,9 +917,9 @@ def test_page_backup_with_alien_wal_segment(self): file = os.path.join(wals_dir, filename) file_destination = os.path.join( os.path.join(backup_dir, 'wal', 'node'), filename) -# file = os.path.join(wals_dir, '000000010000000000000004') - print(file) - print(file_destination) + start = time.time() + while not os.path.exists(file_destination) and time.time() - start < 20: + time.sleep(0.1) os.remove(file_destination) os.rename(file, file_destination) @@ -933,11 +935,7 @@ def test_page_backup_with_alien_wal_segment(self): self.output, self.cmd)) except ProbackupException as e: self.assertTrue( - 'INFO: Wait for LSN' in e.message and - 'in archived WAL segment' in e.message and 'Could not read WAL record at' in e.message and - 'WAL file is from different database system: WAL file database system identifier is' in e.message and - 'pg_control database system identifier is' in e.message and 'Possible WAL corruption. Error has occured during reading WAL segment' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) @@ -958,35 +956,28 @@ def test_page_backup_with_alien_wal_segment(self): "Output: {0} \n CMD: {1}".format( self.output, self.cmd)) except ProbackupException as e: - self.assertTrue( - 'INFO: Wait for LSN' in e.message and - 'in archived WAL segment' in e.message and - 'Could not read WAL record at' in e.message and - 'WAL file is from different database system: WAL file database system identifier is' in e.message and - 'pg_control database system identifier is' in e.message and - 'Possible WAL corruption. Error has occured during reading WAL segment' in e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) + self.assertIn('Could not read WAL record at', e.message) + self.assertIn('WAL file is from different database system: ' + 'WAL file database system identifier is', e.message) + self.assertIn('pg_control database system identifier is', e.message) + self.assertIn('Possible WAL corruption. Error has occured ' + 'during reading WAL segment', e.message) self.assertEqual( 'ERROR', self.show_pb(backup_dir, 'node')[2]['status'], 'Backup {0} should have STATUS "ERROR"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_multithread_page_backup_with_toast(self): """ make node, create toast, do multithread PAGE backup """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1006,9 +997,6 @@ def test_multithread_page_backup_with_toast(self): backup_dir, 'node', node, backup_type='page', options=["-j", "4"]) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_page_create_db(self): """ @@ -1016,16 +1004,14 @@ def test_page_create_db(self): restore database and check it presense """ self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'max_wal_size': '10GB', 'checkpoint_timeout': '5min', - 'autovacuum': 'off' } ) @@ -1058,8 +1044,7 @@ def test_page_create_db(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') - ) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() self.restore_node( @@ -1072,8 +1057,7 @@ def test_page_create_db(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() node_restored.safe_psql('db1', 'select 1') @@ -1102,8 +1086,7 @@ def test_page_create_db(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf(node_restored, {'port': node_restored.port}) node_restored.slow_start() try: @@ -1122,5 +1105,360 @@ def test_page_create_db(self): repr(e.message), self.cmd) ) - # Clean after yourself - self.del_test_dir(module_name, fname) + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_multi_timeline_page(self): + """ + Check that backup in PAGE mode choose + parent backup correctly: + t12 /---P--> + ... + t3 /----> + t2 /----> + t1 -F-----D-> + + P must have F as parent + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql("postgres", "create extension pageinspect") + + try: + node.safe_psql( + "postgres", + "create extension amcheck") + except QueryException as e: + node.safe_psql( + "postgres", + "create extension amcheck_next") + + node.pgbench_init(scale=20) + full_id = self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + self.backup_node(backup_dir, 'node', node, backup_type='delta') + + node.cleanup() + self.restore_node( + backup_dir, 'node', node, backup_id=full_id, + options=[ + '--recovery-target=immediate', + '--recovery-target-action=promote']) + + node.slow_start() + + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + # create timelines + for i in range(2, 7): + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target=latest', + '--recovery-target-action=promote', + '--recovery-target-timeline={0}'.format(i)]) + node.slow_start() + + # at this point there is i+1 timeline + pgbench = node.pgbench(options=['-T', '20', '-c', '1', '--no-vacuum']) + pgbench.wait() + + # create backup at 2, 4 and 6 timeline + if i % 2 == 0: + self.backup_node(backup_dir, 'node', node, backup_type='page') + + page_id = self.backup_node( + backup_dir, 'node', node, backup_type='page', + options=['--log-level-file=VERBOSE']) + + pgdata = self.pgdata_content(node.data_dir) + + result = node.table_checksum("pgbench_accounts") + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node(backup_dir, 'node', node_restored) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() + + result_new = node_restored.table_checksum("pgbench_accounts") + + self.assertEqual(result, result_new) + + self.compare_pgdata(pgdata, pgdata_restored) + + self.checkdb_node( + backup_dir, + 'node', + options=[ + '--amcheck', + '-d', 'postgres', '-p', str(node.port)]) + + self.checkdb_node( + backup_dir, + 'node', + options=[ + '--amcheck', + '-d', 'postgres', '-p', str(node_restored.port)]) + + backup_list = self.show_pb(backup_dir, 'node') + + self.assertEqual( + backup_list[2]['parent-backup-id'], + backup_list[0]['id']) + self.assertEqual(backup_list[2]['current-tli'], 3) + + self.assertEqual( + backup_list[3]['parent-backup-id'], + backup_list[2]['id']) + self.assertEqual(backup_list[3]['current-tli'], 5) + + self.assertEqual( + backup_list[4]['parent-backup-id'], + backup_list[3]['id']) + self.assertEqual(backup_list[4]['current-tli'], 7) + + self.assertEqual( + backup_list[5]['parent-backup-id'], + backup_list[4]['id']) + self.assertEqual(backup_list[5]['current-tli'], 7) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_multitimeline_page_1(self): + """ + Check that backup in PAGE mode choose + parent backup correctly: + t2 /----> + t1 -F--P---D-> + + P must have F as parent + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={'wal_log_hints': 'on'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql("postgres", "create extension pageinspect") + + try: + node.safe_psql( + "postgres", + "create extension amcheck") + except QueryException as e: + node.safe_psql( + "postgres", + "create extension amcheck_next") + + node.pgbench_init(scale=20) + full_id = self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench(options=['-T', '20', '-c', '1']) + pgbench.wait() + + page1 = self.backup_node(backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + page1 = self.backup_node(backup_dir, 'node', node, backup_type='delta') + + node.cleanup() + self.restore_node( + backup_dir, 'node', node, backup_id=page1, + options=[ + '--recovery-target=immediate', + '--recovery-target-action=promote']) + + node.slow_start() + + pgbench = node.pgbench(options=['-T', '20', '-c', '1', '--no-vacuum']) + pgbench.wait() + + print(self.backup_node( + backup_dir, 'node', node, backup_type='page', + options=['--log-level-console=LOG'], return_id=False)) + + pgdata = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node(backup_dir, 'node', node_restored) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start() + + self.compare_pgdata(pgdata, pgdata_restored) + + @unittest.skip("skip") + # @unittest.expectedFailure + def test_page_pg_resetxlog(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'shared_buffers': '512MB', + 'max_wal_size': '3GB'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Create table + node.safe_psql( + "postgres", + "create extension bloom; create sequence t_seq; " + "create table t_heap " + "as select nextval('t_seq')::int as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " +# "from generate_series(0,25600) i") + "from generate_series(0,2560) i") + + self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + 'postgres', + "update t_heap set id = nextval('t_seq'), text = md5(text), " + "tsvector = md5(repeat(tsvector::text, 10))::tsvector") + + self.switch_wal_segment(node) + + # kill the bastard + if self.verbose: + print('Killing postmaster. Losing Ptrack changes') + node.stop(['-m', 'immediate', '-D', node.data_dir]) + + # now smack it with sledgehammer + if node.major_version >= 10: + pg_resetxlog_path = self.get_bin_path('pg_resetwal') + wal_dir = 'pg_wal' + else: + pg_resetxlog_path = self.get_bin_path('pg_resetxlog') + wal_dir = 'pg_xlog' + + self.run_binary( + [ + pg_resetxlog_path, + '-D', + node.data_dir, + '-o 42', + '-f' + ], + asynchronous=False) + + if not node.status(): + node.slow_start() + else: + print("Die! Die! Why won't you die?... Why won't you die?") + exit(1) + + # take ptrack backup +# self.backup_node( +# backup_dir, 'node', node, +# backup_type='page', options=['--stream']) + + try: + self.backup_node( + backup_dir, 'node', node, backup_type='page') + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because instance was brutalized by pg_resetxlog" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd) + ) + except ProbackupException as e: + self.assertIn( + 'Insert error message', + e.message, + '\n Unexpected Error Message: {0}\n' + ' CMD: {1}'.format(repr(e.message), self.cmd)) + +# pgdata = self.pgdata_content(node.data_dir) +# +# node_restored = self.make_simple_node( +# base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) +# node_restored.cleanup() +# +# self.restore_node( +# backup_dir, 'node', node_restored) +# +# pgdata_restored = self.pgdata_content(node_restored.data_dir) +# self.compare_pgdata(pgdata, pgdata_restored) + + def test_page_huge_xlog_record(self): + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'max_locks_per_transaction': '1000', + 'work_mem': '100MB', + 'temp_buffers': '100MB', + 'wal_buffers': '128MB', + 'wal_level' : 'logical', + }) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=3) + + # Do full backup + self.backup_node(backup_dir, 'node', node, backup_type='full') + show_backup = self.show_pb(backup_dir,'node')[0] + + self.assertEqual(show_backup['status'], "OK") + self.assertEqual(show_backup['backup-mode'], "FULL") + + # Originally client had the problem at the transaction that (supposedly) + # deletes a lot of temporary tables (probably it was client disconnect). + # It generated ~40MB COMMIT WAL record. + # + # `pg_logical_emit_message` is much simpler and faster way to generate + # such huge record. + node.safe_psql( + "postgres", + "select pg_logical_emit_message(False, 'z', repeat('o', 60*1000*1000))") + + # Do page backup + self.backup_node(backup_dir, 'node', node, backup_type='page') + + show_backup = self.show_pb(backup_dir,'node')[1] + self.assertEqual(show_backup['status'], "OK") + self.assertEqual(show_backup['backup-mode'], "PAGE") diff --git a/tests/pgpro2068.py b/tests/pgpro2068_test.py similarity index 56% rename from tests/pgpro2068.py rename to tests/pgpro2068_test.py index 0def742a4..04f0eb6fa 100644 --- a/tests/pgpro2068.py +++ b/tests/pgpro2068_test.py @@ -9,18 +9,16 @@ from testgres import ProcessType -module_name = '2068' - - class BugTest(ProbackupTest, unittest.TestCase): def test_minrecpoint_on_replica(self): """ https://jira.postgrespro.ru/browse/PGPRO-2068 """ - fname = self.id().split('.')[3] + self._check_gdb_flag_or_skip_test() + node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ @@ -31,7 +29,7 @@ def test_minrecpoint_on_replica(self): 'bgwriter_lru_multiplier': '4.0', 'max_wal_size': '256MB'}) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -43,7 +41,7 @@ def test_minrecpoint_on_replica(self): # start replica replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) + base_dir=os.path.join(self.module_name, self.fname, 'replica')) replica.cleanup() self.restore_node(backup_dir, 'node', replica, options=['-R']) @@ -51,15 +49,9 @@ def test_minrecpoint_on_replica(self): self.add_instance(backup_dir, 'replica', replica) self.set_archiving(backup_dir, 'replica', replica, replica=True) - replica.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(replica.port)) - replica.append_conf( - 'postgresql.auto.conf', 'restart_after_crash = off') - - # we need those later - node.safe_psql( - "postgres", - "CREATE EXTENSION plpythonu") + self.set_auto_conf( + replica, + {'port': replica.port, 'restart_after_crash': 'off'}) node.safe_psql( "postgres", @@ -83,12 +75,11 @@ def test_minrecpoint_on_replica(self): options=["-c", "4", "-j 4", "-T", "100"]) # wait for shared buffer on replica to be filled with dirty data - sleep(10) + sleep(20) # get pids of replica background workers startup_pid = replica.auxiliary_pids[ProcessType.Startup][0] checkpointer_pid = replica.auxiliary_pids[ProcessType.Checkpointer][0] - bgwriter_pid = replica.auxiliary_pids[ProcessType.BackgroundWriter][0] # break checkpointer on UpdateLastRemovedPtr gdb_checkpointer = self.gdb_attach(checkpointer_pid) @@ -111,7 +102,7 @@ def test_minrecpoint_on_replica(self): pgbench.stdout.close() # kill someone, we need a crash - os.kill(int(bgwriter_pid), 9) + replica.kill(someone=ProcessType.BackgroundWriter) gdb_recovery._execute('detach') gdb_checkpointer._execute('detach') @@ -121,55 +112,51 @@ def test_minrecpoint_on_replica(self): except: pass + # MinRecLSN = replica.get_control_data()['Minimum recovery ending location'] + # Promote replica with 'immediate' target action + if self.get_version(replica) >= self.version_to_num('12.0'): + recovery_config = 'postgresql.auto.conf' + else: + recovery_config = 'recovery.conf' + replica.append_conf( - 'recovery.conf', "recovery_target = 'immediate'") + recovery_config, "recovery_target = 'immediate'") replica.append_conf( - 'recovery.conf', "recovery_target_action = 'promote'") - replica.slow_start() + recovery_config, "recovery_target_action = 'pause'") + replica.slow_start(replica=True) + current_xlog_lsn_query = 'SELECT pg_last_wal_replay_lsn() INTO current_xlog_lsn' if self.get_version(node) < 100000: - script = ''' -DO -$$ -relations = plpy.execute("select class.oid from pg_class class WHERE class.relkind IN ('r', 'i', 't', 'm') and class.relpersistence = 'p'") -current_xlog_lsn = plpy.execute("select pg_last_xlog_replay_location() as lsn")[0]['lsn'] -plpy.notice('CURRENT LSN: {0}'.format(current_xlog_lsn)) -found_corruption = False -for relation in relations: - pages_from_future = plpy.execute("with number_of_blocks as (select blknum from generate_series(0, pg_relation_size({0}) / 8192 -1) as blknum) select blknum, lsn, checksum, flags, lower, upper, special, pagesize, version, prune_xid from number_of_blocks, page_header(get_raw_page('{0}'::oid::regclass::text, number_of_blocks.blknum::int)) where lsn > '{1}'::pg_lsn".format(relation['oid'], current_xlog_lsn)) - - if pages_from_future.nrows() == 0: - continue - - for page in pages_from_future: - plpy.notice('Found page from future. OID: {0}, BLKNUM: {1}, LSN: {2}'.format(relation['oid'], page['blknum'], page['lsn'])) - found_corruption = True -if found_corruption: - plpy.error('Found Corruption') -$$ LANGUAGE plpythonu; -''' - else: - script = ''' + current_xlog_lsn_query = 'SELECT min_recovery_end_location INTO current_xlog_lsn FROM pg_control_recovery()' + + script = f''' DO $$ -relations = plpy.execute("select class.oid from pg_class class WHERE class.relkind IN ('r', 'i', 't', 'm') and class.relpersistence = 'p'") -current_xlog_lsn = plpy.execute("select pg_last_wal_replay_lsn() as lsn")[0]['lsn'] -plpy.notice('CURRENT LSN: {0}'.format(current_xlog_lsn)) -found_corruption = False -for relation in relations: - pages_from_future = plpy.execute("with number_of_blocks as (select blknum from generate_series(0, pg_relation_size({0}) / 8192 -1) as blknum) select blknum, lsn, checksum, flags, lower, upper, special, pagesize, version, prune_xid from number_of_blocks, page_header(get_raw_page('{0}'::oid::regclass::text, number_of_blocks.blknum::int)) where lsn > '{1}'::pg_lsn".format(relation['oid'], current_xlog_lsn)) - - if pages_from_future.nrows() == 0: - continue - - for page in pages_from_future: - plpy.notice('Found page from future. OID: {0}, BLKNUM: {1}, LSN: {2}'.format(relation['oid'], page['blknum'], page['lsn'])) - found_corruption = True -if found_corruption: - plpy.error('Found Corruption') -$$ LANGUAGE plpythonu; -''' +DECLARE + roid oid; + current_xlog_lsn pg_lsn; + pages_from_future RECORD; + found_corruption bool := false; +BEGIN + {current_xlog_lsn_query}; + RAISE NOTICE 'CURRENT LSN: %', current_xlog_lsn; + FOR roid IN select oid from pg_class class where relkind IN ('r', 'i', 't', 'm') and relpersistence = 'p' LOOP + FOR pages_from_future IN + with number_of_blocks as (select blknum from generate_series(0, pg_relation_size(roid) / 8192 -1) as blknum ) + select blknum, lsn, checksum, flags, lower, upper, special, pagesize, version, prune_xid + from number_of_blocks, page_header(get_raw_page(roid::regclass::text, number_of_blocks.blknum::int)) + where lsn > current_xlog_lsn LOOP + RAISE NOTICE 'Found page from future. OID: %, BLKNUM: %, LSN: %', roid, pages_from_future.blknum, pages_from_future.lsn; + found_corruption := true; + END LOOP; + END LOOP; + IF found_corruption THEN + RAISE 'Found Corruption'; + END IF; +END; +$$ LANGUAGE plpgsql; +'''.format(current_xlog_lsn_query=current_xlog_lsn_query) # Find blocks from future replica.safe_psql( @@ -182,6 +169,3 @@ def test_minrecpoint_on_replica(self): # do basebackup # do pg_probackup, expect error - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/pgpro560.py b/tests/pgpro560_test.py similarity index 81% rename from tests/pgpro560.py rename to tests/pgpro560_test.py index 3aada4319..b665fd200 100644 --- a/tests/pgpro560.py +++ b/tests/pgpro560_test.py @@ -3,9 +3,7 @@ from .helpers.ptrack_helpers import ProbackupTest, ProbackupException from datetime import datetime, timedelta import subprocess - - -module_name = 'pgpro560' +from time import sleep class CheckSystemID(ProbackupTest, unittest.TestCase): @@ -19,36 +17,36 @@ def test_pgpro560_control_file_loss(self): make backup check that backup failed """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() file = os.path.join(node.base_dir, 'data', 'global', 'pg_control') - os.remove(file) + # Not delete this file permanently + os.rename(file, os.path.join(node.base_dir, 'data', 'global', 'pg_control_copy')) try: self.backup_node(backup_dir, 'node', node, options=['--stream']) # we should die here because exception is what we expect to happen self.assertEqual( - 1, 0, - "Expecting Error because pg_control was deleted.\n " - "Output: {0} \n CMD: {1}".format(repr(self.output), self.cmd)) + 1, 0, + "Expecting Error because pg_control was deleted.\n " + "Output: {0} \n CMD: {1}".format(repr(self.output), self.cmd)) except ProbackupException as e: self.assertTrue( - 'ERROR: could not open file' in e.message and + 'ERROR: Could not open file' in e.message and 'pg_control' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) + # Return this file to avoid Postger fail + os.rename(os.path.join(node.base_dir, 'data', 'global', 'pg_control_copy'), file) def test_pgpro560_systemid_mismatch(self): """ @@ -57,21 +55,20 @@ def test_pgpro560_systemid_mismatch(self): feed to backup PGDATA from node1 and PGPORT from node2 check that backup failed """ - fname = self.id().split('.')[3] node1 = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node1'), + base_dir=os.path.join(self.module_name, self.fname, 'node1'), set_replication=True, initdb_params=['--data-checksums']) node1.slow_start() node2 = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node2'), + base_dir=os.path.join(self.module_name, self.fname, 'node2'), set_replication=True, initdb_params=['--data-checksums']) node2.slow_start() - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node1', node1) @@ -98,6 +95,8 @@ def test_pgpro560_systemid_mismatch(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) + sleep(1) + try: self.backup_node( backup_dir, 'node1', node2, @@ -122,6 +121,3 @@ def test_pgpro560_systemid_mismatch(self): e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/pgpro589.py b/tests/pgpro589_test.py similarity index 83% rename from tests/pgpro589.py rename to tests/pgpro589_test.py index 122b2793f..8ce8e1f56 100644 --- a/tests/pgpro589.py +++ b/tests/pgpro589_test.py @@ -5,9 +5,6 @@ import subprocess -module_name = 'pgpro589' - - class ArchiveCheck(ProbackupTest, unittest.TestCase): def test_pgpro589(self): @@ -17,18 +14,17 @@ def test_pgpro589(self): check that backup status equal to ERROR check that no files where copied to backup catalogue """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) # make erroneous archive_command - node.append_conf("postgresql.auto.conf", "archive_command = 'exit 0'") + self.set_auto_conf(node, {'archive_command': 'exit 0'}) node.slow_start() node.pgbench_init(scale=5) @@ -57,8 +53,8 @@ def test_pgpro589(self): except ProbackupException as e: self.assertTrue( 'INFO: Wait for WAL segment' in e.message and - 'ERROR: Switched WAL segment' in e.message and - 'could not be archived' in e.message, + 'ERROR: WAL segment' in e.message and + 'could not be archived in 10 seconds' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) @@ -74,6 +70,3 @@ def test_pgpro589(self): "\n Start LSN was not found in archive but datafiles where " "copied to backup catalogue.\n For example: {0}\n " "It is not optimal".format(file)) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/ptrack.py b/tests/ptrack_test.py similarity index 50% rename from tests/ptrack.py rename to tests/ptrack_test.py index 7025d994b..7b5bc416b 100644 --- a/tests/ptrack.py +++ b/tests/ptrack_test.py @@ -3,644 +3,1193 @@ from .helpers.ptrack_helpers import ProbackupTest, ProbackupException, idx_ptrack from datetime import datetime, timedelta import subprocess -from testgres import QueryException +from testgres import QueryException, StartNodeException import shutil import sys -import time +from time import sleep from threading import Thread -module_name = 'ptrack' - - class PtrackTest(ProbackupTest, unittest.TestCase): + def setUp(self): + if self.pg_config_version < self.version_to_num('11.0'): + self.skipTest('You need PostgreSQL >= 11 for this test') + self.fname = self.id().split('.')[3] # @unittest.skip("skip") - # @unittest.expectedFailure - def test_ptrack_enable(self): - """make ptrack without full backup, should result in error""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + def test_drop_rel_during_backup_ptrack(self): + """ + drop relation during ptrack backup + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '30s' - } - ) + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() - # PTRACK BACKUP - try: - self.backup_node( - backup_dir, 'node', node, - backup_type='ptrack', options=["--stream"] - ) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because ptrack disabled.\n" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd - ) - ) - except ProbackupException as e: - self.assertIn( - 'ERROR: Ptrack is disabled\n', - e.message, - '\n Unexpected Error Message: {0}\n' - ' CMD: {1}'.format(repr(e.message), self.cmd) - ) + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + node.safe_psql( + "postgres", + "create table t_heap as select i" + " as id from generate_series(0,100) i") + + relative_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + absolute_path = os.path.join(node.data_dir, relative_path) + + # FULL backup + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + # PTRACK backup + gdb = self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + gdb=True, options=['--log-level-file=LOG']) + + gdb.set_breakpoint('backup_files') + gdb.run_until_break() + + # REMOVE file + os.remove(absolute_path) + + # File removed, we can proceed with backup + gdb.continue_execution_until_exit() + + pgdata = self.pgdata_content(node.data_dir) - # Clean after yourself - self.del_test_dir(module_name, fname) + with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: + log_content = f.read() + self.assertTrue( + 'LOG: File not found: "{0}"'.format(absolute_path) in log_content, + 'File "{0}" should be deleted but it`s not'.format(absolute_path)) + + node.cleanup() + self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) + + # Physical comparison + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") - # @unittest.expectedFailure - def test_ptrack_disable(self): - """ - Take full backup, disable ptrack restart postgresql, - enable ptrack, restart postgresql, take ptrack backup - which should fail - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + def test_ptrack_without_full(self): + """ptrack backup without validated full backup""" node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '30s', - 'ptrack_enable': 'on' - } - ) + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + ptrack_enable=True) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) node.slow_start() - # FULL BACKUP - self.backup_node(backup_dir, 'node', node, options=['--stream']) - - # DISABLE PTRACK - node.safe_psql('postgres', "alter system set ptrack_enable to off") - node.stop() - node.slow_start() - - # ENABLE PTRACK - node.safe_psql('postgres', "alter system set ptrack_enable to on") - node.stop() - node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") - # PTRACK BACKUP try: - self.backup_node( - backup_dir, 'node', node, - backup_type='ptrack', options=["--stream"] - ) + self.backup_node(backup_dir, 'node', node, backup_type="ptrack") # we should die here because exception is what we expect to happen self.assertEqual( 1, 0, - "Expecting Error because ptrack_enable was set to OFF at some" - " point after previous backup.\n" - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd - ) - ) + "Expecting Error because page backup should not be possible " + "without valid full backup.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) except ProbackupException as e: - self.assertIn( - 'ERROR: LSN from ptrack_control', - e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd - ) - ) + self.assertTrue( + "WARNING: Valid full backup on current timeline 1 is not found" in e.message and + "ERROR: Create new full backup before an incremental one" in e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['status'], + "ERROR") # @unittest.skip("skip") - def test_ptrack_uncommitted_xact(self): - """make ptrack backup while there is uncommitted open transaction""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + def test_ptrack_threads(self): + """ptrack multi thread backup mode""" node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'ptrack_enable': 'on' - } - ) - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored'), - ) + ptrack_enable=True) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) - node_restored.cleanup() node.slow_start() - self.backup_node(backup_dir, 'node', node) - con = node.connect("postgres") - con.execute( - "create table t_heap as select i" - " as id from generate_series(0,1) i" - ) + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - options=['--stream'] - ) - pgdata = self.pgdata_content(node.data_dir) + backup_dir, 'node', node, + backup_type="full", options=["-j", "4"]) + self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - options=['--stream'] - ) + backup_dir, 'node', node, + backup_type="ptrack", options=["-j", "4"]) + self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") - self.restore_node( - backup_dir, 'node', node_restored, options=["-j", "4"]) + # @unittest.skip("skip") + def test_ptrack_stop_pg(self): + """ + create node, take full backup, + restart node, check that ptrack backup + can be taken + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) - # Physical comparison - if self.paranoia: - pgdata_restored = self.pgdata_content( - node_restored.data_dir, ignore_ptrack=False) - self.compare_pgdata(pgdata, pgdata_restored) + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") - node_restored.slow_start() + node.pgbench_init(scale=1) - # Clean after yourself - self.del_test_dir(module_name, fname) + # FULL backup + self.backup_node(backup_dir, 'node', node, options=['--stream']) - # @unittest.skip("skip") - def test_ptrack_vacuum_full(self): - """make node, make full and ptrack stream backups, - restore them and check data correctness""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + node.stop() + node.slow_start() + + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['--stream']) + + # @unittest.skip("skip") + def test_ptrack_multi_timeline_backup(self): + """ + t2 /------P2 + t1 ------F---*-----P1 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'ptrack_enable': 'on' - } - ) - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored'), - ) + ptrack_enable=True, + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) - node_restored.cleanup() node.slow_start() - self.create_tblspace_in_node(node, 'somedata') - - self.backup_node(backup_dir, 'node', node) node.safe_psql( "postgres", - "create table t_heap tablespace somedata as select i" - " as id from generate_series(0,1000000) i" - ) + "CREATE EXTENSION ptrack") - pg_connect = node.connect("postgres", autocommit=True) + node.pgbench_init(scale=5) - gdb = self.gdb_attach(pg_connect.pid) - gdb.set_breakpoint('reform_and_rewrite_tuple') + # FULL backup + full_id = self.backup_node(backup_dir, 'node', node) - gdb.continue_execution_until_running() + pgbench = node.pgbench(options=['-T', '30', '-c', '1', '--no-vacuum']) + sleep(15) - process = Thread( - target=pg_connect.execute, args=["VACUUM FULL t_heap"]) - process.start() + xid = node.safe_psql( + 'postgres', + 'SELECT txid_current()').decode('utf-8').rstrip() + pgbench.wait() - while not gdb.stopped_in_breakpoint: - sleep(1) + self.backup_node(backup_dir, 'node', node, backup_type='ptrack') - gdb.continue_execution_until_break(20) + node.cleanup() - self.backup_node( - backup_dir, 'node', node, backup_type='ptrack') + # Restore from full backup to create Timeline 2 + print(self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target-xid={0}'.format(xid), + '--recovery-target-action=promote'])) - self.backup_node( - backup_dir, 'node', node, backup_type='ptrack') + node.slow_start() - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() - gdb.remove_all_breakpoints() - gdb._execute('detach') - process.join() + self.backup_node(backup_dir, 'node', node, backup_type='ptrack') - old_tablespace = self.get_tblspace_path(node, 'somedata') - new_tablespace = self.get_tblspace_path(node_restored, 'somedata_new') + pgdata = self.pgdata_content(node.data_dir) - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "-T", "{0}={1}".format( - old_tablespace, new_tablespace)] - ) + node.cleanup() - # Physical comparison - if self.paranoia: - pgdata_restored = self.pgdata_content( - node_restored.data_dir, ignore_ptrack=False) - self.compare_pgdata(pgdata, pgdata_restored) + self.restore_node(backup_dir, 'node', node) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) - node_restored.slow_start() + node.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) + balance = node.safe_psql( + 'postgres', + 'select (select sum(tbalance) from pgbench_tellers) - ' + '( select sum(bbalance) from pgbench_branches) + ' + '( select sum(abalance) from pgbench_accounts ) - ' + '(select sum(delta) from pgbench_history) as must_be_zero').decode('utf-8').rstrip() - # @unittest.skip("skip") - def test_ptrack_vacuum_truncate(self): - """make node, create table, take full backup, - delete last 3 pages, vacuum relation, - take ptrack backup, take second ptrack backup, - restore last ptrack backup and check data correctness""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self.assertEqual('0', balance) + + # @unittest.skip("skip") + def test_ptrack_multi_timeline_backup_1(self): + """ + t2 /------ + t1 ---F---P1---* + + # delete P1 + t2 /------P2 + t1 ---F--------* + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'ptrack_enable': 'on', - 'autovacuum': 'off' - } - ) - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored'), - ) + ptrack_enable=True, + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) - node_restored.cleanup() node.slow_start() - self.create_tblspace_in_node(node, 'somedata') node.safe_psql( "postgres", - "create sequence t_seq; " - "create table t_heap tablespace somedata as select i as id, " - "md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,1024) i;" - ) - node.safe_psql( - "postgres", - "vacuum t_heap" - ) + "CREATE EXTENSION ptrack") - self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=5) - node.safe_psql( - "postgres", - "delete from t_heap where ctid >= '(11,0)'" - ) - node.safe_psql( - "postgres", - "vacuum t_heap" - ) + # FULL backup + full_id = self.backup_node(backup_dir, 'node', node) - self.backup_node( - backup_dir, 'node', node, backup_type='ptrack') + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() - self.backup_node( - backup_dir, 'node', node, backup_type='ptrack') + ptrack_id = self.backup_node(backup_dir, 'node', node, backup_type='ptrack') + node.cleanup() - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) + self.restore_node(backup_dir, 'node', node) - old_tablespace = self.get_tblspace_path(node, 'somedata') - new_tablespace = self.get_tblspace_path(node_restored, 'somedata_new') + node.slow_start() - self.restore_node( - backup_dir, 'node', node_restored, - options=["-j", "4", "-T", "{0}={1}".format( - old_tablespace, new_tablespace)] - ) + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() - # Physical comparison - if self.paranoia: - pgdata_restored = self.pgdata_content( - node_restored.data_dir, - ignore_ptrack=False - ) - self.compare_pgdata(pgdata, pgdata_restored) + # delete old PTRACK backup + self.delete_pb(backup_dir, 'node', backup_id=ptrack_id) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) - node_restored.slow_start() + # take new PTRACK backup + self.backup_node(backup_dir, 'node', node, backup_type='ptrack') + + pgdata = self.pgdata_content(node.data_dir) + + node.cleanup() + + self.restore_node(backup_dir, 'node', node) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + node.slow_start() + + balance = node.safe_psql( + 'postgres', + 'select (select sum(tbalance) from pgbench_tellers) - ' + '( select sum(bbalance) from pgbench_branches) + ' + '( select sum(abalance) from pgbench_accounts ) - ' + '(select sum(delta) from pgbench_history) as must_be_zero').\ + decode('utf-8').rstrip() - # Clean after yourself - self.del_test_dir(module_name, fname) + self.assertEqual('0', balance) # @unittest.skip("skip") - def test_ptrack_simple(self): - """make node, make full and ptrack stream backups," - " restore them and check data correctness""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + def test_ptrack_eat_my_data(self): + """ + PGPRO-4051 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'ptrack_enable': 'on' - } - ) - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored'), - ) + ptrack_enable=True, + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) - node_restored.cleanup() node.slow_start() - self.backup_node(backup_dir, 'node', node) - node.safe_psql( "postgres", - "create table t_heap as select i" - " as id from generate_series(0,1) i" - ) + "CREATE EXTENSION ptrack") - self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - options=['--stream'] - ) + node.pgbench_init(scale=50) - node.safe_psql( - "postgres", - "update t_heap set id = 100500") + self.backup_node(backup_dir, 'node', node) - self.backup_node( - backup_dir, 'node', node, - backup_type='ptrack', options=['--stream'] - ) + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) + pgbench = node.pgbench(options=['-T', '300', '-c', '1', '--no-vacuum']) - result = node.safe_psql("postgres", "SELECT * FROM t_heap") + for i in range(10): + print("Iteration: {0}".format(i)) - self.restore_node( - backup_dir, 'node', node_restored, options=["-j", "4"]) + sleep(2) - # Physical comparison - if self.paranoia: - pgdata_restored = self.pgdata_content( - node_restored.data_dir, ignore_ptrack=False) - self.compare_pgdata(pgdata, pgdata_restored) + self.backup_node(backup_dir, 'node', node, backup_type='ptrack') +# pgdata = self.pgdata_content(node.data_dir) +# +# node_restored.cleanup() +# +# self.restore_node(backup_dir, 'node', node_restored) +# pgdata_restored = self.pgdata_content(node_restored.data_dir) +# +# self.compare_pgdata(pgdata, pgdata_restored) + + pgbench.terminate() + pgbench.wait() + + self.switch_wal_segment(node) + + result = node.table_checksum("pgbench_accounts") + + node_restored.cleanup() + self.restore_node(backup_dir, 'node', node_restored) + self.set_auto_conf( + node_restored, {'port': node_restored.port}) - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) node_restored.slow_start() + balance = node_restored.safe_psql( + 'postgres', + 'select (select sum(tbalance) from pgbench_tellers) - ' + '( select sum(bbalance) from pgbench_branches) + ' + '( select sum(abalance) from pgbench_accounts ) - ' + '(select sum(delta) from pgbench_history) as must_be_zero').decode('utf-8').rstrip() + + self.assertEqual('0', balance) + # Logical comparison self.assertEqual( result, - node_restored.safe_psql("postgres", "SELECT * FROM t_heap") - ) - - # Clean after yourself - self.del_test_dir(module_name, fname) + node.table_checksum("pgbench_accounts"), + 'Data loss') # @unittest.skip("skip") - def test_ptrack_get_block(self): + def test_ptrack_simple(self): """make node, make full and ptrack stream backups," " restore them and check data correctness""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '300s', - 'ptrack_enable': 'on' - } - ) + ptrack_enable=True, + initdb_params=['--data-checksums']) + self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) node.slow_start() node.safe_psql( "postgres", - "create table t_heap as select i" - " as id from generate_series(0,1) i" - ) + "CREATE EXTENSION ptrack") self.backup_node(backup_dir, 'node', node, options=['--stream']) - gdb = self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - options=['--stream'], - gdb=True - ) - gdb.set_breakpoint('make_pagemap_from_ptrack') - gdb.run_until_break() + node.safe_psql( + "postgres", + "create table t_heap as select i" + " as id from generate_series(0,1) i") + + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + options=['--stream']) node.safe_psql( "postgres", "update t_heap set id = 100500") - gdb.continue_execution_until_exit() - self.backup_node( backup_dir, 'node', node, - backup_type='ptrack', options=['--stream'] - ) + backup_type='ptrack', options=['--stream']) + if self.paranoia: pgdata = self.pgdata_content(node.data_dir) - result = node.safe_psql("postgres", "SELECT * FROM t_heap") - node.cleanup() - self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) + result = node.table_checksum("t_heap") + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, options=["-j", "4"]) # Physical comparison if self.paranoia: pgdata_restored = self.pgdata_content( - node.data_dir, ignore_ptrack=False) + node_restored.data_dir, ignore_ptrack=False) self.compare_pgdata(pgdata, pgdata_restored) - node.slow_start() + self.set_auto_conf( + node_restored, {'port': node_restored.port}) + + node_restored.slow_start() + # Logical comparison self.assertEqual( result, - node.safe_psql("postgres", "SELECT * FROM t_heap") - ) - - # Clean after yourself - self.del_test_dir(module_name, fname) + node_restored.table_checksum("t_heap")) # @unittest.skip("skip") - def test_ptrack_stream(self): - """make node, make full and ptrack stream backups, - restore them and check data correctness""" - self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + def test_ptrack_unprivileged(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'checkpoint_timeout': '30s', - 'ptrack_enable': 'on', - 'autovacuum': 'off' - } - ) + ptrack_enable=True, + initdb_params=['--data-checksums']) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) + # self.set_archiving(backup_dir, 'node', node) node.slow_start() - # FULL BACKUP - node.safe_psql("postgres", "create sequence t_seq") node.safe_psql( "postgres", - "create table t_heap as select i as id, nextval('t_seq')" - " as t_seq, md5(i::text) as text, md5(i::text)::tsvector" - " as tsvector from generate_series(0,100) i" - ) - full_result = node.safe_psql("postgres", "SELECT * FROM t_heap") - full_backup_id = self.backup_node( - backup_dir, 'node', node, options=['--stream']) + "CREATE DATABASE backupdb") - # PTRACK BACKUP - node.safe_psql( - "postgres", - "insert into t_heap select i as id, nextval('t_seq') as t_seq," - " md5(i::text) as text, md5(i::text)::tsvector as tsvector" - " from generate_series(100,200) i" - ) - ptrack_result = node.safe_psql("postgres", "SELECT * FROM t_heap") - ptrack_backup_id = self.backup_node( - backup_dir, 'node', - node, backup_type='ptrack', - options=['--stream'] + # PG 9.5 + if self.get_version(node) < 90600: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;") + # PG 9.6 + elif self.get_version(node) > 90600 and self.get_version(node) < 100000: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_xlog_replay_location() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + # >= 10 && < 15 + elif self.get_version(node) >= 100000 and self.get_version(node) < 150000: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + # >= 15 + else: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_start(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_stop(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" ) - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - # Drop Node - node.cleanup() + node.safe_psql( + "backupdb", + "CREATE SCHEMA ptrack") + node.safe_psql( + "backupdb", + "CREATE EXTENSION ptrack WITH SCHEMA ptrack") + node.safe_psql( + "backupdb", + "GRANT USAGE ON SCHEMA ptrack TO backup") - # Restore and check full backup - self.assertIn( - "INFO: Restore of backup {0} completed.".format(full_backup_id), - self.restore_node( - backup_dir, 'node', node, - backup_id=full_backup_id, - options=["-j", "4", "--recovery-target-action=promote"] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd) - ) - node.slow_start() - full_result_new = node.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(full_result, full_result_new) - node.cleanup() + node.safe_psql( + "backupdb", + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup") - # Restore and check ptrack backup - self.assertIn( - "INFO: Restore of backup {0} completed.".format(ptrack_backup_id), - self.restore_node( - backup_dir, 'node', node, - backup_id=ptrack_backup_id, - options=["-j", "4", "--recovery-target-action=promote"] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd) - ) + if ProbackupTest.enterprise: + node.safe_psql( + "backupdb", + "GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_version() TO backup; " + 'GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_edition() TO backup;') - if self.paranoia: - pgdata_restored = self.pgdata_content( - node.data_dir, ignore_ptrack=False) - self.compare_pgdata(pgdata, pgdata_restored) + self.backup_node( + backup_dir, 'node', node, + datname='backupdb', options=['--stream', "-U", "backup"]) - node.slow_start() - ptrack_result_new = node.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(ptrack_result, ptrack_result_new) + self.backup_node( + backup_dir, 'node', node, datname='backupdb', + backup_type='ptrack', options=['--stream', "-U", "backup"]) - # Clean after yourself - self.del_test_dir(module_name, fname) # @unittest.skip("skip") - def test_ptrack_archive(self): - """make archive node, make full and ptrack backups, - check data correctness in restored instance""" - self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + # @unittest.expectedFailure + def test_ptrack_enable(self): + """make ptrack without full backup, should result in error""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'checkpoint_timeout': '30s', - 'ptrack_enable': 'on', - 'autovacuum': 'off' - } - ) + 'shared_preload_libraries': 'ptrack'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) node.slow_start() - # FULL BACKUP node.safe_psql( "postgres", - "create table t_heap as" - " select i as id," - " md5(i::text) as text," - " md5(i::text)::tsvector as tsvector" - " from generate_series(0,100) i" - ) - full_result = node.safe_psql("postgres", "SELECT * FROM t_heap") - full_backup_id = self.backup_node(backup_dir, 'node', node) - full_target_time = self.show_pb( - backup_dir, 'node', full_backup_id)['recovery-time'] + "CREATE EXTENSION ptrack") # PTRACK BACKUP + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=["--stream"] + ) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because ptrack disabled.\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd + ) + ) + except ProbackupException as e: + self.assertIn( + 'ERROR: Ptrack is disabled\n', + e.message, + '\n Unexpected Error Message: {0}\n' + ' CMD: {1}'.format(repr(e.message), self.cmd) + ) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_ptrack_disable(self): + """ + Take full backup, disable ptrack restart postgresql, + enable ptrack, restart postgresql, take ptrack backup + which should fail + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums'], + pg_options={'checkpoint_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + node.safe_psql( "postgres", - "insert into t_heap select i as id," - " md5(i::text) as text," - " md5(i::text)::tsvector as tsvector" - " from generate_series(100,200) i" - ) - ptrack_result = node.safe_psql("postgres", "SELECT * FROM t_heap") - ptrack_backup_id = self.backup_node( - backup_dir, 'node', node, backup_type='ptrack') - ptrack_target_time = self.show_pb( + "CREATE EXTENSION ptrack") + + # FULL BACKUP + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + # DISABLE PTRACK + node.safe_psql('postgres', "alter system set ptrack.map_size to 0") + node.stop() + node.slow_start() + + # ENABLE PTRACK + node.safe_psql('postgres', "alter system set ptrack.map_size to '128'") + node.safe_psql('postgres', "alter system set shared_preload_libraries to 'ptrack'") + node.stop() + node.slow_start() + + # PTRACK BACKUP + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=["--stream"] + ) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because ptrack_enable was set to OFF at some" + " point after previous backup.\n" + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd + ) + ) + except ProbackupException as e: + self.assertIn( + 'ERROR: LSN from ptrack_control', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd + ) + ) + + # @unittest.skip("skip") + def test_ptrack_uncommitted_xact(self): + """make ptrack backup while there is uncommitted open transaction""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums'], + pg_options={ + 'wal_level': 'replica'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + con = node.connect("postgres") + con.execute( + "create table t_heap as select i" + " as id from generate_series(0,1) i") + + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + options=['--stream']) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored, + node_restored.data_dir, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content( + node_restored.data_dir, ignore_ptrack=False) + + self.set_auto_conf( + node_restored, {'port': node_restored.port}) + + node_restored.slow_start() + + # Physical comparison + if self.paranoia: + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_ptrack_vacuum_full(self): + """make node, make full and ptrack stream backups, + restore them and check data correctness""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.create_tblspace_in_node(node, 'somedata') + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + node.safe_psql( + "postgres", + "create table t_heap tablespace somedata as select i" + " as id from generate_series(0,1000000) i" + ) + + pg_connect = node.connect("postgres", autocommit=True) + + gdb = self.gdb_attach(pg_connect.pid) + gdb.set_breakpoint('reform_and_rewrite_tuple') + + gdb.continue_execution_until_running() + + process = Thread( + target=pg_connect.execute, args=["VACUUM FULL t_heap"]) + process.start() + + while not gdb.stopped_in_breakpoint: + sleep(1) + + gdb.continue_execution_until_break(20) + + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', options=['--stream']) + + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', options=['--stream']) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + gdb.remove_all_breakpoints() + gdb._execute('detach') + process.join() + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + old_tablespace = self.get_tblspace_path(node, 'somedata') + new_tablespace = self.get_tblspace_path(node_restored, 'somedata_new') + + self.restore_node( + backup_dir, 'node', node_restored, + options=["-j", "4", "-T", "{0}={1}".format( + old_tablespace, new_tablespace)] + ) + + # Physical comparison + if self.paranoia: + pgdata_restored = self.pgdata_content( + node_restored.data_dir, ignore_ptrack=False) + self.compare_pgdata(pgdata, pgdata_restored) + + self.set_auto_conf( + node_restored, {'port': node_restored.port}) + + node_restored.slow_start() + + # @unittest.skip("skip") + def test_ptrack_vacuum_truncate(self): + """make node, create table, take full backup, + delete last 3 pages, vacuum relation, + take ptrack backup, take second ptrack backup, + restore last ptrack backup and check data correctness""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.create_tblspace_in_node(node, 'somedata') + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + node.safe_psql( + "postgres", + "create sequence t_seq; " + "create table t_heap tablespace somedata as select i as id, " + "md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,1024) i;") + + node.safe_psql( + "postgres", + "vacuum t_heap") + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + node.safe_psql( + "postgres", + "delete from t_heap where ctid >= '(11,0)'") + + node.safe_psql( + "postgres", + "vacuum t_heap") + + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', options=['--stream']) + + self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', options=['--stream']) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + old_tablespace = self.get_tblspace_path(node, 'somedata') + new_tablespace = self.get_tblspace_path(node_restored, 'somedata_new') + + self.restore_node( + backup_dir, 'node', node_restored, + options=["-j", "4", "-T", "{0}={1}".format( + old_tablespace, new_tablespace)] + ) + + # Physical comparison + if self.paranoia: + pgdata_restored = self.pgdata_content( + node_restored.data_dir, + ignore_ptrack=False + ) + self.compare_pgdata(pgdata, pgdata_restored) + + self.set_auto_conf( + node_restored, {'port': node_restored.port}) + + node_restored.slow_start() + + # @unittest.skip("skip") + def test_ptrack_get_block(self): + """ + make node, make full and ptrack stream backups, + restore them and check data correctness + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + node.safe_psql( + "postgres", + "create table t_heap as select i" + " as id from generate_series(0,1) i") + + self.backup_node(backup_dir, 'node', node, options=['--stream']) + gdb = self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + options=['--stream'], + gdb=True) + + gdb.set_breakpoint('make_pagemap_from_ptrack_2') + gdb.run_until_break() + + node.safe_psql( + "postgres", + "update t_heap set id = 100500") + + gdb.continue_execution_until_exit() + + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['--stream']) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + result = node.table_checksum("t_heap") + node.cleanup() + self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) + + # Physical comparison + if self.paranoia: + pgdata_restored = self.pgdata_content( + node.data_dir, ignore_ptrack=False) + self.compare_pgdata(pgdata, pgdata_restored) + + node.slow_start() + # Logical comparison + self.assertEqual( + result, + node.table_checksum("t_heap")) + + # @unittest.skip("skip") + def test_ptrack_stream(self): + """make node, make full and ptrack stream backups, + restore them and check data correctness""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + # FULL BACKUP + node.safe_psql("postgres", "create sequence t_seq") + node.safe_psql( + "postgres", + "create table t_heap as select i as id, nextval('t_seq')" + " as t_seq, md5(i::text) as text, md5(i::text)::tsvector" + " as tsvector from generate_series(0,100) i") + + full_result = node.table_checksum("t_heap") + full_backup_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + # PTRACK BACKUP + node.safe_psql( + "postgres", + "insert into t_heap select i as id, nextval('t_seq') as t_seq," + " md5(i::text) as text, md5(i::text)::tsvector as tsvector" + " from generate_series(100,200) i") + + ptrack_result = node.table_checksum("t_heap") + ptrack_backup_id = self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['--stream']) + + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + # Drop Node + node.cleanup() + + # Restore and check full backup + self.assertIn( + "INFO: Restore of backup {0} completed.".format(full_backup_id), + self.restore_node( + backup_dir, 'node', node, + backup_id=full_backup_id, options=["-j", "4"] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd) + ) + node.slow_start() + full_result_new = node.table_checksum("t_heap") + self.assertEqual(full_result, full_result_new) + node.cleanup() + + # Restore and check ptrack backup + self.assertIn( + "INFO: Restore of backup {0} completed.".format(ptrack_backup_id), + self.restore_node( + backup_dir, 'node', node, + backup_id=ptrack_backup_id, options=["-j", "4"] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + if self.paranoia: + pgdata_restored = self.pgdata_content( + node.data_dir, ignore_ptrack=False) + self.compare_pgdata(pgdata, pgdata_restored) + + node.slow_start() + ptrack_result_new = node.table_checksum("t_heap") + self.assertEqual(ptrack_result, ptrack_result_new) + + # @unittest.skip("skip") + def test_ptrack_archive(self): + """make archive node, make full and ptrack backups, + check data correctness in restored instance""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + # FULL BACKUP + node.safe_psql( + "postgres", + "create table t_heap as" + " select i as id," + " md5(i::text) as text," + " md5(i::text)::tsvector as tsvector" + " from generate_series(0,100) i") + + full_result = node.table_checksum("t_heap") + full_backup_id = self.backup_node(backup_dir, 'node', node) + full_target_time = self.show_pb( + backup_dir, 'node', full_backup_id)['recovery-time'] + + # PTRACK BACKUP + node.safe_psql( + "postgres", + "insert into t_heap select i as id," + " md5(i::text) as text," + " md5(i::text)::tsvector as tsvector" + " from generate_series(100,200) i") + + ptrack_result = node.table_checksum("t_heap") + ptrack_backup_id = self.backup_node( + backup_dir, 'node', node, backup_type='ptrack') + ptrack_target_time = self.show_pb( backup_dir, 'node', ptrack_backup_id)['recovery-time'] if self.paranoia: pgdata = self.pgdata_content(node.data_dir) + node.safe_psql( + "postgres", + "insert into t_heap select i as id," + " md5(i::text) as text," + " md5(i::text)::tsvector as tsvector" + " from generate_series(200, 300) i") + # Drop Node node.cleanup() @@ -659,7 +1208,7 @@ def test_ptrack_archive(self): ) node.slow_start() - full_result_new = node.safe_psql("postgres", "SELECT * FROM t_heap") + full_result_new = node.table_checksum("t_heap") self.assertEqual(full_result, full_result_new) node.cleanup() @@ -684,31 +1233,26 @@ def test_ptrack_archive(self): self.compare_pgdata(pgdata, pgdata_restored) node.slow_start() - ptrack_result_new = node.safe_psql("postgres", "SELECT * FROM t_heap") + ptrack_result_new = node.table_checksum("t_heap") self.assertEqual(ptrack_result, ptrack_result_new) node.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") + @unittest.skip("skip") def test_ptrack_pgpro417(self): - """Make node, take full backup, take ptrack backup, - delete ptrack backup. Try to take ptrack backup, - which should fail""" - self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + """ + Make node, take full backup, take ptrack backup, + delete ptrack backup. Try to take ptrack backup, + which should fail. Actual only for PTRACK 1.x + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '30s', - 'ptrack_enable': 'on', - 'autovacuum': 'off'} - ) + 'checkpoint_timeout': '30s'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -719,9 +1263,6 @@ def test_ptrack_pgpro417(self): "postgres", "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,100) i") - node.safe_psql( - "postgres", - "SELECT * FROM t_heap") backup_id = self.backup_node( backup_dir, 'node', node, @@ -736,7 +1277,7 @@ def test_ptrack_pgpro417(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector " "from generate_series(100,200) i") - node.safe_psql("postgres", "SELECT * FROM t_heap") + node.table_checksum("t_heap") backup_id = self.backup_node( backup_dir, 'node', node, backup_type='ptrack', options=["--stream"]) @@ -769,27 +1310,21 @@ def test_ptrack_pgpro417(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") + @unittest.skip("skip") def test_page_pgpro417(self): """ Make archive node, take full backup, take page backup, - delete page backup. Try to take ptrack backup, which should fail + delete page backup. Try to take ptrack backup, which should fail. + Actual only for PTRACK 1.x """ - self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '30s', - 'ptrack_enable': 'on', - 'autovacuum': 'off'} - ) + 'checkpoint_timeout': '30s'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -801,8 +1336,7 @@ def test_page_pgpro417(self): "postgres", "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,100) i") - node.safe_psql("postgres", "SELECT * FROM t_heap") - self.backup_node(backup_dir, 'node', node) + node.table_checksum("t_heap") # PAGE BACKUP node.safe_psql( @@ -810,7 +1344,7 @@ def test_page_pgpro417(self): "insert into t_heap select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector " "from generate_series(100,200) i") - node.safe_psql("postgres", "SELECT * FROM t_heap") + node.table_checksum("t_heap") backup_id = self.backup_node( backup_dir, 'node', node, backup_type='page') @@ -839,27 +1373,21 @@ def test_page_pgpro417(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") + @unittest.skip("skip") def test_full_pgpro417(self): """ Make node, take two full backups, delete full second backup. - Try to take ptrack backup, which should fail + Try to take ptrack backup, which should fail. + Relevant only for PTRACK 1.x """ - self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '30s', - 'ptrack_enable': 'on', 'autovacuum': 'off' - } - ) + 'checkpoint_timeout': '30s'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -872,7 +1400,7 @@ def test_full_pgpro417(self): " md5(i::text)::tsvector as tsvector " " from generate_series(0,100) i" ) - node.safe_psql("postgres", "SELECT * FROM t_heap") + node.table_checksum("t_heap") self.backup_node(backup_dir, 'node', node, options=["--stream"]) # SECOND FULL BACKUP @@ -882,7 +1410,7 @@ def test_full_pgpro417(self): " md5(i::text)::tsvector as tsvector" " from generate_series(100,200) i" ) - node.safe_psql("postgres", "SELECT * FROM t_heap") + node.table_checksum("t_heap") backup_id = self.backup_node( backup_dir, 'node', node, options=["--stream"]) @@ -914,42 +1442,36 @@ def test_full_pgpro417(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_create_db(self): """ Make node, take full backup, create database db1, take ptrack backup, restore database and check it presense """ - self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'max_wal_size': '10GB', - 'checkpoint_timeout': '5min', - 'ptrack_enable': 'on', - 'autovacuum': 'off' - } - ) + 'max_wal_size': '10GB'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + # FULL BACKUP node.safe_psql( "postgres", "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,100) i") - node.safe_psql("postgres", "SELECT * FROM t_heap") + node.table_checksum("t_heap") self.backup_node( backup_dir, 'node', node, options=["--stream"]) @@ -964,17 +1486,14 @@ def test_create_db(self): # PTRACK BACKUP backup_id = self.backup_node( backup_dir, 'node', node, - backup_type='ptrack', - options=["--stream"] - ) + backup_type='ptrack', options=["--stream"]) if self.paranoia: pgdata = self.pgdata_content(node.data_dir) # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') - ) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() self.restore_node( @@ -988,8 +1507,8 @@ def test_create_db(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf( + node_restored, {'port': node_restored.port}) node_restored.slow_start() # DROP DATABASE DB1 @@ -1008,8 +1527,7 @@ def test_create_db(self): node_restored.cleanup() self.restore_node( backup_dir, 'node', node_restored, - backup_id=backup_id, options=["-j", "4"] - ) + backup_id=backup_id, options=["-j", "4"]) # COMPARE PHYSICAL CONTENT if self.paranoia: @@ -1018,8 +1536,8 @@ def test_create_db(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) + self.set_auto_conf( + node_restored, {'port': node_restored.port}) node_restored.slow_start() try: @@ -1029,17 +1547,12 @@ def test_create_db(self): 1, 0, "Expecting Error because we are connecting to deleted database" "\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd) - ) + repr(self.output), self.cmd)) except QueryException as e: self.assertTrue( 'FATAL: database "db1" does not exist' in e.message, '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd) - ) - - # Clean after yourself - self.del_test_dir(module_name, fname) + repr(e.message), self.cmd)) # @unittest.skip("skip") def test_create_db_on_replica(self): @@ -1049,24 +1562,23 @@ def test_create_db_on_replica(self): create database db1, take ptrack backup from replica, restore database and check it presense """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '30s', - 'ptrack_enable': 'on', - 'autovacuum': 'off' - } - ) + 'checkpoint_timeout': '30s'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + # FULL BACKUP node.safe_psql( "postgres", @@ -1074,18 +1586,17 @@ def test_create_db_on_replica(self): "md5(i::text)::tsvector as tsvector from generate_series(0,100) i") replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) + base_dir=os.path.join(self.module_name, self.fname, 'replica')) replica.cleanup() self.backup_node( - backup_dir, 'node', node, options=['-j10']) + backup_dir, 'node', node, options=['-j10', '--stream']) self.restore_node(backup_dir, 'node', replica) # Add replica self.add_instance(backup_dir, 'replica', replica) self.set_replica(node, replica, 'replica', synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) replica.slow_start(replica=True) self.backup_node( @@ -1116,6 +1627,7 @@ def test_create_db_on_replica(self): replica, backup_type='ptrack', options=[ '-j10', + '--stream', '--master-host=localhost', '--master-db=postgres', '--master-port={0}'.format(node.port) @@ -1127,8 +1639,7 @@ def test_create_db_on_replica(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') - ) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() self.restore_node( @@ -1141,37 +1652,34 @@ def test_create_db_on_replica(self): node_restored.data_dir) self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_alter_table_set_tablespace_ptrack(self): """Make node, create tablespace with table, take full backup, alter tablespace location, take ptrack backup, restore database.""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, initdb_params=['--data-checksums'], + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '30s', - 'ptrack_enable': 'on', - 'autovacuum': 'off' - } - ) + 'checkpoint_timeout': '30s'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + # FULL BACKUP self.create_tblspace_in_node(node, 'somedata') node.safe_psql( "postgres", "create table t_heap tablespace somedata as select i as id," " md5(i::text) as text, md5(i::text)::tsvector as tsvector" - " from generate_series(0,100) i" - ) + " from generate_series(0,100) i") # FULL backup self.backup_node(backup_dir, 'node', node, options=["--stream"]) @@ -1179,13 +1687,11 @@ def test_alter_table_set_tablespace_ptrack(self): self.create_tblspace_in_node(node, 'somedata_new') node.safe_psql( "postgres", - "alter table t_heap set tablespace somedata_new" - ) + "alter table t_heap set tablespace somedata_new") # sys.exit(1) # PTRACK BACKUP - #result = node.safe_psql( - # "postgres", "select * from t_heap") + #result = node.table_checksum("t_heap") self.backup_node( backup_dir, 'node', node, backup_type='ptrack', @@ -1198,8 +1704,7 @@ def test_alter_table_set_tablespace_ptrack(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored') - ) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() self.restore_node( @@ -1213,8 +1718,7 @@ def test_alter_table_set_tablespace_ptrack(self): "-T", "{0}={1}".format( self.get_tblspace_path(node, 'somedata_new'), self.get_tblspace_path(node_restored, 'somedata_new') - ), - "--recovery-target-action=promote" + ) ] ) @@ -1225,40 +1729,36 @@ def test_alter_table_set_tablespace_ptrack(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - node_restored.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node_restored.port)) + self.set_auto_conf( + node_restored, {'port': node_restored.port}) node_restored.slow_start() -# result_new = node_restored.safe_psql( -# "postgres", "select * from t_heap") +# result_new = node_restored.table_checksum("t_heap") # # self.assertEqual(result, result_new, 'lost some data after restore') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_alter_database_set_tablespace_ptrack(self): """Make node, create tablespace with database," " take full backup, alter tablespace location," " take ptrack backup, restore database.""" - self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '30s', - 'ptrack_enable': 'on', - 'autovacuum': 'off'} - ) + 'checkpoint_timeout': '30s'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + # FULL BACKUP self.backup_node(backup_dir, 'node', node, options=["--stream"]) @@ -1281,7 +1781,7 @@ def test_alter_database_set_tablespace_ptrack(self): # RESTORE node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) node_restored.cleanup() self.restore_node( backup_dir, 'node', @@ -1302,32 +1802,29 @@ def test_alter_database_set_tablespace_ptrack(self): node_restored.port = node.port node_restored.slow_start() - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_drop_tablespace(self): """ Make node, create table, alter table tablespace, take ptrack backup, move table from tablespace, take ptrack backup """ - self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '30s', - 'ptrack_enable': 'on', - 'autovacuum': 'off'} - ) + 'checkpoint_timeout': '30s'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.create_tblspace_in_node(node, 'somedata') # CREATE TABLE @@ -1336,7 +1833,7 @@ def test_drop_tablespace(self): "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,100) i") - result = node.safe_psql("postgres", "select * from t_heap") + result = node.table_checksum("t_heap") # FULL BACKUP self.backup_node(backup_dir, 'node', node, options=["--stream"]) @@ -1364,10 +1861,19 @@ def test_drop_tablespace(self): backup_dir, 'node', node, backup_type='ptrack', options=["--stream"]) + if self.paranoia: + pgdata = self.pgdata_content( + node.data_dir, ignore_ptrack=True) + tblspace = self.get_tblspace_path(node, 'somedata') node.cleanup() shutil.rmtree(tblspace, ignore_errors=True) self.restore_node(backup_dir, 'node', node, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content( + node.data_dir, ignore_ptrack=True) + node.slow_start() tblspc_exist = node.safe_psql( @@ -1381,11 +1887,11 @@ def test_drop_tablespace(self): "Expecting Error because " "tablespace 'somedata' should not be present") - result_new = node.safe_psql("postgres", "select * from t_heap") + result_new = node.table_checksum("t_heap") self.assertEqual(result, result_new) - # Clean after yourself - self.del_test_dir(module_name, fname) + if self.paranoia: + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") def test_ptrack_alter_tablespace(self): @@ -1393,21 +1899,23 @@ def test_ptrack_alter_tablespace(self): Make node, create table, alter table tablespace, take ptrack backup, move table from tablespace, take ptrack backup """ - self.maxDiff = None - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'checkpoint_timeout': '30s', 'ptrack_enable': 'on', - 'autovacuum': 'off'}) + 'checkpoint_timeout': '30s'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.create_tblspace_in_node(node, 'somedata') tblspc_path = self.get_tblspace_path(node, 'somedata') @@ -1417,15 +1925,16 @@ def test_ptrack_alter_tablespace(self): "create table t_heap as select i as id, md5(i::text) as text, " "md5(i::text)::tsvector as tsvector from generate_series(0,100) i") - result = node.safe_psql("postgres", "select * from t_heap") + result = node.table_checksum("t_heap") # FULL BACKUP self.backup_node(backup_dir, 'node', node, options=["--stream"]) # Move table to separate tablespace node.safe_psql( - "postgres", "alter table t_heap set tablespace somedata") + "postgres", + "alter table t_heap set tablespace somedata") # GET LOGICAL CONTENT FROM NODE - result = node.safe_psql("postgres", "select * from t_heap") + result = node.table_checksum("t_heap") # FIRTS PTRACK BACKUP self.backup_node( @@ -1438,13 +1947,12 @@ def test_ptrack_alter_tablespace(self): # Restore ptrack backup restored_node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'restored_node')) + base_dir=os.path.join(self.module_name, self.fname, 'restored_node')) restored_node.cleanup() tblspc_path_new = self.get_tblspace_path( restored_node, 'somedata_restored') self.restore_node(backup_dir, 'node', restored_node, options=[ - "-j", "4", "-T", "{0}={1}".format(tblspc_path, tblspc_path_new), - "--recovery-target-action=promote"]) + "-j", "4", "-T", "{0}={1}".format(tblspc_path, tblspc_path_new)]) # GET PHYSICAL CONTENT FROM RESTORED NODE and COMPARE PHYSICAL CONTENT if self.paranoia: @@ -1453,13 +1961,12 @@ def test_ptrack_alter_tablespace(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - restored_node.append_conf( - "postgresql.auto.conf", "port = {0}".format(restored_node.port)) + self.set_auto_conf( + restored_node, {'port': restored_node.port}) restored_node.slow_start() # COMPARE LOGICAL CONTENT - result_new = restored_node.safe_psql( - "postgres", "select * from t_heap") + result_new = restored_node.table_checksum("t_heap") self.assertEqual(result, result_new) restored_node.cleanup() @@ -1477,9 +1984,10 @@ def test_ptrack_alter_tablespace(self): pgdata = self.pgdata_content(node.data_dir) # Restore second ptrack backup and check table consistency - self.restore_node(backup_dir, 'node', restored_node, options=[ - "-j", "4", "-T", "{0}={1}".format(tblspc_path, tblspc_path_new), - "--recovery-target-action=promote"]) + self.restore_node( + backup_dir, 'node', restored_node, + options=[ + "-j", "4", "-T", "{0}={1}".format(tblspc_path, tblspc_path_new)]) # GET PHYSICAL CONTENT FROM RESTORED NODE and COMPARE PHYSICAL CONTENT if self.paranoia: @@ -1488,142 +1996,133 @@ def test_ptrack_alter_tablespace(self): self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - restored_node.append_conf( - "postgresql.auto.conf", "port = {0}".format(restored_node.port)) + self.set_auto_conf( + restored_node, {'port': restored_node.port}) restored_node.slow_start() - result_new = restored_node.safe_psql( - "postgres", "select * from t_heap") + result_new = restored_node.table_checksum("t_heap") self.assertEqual(result, result_new) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_ptrack_multiple_segments(self): """ Make node, create table, alter table tablespace, take ptrack backup, move table from tablespace, take ptrack backup """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'ptrack_enable': 'on', 'fsync': 'off', - 'autovacuum': 'off', - 'full_page_writes': 'off' - } - ) + 'full_page_writes': 'off'}) self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.create_tblspace_in_node(node, 'somedata') # CREATE TABLE node.pgbench_init(scale=100, options=['--tablespace=somedata']) + result = node.table_checksum("pgbench_accounts") # FULL BACKUP - self.backup_node(backup_dir, 'node', node) + self.backup_node(backup_dir, 'node', node, options=['--stream']) # PTRACK STUFF - idx_ptrack = {'type': 'heap'} - idx_ptrack['path'] = self.get_fork_path(node, 'pgbench_accounts') - idx_ptrack['old_size'] = self.get_fork_size(node, 'pgbench_accounts') - idx_ptrack['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack['path'], idx_ptrack['old_size']) - - pgbench = node.pgbench(options=['-T', '150', '-c', '2', '--no-vacuum']) + if node.major_version < 11: + idx_ptrack = {'type': 'heap'} + idx_ptrack['path'] = self.get_fork_path(node, 'pgbench_accounts') + idx_ptrack['old_size'] = self.get_fork_size(node, 'pgbench_accounts') + idx_ptrack['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack['path'], idx_ptrack['old_size']) + + pgbench = node.pgbench( + options=['-T', '30', '-c', '1', '--no-vacuum']) pgbench.wait() node.safe_psql("postgres", "checkpoint") - idx_ptrack['new_size'] = self.get_fork_size( - node, - 'pgbench_accounts' - ) - idx_ptrack['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack['path'], - idx_ptrack['new_size'] - ) - idx_ptrack['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, - idx_ptrack['path'] - ) + if node.major_version < 11: + idx_ptrack['new_size'] = self.get_fork_size( + node, + 'pgbench_accounts') - if not self.check_ptrack_sanity(idx_ptrack): - self.assertTrue( - False, 'Ptrack has failed to register changes in data files' - ) + idx_ptrack['new_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack['path'], + idx_ptrack['new_size']) + + idx_ptrack['ptrack'] = self.get_ptrack_bits_per_page_for_fork( + node, + idx_ptrack['path']) + + if not self.check_ptrack_sanity(idx_ptrack): + self.assertTrue( + False, 'Ptrack has failed to register changes in data files') # GET LOGICAL CONTENT FROM NODE # it`s stupid, because hint`s are ignored by ptrack - #result = node.safe_psql("postgres", "select * from pgbench_accounts") + result = node.table_checksum("pgbench_accounts") # FIRTS PTRACK BACKUP self.backup_node( - backup_dir, 'node', node, backup_type='ptrack') - - node.safe_psql("postgres", "checkpoint") + backup_dir, 'node', node, backup_type='ptrack', options=['--stream']) # GET PHYSICAL CONTENT FROM NODE pgdata = self.pgdata_content(node.data_dir) # RESTORE NODE restored_node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'restored_node')) + base_dir=os.path.join(self.module_name, self.fname, 'restored_node')) restored_node.cleanup() tblspc_path = self.get_tblspace_path(node, 'somedata') tblspc_path_new = self.get_tblspace_path( restored_node, - 'somedata_restored' - ) + 'somedata_restored') - self.restore_node(backup_dir, 'node', restored_node, options=[ - "-j", "4", "-T", "{0}={1}".format(tblspc_path, tblspc_path_new), - "--recovery-target-action=promote"]) + self.restore_node( + backup_dir, 'node', restored_node, + options=[ + "-j", "4", "-T", "{0}={1}".format( + tblspc_path, tblspc_path_new)]) # GET PHYSICAL CONTENT FROM NODE_RESTORED if self.paranoia: pgdata_restored = self.pgdata_content( restored_node.data_dir, ignore_ptrack=False) - self.compare_pgdata(pgdata, pgdata_restored) # START RESTORED NODE - restored_node.append_conf( - "postgresql.auto.conf", "port = {0}".format(restored_node.port)) + self.set_auto_conf( + restored_node, {'port': restored_node.port}) restored_node.slow_start() - result_new = restored_node.safe_psql( - "postgres", - "select * from pgbench_accounts" - ) + result_new = restored_node.table_checksum("pgbench_accounts") # COMPARE RESTORED FILES - #self.assertEqual(result, result_new, 'data is lost') + self.assertEqual(result, result_new, 'data is lost') - # Clean after yourself - self.del_test_dir(module_name, fname) + if self.paranoia: + self.compare_pgdata(pgdata, pgdata_restored) - # @unittest.skip("skip") - # @unittest.expectedFailure + @unittest.skip("skip") def test_atexit_fail(self): """ - Take backups of every available types and check that PTRACK is clean + Take backups of every available types and check that PTRACK is clean. + Relevant only for PTRACK 1.x """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'ptrack_enable': 'on', 'max_connections': '15'}) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -1635,8 +2134,7 @@ def test_atexit_fail(self): try: self.backup_node( backup_dir, 'node', node, backup_type='ptrack', - options=[ - "--stream", "-j 30"]) + options=["--stream", "-j 30"]) # we should die here because exception is what we expect to happen self.assertEqual( @@ -1659,25 +2157,22 @@ def test_atexit_fail(self): "select * from pg_is_in_backup()").rstrip(), "f") - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") + @unittest.skip("skip") # @unittest.expectedFailure def test_ptrack_clean(self): - """Take backups of every available types and check that PTRACK is clean""" - fname = self.id().split('.')[3] + """ + Take backups of every available types and check that PTRACK is clean + Relevant only for PTRACK 1.x + """ node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) node.slow_start() self.create_tblspace_in_node(node, 'somedata') @@ -1727,8 +2222,8 @@ def test_ptrack_clean(self): # Take PTRACK backup to clean every ptrack backup_id = self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - options=['-j10']) + backup_dir, 'node', node, backup_type='ptrack', options=['-j10', '--stream']) + node.safe_psql('postgres', 'checkpoint') for i in idx_ptrack: @@ -1753,7 +2248,7 @@ def test_ptrack_clean(self): # Take PAGE backup to clean every ptrack self.backup_node( backup_dir, 'node', node, - backup_type='page', options=['-j10']) + backup_type='page', options=['-j10', '--stream']) node.safe_psql('postgres', 'checkpoint') for i in idx_ptrack: @@ -1767,26 +2262,22 @@ def test_ptrack_clean(self): # check that ptrack bits are cleaned self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['size']) - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - # @unittest.expectedFailure + @unittest.skip("skip") def test_ptrack_clean_replica(self): """ Take backups of every available types from - master and check that PTRACK on replica is clean + master and check that PTRACK on replica is clean. + Relevant only for PTRACK 1.x """ - fname = self.id().split('.')[3] master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), + base_dir=os.path.join(self.module_name, self.fname, 'master'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'ptrack_enable': 'on', 'archive_timeout': '30s'}) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'master', master) master.slow_start() @@ -1794,14 +2285,13 @@ def test_ptrack_clean_replica(self): self.backup_node(backup_dir, 'master', master, options=['--stream']) replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) + base_dir=os.path.join(self.module_name, self.fname, 'replica')) replica.cleanup() self.restore_node(backup_dir, 'master', replica) self.add_instance(backup_dir, 'replica', replica) self.set_replica(master, replica, synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) replica.slow_start(replica=True) # Create table and indexes @@ -1883,225 +2373,479 @@ def test_ptrack_clean_replica(self): master.safe_psql('postgres', 'vacuum t_heap') master.safe_psql('postgres', 'checkpoint') - # Take PAGE backup to clean every ptrack - self.backup_node( - backup_dir, - 'replica', - replica, - backup_type='page', - options=[ - '-j10', '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port), - '--stream']) - master.safe_psql('postgres', 'checkpoint') + # Take PAGE backup to clean every ptrack + self.backup_node( + backup_dir, + 'replica', + replica, + backup_type='page', + options=[ + '-j10', '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port), + '--stream']) + master.safe_psql('postgres', 'checkpoint') + + for i in idx_ptrack: + # get new size of heap and indexes and calculate it in pages + idx_ptrack[i]['size'] = self.get_fork_size(replica, i) + # update path to heap and index files in case they`ve changed + idx_ptrack[i]['path'] = self.get_fork_path(replica, i) + # # get ptrack for every idx + idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( + replica, idx_ptrack[i]['path'], [idx_ptrack[i]['size']]) + # check that ptrack bits are cleaned + self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['size']) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_ptrack_cluster_on_btree(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + self.create_tblspace_in_node(node, 'somedata') + + # Create table and indexes + node.safe_psql( + "postgres", + "create extension bloom; create sequence t_seq; " + "create table t_heap tablespace somedata " + "as select i as id, nextval('t_seq') as t_seq, " + "md5(i::text) as text, md5(repeat(i::text,10))::tsvector " + "as tsvector from generate_series(0,2560) i") + for i in idx_ptrack: + if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': + node.safe_psql( + "postgres", + "create index {0} on {1} using {2}({3}) " + "tablespace somedata".format( + i, idx_ptrack[i]['relation'], + idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + + node.safe_psql('postgres', 'vacuum t_heap') + node.safe_psql('postgres', 'checkpoint') + + if node.major_version < 11: + for i in idx_ptrack: + # get size of heap and indexes. size calculated in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(node, i) + # calculate md5sums of pages + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + + self.backup_node( + backup_dir, 'node', node, options=['-j10', '--stream']) + + node.safe_psql('postgres', 'delete from t_heap where id%2 = 1') + node.safe_psql('postgres', 'cluster t_heap using t_btree') + node.safe_psql('postgres', 'checkpoint') + + # CHECK PTRACK SANITY + if node.major_version < 11: + self.check_ptrack_map_sanity(node, idx_ptrack) + + # @unittest.skip("skip") + def test_ptrack_cluster_on_gist(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + # Create table and indexes + node.safe_psql( + "postgres", + "create extension bloom; create sequence t_seq; " + "create table t_heap as select i as id, " + "nextval('t_seq') as t_seq, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,2560) i") + for i in idx_ptrack: + if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': + node.safe_psql( + "postgres", + "create index {0} on {1} using {2}({3})".format( + i, idx_ptrack[i]['relation'], + idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + + node.safe_psql('postgres', 'vacuum t_heap') + node.safe_psql('postgres', 'checkpoint') + + for i in idx_ptrack: + # get size of heap and indexes. size calculated in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(node, i) + # calculate md5sums of pages + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + + self.backup_node( + backup_dir, 'node', node, options=['-j10', '--stream']) + + node.safe_psql('postgres', 'delete from t_heap where id%2 = 1') + node.safe_psql('postgres', 'cluster t_heap using t_gist') + node.safe_psql('postgres', 'checkpoint') + + # CHECK PTRACK SANITY + if node.major_version < 11: + self.check_ptrack_map_sanity(node, idx_ptrack) + + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['-j10', '--stream']) + + pgdata = self.pgdata_content(node.data_dir) + node.cleanup() + + self.restore_node(backup_dir, 'node', node) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_ptrack_cluster_on_btree_replica(self): + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + master.slow_start() + + if master.major_version >= 11: + master.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + self.backup_node(backup_dir, 'master', master, options=['--stream']) + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.restore_node(backup_dir, 'master', replica) + + self.add_instance(backup_dir, 'replica', replica) + self.set_replica(master, replica, synchronous=True) + replica.slow_start(replica=True) + + # Create table and indexes + master.safe_psql( + "postgres", + "create extension bloom; create sequence t_seq; " + "create table t_heap as select i as id, " + "nextval('t_seq') as t_seq, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,2560) i") + + for i in idx_ptrack: + if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': + master.safe_psql( + "postgres", + "create index {0} on {1} using {2}({3})".format( + i, idx_ptrack[i]['relation'], + idx_ptrack[i]['type'], + idx_ptrack[i]['column'])) + + master.safe_psql('postgres', 'vacuum t_heap') + master.safe_psql('postgres', 'checkpoint') + + self.backup_node( + backup_dir, 'replica', replica, options=[ + '-j10', '--stream', '--master-host=localhost', + '--master-db=postgres', '--master-port={0}'.format( + master.port)]) + + for i in idx_ptrack: + # get size of heap and indexes. size calculated in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(replica, i) + # calculate md5sums of pages + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + + master.safe_psql('postgres', 'delete from t_heap where id%2 = 1') + master.safe_psql('postgres', 'cluster t_heap using t_btree') + master.safe_psql('postgres', 'checkpoint') + + # Sync master and replica + self.wait_until_replica_catch_with_master(master, replica) + replica.safe_psql('postgres', 'checkpoint') + + # CHECK PTRACK SANITY + if master.major_version < 11: + self.check_ptrack_map_sanity(replica, idx_ptrack) + + self.backup_node( + backup_dir, 'replica', replica, + backup_type='ptrack', options=['-j10', '--stream']) + + pgdata = self.pgdata_content(replica.data_dir) + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node')) + node.cleanup() + + self.restore_node(backup_dir, 'replica', node) - for i in idx_ptrack: - # get new size of heap and indexes and calculate it in pages - idx_ptrack[i]['size'] = self.get_fork_size(replica, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - replica, idx_ptrack[i]['path'], [idx_ptrack[i]['size']]) - # check that ptrack bits are cleaned - self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['size']) + pgdata_restored = self.pgdata_content(replica.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) - # Clean after yourself - self.del_test_dir(module_name, fname) # @unittest.skip("skip") - # @unittest.expectedFailure - def test_ptrack_cluster_on_btree(self): - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + def test_ptrack_cluster_on_gist_replica(self): + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) + ptrack_enable=True) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() + self.add_instance(backup_dir, 'master', master) + master.slow_start() - self.create_tblspace_in_node(node, 'somedata') + if master.major_version >= 11: + master.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + self.backup_node(backup_dir, 'master', master, options=['--stream']) + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.restore_node(backup_dir, 'master', replica) + + self.add_instance(backup_dir, 'replica', replica) + self.set_replica(master, replica, 'replica', synchronous=True) + replica.slow_start(replica=True) # Create table and indexes - node.safe_psql( + master.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap tablespace somedata " - "as select i as id, nextval('t_seq') as t_seq, " - "md5(i::text) as text, md5(repeat(i::text,10))::tsvector " - "as tsvector from generate_series(0,2560) i") + "create table t_heap as select i as id, " + "nextval('t_seq') as t_seq, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,2560) i") + for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': - node.safe_psql( + master.safe_psql( "postgres", - "create index {0} on {1} using {2}({3}) " - "tablespace somedata".format( + "create index {0} on {1} using {2}({3})".format( i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + idx_ptrack[i]['type'], + idx_ptrack[i]['column'])) - node.safe_psql('postgres', 'vacuum t_heap') - node.safe_psql('postgres', 'checkpoint') + master.safe_psql('postgres', 'vacuum t_heap') + master.safe_psql('postgres', 'checkpoint') + + # Sync master and replica + self.wait_until_replica_catch_with_master(master, replica) + replica.safe_psql('postgres', 'checkpoint') + + self.backup_node( + backup_dir, 'replica', replica, options=[ + '-j10', '--stream', '--master-host=localhost', + '--master-db=postgres', '--master-port={0}'.format( + master.port)]) for i in idx_ptrack: # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) + idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(node, i) + idx_ptrack[i]['path'] = self.get_fork_path(replica, i) # calculate md5sums of pages idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - self.backup_node( - backup_dir, 'node', node, options=['-j10', '--stream']) + master.safe_psql('postgres', 'DELETE FROM t_heap WHERE id%2 = 1') + master.safe_psql('postgres', 'CLUSTER t_heap USING t_gist') - node.safe_psql('postgres', 'delete from t_heap where id%2 = 1') - node.safe_psql('postgres', 'cluster t_heap using t_btree') - node.safe_psql('postgres', 'checkpoint') + if master.major_version < 11: + master.safe_psql('postgres', 'CHECKPOINT') - # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(node, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + # Sync master and replica + self.wait_until_replica_catch_with_master(master, replica) - # compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + if master.major_version < 11: + replica.safe_psql('postgres', 'CHECKPOINT') + self.check_ptrack_map_sanity(replica, idx_ptrack) - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + self.backup_node( + backup_dir, 'replica', replica, + backup_type='ptrack', options=['-j10', '--stream']) + + if self.paranoia: + pgdata = self.pgdata_content(replica.data_dir) + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node')) + node.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) + self.restore_node(backup_dir, 'replica', node) + + if self.paranoia: + pgdata_restored = self.pgdata_content(replica.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") - def test_ptrack_cluster_on_gist(self): - fname = self.id().split('.')[3] + # @unittest.expectedFailure + def test_ptrack_empty(self): + """Take backups of every available types and check that PTRACK is clean""" node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() - # Create table and indexes + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + self.create_tblspace_in_node(node, 'somedata') + + # Create table node.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap as select i as id, " - "nextval('t_seq') as t_seq, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,2560) i") + "create table t_heap " + "(id int DEFAULT nextval('t_seq'), text text, tsvector tsvector) " + "tablespace somedata") + + # Take FULL backup to clean every ptrack + self.backup_node( + backup_dir, 'node', node, + options=['-j10', '--stream']) + + # Create indexes for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': node.safe_psql( "postgres", - "create index {0} on {1} using {2}({3})".format( + "create index {0} on {1} using {2}({3}) " + "tablespace somedata".format( i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + idx_ptrack[i]['type'], + idx_ptrack[i]['column'])) - node.safe_psql('postgres', 'vacuum t_heap') node.safe_psql('postgres', 'checkpoint') - for i in idx_ptrack: - # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate md5sums of pages - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - - self.backup_node( - backup_dir, 'node', node, options=['-j10', '--stream']) + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() - node.safe_psql('postgres', 'delete from t_heap where id%2 = 1') - node.safe_psql('postgres', 'cluster t_heap using t_gist') - node.safe_psql('postgres', 'checkpoint') + tblspace1 = self.get_tblspace_path(node, 'somedata') + tblspace2 = self.get_tblspace_path(node_restored, 'somedata') - # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(node, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + # Take PTRACK backup + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type='ptrack', + options=['-j10', '--stream']) - # Compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + self.restore_node( + backup_dir, 'node', node_restored, + backup_id=backup_id, + options=[ + "-j", "4", + "-T{0}={1}".format(tblspace1, tblspace2)]) - # Clean after yourself - self.del_test_dir(module_name, fname) + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") - def test_ptrack_cluster_on_btree_replica(self): - fname = self.id().split('.')[3] + # @unittest.expectedFailure + def test_ptrack_empty_replica(self): + """ + Take backups of every available types from master + and check that PTRACK on replica is clean + """ master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), + base_dir=os.path.join(self.module_name, self.fname, 'master'), set_replication=True, initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) + ptrack_enable=True) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'master', master) master.slow_start() + if master.major_version >= 11: + master.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.backup_node(backup_dir, 'master', master, options=['--stream']) replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) + base_dir=os.path.join(self.module_name, self.fname, 'replica')) replica.cleanup() self.restore_node(backup_dir, 'master', replica) self.add_instance(backup_dir, 'replica', replica) self.set_replica(master, replica, synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) replica.slow_start(replica=True) - # Create table and indexes + # Create table master.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap as select i as id, " - "nextval('t_seq') as t_seq, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,2560) i") + "create table t_heap " + "(id int DEFAULT nextval('t_seq'), text text, tsvector tsvector)") + self.wait_until_replica_catch_with_master(master, replica) + + # Take FULL backup + self.backup_node( + backup_dir, + 'replica', + replica, + options=[ + '-j10', '--stream', + '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port)]) + # Create indexes for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': master.safe_psql( @@ -2111,92 +2855,153 @@ def test_ptrack_cluster_on_btree_replica(self): idx_ptrack[i]['type'], idx_ptrack[i]['column'])) - master.safe_psql('postgres', 'vacuum t_heap') - master.safe_psql('postgres', 'checkpoint') + self.wait_until_replica_catch_with_master(master, replica) + + # Take PTRACK backup + backup_id = self.backup_node( + backup_dir, + 'replica', + replica, + backup_type='ptrack', + options=[ + '-j1', '--stream', + '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port)]) + + if self.paranoia: + pgdata = self.pgdata_content(replica.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node( + backup_dir, 'replica', node_restored, + backup_id=backup_id, options=["-j", "4"]) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_ptrack_truncate(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + self.create_tblspace_in_node(node, 'somedata') + + # Create table and indexes + node.safe_psql( + "postgres", + "create extension bloom; create sequence t_seq; " + "create table t_heap tablespace somedata " + "as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,2560) i") + + if node.major_version < 11: + for i in idx_ptrack: + if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': + node.safe_psql( + "postgres", + "create index {0} on {1} using {2}({3}) " + "tablespace somedata".format( + i, idx_ptrack[i]['relation'], + idx_ptrack[i]['type'], idx_ptrack[i]['column'])) self.backup_node( - backup_dir, 'replica', replica, options=[ - '-j10', '--stream', '--master-host=localhost', - '--master-db=postgres', '--master-port={0}'.format( - master.port)]) - - for i in idx_ptrack: - # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate md5sums of pages - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + backup_dir, 'node', node, options=['--stream']) - master.safe_psql('postgres', 'delete from t_heap where id%2 = 1') - master.safe_psql('postgres', 'cluster t_heap using t_btree') - master.safe_psql('postgres', 'checkpoint') + node.safe_psql('postgres', 'truncate t_heap') + node.safe_psql('postgres', 'checkpoint') - # Sync master and replica - self.wait_until_replica_catch_with_master(master, replica) - replica.safe_psql('postgres', 'checkpoint') + if node.major_version < 11: + for i in idx_ptrack: + # get fork size and calculate it in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(node, i) + # calculate md5sums for every page of this fork + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + + # Make backup to clean every ptrack + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['-j10', '--stream']) - # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(replica, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - replica, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + pgdata = self.pgdata_content(node.data_dir) - # compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + if node.major_version < 11: + for i in idx_ptrack: + idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( + node, idx_ptrack[i]['path'], [idx_ptrack[i]['old_size']]) + self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['old_size']) - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + node.cleanup() + shutil.rmtree( + self.get_tblspace_path(node, 'somedata'), + ignore_errors=True) + + self.restore_node(backup_dir, 'node', node) - # Clean after yourself - self.del_test_dir(module_name, fname) + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") - def test_ptrack_cluster_on_gist_replica(self): - fname = self.id().split('.')[3] + def test_basic_ptrack_truncate_replica(self): master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), + base_dir=os.path.join(self.module_name, self.fname, 'master'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'ptrack_enable': 'on'}) + 'max_wal_size': '32MB', + 'archive_timeout': '10s', + 'checkpoint_timeout': '5min'}) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'master', master) master.slow_start() + if master.major_version >= 11: + master.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.backup_node(backup_dir, 'master', master, options=['--stream']) replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) + base_dir=os.path.join(self.module_name, self.fname, 'replica')) replica.cleanup() self.restore_node(backup_dir, 'master', replica) self.add_instance(backup_dir, 'replica', replica) self.set_replica(master, replica, 'replica', synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) replica.slow_start(replica=True) # Create table and indexes master.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap as select i as id, " - "nextval('t_seq') as t_seq, md5(i::text) as text, " + "create table t_heap " + "as select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,2560) i") @@ -2204,101 +3009,112 @@ def test_ptrack_cluster_on_gist_replica(self): if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': master.safe_psql( "postgres", - "create index {0} on {1} using {2}({3})".format( + "create index {0} on {1} using {2}({3}) ".format( i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], - idx_ptrack[i]['column'])) - - master.safe_psql('postgres', 'vacuum t_heap') - master.safe_psql('postgres', 'checkpoint') + idx_ptrack[i]['type'], idx_ptrack[i]['column'])) # Sync master and replica self.wait_until_replica_catch_with_master(master, replica) replica.safe_psql('postgres', 'checkpoint') + if replica.major_version < 11: + for i in idx_ptrack: + # get fork size and calculate it in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(replica, i) + # calculate md5sums for every page of this fork + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + + # Make backup to clean every ptrack self.backup_node( - backup_dir, 'replica', replica, options=[ - '-j10', '--stream', '--master-host=localhost', - '--master-db=postgres', '--master-port={0}'.format( - master.port)]) + backup_dir, 'replica', replica, + options=[ + '-j10', + '--stream', + '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port)]) - for i in idx_ptrack: - # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate md5sums of pages - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + if replica.major_version < 11: + for i in idx_ptrack: + idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( + replica, idx_ptrack[i]['path'], [idx_ptrack[i]['old_size']]) + self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['old_size']) - master.safe_psql('postgres', 'delete from t_heap where id%2 = 1') - master.safe_psql('postgres', 'cluster t_heap using t_gist') - master.safe_psql('postgres', 'checkpoint') + master.safe_psql('postgres', 'truncate t_heap') # Sync master and replica self.wait_until_replica_catch_with_master(master, replica) - replica.safe_psql('postgres', 'checkpoint') - # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(replica, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - replica, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + if replica.major_version < 10: + replica.safe_psql( + "postgres", + "select pg_xlog_replay_pause()") + else: + replica.safe_psql( + "postgres", + "select pg_wal_replay_pause()") + + self.backup_node( + backup_dir, 'replica', replica, backup_type='ptrack', + options=[ + '-j10', + '--stream', + '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port)]) - # Compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + pgdata = self.pgdata_content(replica.data_dir) - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node')) + node.cleanup() + + self.restore_node(backup_dir, 'replica', node, data_dir=node.data_dir) + + pgdata_restored = self.pgdata_content(node.data_dir) - # Clean after yourself - self.del_test_dir(module_name, fname) + if self.paranoia: + self.compare_pgdata(pgdata, pgdata_restored) + + self.set_auto_conf(node, {'port': node.port}) + + node.slow_start() + + node.safe_psql( + 'postgres', + 'select 1') # @unittest.skip("skip") # @unittest.expectedFailure - def test_ptrack_empty(self): - """Take backups of every available types and check that PTRACK is clean""" - fname = self.id().split('.')[3] + def test_ptrack_vacuum(self): node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on', - 'autovacuum': 'off'}) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + ptrack_enable=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.create_tblspace_in_node(node, 'somedata') - # Create table + # Create table and indexes node.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap " - "(id int DEFAULT nextval('t_seq'), text text, tsvector tsvector) " - "tablespace somedata") - - # Take FULL backup to clean every ptrack - self.backup_node( - backup_dir, 'node', node, - options=['-j10', '--stream']) - - # Create indexes + "create table t_heap tablespace somedata " + "as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,2560) i") for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': node.safe_psql( @@ -2309,162 +3125,185 @@ def test_ptrack_empty(self): idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + comparision_exclusion = self.get_known_bugs_comparision_exclusion_dict(node) + + node.safe_psql('postgres', 'vacuum t_heap') node.safe_psql('postgres', 'checkpoint') - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - node_restored.cleanup() + # Make full backup to clean every ptrack + self.backup_node( + backup_dir, 'node', node, options=['-j10', '--stream']) - tblspace1 = self.get_tblspace_path(node, 'somedata') - tblspace2 = self.get_tblspace_path(node_restored, 'somedata') + if node.major_version < 11: + for i in idx_ptrack: + # get fork size and calculate it in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(node, i) + # calculate md5sums for every page of this fork + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( + node, idx_ptrack[i]['path'], [idx_ptrack[i]['old_size']]) + self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['old_size']) - # Take PTRACK backup - backup_id = self.backup_node( - backup_dir, 'node', node, backup_type='ptrack', - options=['-j10']) + # Delete some rows, vacuum it and make checkpoint + node.safe_psql('postgres', 'delete from t_heap where id%2 = 1') + node.safe_psql('postgres', 'vacuum t_heap') + node.safe_psql('postgres', 'checkpoint') - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) + # CHECK PTRACK SANITY + if node.major_version < 11: + self.check_ptrack_map_sanity(node, idx_ptrack) - self.restore_node( - backup_dir, 'node', node_restored, - backup_id=backup_id, - options=[ - "-j", "4", - "-T{0}={1}".format(tblspace1, tblspace2)] - ) + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['-j10', '--stream']) - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) + pgdata = self.pgdata_content(node.data_dir) + node.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) + shutil.rmtree( + self.get_tblspace_path(node, 'somedata'), + ignore_errors=True) + + self.restore_node(backup_dir, 'node', node) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored, comparision_exclusion) # @unittest.skip("skip") - # @unittest.expectedFailure - def test_ptrack_empty_replica(self): - """ - Take backups of every available types from master - and check that PTRACK on replica is clean - """ - fname = self.id().split('.')[3] + def test_ptrack_vacuum_replica(self): master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), + base_dir=os.path.join(self.module_name, self.fname, 'master'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'ptrack_enable': 'on'}) + 'checkpoint_timeout': '30'}) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'master', master) master.slow_start() + if master.major_version >= 11: + master.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.backup_node(backup_dir, 'master', master, options=['--stream']) replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) + base_dir=os.path.join(self.module_name, self.fname, 'replica')) replica.cleanup() self.restore_node(backup_dir, 'master', replica) self.add_instance(backup_dir, 'replica', replica) - self.set_replica(master, replica, synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) + self.set_replica(master, replica, 'replica', synchronous=True) replica.slow_start(replica=True) - # Create table + # Create table and indexes master.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap " - "(id int DEFAULT nextval('t_seq'), text text, tsvector tsvector)") - self.wait_until_replica_catch_with_master(master, replica) - - # Take FULL backup - self.backup_node( - backup_dir, - 'replica', - replica, - options=[ - '-j10', '--stream', - '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port)]) + "create table t_heap as select i as id, " + "md5(i::text) as text, md5(repeat(i::text,10))::tsvector " + "as tsvector from generate_series(0,2560) i") - # Create indexes for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': master.safe_psql( "postgres", "create index {0} on {1} using {2}({3})".format( i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], - idx_ptrack[i]['column'])) + idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + master.safe_psql('postgres', 'vacuum t_heap') + master.safe_psql('postgres', 'checkpoint') + # Sync master and replica self.wait_until_replica_catch_with_master(master, replica) + replica.safe_psql('postgres', 'checkpoint') - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - node_restored.cleanup() - - # Take PTRACK backup - backup_id = self.backup_node( - backup_dir, - 'replica', - replica, - backup_type='ptrack', - options=[ - '-j10', '--stream', - '--master-host=localhost', + # Make FULL backup to clean every ptrack + self.backup_node( + backup_dir, 'replica', replica, options=[ + '-j10', '--master-host=localhost', '--master-db=postgres', - '--master-port={0}'.format(master.port)]) - - if self.paranoia: - pgdata = self.pgdata_content(replica.data_dir) + '--master-port={0}'.format(master.port), + '--stream']) - self.restore_node( - backup_dir, 'replica', node_restored, - backup_id=backup_id, - options=["-j", "4"] - ) + if replica.major_version < 11: + for i in idx_ptrack: + # get fork size and calculate it in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(replica, i) + # calculate md5sums for every page of this fork + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( + replica, idx_ptrack[i]['path'], [idx_ptrack[i]['old_size']]) + self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['old_size']) - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) + # Delete some rows, vacuum it and make checkpoint + master.safe_psql('postgres', 'delete from t_heap where id%2 = 1') + master.safe_psql('postgres', 'vacuum t_heap') + master.safe_psql('postgres', 'checkpoint') + + # Sync master and replica + self.wait_until_replica_catch_with_master(master, replica) + replica.safe_psql('postgres', 'checkpoint') + + # CHECK PTRACK SANITY + if replica.major_version < 11: + self.check_ptrack_map_sanity(master, idx_ptrack) - # Clean after yourself - self.del_test_dir(module_name, fname) + self.backup_node( + backup_dir, 'replica', replica, + backup_type='ptrack', options=['-j10', '--stream']) + + pgdata = self.pgdata_content(replica.data_dir) + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node')) + node.cleanup() + + self.restore_node(backup_dir, 'replica', node, data_dir=node.data_dir) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") # @unittest.expectedFailure - def test_ptrack_truncate(self): - fname = self.id().split('.')[3] + def test_ptrack_vacuum_bits_frozen(self): node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.create_tblspace_in_node(node, 'somedata') # Create table and indexes - node.safe_psql( + res = node.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " "create table t_heap tablespace somedata " "as select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,2560) i") - for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': node.safe_psql( @@ -2472,109 +3311,120 @@ def test_ptrack_truncate(self): "create index {0} on {1} using {2}({3}) " "tablespace somedata".format( i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + idx_ptrack[i]['type'], + idx_ptrack[i]['column'])) - node.safe_psql('postgres', 'truncate t_heap') + comparision_exclusion = self.get_known_bugs_comparision_exclusion_dict(node) node.safe_psql('postgres', 'checkpoint') - for i in idx_ptrack: - # get fork size and calculate it in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate md5sums for every page of this fork - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - - # Make full backup to clean every ptrack self.backup_node( backup_dir, 'node', node, options=['-j10', '--stream']) - for i in idx_ptrack: - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, idx_ptrack[i]['path'], [idx_ptrack[i]['old_size']]) - self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['old_size']) + node.safe_psql('postgres', 'vacuum freeze t_heap') + node.safe_psql('postgres', 'checkpoint') - # Clean after yourself - self.del_test_dir(module_name, fname) + if node.major_version < 11: + for i in idx_ptrack: + # get size of heap and indexes. size calculated in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(node, i) + # calculate md5sums of pages + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - @unittest.skip("skip") - def test_ptrack_truncate_replica(self): - fname = self.id().split('.')[3] + # CHECK PTRACK SANITY + if node.major_version < 11: + self.check_ptrack_map_sanity(node, idx_ptrack) + + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['-j10', '--stream']) + + pgdata = self.pgdata_content(node.data_dir) + node.cleanup() + shutil.rmtree( + self.get_tblspace_path(node, 'somedata'), + ignore_errors=True) + + self.restore_node(backup_dir, 'node', node) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored, comparision_exclusion) + + # @unittest.skip("skip") + def test_ptrack_vacuum_bits_frozen_replica(self): master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), + base_dir=os.path.join(self.module_name, self.fname, 'master'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on', - 'max_wal_size': '32MB', - 'archive_timeout': '10s', - 'checkpoint_timeout': '30s'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'master', master) master.slow_start() + if master.major_version >= 11: + master.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.backup_node(backup_dir, 'master', master, options=['--stream']) replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) + base_dir=os.path.join(self.module_name, self.fname, 'replica')) replica.cleanup() self.restore_node(backup_dir, 'master', replica) self.add_instance(backup_dir, 'replica', replica) - self.set_replica(master, replica, 'replica', synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) + self.set_replica(master, replica, synchronous=True) replica.slow_start(replica=True) # Create table and indexes - self.create_tblspace_in_node(master, 'somedata') master.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap tablespace somedata " - "as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,2560) i") + "create table t_heap as select i as id, " + "md5(i::text) as text, md5(repeat(i::text,10))::tsvector " + "as tsvector from generate_series(0,2560) i") for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': master.safe_psql( - "postgres", "create index {0} on {1} using {2}({3}) " - "tablespace somedata".format( + "postgres", + "create index {0} on {1} using {2}({3})".format( i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + idx_ptrack[i]['type'], + idx_ptrack[i]['column'])) + + master.safe_psql('postgres', 'checkpoint') # Sync master and replica self.wait_until_replica_catch_with_master(master, replica) replica.safe_psql('postgres', 'checkpoint') - for i in idx_ptrack: - # get fork size and calculate it in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate md5sums for every page of this fork - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - - # Make full backup to clean every ptrack + # Take backup to clean every ptrack self.backup_node( backup_dir, 'replica', replica, options=[ '-j10', - '--stream', '--master-host=localhost', '--master-db=postgres', - '--master-port={0}'.format(master.port)]) + '--master-port={0}'.format(master.port), + '--stream']) - for i in idx_ptrack: - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - replica, idx_ptrack[i]['path'], [idx_ptrack[i]['old_size']]) - self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['old_size']) + if replica.major_version < 11: + for i in idx_ptrack: + # get size of heap and indexes. size calculated in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(replica, i) + # calculate md5sums of pages + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - master.safe_psql('postgres', 'truncate t_heap') + master.safe_psql('postgres', 'vacuum freeze t_heap') master.safe_psql('postgres', 'checkpoint') # Sync master and replica @@ -2582,56 +3432,50 @@ def test_ptrack_truncate_replica(self): replica.safe_psql('postgres', 'checkpoint') # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes and calculate it in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(replica, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - replica, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + if replica.major_version < 11: + self.check_ptrack_map_sanity(master, idx_ptrack) - # compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + self.backup_node( + backup_dir, 'replica', replica, backup_type='ptrack', + options=['-j10', '--stream']) - self.assertTrue( - success, 'Ptrack has failed to register changes in data files') + pgdata = self.pgdata_content(replica.data_dir) + replica.cleanup() + + self.restore_node(backup_dir, 'replica', replica) - # Clean after yourself - self.del_test_dir(module_name, fname) + pgdata_restored = self.pgdata_content(replica.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") # @unittest.expectedFailure - def test_ptrack_vacuum(self): - fname = self.id().split('.')[3] + def test_ptrack_vacuum_bits_visibility(self): node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.create_tblspace_in_node(node, 'somedata') # Create table and indexes - node.safe_psql( + res = node.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " "create table t_heap tablespace somedata " "as select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,2560) i") + for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': node.safe_psql( @@ -2639,86 +3483,149 @@ def test_ptrack_vacuum(self): "create index {0} on {1} using {2}({3}) " "tablespace somedata".format( i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], - idx_ptrack[i]['column'])) + idx_ptrack[i]['type'], idx_ptrack[i]['column'])) - node.safe_psql('postgres', 'vacuum t_heap') + comparision_exclusion = self.get_known_bugs_comparision_exclusion_dict(node) node.safe_psql('postgres', 'checkpoint') - # Make full backup to clean every ptrack self.backup_node( backup_dir, 'node', node, options=['-j10', '--stream']) - for i in idx_ptrack: - # get fork size and calculate it in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate md5sums for every page of this fork - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, idx_ptrack[i]['path'], [idx_ptrack[i]['old_size']]) - self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['old_size']) + if node.major_version < 11: + for i in idx_ptrack: + # get size of heap and indexes. size calculated in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(node, i) + # calculate md5sums of pages + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - # Delete some rows, vacuum it and make checkpoint - node.safe_psql('postgres', 'delete from t_heap where id%2 = 1') node.safe_psql('postgres', 'vacuum t_heap') node.safe_psql('postgres', 'checkpoint') # CHECK PTRACK SANITY - success = True + if node.major_version < 11: + self.check_ptrack_map_sanity(node, idx_ptrack) + + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['-j10', '--stream']) + + pgdata = self.pgdata_content(node.data_dir) + node.cleanup() + shutil.rmtree( + self.get_tblspace_path(node, 'somedata'), + ignore_errors=True) + + self.restore_node(backup_dir, 'node', node) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored, comparision_exclusion) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_ptrack_vacuum_full_2(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + pg_options={ 'wal_log_hints': 'on' }) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + self.create_tblspace_in_node(node, 'somedata') + + # Create table and indexes + res = node.safe_psql( + "postgres", + "create extension bloom; create sequence t_seq; " + "create table t_heap tablespace somedata " + "as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,2560) i") for i in idx_ptrack: - # get new size of heap and indexes and calculate it in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(node, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': + node.safe_psql( + "postgres", "create index {0} on {1} " + "using {2}({3}) tablespace somedata".format( + i, idx_ptrack[i]['relation'], + idx_ptrack[i]['type'], idx_ptrack[i]['column'])) - # compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + node.safe_psql('postgres', 'vacuum t_heap') + node.safe_psql('postgres', 'checkpoint') - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + self.backup_node( + backup_dir, 'node', node, options=['-j10', '--stream']) - # Clean after yourself - self.del_test_dir(module_name, fname) + if node.major_version < 11: + for i in idx_ptrack: + # get size of heap and indexes. size calculated in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(node, i) + # calculate md5sums of pages + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + + node.safe_psql('postgres', 'delete from t_heap where id%2 = 1') + node.safe_psql('postgres', 'vacuum full t_heap') + node.safe_psql('postgres', 'checkpoint') + + if node.major_version < 11: + self.check_ptrack_map_sanity(node, idx_ptrack) + + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['-j10', '--stream']) + + pgdata = self.pgdata_content(node.data_dir) + node.cleanup() + + shutil.rmtree( + self.get_tblspace_path(node, 'somedata'), + ignore_errors=True) + + self.restore_node(backup_dir, 'node', node) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") - def test_ptrack_vacuum_replica(self): - fname = self.id().split('.')[3] + # @unittest.expectedFailure + def test_ptrack_vacuum_full_replica(self): master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), + base_dir=os.path.join(self.module_name, self.fname, 'master'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on', - 'checkpoint_timeout': '30'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'master', master) master.slow_start() - self.backup_node(backup_dir, 'master', master, options=['--stream']) + if master.major_version >= 11: + master.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.backup_node(backup_dir, 'master', master, options=['--stream']) replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) + base_dir=os.path.join(self.module_name, self.fname, 'replica')) replica.cleanup() self.restore_node(backup_dir, 'master', replica) self.add_instance(backup_dir, 'replica', replica) self.set_replica(master, replica, 'replica', synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) replica.slow_start(replica=True) # Create table and indexes @@ -2726,16 +3633,18 @@ def test_ptrack_vacuum_replica(self): "postgres", "create extension bloom; create sequence t_seq; " "create table t_heap as select i as id, " - "md5(i::text) as text, md5(repeat(i::text,10))::tsvector " - "as tsvector from generate_series(0,2560) i") + "md5(i::text) as text, md5(repeat(i::text,10))::tsvector as " + "tsvector from generate_series(0,256000) i") - for i in idx_ptrack: - if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': - master.safe_psql( - "postgres", - "create index {0} on {1} using {2}({3})".format( - i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + if master.major_version < 11: + for i in idx_ptrack: + if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': + master.safe_psql( + "postgres", + "create index {0} on {1} using {2}({3})".format( + i, idx_ptrack[i]['relation'], + idx_ptrack[i]['type'], + idx_ptrack[i]['column'])) master.safe_psql('postgres', 'vacuum t_heap') master.safe_psql('postgres', 'checkpoint') @@ -2744,166 +3653,151 @@ def test_ptrack_vacuum_replica(self): self.wait_until_replica_catch_with_master(master, replica) replica.safe_psql('postgres', 'checkpoint') - # Make FULL backup to clean every ptrack + # Take FULL backup to clean every ptrack self.backup_node( - backup_dir, 'replica', replica, options=[ - '-j10', '--master-host=localhost', + backup_dir, 'replica', replica, + options=[ + '-j10', + '--master-host=localhost', '--master-db=postgres', '--master-port={0}'.format(master.port), '--stream']) - for i in idx_ptrack: - # get fork size and calculate it in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate md5sums for every page of this fork - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - replica, idx_ptrack[i]['path'], [idx_ptrack[i]['old_size']]) - self.check_ptrack_clean(idx_ptrack[i], idx_ptrack[i]['old_size']) + if replica.major_version < 11: + for i in idx_ptrack: + # get size of heap and indexes. size calculated in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(replica, i) + # calculate md5sums of pages + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - # Delete some rows, vacuum it and make checkpoint master.safe_psql('postgres', 'delete from t_heap where id%2 = 1') - master.safe_psql('postgres', 'vacuum t_heap') + master.safe_psql('postgres', 'vacuum full t_heap') master.safe_psql('postgres', 'checkpoint') # Sync master and replica self.wait_until_replica_catch_with_master(master, replica) replica.safe_psql('postgres', 'checkpoint') - # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes and calculate it in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(replica, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - replica, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + if replica.major_version < 11: + self.check_ptrack_map_sanity(master, idx_ptrack) - # compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + self.backup_node( + backup_dir, 'replica', replica, + backup_type='ptrack', options=['-j10', '--stream']) - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + pgdata = self.pgdata_content(replica.data_dir) + replica.cleanup() + + self.restore_node(backup_dir, 'replica', replica) - # Clean after yourself - self.del_test_dir(module_name, fname) + pgdata_restored = self.pgdata_content(replica.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") # @unittest.expectedFailure - def test_ptrack_vacuum_bits_frozen(self): - fname = self.id().split('.')[3] + def test_ptrack_vacuum_truncate_2(self): node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() - self.create_tblspace_in_node(node, 'somedata') + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") # Create table and indexes res = node.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap tablespace somedata " + "create table t_heap " "as select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,2560) i") - for i in idx_ptrack: - if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': - node.safe_psql( - "postgres", - "create index {0} on {1} using {2}({3}) " - "tablespace somedata".format( - i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], - idx_ptrack[i]['column'])) - node.safe_psql('postgres', 'checkpoint') - - self.backup_node( - backup_dir, 'node', node, options=['-j10', '--stream']) - - node.safe_psql('postgres', 'vacuum freeze t_heap') - node.safe_psql('postgres', 'checkpoint') - - for i in idx_ptrack: - # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate md5sums of pages - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + if node.major_version < 11: + for i in idx_ptrack: + if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': + node.safe_psql( + "postgres", "create index {0} on {1} using {2}({3})".format( + i, idx_ptrack[i]['relation'], + idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + + node.safe_psql('postgres', 'VACUUM t_heap') + + self.backup_node( + backup_dir, 'node', node, options=['-j10', '--stream']) + + if node.major_version < 11: + for i in idx_ptrack: + # get size of heap and indexes. size calculated in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(node, i) + # calculate md5sums of pages + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + + node.safe_psql('postgres', 'DELETE FROM t_heap WHERE id > 128') + node.safe_psql('postgres', 'VACUUM t_heap') + node.safe_psql('postgres', 'CHECKPOINT') # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(node, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + if node.major_version < 11: + self.check_ptrack_map_sanity(node, idx_ptrack) + + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['--stream']) - # compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + pgdata = self.pgdata_content(node.data_dir) - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() - # Clean after yourself - self.del_test_dir(module_name, fname) + self.restore_node(backup_dir, 'node', node_restored) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") - def test_ptrack_vacuum_bits_frozen_replica(self): - fname = self.id().split('.')[3] + # @unittest.expectedFailure + def test_ptrack_vacuum_truncate_replica(self): master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), + base_dir=os.path.join(self.module_name, self.fname, 'master'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'master', master) master.slow_start() + if master.major_version >= 11: + master.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + self.backup_node(backup_dir, 'master', master, options=['--stream']) replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) + base_dir=os.path.join(self.module_name, self.fname, 'replica')) replica.cleanup() self.restore_node(backup_dir, 'master', replica) self.add_instance(backup_dir, 'replica', replica) - self.set_replica(master, replica, synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) + self.set_replica(master, replica, 'replica', synchronous=True) replica.slow_start(replica=True) # Create table and indexes @@ -2913,93 +3807,91 @@ def test_ptrack_vacuum_bits_frozen_replica(self): "create table t_heap as select i as id, " "md5(i::text) as text, md5(repeat(i::text,10))::tsvector " "as tsvector from generate_series(0,2560) i") + for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': master.safe_psql( - "postgres", - "create index {0} on {1} using {2}({3})".format( + "postgres", "create index {0} on {1} " + "using {2}({3})".format( i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], - idx_ptrack[i]['column'])) + idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + master.safe_psql('postgres', 'vacuum t_heap') master.safe_psql('postgres', 'checkpoint') - # Sync master and replica - self.wait_until_replica_catch_with_master(master, replica) - replica.safe_psql('postgres', 'checkpoint') - - # Take PTRACK backup to clean every ptrack + # Take FULL backup to clean every ptrack self.backup_node( backup_dir, 'replica', replica, options=[ '-j10', + '--stream', '--master-host=localhost', '--master-db=postgres', - '--master-port={0}'.format(master.port), - '--stream']) + '--master-port={0}'.format(master.port) + ] + ) - for i in idx_ptrack: - # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate md5sums of pages - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + if master.major_version < 11: + for i in idx_ptrack: + # get size of heap and indexes. size calculated in pages + idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) + # get path to heap and index files + idx_ptrack[i]['path'] = self.get_fork_path(replica, i) + # calculate md5sums of pages + idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( + idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - master.safe_psql('postgres', 'vacuum freeze t_heap') - master.safe_psql('postgres', 'checkpoint') + master.safe_psql('postgres', 'DELETE FROM t_heap WHERE id > 128;') + master.safe_psql('postgres', 'VACUUM t_heap') + master.safe_psql('postgres', 'CHECKPOINT') # Sync master and replica self.wait_until_replica_catch_with_master(master, replica) - replica.safe_psql('postgres', 'checkpoint') + replica.safe_psql('postgres', 'CHECKPOINT') # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(replica, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - replica, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + if master.major_version < 11: + self.check_ptrack_map_sanity(master, idx_ptrack) - # compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + self.backup_node( + backup_dir, 'replica', replica, backup_type='ptrack', + options=[ + '--stream', + '--log-level-file=INFO', + '--archive-timeout=30']) - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + pgdata = self.pgdata_content(replica.data_dir) - # Clean after yourself - self.del_test_dir(module_name, fname) + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_ptrack_vacuum_bits_visibility(self): - fname = self.id().split('.')[3] + self.restore_node(backup_dir, 'replica', node_restored) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + @unittest.skip("skip") + def test_ptrack_recovery(self): + """ + Check that ptrack map contain correct bits after recovery. + Actual only for PTRACK 1.x + """ node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() self.create_tblspace_in_node(node, 'somedata') - # Create table and indexes - res = node.safe_psql( + # Create table + node.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " "create table t_heap tablespace somedata " @@ -3007,480 +3899,499 @@ def test_ptrack_vacuum_bits_visibility(self): "md5(repeat(i::text,10))::tsvector as tsvector " "from generate_series(0,2560) i") + # Create indexes for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': node.safe_psql( - "postgres", - "create index {0} on {1} using {2}({3}) " + "postgres", "create index {0} on {1} using {2}({3}) " "tablespace somedata".format( i, idx_ptrack[i]['relation'], idx_ptrack[i]['type'], idx_ptrack[i]['column'])) - node.safe_psql('postgres', 'checkpoint') - - self.backup_node( - backup_dir, 'node', node, options=['-j10', '--stream']) - - for i in idx_ptrack: # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) + idx_ptrack[i]['size'] = int(self.get_fork_size(node, i)) # get path to heap and index files idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate md5sums of pages - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - node.safe_psql('postgres', 'vacuum t_heap') - node.safe_psql('postgres', 'checkpoint') + if self.verbose: + print('Killing postmaster. Losing Ptrack changes') + node.stop(['-m', 'immediate', '-D', node.data_dir]) + if not node.status(): + node.slow_start() + else: + print("Die! Die! Why won't you die?... Why won't you die?") + exit(1) - # CHECK PTRACK SANITY - success = True for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(node, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) # get ptrack for every idx idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) - - # compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False - - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) - - # Clean after yourself - self.del_test_dir(module_name, fname) + node, idx_ptrack[i]['path'], [idx_ptrack[i]['size']]) + # check that ptrack has correct bits after recovery + self.check_ptrack_recovery(idx_ptrack[i]) # @unittest.skip("skip") # @unittest.expectedFailure - def test_ptrack_vacuum_full(self): - fname = self.id().split('.')[3] + def test_ptrack_recovery_1(self): node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'ptrack_enable': 'on'}) + 'shared_buffers': '512MB', + 'max_wal_size': '3GB'}) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() - self.create_tblspace_in_node(node, 'somedata') + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") - # Create table and indexes - res = node.safe_psql( + # Create table + node.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap tablespace somedata " - "as select i as id, md5(i::text) as text, " + "create table t_heap " + "as select nextval('t_seq')::int as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " +# "from generate_series(0,25600) i") "from generate_series(0,2560) i") + + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + # Create indexes for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': node.safe_psql( - "postgres", "create index {0} on {1} " - "using {2}({3}) tablespace somedata".format( + "postgres", + "CREATE INDEX {0} ON {1} USING {2}({3})".format( i, idx_ptrack[i]['relation'], idx_ptrack[i]['type'], idx_ptrack[i]['column'])) - node.safe_psql('postgres', 'vacuum t_heap') - node.safe_psql('postgres', 'checkpoint') + node.safe_psql( + 'postgres', + "update t_heap set id = nextval('t_seq'), text = md5(text), " + "tsvector = md5(repeat(tsvector::text, 10))::tsvector") - self.backup_node( - backup_dir, 'node', node, options=['-j10', '--stream']) + node.safe_psql( + 'postgres', + "create extension pg_buffercache") - for i in idx_ptrack: - # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate md5sums of pages - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + #print(node.safe_psql( + # 'postgres', + # "SELECT count(*) FROM pg_buffercache WHERE isdirty")) - node.safe_psql('postgres', 'delete from t_heap where id%2 = 1') - node.safe_psql('postgres', 'vacuum full t_heap') - node.safe_psql('postgres', 'checkpoint') + if self.verbose: + print('Killing postmaster. Losing Ptrack changes') + node.stop(['-m', 'immediate', '-D', node.data_dir]) - # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(node, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + if not node.status(): + node.slow_start() + else: + print("Die! Die! Why won't you die?... Why won't you die?") + exit(1) - # compare pages and check ptrack sanity, the most important part - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['--stream']) - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + pgdata = self.pgdata_content(node.data_dir) - # Clean after yourself - self.del_test_dir(module_name, fname) + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") # @unittest.expectedFailure - def test_ptrack_vacuum_full_replica(self): - fname = self.id().split('.')[3] - master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), + def test_ptrack_zero_changes(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on', - 'autovacuum': 'off', - 'archive_timeout': '30s'} - ) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'master', master) - master.slow_start() - - self.backup_node(backup_dir, 'master', master, options=['--stream']) - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() - - self.restore_node(backup_dir, 'master', replica) + ptrack_enable=True, + initdb_params=['--data-checksums']) - self.add_instance(backup_dir, 'replica', replica) - self.set_replica(master, replica, 'replica', synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) - replica.slow_start(replica=True) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() - # Create table and indexes - master.safe_psql( + node.safe_psql( "postgres", - "create extension bloom; create sequence t_seq; " - "create table t_heap as select i as id, " - "md5(i::text) as text, md5(repeat(i::text,10))::tsvector as " - "tsvector from generate_series(0,256000) i") - for i in idx_ptrack: - if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': - master.safe_psql( - "postgres", - "create index {0} on {1} using {2}({3})".format( - i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], - idx_ptrack[i]['column'])) - - master.safe_psql('postgres', 'vacuum t_heap') - master.safe_psql('postgres', 'checkpoint') + "CREATE EXTENSION ptrack") - # Sync master and replica - self.wait_until_replica_catch_with_master(master, replica) - replica.safe_psql('postgres', 'checkpoint') + # Create table + node.safe_psql( + "postgres", + "create table t_heap " + "as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,2560) i") - # Take FULL backup to clean every ptrack self.backup_node( - backup_dir, 'replica', replica, - options=[ - '-j10', - '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port), - '--stream' - ] - ) - # TODO: check that all ptrack are nullified - for i in idx_ptrack: - # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate md5sums of pages - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) - - master.safe_psql('postgres', 'delete from t_heap where id%2 = 1') - master.safe_psql('postgres', 'vacuum full t_heap') - master.safe_psql('postgres', 'checkpoint') - - # Sync master and replica - self.wait_until_replica_catch_with_master(master, replica) - replica.safe_psql('postgres', 'checkpoint') + backup_dir, 'node', node, options=['--stream']) - # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(replica, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - replica, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['--stream']) - # compare pages and check ptrack sanity, the most important part - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + pgdata = self.pgdata_content(node.data_dir) + node.cleanup() - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + self.restore_node(backup_dir, 'node', node) - # Clean after yourself - self.del_test_dir(module_name, fname) + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") # @unittest.expectedFailure - def test_ptrack_vacuum_truncate(self): - fname = self.id().split('.')[3] + def test_ptrack_pg_resetxlog(self): node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, + ptrack_enable=True, initdb_params=['--data-checksums'], pg_options={ - 'ptrack_enable': 'on'}) + 'shared_buffers': '512MB', + 'max_wal_size': '3GB'}) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() - self.create_tblspace_in_node(node, 'somedata') + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") - # Create table and indexes - res = node.safe_psql( + # Create table + node.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap tablespace somedata " - "as select i as id, md5(i::text) as text, " + "create table t_heap " + "as select nextval('t_seq')::int as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " +# "from generate_series(0,25600) i") "from generate_series(0,2560) i") + + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + # Create indexes for i in idx_ptrack: if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': node.safe_psql( - "postgres", "create index {0} on {1} using {2}({3}) " - "tablespace somedata".format( + "postgres", + "CREATE INDEX {0} ON {1} USING {2}({3})".format( i, idx_ptrack[i]['relation'], idx_ptrack[i]['type'], idx_ptrack[i]['column'])) - node.safe_psql('postgres', 'vacuum t_heap') - node.safe_psql('postgres', 'checkpoint') + node.safe_psql( + 'postgres', + "update t_heap set id = nextval('t_seq'), text = md5(text), " + "tsvector = md5(repeat(tsvector::text, 10))::tsvector") - self.backup_node( - backup_dir, 'node', node, options=['-j10', '--stream']) +# node.safe_psql( +# 'postgres', +# "create extension pg_buffercache") +# +# print(node.safe_psql( +# 'postgres', +# "SELECT count(*) FROM pg_buffercache WHERE isdirty")) - for i in idx_ptrack: - # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(node, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate md5sums of pages - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + # kill the bastard + if self.verbose: + print('Killing postmaster. Losing Ptrack changes') + node.stop(['-m', 'immediate', '-D', node.data_dir]) - node.safe_psql('postgres', 'delete from t_heap where id > 128;') - node.safe_psql('postgres', 'vacuum t_heap') - node.safe_psql('postgres', 'checkpoint') + # now smack it with sledgehammer + if node.major_version >= 10: + pg_resetxlog_path = self.get_bin_path('pg_resetwal') + wal_dir = 'pg_wal' + else: + pg_resetxlog_path = self.get_bin_path('pg_resetxlog') + wal_dir = 'pg_xlog' + + self.run_binary( + [ + pg_resetxlog_path, + '-D', + node.data_dir, + '-o 42', + '-f' + ], + asynchronous=False) - # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(node, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(node, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + if not node.status(): + node.slow_start() + else: + print("Die! Die! Why won't you die?... Why won't you die?") + exit(1) - # compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + # take ptrack backup +# self.backup_node( +# backup_dir, 'node', node, +# backup_type='ptrack', options=['--stream']) - self.assertTrue( - success, 'Ptrack has failed to register changes in data files' - ) + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['--stream']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because instance was brutalized by pg_resetxlog" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd) + ) + except ProbackupException as e: + self.assertTrue( + 'ERROR: LSN from ptrack_control ' in e.message and + 'is greater than Start LSN of previous backup' in e.message, + '\n Unexpected Error Message: {0}\n' + ' CMD: {1}'.format(repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) +# pgdata = self.pgdata_content(node.data_dir) +# +# node_restored = self.make_simple_node( +# base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) +# node_restored.cleanup() +# +# self.restore_node( +# backup_dir, 'node', node_restored) +# +# pgdata_restored = self.pgdata_content(node_restored.data_dir) +# self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") # @unittest.expectedFailure - def test_ptrack_vacuum_truncate_replica(self): - fname = self.id().split('.')[3] - master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), + def test_corrupt_ptrack_map(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on', - 'checkpoint_timeout': '30'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) - self.add_instance(backup_dir, 'master', master) - master.slow_start() - - self.backup_node(backup_dir, 'master', master, options=['--stream']) - - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() + self.add_instance(backup_dir, 'node', node) + node.slow_start() - self.restore_node(backup_dir, 'master', replica) + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") - self.add_instance(backup_dir, 'replica', replica) - self.set_replica(master, replica, 'replica', synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) - replica.slow_start(replica=True) + ptrack_version = self.get_ptrack_version(node) - # Create table and indexes - master.safe_psql( + # Create table + node.safe_psql( "postgres", "create extension bloom; create sequence t_seq; " - "create table t_heap as select i as id, " - "md5(i::text) as text, md5(repeat(i::text,10))::tsvector " - "as tsvector from generate_series(0,2560) i") + "create table t_heap " + "as select nextval('t_seq')::int as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,2560) i") - for i in idx_ptrack: - if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': - master.safe_psql( - "postgres", "create index {0} on {1} " - "using {2}({3})".format( - i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + self.backup_node( + backup_dir, 'node', node, options=['--stream']) - master.safe_psql('postgres', 'vacuum t_heap') - master.safe_psql('postgres', 'checkpoint') + node.safe_psql( + 'postgres', + "update t_heap set id = nextval('t_seq'), text = md5(text), " + "tsvector = md5(repeat(tsvector::text, 10))::tsvector") + + # kill the bastard + if self.verbose: + print('Killing postmaster. Losing Ptrack changes') + + node.stop(['-m', 'immediate', '-D', node.data_dir]) + + ptrack_map = os.path.join(node.data_dir, 'global', 'ptrack.map') + + # Let`s do index corruption. ptrack.map + with open(ptrack_map, "rb+", 0) as f: + f.seek(42) + f.write(b"blablahblahs") + f.flush() + f.close + +# os.remove(os.path.join(node.logs_dir, node.pg_log_name)) + + if self.verbose: + print('Ptrack version:', ptrack_version) + if ptrack_version >= self.version_to_num("2.3"): + node.slow_start() + + log_file = os.path.join(node.logs_dir, 'postgresql.log') + with open(log_file, 'r') as f: + log_content = f.read() + + self.assertIn( + 'WARNING: ptrack read map: incorrect checksum of file "{0}"'.format(ptrack_map), + log_content) + + node.stop(['-D', node.data_dir]) + else: + try: + node.slow_start() + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because ptrack.map is corrupted" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except StartNodeException as e: + self.assertIn( + 'Cannot start node', + e.message, + '\n Unexpected Error Message: {0}\n' + ' CMD: {1}'.format(repr(e.message), self.cmd)) + + log_file = os.path.join(node.logs_dir, 'postgresql.log') + with open(log_file, 'r') as f: + log_content = f.read() + + self.assertIn( + 'FATAL: ptrack init: incorrect checksum of file "{0}"'.format(ptrack_map), + log_content) + + self.set_auto_conf(node, {'ptrack.map_size': '0'}) + node.slow_start() + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['--stream']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because instance ptrack is disabled" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Ptrack is disabled', + e.message, + '\n Unexpected Error Message: {0}\n' + ' CMD: {1}'.format(repr(e.message), self.cmd)) + + node.safe_psql( + 'postgres', + "update t_heap set id = nextval('t_seq'), text = md5(text), " + "tsvector = md5(repeat(tsvector::text, 10))::tsvector") + + node.stop(['-m', 'immediate', '-D', node.data_dir]) + + self.set_auto_conf(node, {'ptrack.map_size': '32', 'shared_preload_libraries': 'ptrack'}) + node.slow_start() + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['--stream']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because ptrack map is from future" + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: LSN from ptrack_control', + e.message, + '\n Unexpected Error Message: {0}\n' + ' CMD: {1}'.format(repr(e.message), self.cmd)) - # Take PTRACK backup to clean every ptrack self.backup_node( - backup_dir, 'replica', replica, - options=[ - '-j10', - '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port), - '--stream' - ] - ) + backup_dir, 'node', node, + backup_type='delta', options=['--stream']) - for i in idx_ptrack: - # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['old_size'] = self.get_fork_size(replica, i) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate md5sums of pages - idx_ptrack[i]['old_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['old_size']) + node.safe_psql( + 'postgres', + "update t_heap set id = nextval('t_seq'), text = md5(text), " + "tsvector = md5(repeat(tsvector::text, 10))::tsvector") - master.safe_psql('postgres', 'delete from t_heap where id > 128;') - master.safe_psql('postgres', 'vacuum t_heap') - master.safe_psql('postgres', 'checkpoint') + self.backup_node( + backup_dir, 'node', node, + backup_type='ptrack', options=['--stream']) - # CHECK PTRACK SANITY - success = True - for i in idx_ptrack: - # get new size of heap and indexes. size calculated in pages - idx_ptrack[i]['new_size'] = self.get_fork_size(replica, i) - # update path to heap and index files in case they`ve changed - idx_ptrack[i]['path'] = self.get_fork_path(replica, i) - # calculate new md5sums for pages - idx_ptrack[i]['new_pages'] = self.get_md5_per_page_for_fork( - idx_ptrack[i]['path'], idx_ptrack[i]['new_size']) - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - replica, idx_ptrack[i]['path'], - [idx_ptrack[i]['old_size'], idx_ptrack[i]['new_size']]) + pgdata = self.pgdata_content(node.data_dir) - # compare pages and check ptrack sanity - if not self.check_ptrack_sanity(idx_ptrack[i]): - success = False + node.cleanup() - self.assertTrue( - success, 'Ptrack has failed to register changes in data files') + self.restore_node(backup_dir, 'node', node) - # Clean after yourself - self.del_test_dir(module_name, fname) + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") - # @unittest.expectedFailure - def test_ptrack_recovery(self): - fname = self.id().split('.')[3] + def test_horizon_lsn_ptrack(self): + """ + https://github.com/postgrespro/pg_probackup/pull/386 + """ + if not self.probackup_old_path: + self.skipTest("You must specify PGPROBACKUPBIN_OLD" + " for run this test") + self.assertLessEqual( + self.version_to_num(self.old_probackup_version), + self.version_to_num('2.4.15'), + 'You need pg_probackup old_binary =< 2.4.15 for this test') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) + ptrack_enable=True, + initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() - self.create_tblspace_in_node(node, 'somedata') - - # Create table node.safe_psql( "postgres", - "create extension bloom; create sequence t_seq; " - "create table t_heap tablespace somedata " - "as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,2560) i") + "CREATE EXTENSION ptrack") - # Create indexes - for i in idx_ptrack: - if idx_ptrack[i]['type'] != 'heap' and idx_ptrack[i]['type'] != 'seq': - node.safe_psql( - "postgres", "create index {0} on {1} using {2}({3}) " - "tablespace somedata".format( - i, idx_ptrack[i]['relation'], - idx_ptrack[i]['type'], idx_ptrack[i]['column'])) + self.assertGreaterEqual( + self.get_ptrack_version(node), + self.version_to_num("2.1"), + "You need ptrack >=2.1 for this test") - # get size of heap and indexes. size calculated in pages - idx_ptrack[i]['size'] = int(self.get_fork_size(node, i)) - # get path to heap and index files - idx_ptrack[i]['path'] = self.get_fork_path(node, i) + # set map_size to a minimal value + self.set_auto_conf(node, {'ptrack.map_size': '1'}) + node.restart() - if self.verbose: - print('Killing postmaster. Losing Ptrack changes') - node.stop(['-m', 'immediate', '-D', node.data_dir]) - if not node.status(): - node.slow_start() - else: - print("Die! Die! Why won't you die?... Why won't you die?") - exit(1) + node.pgbench_init(scale=100) - for i in idx_ptrack: - # get ptrack for every idx - idx_ptrack[i]['ptrack'] = self.get_ptrack_bits_per_page_for_fork( - node, idx_ptrack[i]['path'], [idx_ptrack[i]['size']]) - # check that ptrack has correct bits after recovery - self.check_ptrack_recovery(idx_ptrack[i]) + # FULL backup + full_id = self.backup_node(backup_dir, 'node', node, options=['--stream'], old_binary=True) + + # enable archiving so the WAL size to do interfere with data bytes comparison later + self.set_archiving(backup_dir, 'node', node) + node.restart() + + # change data + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + # DELTA is exemplar + delta_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta') + delta_bytes = self.show_pb(backup_dir, 'node', backup_id=delta_id)["data-bytes"] + self.delete_pb(backup_dir, 'node', backup_id=delta_id) - # Clean after yourself - self.del_test_dir(module_name, fname) + # PTRACK with current binary + ptrack_id = self.backup_node(backup_dir, 'node', node, backup_type='ptrack') + ptrack_bytes = self.show_pb(backup_dir, 'node', backup_id=ptrack_id)["data-bytes"] + # make sure that backup size is exactly the same + self.assertEqual(delta_bytes, ptrack_bytes) diff --git a/tests/remote.py b/tests/remote.py deleted file mode 100644 index 064ee219f..000000000 --- a/tests/remote.py +++ /dev/null @@ -1,44 +0,0 @@ -import unittest -import os -from time import sleep -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException -from .helpers.cfs_helpers import find_by_name - - -module_name = 'remote' - - -class RemoteTest(ProbackupTest, unittest.TestCase): - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_remote_sanity(self): - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - try: - self.backup_node( - backup_dir, 'node', - node, options=['--remote-proto=ssh', '--stream']) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because remote-host option is missing." - "\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - "Insert correct error", - e.message, - "\n Unexpected Error Message: {0}\n CMD: {1}".format( - repr(e.message), self.cmd)) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/remote_test.py b/tests/remote_test.py new file mode 100644 index 000000000..2d36d7346 --- /dev/null +++ b/tests/remote_test.py @@ -0,0 +1,43 @@ +import unittest +import os +from time import sleep +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +from .helpers.cfs_helpers import find_by_name + + +class RemoteTest(ProbackupTest, unittest.TestCase): + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_remote_sanity(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + output = self.backup_node( + backup_dir, 'node', node, + options=['--stream'], no_remote=True, return_id=False) + self.assertIn('remote: false', output) + + # try: + # self.backup_node( + # backup_dir, 'node', + # node, options=['--remote-proto=ssh', '--stream'], no_remote=True) + # # we should die here because exception is what we expect to happen + # self.assertEqual( + # 1, 0, + # "Expecting Error because remote-host option is missing." + # "\n Output: {0} \n CMD: {1}".format( + # repr(self.output), self.cmd)) + # except ProbackupException as e: + # self.assertIn( + # "Insert correct error", + # e.message, + # "\n Unexpected Error Message: {0}\n CMD: {1}".format( + # repr(e.message), self.cmd)) diff --git a/tests/replica.py b/tests/replica.py deleted file mode 100644 index 1d38d2659..000000000 --- a/tests/replica.py +++ /dev/null @@ -1,530 +0,0 @@ -import os -import unittest -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException, idx_ptrack -from datetime import datetime, timedelta -import subprocess -import time - - -module_name = 'replica' - - -class ReplicaTest(ProbackupTest, unittest.TestCase): - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_replica_stream_ptrack_backup(self): - """ - make node, take full backup, restore it and make replica from it, - take full stream backup from replica - """ - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) - - master.slow_start() - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'master', master) - - # CREATE TABLE - master.psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,256) i") - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - - # take full backup and restore it - self.backup_node(backup_dir, 'master', master, options=['--stream']) - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() - self.restore_node(backup_dir, 'master', replica) - self.set_replica(master, replica) - - # Check data correctness on replica - replica.slow_start(replica=True) - after = replica.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(before, after) - - # Change data on master, take FULL backup from replica, - # restore taken backup and check that restored data equal - # to original data - master.psql( - "postgres", - "insert into t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(256,512) i") - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - self.add_instance(backup_dir, 'replica', replica) - - backup_id = self.backup_node( - backup_dir, 'replica', replica, - options=[ - '--stream', - '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port)]) - self.validate_pb(backup_dir, 'replica') - self.assertEqual( - 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) - - # RESTORE FULL BACKUP TAKEN FROM PREVIOUS STEP - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node')) - node.cleanup() - self.restore_node(backup_dir, 'replica', data_dir=node.data_dir) - - node.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node.port)) - node.slow_start() - - # CHECK DATA CORRECTNESS - after = node.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(before, after) - - # Change data on master, take PTRACK backup from replica, - # restore taken backup and check that restored data equal - # to original data - master.psql( - "postgres", - "insert into t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(512,768) i") - - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - - backup_id = self.backup_node( - backup_dir, 'replica', replica, backup_type='ptrack', - options=[ - '--stream', - '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port)]) - self.validate_pb(backup_dir, 'replica') - self.assertEqual( - 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) - - # RESTORE PTRACK BACKUP TAKEN FROM replica - node.cleanup() - self.restore_node( - backup_dir, 'replica', data_dir=node.data_dir, backup_id=backup_id) - - node.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node.port)) - node.slow_start() - - # CHECK DATA CORRECTNESS - after = node.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(before, after) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_replica_archive_page_backup(self): - """ - make archive master, take full and page archive backups from master, - set replica, make archive backup from replica - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'archive_timeout': '10s', - 'checkpoint_timeout': '30s', - 'max_wal_size': '32MB'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'master', master) - self.set_archiving(backup_dir, 'master', master) - master.slow_start() - - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() - - self.backup_node(backup_dir, 'master', master) - - master.psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,2560) i") - - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - - backup_id = self.backup_node( - backup_dir, 'master', master, backup_type='page') - self.restore_node(backup_dir, 'master', replica) - - # Settings for Replica - self.add_instance(backup_dir, 'replica', replica) - self.set_replica(master, replica, synchronous=True) - self.set_archiving(backup_dir, 'replica', replica, replica=True) - - replica.slow_start(replica=True) - - # Check data correctness on replica - after = replica.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(before, after) - - # Change data on master, take FULL backup from replica, - # restore taken backup and check that restored data - # equal to original data - master.psql( - "postgres", - "insert into t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(256,25120) i") - - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - - master.psql( - "postgres", - "CHECKPOINT") - - self.wait_until_replica_catch_with_master(master, replica) - - backup_id = self.backup_node( - backup_dir, 'replica', replica, - options=[ - '--archive-timeout=60', - '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port)]) - - self.validate_pb(backup_dir, 'replica') - self.assertEqual( - 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) - - # RESTORE FULL BACKUP TAKEN FROM replica - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node')) - node.cleanup() - self.restore_node(backup_dir, 'replica', data_dir=node.data_dir) - - node.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node.port)) - - node.append_conf( - 'postgresql.auto.conf', 'archive_mode = off'.format(node.port)) - - node.slow_start() - - # CHECK DATA CORRECTNESS - after = node.safe_psql("postgres", "SELECT * FROM t_heap") - self.assertEqual(before, after) - node.cleanup() - - # Change data on master, make PAGE backup from replica, - # restore taken backup and check that restored data equal - # to original data - master.pgbench_init(scale=5) - - pgbench = master.pgbench( - options=['-T', '30', '-c', '2', '--no-vacuum']) - - backup_id = self.backup_node( - backup_dir, 'replica', - replica, backup_type='page', - options=[ - '--archive-timeout=60', - '--master-host=localhost', - '--master-db=postgres', - '--master-port={0}'.format(master.port)]) - - pgbench.wait() - - self.switch_wal_segment(master) - - before = master.safe_psql("postgres", "SELECT * FROM pgbench_accounts") - - self.validate_pb(backup_dir, 'replica') - self.assertEqual( - 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) - - # RESTORE PAGE BACKUP TAKEN FROM replica - self.restore_node( - backup_dir, 'replica', data_dir=node.data_dir, - backup_id=backup_id) - - node.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node.port)) - - node.append_conf( - 'postgresql.auto.conf', 'archive_mode = off') - - node.slow_start() - - # CHECK DATA CORRECTNESS - after = node.safe_psql("postgres", "SELECT * FROM pgbench_accounts") - self.assertEqual( - before, after, 'Restored data is not equal to original') - - self.add_instance(backup_dir, 'node', node) - self.backup_node( - backup_dir, 'node', node, options=['--stream']) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_basic_make_replica_via_restore(self): - """ - make archive master, take full and page archive backups from master, - set replica, make archive backup from replica - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'archive_timeout': '10s'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'master', master) - self.set_archiving(backup_dir, 'master', master) - # force more frequent wal switch - master.append_conf('postgresql.auto.conf', 'archive_timeout = 10') - master.slow_start() - - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() - - self.backup_node(backup_dir, 'master', master) - - master.psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,8192) i") - - before = master.safe_psql("postgres", "SELECT * FROM t_heap") - - backup_id = self.backup_node( - backup_dir, 'master', master, backup_type='page') - self.restore_node( - backup_dir, 'master', replica, options=['-R']) - - # Settings for Replica - self.add_instance(backup_dir, 'replica', replica) - self.set_archiving(backup_dir, 'replica', replica, replica=True) - replica.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(replica.port)) - replica.append_conf( - 'postgresql.auto.conf', 'hot_standby = on') - - replica.slow_start(replica=True) - - self.backup_node( - backup_dir, 'replica', replica, - options=['--archive-timeout=30s', '--stream']) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_take_backup_from_delayed_replica(self): - """ - make archive master, take full backups from master, - restore full backup as delayed replica, launch pgbench, - take FULL, PAGE and DELTA backups from replica - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'archive_timeout': '10s'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'master', master) - self.set_archiving(backup_dir, 'master', master) - master.slow_start() - - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() - - self.backup_node(backup_dir, 'master', master) - - master.psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,165000) i") - - master.psql( - "postgres", - "CHECKPOINT") - - master.psql( - "postgres", - "create table t_heap_1 as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,165000) i") - - self.restore_node( - backup_dir, 'master', replica, options=['-R']) - - # Settings for Replica - self.add_instance(backup_dir, 'replica', replica) - self.set_archiving(backup_dir, 'replica', replica, replica=True) - - replica.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(replica.port)) - - replica.slow_start(replica=True) - - self.wait_until_replica_catch_with_master(master, replica) - - replica.append_conf( - 'recovery.conf', "recovery_min_apply_delay = '300s'") - - replica.stop() - replica.slow_start(replica=True) - - master.pgbench_init(scale=10) - - pgbench = master.pgbench( - options=['-T', '60', '-c', '2', '--no-vacuum']) - - self.backup_node( - backup_dir, 'replica', - replica, options=['--archive-timeout=60s']) - - self.backup_node( - backup_dir, 'replica', replica, - data_dir=replica.data_dir, - backup_type='page', options=['--archive-timeout=60s']) - - self.backup_node( - backup_dir, 'replica', replica, - backup_type='delta', options=['--archive-timeout=60s']) - - pgbench.wait() - - pgbench = master.pgbench( - options=['-T', '30', '-c', '2', '--no-vacuum']) - - self.backup_node( - backup_dir, 'replica', replica, - options=['--stream']) - - self.backup_node( - backup_dir, 'replica', replica, - backup_type='page', options=['--stream']) - - self.backup_node( - backup_dir, 'replica', replica, - backup_type='delta', options=['--stream']) - - pgbench.wait() - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_replica_promote(self): - """ - start backup from replica, during backup promote replica - check that backup is failed - """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - master = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'master'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'archive_timeout': '10s', - 'checkpoint_timeout': '30s', - 'max_wal_size': '32MB'}) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'master', master) - self.set_archiving(backup_dir, 'master', master) - master.slow_start() - - replica = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'replica')) - replica.cleanup() - - self.backup_node(backup_dir, 'master', master) - - master.psql( - "postgres", - "create table t_heap as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,165000) i") - - self.restore_node( - backup_dir, 'master', replica, options=['-R']) - - # Settings for Replica - self.add_instance(backup_dir, 'replica', replica) - self.set_archiving(backup_dir, 'replica', replica, replica=True) - self.set_replica( - master, replica, - replica_name='replica', synchronous=True) - - replica.slow_start(replica=True) - - master.psql( - "postgres", - "create table t_heap_1 as select i as id, md5(i::text) as text, " - "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(0,165000) i") - - self.wait_until_replica_catch_with_master(master, replica) - - # start backup from replica - gdb = self.backup_node( - backup_dir, 'replica', replica, gdb=True, - options=['--log-level-file=verbose']) - - gdb.set_breakpoint('backup_data_file') - gdb.run_until_break() - gdb.continue_execution_until_break(20) - - replica.promote() - - gdb.remove_all_breakpoints() - gdb.continue_execution_until_exit() - - backup_id = self.show_pb( - backup_dir, 'replica')[0]["id"] - - # read log file content - with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: - log_content = f.read() - f.close - - self.assertIn( - 'ERROR: the standby was promoted during online backup', - log_content) - - self.assertIn( - 'WARNING: Backup {0} is running, ' - 'setting its status to ERROR'.format(backup_id), - log_content) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/replica_test.py b/tests/replica_test.py new file mode 100644 index 000000000..17fc5a823 --- /dev/null +++ b/tests/replica_test.py @@ -0,0 +1,1654 @@ +import os +import unittest +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException, idx_ptrack +from datetime import datetime, timedelta +import subprocess +import time +from distutils.dir_util import copy_tree +from testgres import ProcessType +from time import sleep + + +class ReplicaTest(ProbackupTest, unittest.TestCase): + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_replica_switchover(self): + """ + check that archiving on replica works correctly + over the course of several switchovers + https://www.postgresql.org/message-id/54b059d4-2b48-13a4-6f43-95a087c92367%40postgrespro.ru + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1'), + set_replication=True, + initdb_params=['--data-checksums']) + + if self.get_version(node1) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node1', node1) + + node1.slow_start() + + # take full backup and restore it + self.backup_node(backup_dir, 'node1', node1, options=['--stream']) + node2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node2')) + node2.cleanup() + + # create replica + self.restore_node(backup_dir, 'node1', node2) + + # setup replica + self.add_instance(backup_dir, 'node2', node2) + self.set_archiving(backup_dir, 'node2', node2, replica=True) + self.set_replica(node1, node2, synchronous=False) + self.set_auto_conf(node2, {'port': node2.port}) + + node2.slow_start(replica=True) + + # generate some data + node1.pgbench_init(scale=5) + + # take full backup on replica + self.backup_node(backup_dir, 'node2', node2, options=['--stream']) + + # first switchover + node1.stop() + node2.promote() + + self.set_replica(node2, node1, synchronous=False) + node2.reload() + node1.slow_start(replica=True) + + # take incremental backup from new master + self.backup_node( + backup_dir, 'node2', node2, + backup_type='delta', options=['--stream']) + + # second switchover + node2.stop() + node1.promote() + self.set_replica(node1, node2, synchronous=False) + node1.reload() + node2.slow_start(replica=True) + + # generate some more data + node1.pgbench_init(scale=5) + + # take incremental backup from replica + self.backup_node( + backup_dir, 'node2', node2, + backup_type='delta', options=['--stream']) + + # https://github.com/postgrespro/pg_probackup/issues/251 + self.validate_pb(backup_dir) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_replica_stream_ptrack_backup(self): + """ + make node, take full backup, restore it and make replica from it, + take full stream backup from replica + """ + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + if self.pg_config_version > self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + + master.slow_start() + + if master.major_version >= 12: + master.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + # CREATE TABLE + master.psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,256) i") + before = master.table_checksum("t_heap") + + # take full backup and restore it + self.backup_node(backup_dir, 'master', master, options=['--stream']) + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'master', replica) + self.set_replica(master, replica) + + # Check data correctness on replica + replica.slow_start(replica=True) + after = replica.table_checksum("t_heap") + self.assertEqual(before, after) + + # Change data on master, take FULL backup from replica, + # restore taken backup and check that restored data equal + # to original data + master.psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(256,512) i") + before = master.table_checksum("t_heap") + self.add_instance(backup_dir, 'replica', replica) + + backup_id = self.backup_node( + backup_dir, 'replica', replica, + options=[ + '--stream', + '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port)]) + self.validate_pb(backup_dir, 'replica') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) + + # RESTORE FULL BACKUP TAKEN FROM PREVIOUS STEP + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node')) + node.cleanup() + self.restore_node(backup_dir, 'replica', data_dir=node.data_dir) + + self.set_auto_conf(node, {'port': node.port}) + + node.slow_start() + + # CHECK DATA CORRECTNESS + after = node.table_checksum("t_heap") + self.assertEqual(before, after) + + # Change data on master, take PTRACK backup from replica, + # restore taken backup and check that restored data equal + # to original data + master.psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(512,768) i") + + before = master.table_checksum("t_heap") + + backup_id = self.backup_node( + backup_dir, 'replica', replica, backup_type='ptrack', + options=[ + '--stream', + '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port)]) + self.validate_pb(backup_dir, 'replica') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) + + # RESTORE PTRACK BACKUP TAKEN FROM replica + node.cleanup() + self.restore_node( + backup_dir, 'replica', data_dir=node.data_dir, backup_id=backup_id) + + self.set_auto_conf(node, {'port': node.port}) + + node.slow_start() + + # CHECK DATA CORRECTNESS + after = node.table_checksum("t_heap") + self.assertEqual(before, after) + + # @unittest.skip("skip") + def test_replica_archive_page_backup(self): + """ + make archive master, take full and page archive backups from master, + set replica, make archive backup from replica + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_timeout': '10s', + 'checkpoint_timeout': '30s', + 'max_wal_size': '32MB'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + self.set_archiving(backup_dir, 'master', master) + master.slow_start() + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.backup_node(backup_dir, 'master', master) + + master.psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,2560) i") + + before = master.table_checksum("t_heap") + + backup_id = self.backup_node( + backup_dir, 'master', master, backup_type='page') + self.restore_node(backup_dir, 'master', replica) + + # Settings for Replica + self.add_instance(backup_dir, 'replica', replica) + self.set_replica(master, replica, synchronous=True) + self.set_archiving(backup_dir, 'replica', replica, replica=True) + + replica.slow_start(replica=True) + + # Check data correctness on replica + after = replica.table_checksum("t_heap") + self.assertEqual(before, after) + + # Change data on master, take FULL backup from replica, + # restore taken backup and check that restored data + # equal to original data + master.psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(256,25120) i") + + before = master.table_checksum("t_heap") + + self.wait_until_replica_catch_with_master(master, replica) + + backup_id = self.backup_node( + backup_dir, 'replica', replica, + options=[ + '--archive-timeout=60', + '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port)]) + + self.validate_pb(backup_dir, 'replica') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) + + # RESTORE FULL BACKUP TAKEN FROM replica + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node')) + node.cleanup() + self.restore_node(backup_dir, 'replica', data_dir=node.data_dir) + + self.set_auto_conf(node, {'port': node.port, 'archive_mode': 'off'}) + + node.slow_start() + + # CHECK DATA CORRECTNESS + after = node.table_checksum("t_heap") + self.assertEqual(before, after) + node.cleanup() + + # Change data on master, make PAGE backup from replica, + # restore taken backup and check that restored data equal + # to original data + master.pgbench_init(scale=5) + + pgbench = master.pgbench( + options=['-T', '30', '-c', '2', '--no-vacuum']) + + backup_id = self.backup_node( + backup_dir, 'replica', + replica, backup_type='page', + options=[ + '--archive-timeout=60', + '--master-host=localhost', + '--master-db=postgres', + '--master-port={0}'.format(master.port)]) + + pgbench.wait() + + self.switch_wal_segment(master) + + before = master.table_checksum("pgbench_accounts") + + self.validate_pb(backup_dir, 'replica') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'replica', backup_id)['status']) + + # RESTORE PAGE BACKUP TAKEN FROM replica + self.restore_node( + backup_dir, 'replica', data_dir=node.data_dir, + backup_id=backup_id) + + self.set_auto_conf(node, {'port': node.port, 'archive_mode': 'off'}) + + node.slow_start() + + # CHECK DATA CORRECTNESS + after = master.table_checksum("pgbench_accounts") + self.assertEqual( + before, after, 'Restored data is not equal to original') + + self.add_instance(backup_dir, 'node', node) + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + # @unittest.skip("skip") + def test_basic_make_replica_via_restore(self): + """ + make archive master, take full and page archive backups from master, + set replica, make archive backup from replica + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_timeout': '10s'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + self.set_archiving(backup_dir, 'master', master) + master.slow_start() + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.backup_node(backup_dir, 'master', master) + + master.psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,8192) i") + + before = master.table_checksum("t_heap") + + backup_id = self.backup_node( + backup_dir, 'master', master, backup_type='page') + self.restore_node( + backup_dir, 'master', replica, options=['-R']) + + # Settings for Replica + self.add_instance(backup_dir, 'replica', replica) + self.set_archiving(backup_dir, 'replica', replica, replica=True) + self.set_replica(master, replica, synchronous=True) + + replica.slow_start(replica=True) + + self.backup_node( + backup_dir, 'replica', replica, + options=['--archive-timeout=30s', '--stream']) + + # @unittest.skip("skip") + def test_take_backup_from_delayed_replica(self): + """ + make archive master, take full backups from master, + restore full backup as delayed replica, launch pgbench, + take FULL, PAGE and DELTA backups from replica + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={'archive_timeout': '10s'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + self.set_archiving(backup_dir, 'master', master) + master.slow_start() + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.backup_node(backup_dir, 'master', master) + + master.psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,165000) i") + + master.psql( + "postgres", + "create table t_heap_1 as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,165000) i") + + self.restore_node( + backup_dir, 'master', replica, options=['-R']) + + # Settings for Replica + self.add_instance(backup_dir, 'replica', replica) + self.set_archiving(backup_dir, 'replica', replica, replica=True) + + self.set_auto_conf(replica, {'port': replica.port}) + + replica.slow_start(replica=True) + + self.wait_until_replica_catch_with_master(master, replica) + + if self.get_version(master) >= self.version_to_num('12.0'): + self.set_auto_conf( + replica, {'recovery_min_apply_delay': '300s'}) + else: + replica.append_conf( + 'recovery.conf', + 'recovery_min_apply_delay = 300s') + + replica.stop() + replica.slow_start(replica=True) + + master.pgbench_init(scale=10) + + pgbench = master.pgbench( + options=['-T', '60', '-c', '2', '--no-vacuum']) + + self.backup_node( + backup_dir, 'replica', + replica, options=['--archive-timeout=60s']) + + self.backup_node( + backup_dir, 'replica', replica, + data_dir=replica.data_dir, + backup_type='page', options=['--archive-timeout=60s']) + + sleep(1) + + self.backup_node( + backup_dir, 'replica', replica, + backup_type='delta', options=['--archive-timeout=60s']) + + pgbench.wait() + + pgbench = master.pgbench( + options=['-T', '30', '-c', '2', '--no-vacuum']) + + self.backup_node( + backup_dir, 'replica', replica, + options=['--stream']) + + self.backup_node( + backup_dir, 'replica', replica, + backup_type='page', options=['--stream']) + + self.backup_node( + backup_dir, 'replica', replica, + backup_type='delta', options=['--stream']) + + pgbench.wait() + + # @unittest.skip("skip") + def test_replica_promote(self): + """ + start backup from replica, during backup promote replica + check that backup is failed + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_timeout': '10s', + 'checkpoint_timeout': '30s', + 'max_wal_size': '32MB'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + self.set_archiving(backup_dir, 'master', master) + master.slow_start() + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + self.backup_node(backup_dir, 'master', master) + + master.psql( + "postgres", + "create table t_heap as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,165000) i") + + self.restore_node( + backup_dir, 'master', replica, options=['-R']) + + # Settings for Replica + self.add_instance(backup_dir, 'replica', replica) + self.set_archiving(backup_dir, 'replica', replica, replica=True) + self.set_replica( + master, replica, replica_name='replica', synchronous=True) + + replica.slow_start(replica=True) + + master.psql( + "postgres", + "create table t_heap_1 as select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,165000) i") + + self.wait_until_replica_catch_with_master(master, replica) + + # start backup from replica + gdb = self.backup_node( + backup_dir, 'replica', replica, gdb=True, + options=['--log-level-file=verbose']) + + gdb.set_breakpoint('backup_data_file') + gdb.run_until_break() + gdb.continue_execution_until_break(20) + + replica.promote() + + gdb.remove_all_breakpoints() + gdb.continue_execution_until_exit() + + backup_id = self.show_pb( + backup_dir, 'replica')[0]["id"] + + # read log file content + with open(os.path.join(backup_dir, 'log', 'pg_probackup.log')) as f: + log_content = f.read() + f.close + + self.assertIn( + 'ERROR: the standby was promoted during online backup', + log_content) + + self.assertIn( + 'WARNING: Backup {0} is running, ' + 'setting its status to ERROR'.format(backup_id), + log_content) + + # @unittest.skip("skip") + def test_replica_stop_lsn_null_offset(self): + """ + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '1h', + 'wal_level': 'replica'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', master) + self.set_archiving(backup_dir, 'node', master) + master.slow_start() + + # freeze bgwriter to get rid of RUNNING XACTS records + bgwriter_pid = master.auxiliary_pids[ProcessType.BackgroundWriter][0] + gdb_checkpointer = self.gdb_attach(bgwriter_pid) + + self.backup_node(backup_dir, 'node', master) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'node', replica) + + # Settings for Replica + self.set_replica(master, replica, synchronous=True) + self.set_archiving(backup_dir, 'node', replica, replica=True) + + replica.slow_start(replica=True) + + self.switch_wal_segment(master) + self.switch_wal_segment(master) + + output = self.backup_node( + backup_dir, 'node', replica, replica.data_dir, + options=[ + '--archive-timeout=30', + '--log-level-console=LOG', + '--no-validate', + '--stream'], + return_id=False) + + self.assertIn( + 'LOG: Invalid offset in stop_lsn value 0/4000000', + output) + + self.assertIn( + 'WARNING: WAL segment 000000010000000000000004 could not be streamed in 30 seconds', + output) + + self.assertIn( + 'WARNING: Failed to get next WAL record after 0/4000000, looking for previous WAL record', + output) + + self.assertIn( + 'LOG: Looking for LSN 0/4000000 in segment: 000000010000000000000003', + output) + + self.assertIn( + 'has endpoint 0/4000000 which is ' + 'equal or greater than requested LSN 0/4000000', + output) + + self.assertIn( + 'LOG: Found prior LSN:', + output) + + # Clean after yourself + gdb_checkpointer.kill() + + # @unittest.skip("skip") + def test_replica_stop_lsn_null_offset_next_record(self): + """ + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '1h', + 'wal_level': 'replica'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + self.set_archiving(backup_dir, 'master', master) + master.slow_start() + + # freeze bgwriter to get rid of RUNNING XACTS records + bgwriter_pid = master.auxiliary_pids[ProcessType.BackgroundWriter][0] + + self.backup_node(backup_dir, 'master', master) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'master', replica) + + # Settings for Replica + self.add_instance(backup_dir, 'replica', replica) + self.set_replica(master, replica, synchronous=True) + self.set_archiving(backup_dir, 'replica', replica, replica=True) + + copy_tree( + os.path.join(backup_dir, 'wal', 'master'), + os.path.join(backup_dir, 'wal', 'replica')) + + replica.slow_start(replica=True) + + self.switch_wal_segment(master) + self.switch_wal_segment(master) + + # open connection to master + conn = master.connect() + + gdb = self.backup_node( + backup_dir, 'replica', replica, + options=[ + '--archive-timeout=40', + '--log-level-file=LOG', + '--no-validate', + '--stream'], + gdb=True) + + # Attention! this breakpoint is set to a probackup internal function, not a postgres core one + gdb.set_breakpoint('pg_stop_backup') + gdb.run_until_break() + gdb.remove_all_breakpoints() + gdb.continue_execution_until_running() + + sleep(5) + + conn.execute("create table t1()") + conn.commit() + + while 'RUNNING' in self.show_pb(backup_dir, 'replica')[0]['status']: + sleep(5) + + file = os.path.join(backup_dir, 'log', 'pg_probackup.log') + + with open(file) as f: + log_content = f.read() + + self.assertIn( + 'LOG: Invalid offset in stop_lsn value 0/4000000', + log_content) + + self.assertIn( + 'LOG: Looking for segment: 000000010000000000000004', + log_content) + + self.assertIn( + 'LOG: First record in WAL segment "000000010000000000000004": 0/4000028', + log_content) + + self.assertIn( + 'INFO: stop_lsn: 0/4000000', + log_content) + + self.assertTrue(self.show_pb(backup_dir, 'replica')[0]['status'] == 'DONE') + + # @unittest.skip("skip") + def test_archive_replica_null_offset(self): + """ + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '1h', + 'wal_level': 'replica'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', master) + self.set_archiving(backup_dir, 'node', master) + master.slow_start() + + self.backup_node(backup_dir, 'node', master) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'node', replica) + + # Settings for Replica + self.set_replica(master, replica, synchronous=True) + self.set_archiving(backup_dir, 'node', replica, replica=True) + + # freeze bgwriter to get rid of RUNNING XACTS records + bgwriter_pid = master.auxiliary_pids[ProcessType.BackgroundWriter][0] + gdb_checkpointer = self.gdb_attach(bgwriter_pid) + + replica.slow_start(replica=True) + + self.switch_wal_segment(master) + self.switch_wal_segment(master) + + # take backup from replica + output = self.backup_node( + backup_dir, 'node', replica, replica.data_dir, + options=[ + '--archive-timeout=30', + '--log-level-console=LOG', + '--no-validate'], + return_id=False) + + self.assertIn( + 'LOG: Invalid offset in stop_lsn value 0/4000000', + output) + + self.assertIn( + 'WARNING: WAL segment 000000010000000000000004 could not be archived in 30 seconds', + output) + + self.assertIn( + 'WARNING: Failed to get next WAL record after 0/4000000, looking for previous WAL record', + output) + + self.assertIn( + 'LOG: Looking for LSN 0/4000000 in segment: 000000010000000000000003', + output) + + self.assertIn( + 'has endpoint 0/4000000 which is ' + 'equal or greater than requested LSN 0/4000000', + output) + + self.assertIn( + 'LOG: Found prior LSN:', + output) + + print(output) + + # @unittest.skip("skip") + def test_archive_replica_not_null_offset(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '1h', + 'wal_level': 'replica'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', master) + self.set_archiving(backup_dir, 'node', master) + master.slow_start() + + self.backup_node(backup_dir, 'node', master) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'node', replica) + + # Settings for Replica + self.set_replica(master, replica, synchronous=True) + self.set_archiving(backup_dir, 'node', replica, replica=True) + + replica.slow_start(replica=True) + + # take backup from replica + self.backup_node( + backup_dir, 'node', replica, replica.data_dir, + options=[ + '--archive-timeout=30', + '--log-level-console=LOG', + '--no-validate'], + return_id=False) + + try: + self.backup_node( + backup_dir, 'node', replica, replica.data_dir, + options=[ + '--archive-timeout=30', + '--log-level-console=LOG', + '--no-validate']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of archive timeout. " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + # vanilla -- 0/4000060 + # pgproee -- 0/4000078 + self.assertRegex( + e.message, + r'LOG: Looking for LSN (0/4000060|0/4000078) in segment: 000000010000000000000004', + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertRegex( + e.message, + r'INFO: Wait for LSN (0/4000060|0/4000078) in archived WAL segment', + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.assertIn( + 'ERROR: WAL segment 000000010000000000000004 could not be archived in 30 seconds', + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_replica_toast(self): + """ + make archive master, take full and page archive backups from master, + set replica, make archive backup from replica + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '1h', + 'wal_level': 'replica', + 'shared_buffers': '128MB'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + self.set_archiving(backup_dir, 'master', master) + master.slow_start() + + # freeze bgwriter to get rid of RUNNING XACTS records + bgwriter_pid = master.auxiliary_pids[ProcessType.BackgroundWriter][0] + gdb_checkpointer = self.gdb_attach(bgwriter_pid) + + self.backup_node(backup_dir, 'master', master) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'master', replica) + + # Settings for Replica + self.add_instance(backup_dir, 'replica', replica) + self.set_replica(master, replica, synchronous=True) + self.set_archiving(backup_dir, 'replica', replica, replica=True) + + copy_tree( + os.path.join(backup_dir, 'wal', 'master'), + os.path.join(backup_dir, 'wal', 'replica')) + + replica.slow_start(replica=True) + + self.switch_wal_segment(master) + self.switch_wal_segment(master) + + master.safe_psql( + 'postgres', + 'CREATE TABLE t1 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,10) i') + + self.wait_until_replica_catch_with_master(master, replica) + + output = self.backup_node( + backup_dir, 'replica', replica, + options=[ + '--archive-timeout=30', + '--log-level-console=LOG', + '--no-validate', + '--stream'], + return_id=False) + + pgdata = self.pgdata_content(replica.data_dir) + + self.assertIn( + 'WARNING: Could not read WAL record at', + output) + + self.assertIn( + 'LOG: Found prior LSN:', + output) + + res1 = replica.safe_psql( + 'postgres', + 'select md5(fat_attr) from t1') + + replica.cleanup() + + self.restore_node(backup_dir, 'replica', replica) + pgdata_restored = self.pgdata_content(replica.data_dir) + + replica.slow_start() + + res2 = replica.safe_psql( + 'postgres', + 'select md5(fat_attr) from t1') + + self.assertEqual(res1, res2) + + self.compare_pgdata(pgdata, pgdata_restored) + + # Clean after yourself + gdb_checkpointer.kill() + + # @unittest.skip("skip") + def test_start_stop_lsn_in_the_same_segno(self): + """ + """ + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '1h', + 'wal_level': 'replica', + 'shared_buffers': '128MB'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + master.slow_start() + + # freeze bgwriter to get rid of RUNNING XACTS records + bgwriter_pid = master.auxiliary_pids[ProcessType.BackgroundWriter][0] + + self.backup_node(backup_dir, 'master', master, options=['--stream']) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'master', replica) + + # Settings for Replica + self.add_instance(backup_dir, 'replica', replica) + self.set_replica(master, replica, synchronous=True) + + replica.slow_start(replica=True) + + self.switch_wal_segment(master) + self.switch_wal_segment(master) + + master.safe_psql( + 'postgres', + 'CREATE TABLE t1 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,10) i') + + master.safe_psql( + 'postgres', + 'CHECKPOINT') + + self.wait_until_replica_catch_with_master(master, replica) + + sleep(60) + + self.backup_node( + backup_dir, 'replica', replica, + options=[ + '--archive-timeout=30', + '--log-level-console=LOG', + '--no-validate', + '--stream'], + return_id=False) + + self.backup_node( + backup_dir, 'replica', replica, + options=[ + '--archive-timeout=30', + '--log-level-console=LOG', + '--no-validate', + '--stream'], + return_id=False) + + @unittest.skip("skip") + def test_replica_promote_1(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '1h', + 'wal_level': 'replica'}) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + # set replica True, so archive_mode 'always' is used. + self.set_archiving(backup_dir, 'master', master, replica=True) + master.slow_start() + + self.backup_node(backup_dir, 'master', master) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'master', replica) + + # Settings for Replica + self.set_replica(master, replica) + + replica.slow_start(replica=True) + + master.safe_psql( + 'postgres', + 'CREATE TABLE t1 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,10) i') + + self.wait_until_replica_catch_with_master(master, replica) + + wal_file = os.path.join( + backup_dir, 'wal', 'master', '000000010000000000000004') + + wal_file_partial = os.path.join( + backup_dir, 'wal', 'master', '000000010000000000000004.partial') + + self.assertFalse(os.path.exists(wal_file)) + + replica.promote() + + while not os.path.exists(wal_file_partial): + sleep(1) + + self.switch_wal_segment(master) + + # sleep to be sure, that any partial timeout is expired + sleep(70) + + self.assertTrue( + os.path.exists(wal_file_partial), + "File {0} disappeared".format(wal_file)) + + self.assertTrue( + os.path.exists(wal_file_partial), + "File {0} disappeared".format(wal_file_partial)) + + # @unittest.skip("skip") + def test_replica_promote_2(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + # set replica True, so archive_mode 'always' is used. + self.set_archiving( + backup_dir, 'master', master, replica=True) + master.slow_start() + + self.backup_node(backup_dir, 'master', master) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'master', replica) + + # Settings for Replica + self.set_replica(master, replica) + self.set_auto_conf(replica, {'port': replica.port}) + + replica.slow_start(replica=True) + + master.safe_psql( + 'postgres', + 'CREATE TABLE t1 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,1) i') + + self.wait_until_replica_catch_with_master(master, replica) + + replica.promote() + + self.backup_node( + backup_dir, 'master', replica, data_dir=replica.data_dir, + backup_type='page') + + # @unittest.skip("skip") + def test_replica_promote_archive_delta(self): + """ + t3 /---D3--> + t2 /-------> + t1 --F---D1--D2-- + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s', + 'archive_timeout': '30s'}) + + if self.get_version(node1) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node1) + self.set_config( + backup_dir, 'node', options=['--archive-timeout=60s']) + self.set_archiving(backup_dir, 'node', node1) + + node1.slow_start() + + self.backup_node(backup_dir, 'node', node1, options=['--stream']) + + # Create replica + node2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node2')) + node2.cleanup() + self.restore_node(backup_dir, 'node', node2, node2.data_dir) + + # Settings for Replica + self.set_replica(node1, node2) + self.set_auto_conf(node2, {'port': node2.port}) + self.set_archiving(backup_dir, 'node', node2, replica=True) + + node2.slow_start(replica=True) + + node1.safe_psql( + 'postgres', + 'CREATE TABLE t1 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,20) i') + self.wait_until_replica_catch_with_master(node1, node2) + + node1.safe_psql( + 'postgres', + 'CREATE TABLE t2 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,20) i') + self.wait_until_replica_catch_with_master(node1, node2) + + # delta backup on replica on timeline 1 + delta1_id = self.backup_node( + backup_dir, 'node', node2, node2.data_dir, + 'delta', options=['--stream']) + + # delta backup on replica on timeline 1 + delta2_id = self.backup_node( + backup_dir, 'node', node2, node2.data_dir, 'delta') + + self.change_backup_status( + backup_dir, 'node', delta2_id, 'ERROR') + + # node2 is now master + node2.promote() + + node2.safe_psql( + 'postgres', + 'CREATE TABLE t3 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,20) i') + + # node1 is now replica + node1.cleanup() + # kludge "backup_id=delta1_id" + self.restore_node( + backup_dir, 'node', node1, node1.data_dir, + backup_id=delta1_id, + options=[ + '--recovery-target-timeline=2', + '--recovery-target=latest']) + + # Settings for Replica + self.set_replica(node2, node1) + self.set_auto_conf(node1, {'port': node1.port}) + self.set_archiving(backup_dir, 'node', node1, replica=True) + + node1.slow_start(replica=True) + + node2.safe_psql( + 'postgres', + 'CREATE TABLE t4 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,30) i') + self.wait_until_replica_catch_with_master(node2, node1) + + # node1 is back to be a master + node1.promote() + + sleep(5) + + # delta backup on timeline 3 + self.backup_node( + backup_dir, 'node', node1, node1.data_dir, 'delta', + options=['--archive-timeout=60']) + + pgdata = self.pgdata_content(node1.data_dir) + + node1.cleanup() + self.restore_node(backup_dir, 'node', node1, node1.data_dir) + + pgdata_restored = self.pgdata_content(node1.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_replica_promote_archive_page(self): + """ + t3 /---P3--> + t2 /-------> + t1 --F---P1--P2-- + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node1'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'checkpoint_timeout': '30s', + 'archive_timeout': '30s'}) + + if self.get_version(node1) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node1) + self.set_archiving(backup_dir, 'node', node1) + self.set_config( + backup_dir, 'node', options=['--archive-timeout=60s']) + + node1.slow_start() + + self.backup_node(backup_dir, 'node', node1, options=['--stream']) + + # Create replica + node2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node2')) + node2.cleanup() + self.restore_node(backup_dir, 'node', node2, node2.data_dir) + + # Settings for Replica + self.set_replica(node1, node2) + self.set_auto_conf(node2, {'port': node2.port}) + self.set_archiving(backup_dir, 'node', node2, replica=True) + + node2.slow_start(replica=True) + + node1.safe_psql( + 'postgres', + 'CREATE TABLE t1 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,20) i') + self.wait_until_replica_catch_with_master(node1, node2) + + node1.safe_psql( + 'postgres', + 'CREATE TABLE t2 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,20) i') + self.wait_until_replica_catch_with_master(node1, node2) + + # page backup on replica on timeline 1 + page1_id = self.backup_node( + backup_dir, 'node', node2, node2.data_dir, + 'page', options=['--stream']) + + # page backup on replica on timeline 1 + page2_id = self.backup_node( + backup_dir, 'node', node2, node2.data_dir, 'page') + + self.change_backup_status( + backup_dir, 'node', page2_id, 'ERROR') + + # node2 is now master + node2.promote() + + node2.safe_psql( + 'postgres', + 'CREATE TABLE t3 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,20) i') + + # node1 is now replica + node1.cleanup() + # kludge "backup_id=page1_id" + self.restore_node( + backup_dir, 'node', node1, node1.data_dir, + backup_id=page1_id, + options=[ + '--recovery-target-timeline=2', + '--recovery-target=latest']) + + # Settings for Replica + self.set_replica(node2, node1) + self.set_auto_conf(node1, {'port': node1.port}) + self.set_archiving(backup_dir, 'node', node1, replica=True) + + node1.slow_start(replica=True) + + node2.safe_psql( + 'postgres', + 'CREATE TABLE t4 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,30) i') + self.wait_until_replica_catch_with_master(node2, node1) + + # node1 is back to be a master + node1.promote() + self.switch_wal_segment(node1) + + sleep(5) + + # delta3_id = self.backup_node( + # backup_dir, 'node', node2, node2.data_dir, 'delta') + # page backup on timeline 3 + page3_id = self.backup_node( + backup_dir, 'node', node1, node1.data_dir, 'page', + options=['--archive-timeout=60']) + + pgdata = self.pgdata_content(node1.data_dir) + + node1.cleanup() + self.restore_node(backup_dir, 'node', node1, node1.data_dir) + + pgdata_restored = self.pgdata_content(node1.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_parent_choosing(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + master = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'master'), + set_replication=True, + initdb_params=['--data-checksums']) + + if self.get_version(master) < self.version_to_num('9.6.0'): + self.skipTest( + 'Skipped because backup from replica is not supported in PG 9.5') + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'master', master) + + master.slow_start() + + self.backup_node(backup_dir, 'master', master, options=['--stream']) + + # Create replica + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + self.restore_node(backup_dir, 'master', replica) + + # Settings for Replica + self.set_replica(master, replica) + self.set_auto_conf(replica, {'port': replica.port}) + + replica.slow_start(replica=True) + + master.safe_psql( + 'postgres', + 'CREATE TABLE t1 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,20) i') + self.wait_until_replica_catch_with_master(master, replica) + + self.add_instance(backup_dir, 'replica', replica) + + full_id = self.backup_node( + backup_dir, 'replica', + replica, options=['--stream']) + + master.safe_psql( + 'postgres', + 'CREATE TABLE t2 AS ' + 'SELECT i, repeat(md5(i::text),5006056) AS fat_attr ' + 'FROM generate_series(0,20) i') + self.wait_until_replica_catch_with_master(master, replica) + + self.backup_node( + backup_dir, 'replica', replica, + backup_type='delta', options=['--stream']) + + replica.promote() + + # failing, because without archving, it is impossible to + # take multi-timeline backup. + self.backup_node( + backup_dir, 'replica', replica, + backup_type='delta', options=['--stream']) + + # @unittest.skip("skip") + def test_instance_from_the_past(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + node.slow_start() + + full_id = self.backup_node(backup_dir, 'node', node, options=['--stream']) + + node.pgbench_init(scale=10) + self.backup_node(backup_dir, 'node', node, options=['--stream']) + node.cleanup() + + self.restore_node(backup_dir, 'node', node, backup_id=full_id) + node.slow_start() + + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['--stream']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because instance is from the past " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertTrue( + 'ERROR: Current START LSN' in e.message and + 'is lower than START LSN' in e.message and + 'It may indicate that we are trying to backup ' + 'PostgreSQL instance from the past' in e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_replica_via_basebackup(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={'hot_standby': 'on'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + node.slow_start() + + node.pgbench_init(scale=10) + + #FULL backup + full_id = self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=['--recovery-target=latest', '--recovery-target-action=promote']) + node.slow_start() + + # Timeline 2 + # Take stream page backup from instance in timeline2 + self.backup_node( + backup_dir, 'node', node, backup_type='full', + options=['--stream', '--log-level-file=verbose']) + + node.cleanup() + + # restore stream backup + self.restore_node(backup_dir, 'node', node) + + xlog_dir = 'pg_wal' + if self.get_version(node) < 100000: + xlog_dir = 'pg_xlog' + + filepath = os.path.join(node.data_dir, xlog_dir, "00000002.history") + self.assertTrue( + os.path.exists(filepath), + "History file do not exists: {0}".format(filepath)) + + node.slow_start() + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + pg_basebackup_path = self.get_bin_path('pg_basebackup') + + self.run_binary( + [ + pg_basebackup_path, '-p', str(node.port), '-h', 'localhost', + '-R', '-X', 'stream', '-D', node_restored.data_dir + ]) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + node_restored.slow_start(replica=True) + +# TODO: +# null offset STOP LSN and latest record in previous segment is conrecord (manual only) +# archiving from promoted delayed replica diff --git a/tests/requirements.txt b/tests/requirements.txt new file mode 100644 index 000000000..e2ac18bea --- /dev/null +++ b/tests/requirements.txt @@ -0,0 +1,13 @@ +# Testgres can be installed in the following ways: +# 1. From a pip package (recommended) +# testgres==1.8.5 +# 2. From a specific Git branch, tag or commit +# git+https://github.com/postgrespro/testgres.git@ +# 3. From a local directory +# /path/to/local/directory/testgres +git+https://github.com/postgrespro/testgres.git@archive-command-exec#egg=testgres-pg_probackup2&subdirectory=testgres/plugins/pg_probackup2 +allure-pytest +deprecation +pexpect +pytest==7.4.3 +pytest-xdist diff --git a/tests/restore.py b/tests/restore.py deleted file mode 100644 index 583a99a0a..000000000 --- a/tests/restore.py +++ /dev/null @@ -1,2359 +0,0 @@ -import os -import unittest -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException -import subprocess -from datetime import datetime -import sys -from time import sleep -from datetime import datetime, timedelta -import hashlib -import shutil - - -module_name = 'restore' - - -class RestoreTest(ProbackupTest, unittest.TestCase): - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_restore_full_to_latest(self): - """recovery to latest from full backup""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - before = node.execute("postgres", "SELECT * FROM pgbench_branches") - backup_id = self.backup_node(backup_dir, 'node', node) - - node.stop() - node.cleanup() - - # 1 - Test recovery from latest - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=["-j", "4", "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - # 2 - Test that recovery.conf was created - recovery_conf = os.path.join(node.data_dir, "recovery.conf") - self.assertEqual(os.path.isfile(recovery_conf), True) - - node.slow_start() - - after = node.execute("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_full_page_to_latest(self): - """recovery to latest from full + page backups""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - - self.backup_node(backup_dir, 'node', node) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - backup_id = self.backup_node( - backup_dir, 'node', node, backup_type="page") - - before = node.execute("postgres", "SELECT * FROM pgbench_branches") - - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=["-j", "4", "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - - after = node.execute("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_to_specific_timeline(self): - """recovery to target timeline""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - - before = node.execute("postgres", "SELECT * FROM pgbench_branches") - - backup_id = self.backup_node(backup_dir, 'node', node) - - target_tli = int( - node.get_control_data()["Latest checkpoint's TimeLineID"]) - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=["-j", "4", "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT, - options=['-T', '10', '-c', '2', '--no-vacuum']) - pgbench.wait() - pgbench.stdout.close() - - self.backup_node(backup_dir, 'node', node) - - node.stop() - node.cleanup() - - # Correct Backup must be choosen for restore - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-j", "4", "--timeline={0}".format(target_tli), - "--recovery-target-action=promote"] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - recovery_target_timeline = self.get_recovery_conf( - node)["recovery_target_timeline"] - self.assertEqual(int(recovery_target_timeline), target_tli) - - node.slow_start() - after = node.execute("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_to_time(self): - """recovery to target time""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.append_conf("postgresql.auto.conf", "TimeZone = Europe/Moscow") - node.slow_start() - - node.pgbench_init(scale=2) - before = node.execute("postgres", "SELECT * FROM pgbench_branches") - - backup_id = self.backup_node(backup_dir, 'node', node) - - target_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S") - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-j", "4", '--time={0}'.format(target_time), - "--recovery-target-action=promote" - ] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - after = node.execute("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_to_xid_inclusive(self): - """recovery to target xid""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - with node.connect("postgres") as con: - con.execute("CREATE TABLE tbl0005 (a text)") - con.commit() - - backup_id = self.backup_node(backup_dir, 'node', node) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - before = node.safe_psql("postgres", "SELECT * FROM pgbench_branches") - with node.connect("postgres") as con: - res = con.execute("INSERT INTO tbl0005 VALUES ('inserted') RETURNING (xmin)") - con.commit() - target_xid = res[0][0] - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-j", "4", '--xid={0}'.format(target_xid), - "--recovery-target-action=promote"] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - after = node.safe_psql("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - self.assertEqual( - len(node.execute("postgres", "SELECT * FROM tbl0005")), 1) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_to_xid_not_inclusive(self): - """recovery with target inclusive false""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - with node.connect("postgres") as con: - con.execute("CREATE TABLE tbl0005 (a text)") - con.commit() - - backup_id = self.backup_node(backup_dir, 'node', node) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - before = node.execute("postgres", "SELECT * FROM pgbench_branches") - with node.connect("postgres") as con: - result = con.execute("INSERT INTO tbl0005 VALUES ('inserted') RETURNING (xmin)") - con.commit() - target_xid = result[0][0] - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-j", "4", - '--xid={0}'.format(target_xid), - "--inclusive=false", - "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - after = node.execute("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - self.assertEqual( - len(node.execute("postgres", "SELECT * FROM tbl0005")), 0) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_to_lsn_inclusive(self): - """recovery to target lsn""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - if self.get_version(node) < self.version_to_num('10.0'): - self.del_test_dir(module_name, fname) - return - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - with node.connect("postgres") as con: - con.execute("CREATE TABLE tbl0005 (a int)") - con.commit() - - backup_id = self.backup_node(backup_dir, 'node', node) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - before = node.safe_psql("postgres", "SELECT * FROM pgbench_branches") - with node.connect("postgres") as con: - con.execute("INSERT INTO tbl0005 VALUES (1)") - con.commit() - res = con.execute("SELECT pg_current_wal_lsn()") - con.commit() - con.execute("INSERT INTO tbl0005 VALUES (2)") - con.commit() - xlogid, xrecoff = res[0][0].split('/') - xrecoff = hex(int(xrecoff, 16) + 1)[2:] - target_lsn = "{0}/{1}".format(xlogid, xrecoff) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-j", "4", '--lsn={0}'.format(target_lsn), - "--recovery-target-action=promote"] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - - after = node.safe_psql("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - self.assertEqual( - len(node.execute("postgres", "SELECT * FROM tbl0005")), 2) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_to_lsn_not_inclusive(self): - """recovery to target lsn""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - if self.get_version(node) < self.version_to_num('10.0'): - self.del_test_dir(module_name, fname) - return - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - with node.connect("postgres") as con: - con.execute("CREATE TABLE tbl0005 (a int)") - con.commit() - - backup_id = self.backup_node(backup_dir, 'node', node) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - before = node.safe_psql("postgres", "SELECT * FROM pgbench_branches") - with node.connect("postgres") as con: - con.execute("INSERT INTO tbl0005 VALUES (1)") - con.commit() - res = con.execute("SELECT pg_current_wal_lsn()") - con.commit() - con.execute("INSERT INTO tbl0005 VALUES (2)") - con.commit() - xlogid, xrecoff = res[0][0].split('/') - xrecoff = hex(int(xrecoff, 16) + 1)[2:] - target_lsn = "{0}/{1}".format(xlogid, xrecoff) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "--inclusive=false", - "-j", "4", '--lsn={0}'.format(target_lsn), - "--recovery-target-action=promote"] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - - after = node.safe_psql("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - self.assertEqual( - len(node.execute("postgres", "SELECT * FROM tbl0005")), 1) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_full_ptrack_archive(self): - """recovery to latest from archive full+ptrack backups""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums'], - pg_options={'ptrack_enable': 'on'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - - self.backup_node(backup_dir, 'node', node) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - backup_id = self.backup_node( - backup_dir, 'node', node, backup_type="ptrack") - - before = node.execute("postgres", "SELECT * FROM pgbench_branches") - - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-j", "4", "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - after = node.execute("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_ptrack(self): - """recovery to latest from archive full+ptrack+ptrack backups""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums'], - pg_options={'ptrack_enable': 'on'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - - self.backup_node(backup_dir, 'node', node) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - self.backup_node(backup_dir, 'node', node, backup_type="ptrack") - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - backup_id = self.backup_node( - backup_dir, 'node', node, backup_type="ptrack") - - before = node.execute("postgres", "SELECT * FROM pgbench_branches") - - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-j", "4", "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - after = node.execute("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_full_ptrack_stream(self): - """recovery in stream mode to latest from full + ptrack backups""" - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={'ptrack_enable': 'on'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - - self.backup_node(backup_dir, 'node', node, options=["--stream"]) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - backup_id = self.backup_node( - backup_dir, 'node', node, - backup_type="ptrack", options=["--stream"]) - - before = node.execute("postgres", "SELECT * FROM pgbench_branches") - - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=["-j", "4", "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - after = node.execute("postgres", "SELECT * FROM pgbench_branches") - self.assertEqual(before, after) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_full_ptrack_under_load(self): - """ - recovery to latest from full + ptrack backups - with loads when ptrack backup do - """ - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=2) - - self.backup_node(backup_dir, 'node', node) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "8"] - ) - - backup_id = self.backup_node( - backup_dir, 'node', node, - backup_type="ptrack", options=["--stream"]) - - pgbench.wait() - pgbench.stdout.close() - - bbalance = node.execute( - "postgres", "SELECT sum(bbalance) FROM pgbench_branches") - delta = node.execute( - "postgres", "SELECT sum(delta) FROM pgbench_history") - - self.assertEqual(bbalance, delta) - node.stop() - node.cleanup() - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=["-j", "4", "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - bbalance = node.execute( - "postgres", "SELECT sum(bbalance) FROM pgbench_branches") - delta = node.execute( - "postgres", "SELECT sum(delta) FROM pgbench_history") - self.assertEqual(bbalance, delta) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_full_under_load_ptrack(self): - """ - recovery to latest from full + page backups - with loads when full backup do - """ - if not self.ptrack: - return unittest.skip('Skipped because ptrack support is disabled') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'ptrack_enable': 'on'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # wal_segment_size = self.guc_wal_segment_size(node) - node.pgbench_init(scale=2) - - pgbench = node.pgbench( - stdout=subprocess.PIPE, - stderr=subprocess.STDOUT, - options=["-c", "4", "-T", "8"] - ) - - self.backup_node(backup_dir, 'node', node) - - pgbench.wait() - pgbench.stdout.close() - - backup_id = self.backup_node( - backup_dir, 'node', node, - backup_type="ptrack", options=["--stream"]) - - bbalance = node.execute( - "postgres", "SELECT sum(bbalance) FROM pgbench_branches") - delta = node.execute( - "postgres", "SELECT sum(delta) FROM pgbench_history") - - self.assertEqual(bbalance, delta) - - node.stop() - node.cleanup() - # self.wrong_wal_clean(node, wal_segment_size) - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=["-j", "4", "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - node.slow_start() - bbalance = node.execute( - "postgres", "SELECT sum(bbalance) FROM pgbench_branches") - delta = node.execute( - "postgres", "SELECT sum(delta) FROM pgbench_history") - self.assertEqual(bbalance, delta) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_with_tablespace_mapping_1(self): - """recovery using tablespace-mapping option""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # Create tablespace - tblspc_path = os.path.join(node.base_dir, "tblspc") - os.makedirs(tblspc_path) - with node.connect("postgres") as con: - con.connection.autocommit = True - con.execute("CREATE TABLESPACE tblspc LOCATION '%s'" % tblspc_path) - con.connection.autocommit = False - con.execute("CREATE TABLE test (id int) TABLESPACE tblspc") - con.execute("INSERT INTO test VALUES (1)") - con.commit() - - backup_id = self.backup_node(backup_dir, 'node', node) - self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") - - # 1 - Try to restore to existing directory - node.stop() - try: - self.restore_node(backup_dir, 'node', node) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because restore destination is not empty.\n " - "Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - 'ERROR: restore destination is not empty:', - e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - # 2 - Try to restore to existing tablespace directory - tblspc_path_tmp = os.path.join(node.base_dir, "tblspc_tmp") - os.rename(tblspc_path, tblspc_path_tmp) - node.cleanup() - os.rename(tblspc_path_tmp, tblspc_path) - try: - self.restore_node(backup_dir, 'node', node) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because restore tablespace destination is " - "not empty.\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - 'ERROR: restore tablespace destination is not empty:', - e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - # 3 - Restore using tablespace-mapping to not empty directory - tblspc_path_temp = os.path.join(node.base_dir, "tblspc_temp") - os.mkdir(tblspc_path_temp) - with open(os.path.join(tblspc_path_temp, 'file'), 'w+') as f: - f.close() - - try: - self.restore_node( - backup_dir, 'node', node, - options=["-T", "%s=%s" % (tblspc_path, tblspc_path_temp)]) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because restore tablespace destination is " - "not empty.\n Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - 'ERROR: restore tablespace destination is not empty:', - e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - # 4 - Restore using tablespace-mapping - tblspc_path_new = os.path.join(node.base_dir, "tblspc_new") - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-T", "%s=%s" % (tblspc_path, tblspc_path_new), - "--recovery-target-action=promote"] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - - result = node.execute("postgres", "SELECT id FROM test") - self.assertEqual(result[0][0], 1) - - # 4 - Restore using tablespace-mapping using page backup - self.backup_node(backup_dir, 'node', node) - with node.connect("postgres") as con: - con.execute("INSERT INTO test VALUES (2)") - con.commit() - backup_id = self.backup_node( - backup_dir, 'node', node, backup_type="page") - - show_pb = self.show_pb(backup_dir, 'node') - self.assertEqual(show_pb[1]['status'], "OK") - self.assertEqual(show_pb[2]['status'], "OK") - - node.stop() - node.cleanup() - tblspc_path_page = os.path.join(node.base_dir, "tblspc_page") - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-T", "%s=%s" % (tblspc_path_new, tblspc_path_page), - "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - result = node.execute("postgres", "SELECT id FROM test OFFSET 1") - self.assertEqual(result[0][0], 2) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_with_tablespace_mapping_2(self): - """recovery using tablespace-mapping option and page backup""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # Full backup - self.backup_node(backup_dir, 'node', node) - self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") - - # Create tablespace - tblspc_path = os.path.join(node.base_dir, "tblspc") - os.makedirs(tblspc_path) - with node.connect("postgres") as con: - con.connection.autocommit = True - con.execute("CREATE TABLESPACE tblspc LOCATION '%s'" % tblspc_path) - con.connection.autocommit = False - con.execute( - "CREATE TABLE tbl AS SELECT * " - "FROM generate_series(0,3) AS integer") - con.commit() - - # First page backup - self.backup_node(backup_dir, 'node', node, backup_type="page") - self.assertEqual(self.show_pb(backup_dir, 'node')[1]['status'], "OK") - self.assertEqual( - self.show_pb(backup_dir, 'node')[1]['backup-mode'], "PAGE") - - # Create tablespace table - with node.connect("postgres") as con: - con.connection.autocommit = True - con.execute("CHECKPOINT") - con.connection.autocommit = False - con.execute("CREATE TABLE tbl1 (a int) TABLESPACE tblspc") - con.execute( - "INSERT INTO tbl1 SELECT * " - "FROM generate_series(0,3) AS integer") - con.commit() - - # Second page backup - backup_id = self.backup_node( - backup_dir, 'node', node, backup_type="page") - self.assertEqual(self.show_pb(backup_dir, 'node')[2]['status'], "OK") - self.assertEqual( - self.show_pb(backup_dir, 'node')[2]['backup-mode'], "PAGE") - - node.stop() - node.cleanup() - - tblspc_path_new = os.path.join(node.base_dir, "tblspc_new") - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-T", "%s=%s" % (tblspc_path, tblspc_path_new), - "--recovery-target-action=promote"]), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - node.slow_start() - - count = node.execute("postgres", "SELECT count(*) FROM tbl") - self.assertEqual(count[0][0], 4) - count = node.execute("postgres", "SELECT count(*) FROM tbl1") - self.assertEqual(count[0][0], 4) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_archive_node_backup_stream_restore_to_recovery_time(self): - """ - make node with archiving, make stream backup, - make PITR to Recovery Time - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node( - backup_dir, 'node', node, options=["--stream"]) - node.safe_psql("postgres", "create table t_heap(a int)") - - node.stop() - node.cleanup() - - recovery_time = self.show_pb( - backup_dir, 'node', backup_id)['recovery-time'] - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-j", "4", '--time={0}'.format(recovery_time), - "--recovery-target-action=promote" - ] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - - result = node.psql("postgres", 'select * from t_heap') - self.assertTrue('does not exist' in result[2].decode("utf-8")) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_archive_node_backup_stream_restore_to_recovery_time(self): - """ - make node with archiving, make stream backup, - make PITR to Recovery Time - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node( - backup_dir, 'node', node, options=["--stream"]) - node.safe_psql("postgres", "create table t_heap(a int)") - node.stop() - node.cleanup() - - recovery_time = self.show_pb( - backup_dir, 'node', backup_id)['recovery-time'] - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-j", "4", '--time={0}'.format(recovery_time), - "--recovery-target-action=promote" - ] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - result = node.psql("postgres", 'select * from t_heap') - self.assertTrue('does not exist' in result[2].decode("utf-8")) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_archive_node_backup_stream_pitr(self): - """ - make node with archiving, make stream backup, - create table t_heap, make pitr to Recovery Time, - check that t_heap do not exists - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node( - backup_dir, 'node', node, options=["--stream"]) - node.safe_psql("postgres", "create table t_heap(a int)") - node.cleanup() - - recovery_time = self.show_pb( - backup_dir, 'node', backup_id)['recovery-time'] - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node, - options=[ - "-j", "4", '--time={0}'.format(recovery_time), - "--recovery-target-action=promote" - ] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - node.slow_start() - - result = node.psql("postgres", 'select * from t_heap') - self.assertEqual(True, 'does not exist' in result[2].decode("utf-8")) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_archive_node_backup_archive_pitr_2(self): - """ - make node with archiving, make archive backup, - create table t_heap, make pitr to Recovery Time, - check that t_heap do not exists - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node(backup_dir, 'node', node) - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) - - node.safe_psql("postgres", "create table t_heap(a int)") - node.stop() - - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - node_restored.cleanup() - - recovery_time = self.show_pb( - backup_dir, 'node', backup_id)['recovery-time'] - - self.assertIn( - "INFO: Restore of backup {0} completed.".format(backup_id), - self.restore_node( - backup_dir, 'node', node_restored, - options=[ - "-j", "4", '--time={0}'.format(recovery_time), - "--recovery-target-action=promote"] - ), - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(self.output), self.cmd)) - - if self.paranoia: - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) - - node_restored.slow_start() - - result = node_restored.psql("postgres", 'select * from t_heap') - self.assertTrue('does not exist' in result[2].decode("utf-8")) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_archive_restore_to_restore_point(self): - """ - make node with archiving, make archive backup, - create table t_heap, make pitr to Recovery Time, - check that t_heap do not exists - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.backup_node(backup_dir, 'node', node) - - node.safe_psql( - "postgres", - "create table t_heap as select generate_series(0,10000)") - result = node.safe_psql( - "postgres", - "select * from t_heap") - node.safe_psql( - "postgres", "select pg_create_restore_point('savepoint')") - node.safe_psql( - "postgres", - "create table t_heap_1 as select generate_series(0,10000)") - node.cleanup() - - self.restore_node( - backup_dir, 'node', node, - options=[ - "--recovery-target-name=savepoint", - "--recovery-target-action=promote"]) - - node.slow_start() - - result_new = node.safe_psql("postgres", "select * from t_heap") - res = node.psql("postgres", "select * from t_heap_1") - self.assertEqual( - res[0], 1, - "Table t_heap_1 should not exist in restored instance") - - self.assertEqual(result, result_new) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - @unittest.skip("skip") - # @unittest.expectedFailure - def test_zags_block_corrupt(self): - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.backup_node(backup_dir, 'node', node) - - conn = node.connect() - with node.connect("postgres") as conn: - - conn.execute( - "create table tbl(i int)") - conn.commit() - conn.execute( - "create index idx ON tbl (i)") - conn.commit() - conn.execute( - "insert into tbl select i from generate_series(0,400) as i") - conn.commit() - conn.execute( - "select pg_relation_size('idx')") - conn.commit() - conn.execute( - "delete from tbl where i < 100") - conn.commit() - conn.execute( - "explain analyze select i from tbl order by i") - conn.commit() - conn.execute( - "select i from tbl order by i") - conn.commit() - conn.execute( - "create extension pageinspect") - conn.commit() - print(conn.execute( - "select * from bt_page_stats('idx',1)")) - conn.commit() - conn.execute( - "insert into tbl select i from generate_series(0,100) as i") - conn.commit() - conn.execute( - "insert into tbl select i from generate_series(0,100) as i") - conn.commit() - conn.execute( - "insert into tbl select i from generate_series(0,100) as i") - conn.commit() - conn.execute( - "insert into tbl select i from generate_series(0,100) as i") - - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored'), - initdb_params=['--data-checksums']) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored) - - node_restored.append_conf("postgresql.auto.conf", "archive_mode = 'off'") - node_restored.append_conf("postgresql.auto.conf", "hot_standby = 'on'") - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) - - node_restored.slow_start() - - @unittest.skip("skip") - # @unittest.expectedFailure - def test_zags_block_corrupt_1(self): - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off', - 'full_page_writes': 'on'} - ) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.backup_node(backup_dir, 'node', node) - - node.safe_psql('postgres', 'create table tbl(i int)') - - node.safe_psql('postgres', 'create index idx ON tbl (i)') - - node.safe_psql( - 'postgres', - 'insert into tbl select i from generate_series(0,100000) as i') - - node.safe_psql( - 'postgres', - 'delete from tbl where i%2 = 0') - - node.safe_psql( - 'postgres', - 'explain analyze select i from tbl order by i') - - node.safe_psql( - 'postgres', - 'select i from tbl order by i') - - node.safe_psql( - 'postgres', - 'create extension pageinspect') - - node.safe_psql( - 'postgres', - 'checkpoint') - - node.safe_psql( - 'postgres', - 'insert into tbl select i from generate_series(0,100) as i') - - node.safe_psql( - 'postgres', - 'insert into tbl select i from generate_series(0,100) as i') - - node.safe_psql( - 'postgres', - 'insert into tbl select i from generate_series(0,100) as i') - - node.safe_psql( - 'postgres', - 'insert into tbl select i from generate_series(0,100) as i') - - self.switch_wal_segment(node) - - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored'), - initdb_params=['--data-checksums']) - - pgdata = self.pgdata_content(node.data_dir) - - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored) - - node_restored.append_conf("postgresql.auto.conf", "archive_mode = 'off'") - node_restored.append_conf("postgresql.auto.conf", "hot_standby = 'on'") - node_restored.append_conf( - "postgresql.auto.conf", "port = {0}".format(node_restored.port)) - - node_restored.slow_start() - - while True: - with open(node_restored.pg_log_file, 'r') as f: - if 'selected new timeline ID' in f.read(): - break - - # with open(node_restored.pg_log_file, 'r') as f: - # print(f.read()) - - pgdata_restored = self.pgdata_content(node_restored.data_dir) - - self.compare_pgdata(pgdata, pgdata_restored) - -# pg_xlogdump_path = self.get_bin_path('pg_xlogdump') - -# pg_xlogdump = self.run_binary( -# [ -# pg_xlogdump_path, '-b', -# os.path.join(backup_dir, 'wal', 'node', '000000010000000000000003'), -# ' | ', 'grep', 'Btree', '' -# ], async=False) - - if pg_xlogdump.returncode: - self.assertFalse( - True, - 'Failed to start pg_wal_dump: {0}'.format( - pg_receivexlog.communicate()[1])) - - # @unittest.skip("skip") - def test_restore_chain(self): - """ - make node, take full backup, take several - ERROR delta backups, take valid delta backup, - restore must be successfull - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # Take FULL - self.backup_node( - backup_dir, 'node', node) - - # Take DELTA - self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - # Take ERROR DELTA - try: - self.backup_node( - backup_dir, 'node', node, - backup_type='delta', options=['--archive-timeout=0s']) - except ProbackupException as e: - pass - - # Take ERROR DELTA - try: - self.backup_node( - backup_dir, 'node', node, - backup_type='delta', options=['--archive-timeout=0s']) - except ProbackupException as e: - pass - - # Take DELTA - self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - # Take ERROR DELTA - try: - self.backup_node( - backup_dir, 'node', node, - backup_type='delta', options=['--archive-timeout=0s']) - except ProbackupException as e: - pass - - self.assertEqual( - 'OK', - self.show_pb(backup_dir, 'node')[0]['status'], - 'Backup STATUS should be "OK"') - - self.assertEqual( - 'OK', - self.show_pb(backup_dir, 'node')[1]['status'], - 'Backup STATUS should be "OK"') - - self.assertEqual( - 'ERROR', - self.show_pb(backup_dir, 'node')[2]['status'], - 'Backup STATUS should be "ERROR"') - - self.assertEqual( - 'ERROR', - self.show_pb(backup_dir, 'node')[3]['status'], - 'Backup STATUS should be "ERROR"') - - self.assertEqual( - 'OK', - self.show_pb(backup_dir, 'node')[4]['status'], - 'Backup STATUS should be "OK"') - - self.assertEqual( - 'ERROR', - self.show_pb(backup_dir, 'node')[5]['status'], - 'Backup STATUS should be "ERROR"') - - node.cleanup() - - self.restore_node(backup_dir, 'node', node) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_chain_with_corrupted_backup(self): - """more complex test_restore_chain()""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # Take FULL - self.backup_node( - backup_dir, 'node', node) - - # Take DELTA - self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # Take ERROR DELTA - try: - self.backup_node( - backup_dir, 'node', node, - backup_type='page', options=['--archive-timeout=0s']) - except ProbackupException as e: - pass - - # Take 1 DELTA - self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - # Take ERROR DELTA - try: - self.backup_node( - backup_dir, 'node', node, - backup_type='delta', options=['--archive-timeout=0s']) - except ProbackupException as e: - pass - - # Take 2 DELTA - self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - # Take ERROR DELTA - try: - self.backup_node( - backup_dir, 'node', node, - backup_type='delta', options=['--archive-timeout=0s']) - except ProbackupException as e: - pass - - # Take 3 DELTA - self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - # Corrupted 4 DELTA - corrupt_id = self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - # ORPHAN 5 DELTA - restore_target_id = self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - # ORPHAN 6 DELTA - self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - # NEXT FULL BACKUP - self.backup_node( - backup_dir, 'node', node, backup_type='full') - - # Next Delta - self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - # do corrupt 6 DELTA backup - file = os.path.join( - backup_dir, 'backups', 'node', - corrupt_id, 'database', 'global', 'pg_control') - - file_new = os.path.join(backup_dir, 'pg_control') - os.rename(file, file_new) - - # RESTORE BACKUP - node.cleanup() - - try: - self.restore_node( - backup_dir, 'node', node, backup_id=restore_target_id) - self.assertEqual( - 1, 0, - "Expecting Error because restore backup is corrupted.\n " - "Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd)) - except ProbackupException as e: - self.assertIn( - 'ERROR: Backup {0} is orphan'.format(restore_target_id), - e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - self.assertEqual( - 'OK', - self.show_pb(backup_dir, 'node')[0]['status'], - 'Backup STATUS should be "OK"') - - self.assertEqual( - 'OK', - self.show_pb(backup_dir, 'node')[1]['status'], - 'Backup STATUS should be "OK"') - - self.assertEqual( - 'ERROR', - self.show_pb(backup_dir, 'node')[2]['status'], - 'Backup STATUS should be "ERROR"') - - self.assertEqual( - 'OK', - self.show_pb(backup_dir, 'node')[3]['status'], - 'Backup STATUS should be "OK"') - - self.assertEqual( - 'ERROR', - self.show_pb(backup_dir, 'node')[4]['status'], - 'Backup STATUS should be "ERROR"') - - self.assertEqual( - 'OK', - self.show_pb(backup_dir, 'node')[5]['status'], - 'Backup STATUS should be "OK"') - - self.assertEqual( - 'ERROR', - self.show_pb(backup_dir, 'node')[6]['status'], - 'Backup STATUS should be "ERROR"') - - self.assertEqual( - 'OK', - self.show_pb(backup_dir, 'node')[7]['status'], - 'Backup STATUS should be "OK"') - - # corruption victim - self.assertEqual( - 'CORRUPT', - self.show_pb(backup_dir, 'node')[8]['status'], - 'Backup STATUS should be "CORRUPT"') - - # orphaned child - self.assertEqual( - 'ORPHAN', - self.show_pb(backup_dir, 'node')[9]['status'], - 'Backup STATUS should be "ORPHAN"') - - # orphaned child - self.assertEqual( - 'ORPHAN', - self.show_pb(backup_dir, 'node')[10]['status'], - 'Backup STATUS should be "ORPHAN"') - - # next FULL - self.assertEqual( - 'OK', - self.show_pb(backup_dir, 'node')[11]['status'], - 'Backup STATUS should be "OK"') - - # next DELTA - self.assertEqual( - 'OK', - self.show_pb(backup_dir, 'node')[12]['status'], - 'Backup STATUS should be "OK"') - - node.cleanup() - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_backup_from_future(self): - """more complex test_restore_chain()""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums'], - pg_options={ - 'autovacuum': 'off'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # Take FULL - self.backup_node(backup_dir, 'node', node) - - node.pgbench_init(scale=3) - # pgbench = node.pgbench(options=['-T', '20', '-c', '2']) - # pgbench.wait() - - # Take PAGE from future - backup_id = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - with open( - os.path.join( - backup_dir, 'backups', 'node', - backup_id, "backup.control"), "a") as conf: - conf.write("start-time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() + timedelta(days=3))) - - # rename directory - new_id = self.show_pb(backup_dir, 'node')[1]['id'] - - os.rename( - os.path.join(backup_dir, 'backups', 'node', backup_id), - os.path.join(backup_dir, 'backups', 'node', new_id)) - - pgbench = node.pgbench(options=['-T', '3', '-c', '2', '--no-vacuum']) - pgbench.wait() - - backup_id = self.backup_node( - backup_dir, 'node', node, backup_type='page') - pgdata = self.pgdata_content(node.data_dir) - - node.cleanup() - self.restore_node(backup_dir, 'node', node, backup_id=backup_id) - - pgdata_restored = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_target_immediate_stream(self): - """ - correct handling of immediate recovery target - for STREAM backups - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - # Take FULL - self.backup_node( - backup_dir, 'node', node, options=['--stream']) - - # Take delta - backup_id = self.backup_node( - backup_dir, 'node', node, - backup_type='delta', options=['--stream']) - - pgdata = self.pgdata_content(node.data_dir) - - recovery_conf = os.path.join(node.data_dir, 'recovery.conf') - - # restore page backup - node.cleanup() - self.restore_node( - backup_dir, 'node', node, options=['--immediate']) - - # For stream backup with immediate recovery target there is no need to - # create recovery.conf. Is it wise? - self.assertFalse( - os.path.isfile(recovery_conf)) - - # restore page backup - node.cleanup() - self.restore_node( - backup_dir, 'node', node, options=['--recovery-target=immediate']) - - # For stream backup with immediate recovery target there is no need to - # create recovery.conf. Is it wise? - self.assertFalse( - os.path.isfile(recovery_conf)) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_target_immediate_archive(self): - """ - correct handling of immediate recovery target - for ARCHIVE backups - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # Take FULL - self.backup_node( - backup_dir, 'node', node) - - # Take delta - backup_id = self.backup_node( - backup_dir, 'node', node, - backup_type='delta') - - pgdata = self.pgdata_content(node.data_dir) - - recovery_conf = os.path.join(node.data_dir, 'recovery.conf') - - # restore page backup - node.cleanup() - self.restore_node( - backup_dir, 'node', node, options=['--immediate']) - - # For archive backup with immediate recovery target - # recovery.conf is mandatory - with open(recovery_conf, 'r') as f: - self.assertIn("recovery_target = 'immediate'", f.read()) - - # restore page backup - node.cleanup() - self.restore_node( - backup_dir, 'node', node, options=['--recovery-target=immediate']) - - # For archive backup with immediate recovery target - # recovery.conf is mandatory - with open(recovery_conf, 'r') as f: - self.assertIn("recovery_target = 'immediate'", f.read()) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_target_latest_archive(self): - """ - make sure that recovery_target 'latest' - is default recovery target - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # Take FULL - self.backup_node( - backup_dir, 'node', node) - - recovery_conf = os.path.join(node.data_dir, 'recovery.conf') - - # restore - node.cleanup() - self.restore_node( - backup_dir, 'node', node) - - # with open(recovery_conf, 'r') as f: - # print(f.read()) - - hash_1 = hashlib.md5( - open(recovery_conf, 'rb').read()).hexdigest() - - # restore - node.cleanup() - self.restore_node( - backup_dir, 'node', node, options=['--recovery-target=latest']) - - # with open(recovery_conf, 'r') as f: - # print(f.read()) - - hash_2 = hashlib.md5( - open(recovery_conf, 'rb').read()).hexdigest() - - self.assertEqual(hash_1, hash_2) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_target_new_options(self): - """ - check that new --recovery-target-* - options are working correctly - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # Take FULL - self.backup_node(backup_dir, 'node', node) - - recovery_conf = os.path.join(node.data_dir, 'recovery.conf') - node.pgbench_init(scale=2) - pgbench = node.pgbench( - stdout=subprocess.PIPE, stderr=subprocess.STDOUT) - pgbench.wait() - pgbench.stdout.close() - - node.safe_psql( - "postgres", - "CREATE TABLE tbl0005 (a text)") - - node.safe_psql( - "postgres", "select pg_create_restore_point('savepoint')") - - target_name = 'savepoint' - - target_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S") - with node.connect("postgres") as con: - res = con.execute( - "INSERT INTO tbl0005 VALUES ('inserted') RETURNING (xmin)") - con.commit() - target_xid = res[0][0] - - with node.connect("postgres") as con: - con.execute("INSERT INTO tbl0005 VALUES (1)") - con.commit() - if self.get_version(node) > self.version_to_num('10.0'): - res = con.execute("SELECT pg_current_wal_lsn()") - else: - res = con.execute("SELECT pg_current_xlog_location()") - - con.commit() - con.execute("INSERT INTO tbl0005 VALUES (2)") - con.commit() - xlogid, xrecoff = res[0][0].split('/') - xrecoff = hex(int(xrecoff, 16) + 1)[2:] - target_lsn = "{0}/{1}".format(xlogid, xrecoff) - - # Restore with recovery target time - node.cleanup() - self.restore_node( - backup_dir, 'node', node, - options=[ - '--recovery-target-time={0}'.format(target_time), - "--recovery-target-action=promote", - '--recovery-target-timeline=1', - ]) - - with open(recovery_conf, 'r') as f: - recovery_conf_content = f.read() - - self.assertIn( - "recovery_target_time = '{0}'".format(target_time), - recovery_conf_content) - - self.assertIn( - "recovery_target_action = 'promote'", - recovery_conf_content) - - self.assertIn( - "recovery_target_timeline = '1'", - recovery_conf_content) - - node.slow_start() - - # Restore with recovery target xid - node.cleanup() - self.restore_node( - backup_dir, 'node', node, - options=[ - '--recovery-target-xid={0}'.format(target_xid), - "--recovery-target-action=promote", - '--recovery-target-timeline=1', - ]) - - with open(recovery_conf, 'r') as f: - recovery_conf_content = f.read() - - self.assertIn( - "recovery_target_xid = '{0}'".format(target_xid), - recovery_conf_content) - - self.assertIn( - "recovery_target_action = 'promote'", - recovery_conf_content) - - self.assertIn( - "recovery_target_timeline = '1'", - recovery_conf_content) - - node.slow_start() - - # Restore with recovery target name - node.cleanup() - self.restore_node( - backup_dir, 'node', node, - options=[ - '--recovery-target-name={0}'.format(target_name), - "--recovery-target-action=promote", - '--recovery-target-timeline=1', - ]) - - with open(recovery_conf, 'r') as f: - recovery_conf_content = f.read() - - self.assertIn( - "recovery_target_name = '{0}'".format(target_name), - recovery_conf_content) - - self.assertIn( - "recovery_target_action = 'promote'", - recovery_conf_content) - - self.assertIn( - "recovery_target_timeline = '1'", - recovery_conf_content) - - node.slow_start() - - # Restore with recovery target lsn - if self.get_version(node) >= 100000: - - node.cleanup() - self.restore_node( - backup_dir, 'node', node, - options=[ - '--recovery-target-lsn={0}'.format(target_lsn), - "--recovery-target-action=promote", - '--recovery-target-timeline=1', - ]) - - with open(recovery_conf, 'r') as f: - recovery_conf_content = f.read() - - self.assertIn( - "recovery_target_lsn = '{0}'".format(target_lsn), - recovery_conf_content) - - self.assertIn( - "recovery_target_action = 'promote'", - recovery_conf_content) - - self.assertIn( - "recovery_target_timeline = '1'", - recovery_conf_content) - - node.slow_start() - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_smart_restore(self): - """ - make node, create database, take full backup, drop database, - take incremental backup and restore it, - make sure that files from dropped database are not - copied during restore - https://github.com/postgrespro/pg_probackup/issues/63 - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # create database - node.safe_psql( - "postgres", - "CREATE DATABASE testdb") - - # take FULL backup - full_id = self.backup_node(backup_dir, 'node', node) - - # drop database - node.safe_psql( - "postgres", - "DROP DATABASE testdb") - - # take PAGE backup - page_id = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # restore PAGE backup - node.cleanup() - self.restore_node( - backup_dir, 'node', node, backup_id=page_id, - options=['--no-validate', '--log-level-file=VERBOSE']) - - logfile = os.path.join(backup_dir, 'log', 'pg_probackup.log') - with open(logfile, 'r') as f: - logfile_content = f.read() - - # get delta between FULL and PAGE filelists - filelist_full = self.get_backup_filelist( - backup_dir, 'node', full_id) - - filelist_page = self.get_backup_filelist( - backup_dir, 'node', page_id) - - filelist_diff = self.get_backup_filelist_diff( - filelist_full, filelist_page) - - for file in filelist_diff: - self.assertNotIn(file, logfile_content) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_pg_11_group_access(self): - """ - test group access for PG >= 11 - """ - if self.pg_config_version < self.version_to_num('11.0'): - return unittest.skip('You need PostgreSQL >= 11 for this test') - - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=[ - '--data-checksums', - '--allow-group-access']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - # take FULL backup - self.backup_node(backup_dir, 'node', node, options=['--stream']) - - pgdata = self.pgdata_content(node.data_dir) - - # restore backup - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored) - - # compare pgdata permissions - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_pg_10_waldir(self): - """ - test group access for PG >= 11 - """ - if self.pg_config_version < self.version_to_num('10.0'): - return unittest.skip('You need PostgreSQL >= 10 for this test') - - fname = self.id().split('.')[3] - wal_dir = os.path.join( - os.path.join(self.tmp_path, module_name, fname), 'wal_dir') - shutil.rmtree(wal_dir, ignore_errors=True) - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=[ - '--data-checksums', - '--waldir={0}'.format(wal_dir)]) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - # take FULL backup - self.backup_node( - backup_dir, 'node', node, options=['--stream']) - - pgdata = self.pgdata_content(node.data_dir) - - # restore backup - node_restored = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node_restored')) - node_restored.cleanup() - - self.restore_node( - backup_dir, 'node', node_restored) - - # compare pgdata permissions - pgdata_restored = self.pgdata_content(node_restored.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - self.assertTrue( - os.path.islink(os.path.join(node_restored.data_dir, 'pg_wal')), - 'pg_wal should be symlink') - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_restore_concurrent_drop_table(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=1) - - # FULL backup - self.backup_node( - backup_dir, 'node', node, - options=['--stream', '--compress']) - - # DELTA backup - gdb = self.backup_node( - backup_dir, 'node', node, backup_type='delta', - options=['--stream', '--compress', '--no-validate'], - gdb=True) - - gdb.set_breakpoint('backup_data_file') - gdb.run_until_break() - - node.safe_psql( - 'postgres', - 'DROP TABLE pgbench_accounts') - - # do checkpoint to guarantee filenode removal - node.safe_psql( - 'postgres', - 'CHECKPOINT') - - gdb.remove_all_breakpoints() - gdb.continue_execution_until_exit() - - pgdata = self.pgdata_content(node.data_dir) - node.cleanup() - - self.restore_node( - backup_dir, 'node', node, options=['--no-validate']) - - pgdata_restored = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_lost_non_data_file(self): - """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - set_replication=True, - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - node.slow_start() - - # FULL backup - backup_id = self.backup_node( - backup_dir, 'node', node, options=['--stream']) - - file = os.path.join( - backup_dir, 'backups', 'node', - backup_id, 'database', 'postgresql.auto.conf') - - os.remove(file) - - node.cleanup() - - try: - self.restore_node( - backup_dir, 'node', node, options=['--no-validate']) - self.assertEqual( - 1, 0, - "Expecting Error because of non-data file dissapearance.\n " - "Output: {0} \n CMD: {1}".format( - self.output, self.cmd)) - except ProbackupException as e: - self.assertIn( - 'is not found', e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - self.assertIn( - 'ERROR: Data files restoring failed', e.message, - '\n Unexpected Error Message: {0}\n CMD: {1}'.format( - repr(e.message), self.cmd)) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/restore_test.py b/tests/restore_test.py new file mode 100644 index 000000000..b6664252e --- /dev/null +++ b/tests/restore_test.py @@ -0,0 +1,3932 @@ +import os +import unittest +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +import subprocess +import sys +from datetime import datetime, timedelta, timezone +import hashlib +import shutil +import json +import stat +from shutil import copyfile +from testgres import QueryException, StartNodeException +from stat import S_ISDIR + + +class RestoreTest(ProbackupTest, unittest.TestCase): + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_restore_full_to_latest(self): + """recovery to latest from full backup""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=2) + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + before = node.table_checksum("pgbench_branches") + backup_id = self.backup_node(backup_dir, 'node', node) + + node.stop() + node.cleanup() + + # 1 - Test recovery from latest + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + # 2 - Test that recovery.conf was created + # TODO update test + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_conf = os.path.join(node.data_dir, 'postgresql.auto.conf') + with open(recovery_conf, 'r') as f: + print(f.read()) + else: + recovery_conf = os.path.join(node.data_dir, 'recovery.conf') + self.assertEqual(os.path.isfile(recovery_conf), True) + + node.slow_start() + + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + + # @unittest.skip("skip") + def test_restore_full_page_to_latest(self): + """recovery to latest from full + page backups""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=2) + + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type="page") + + before = node.table_checksum("pgbench_branches") + + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + + # @unittest.skip("skip") + def test_restore_to_specific_timeline(self): + """recovery to target timeline""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=2) + + before = node.table_checksum("pgbench_branches") + + backup_id = self.backup_node(backup_dir, 'node', node) + + target_tli = int( + node.get_control_data()["Latest checkpoint's TimeLineID"]) + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '2', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node(backup_dir, 'node', node) + + node.stop() + node.cleanup() + + # Correct Backup must be choosen for restore + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", "--timeline={0}".format(target_tli)] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + recovery_target_timeline = self.get_recovery_conf( + node)["recovery_target_timeline"] + self.assertEqual(int(recovery_target_timeline), target_tli) + + node.slow_start() + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + + # @unittest.skip("skip") + def test_restore_to_time(self): + """recovery to target time""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + pg_options={'TimeZone': 'GMT'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=2) + before = node.table_checksum("pgbench_branches") + + backup_id = self.backup_node(backup_dir, 'node', node) + + target_time = node.execute( + "postgres", "SELECT to_char(now(), 'YYYY-MM-DD HH24:MI:SS+00')" + )[0][0] + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", '--time={0}'.format(target_time), + "--recovery-target-action=promote" + ] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + + # @unittest.skip("skip") + def test_restore_to_xid_inclusive(self): + """recovery to target xid""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=2) + with node.connect("postgres") as con: + con.execute("CREATE TABLE tbl0005 (a text)") + con.commit() + + backup_id = self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + before = node.table_checksum("pgbench_branches") + with node.connect("postgres") as con: + res = con.execute("INSERT INTO tbl0005 VALUES ('inserted') RETURNING (xmin)") + con.commit() + target_xid = res[0][0] + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", '--xid={0}'.format(target_xid), + "--recovery-target-action=promote"] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + self.assertEqual( + len(node.execute("postgres", "SELECT * FROM tbl0005")), 1) + + # @unittest.skip("skip") + def test_restore_to_xid_not_inclusive(self): + """recovery with target inclusive false""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=2) + with node.connect("postgres") as con: + con.execute("CREATE TABLE tbl0005 (a text)") + con.commit() + + backup_id = self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + before = node.table_checksum("pgbench_branches") + with node.connect("postgres") as con: + result = con.execute("INSERT INTO tbl0005 VALUES ('inserted') RETURNING (xmin)") + con.commit() + target_xid = result[0][0] + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", + '--xid={0}'.format(target_xid), + "--inclusive=false", + "--recovery-target-action=promote"]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + self.assertEqual( + len(node.execute("postgres", "SELECT * FROM tbl0005")), 0) + + # @unittest.skip("skip") + def test_restore_to_lsn_inclusive(self): + """recovery to target lsn""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + if self.get_version(node) < self.version_to_num('10.0'): + return + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=2) + with node.connect("postgres") as con: + con.execute("CREATE TABLE tbl0005 (a int)") + con.commit() + + backup_id = self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + before = node.table_checksum("pgbench_branches") + with node.connect("postgres") as con: + con.execute("INSERT INTO tbl0005 VALUES (1)") + con.commit() + res = con.execute("SELECT pg_current_wal_lsn()") + con.commit() + con.execute("INSERT INTO tbl0005 VALUES (2)") + con.commit() + xlogid, xrecoff = res[0][0].split('/') + xrecoff = hex(int(xrecoff, 16) + 1)[2:] + target_lsn = "{0}/{1}".format(xlogid, xrecoff) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", '--lsn={0}'.format(target_lsn), + "--recovery-target-action=promote"] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + self.assertEqual( + len(node.execute("postgres", "SELECT * FROM tbl0005")), 2) + + # @unittest.skip("skip") + def test_restore_to_lsn_not_inclusive(self): + """recovery to target lsn""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + if self.get_version(node) < self.version_to_num('10.0'): + return + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=2) + with node.connect("postgres") as con: + con.execute("CREATE TABLE tbl0005 (a int)") + con.commit() + + backup_id = self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + before = node.table_checksum("pgbench_branches") + with node.connect("postgres") as con: + con.execute("INSERT INTO tbl0005 VALUES (1)") + con.commit() + res = con.execute("SELECT pg_current_wal_lsn()") + con.commit() + con.execute("INSERT INTO tbl0005 VALUES (2)") + con.commit() + xlogid, xrecoff = res[0][0].split('/') + xrecoff = hex(int(xrecoff, 16) + 1)[2:] + target_lsn = "{0}/{1}".format(xlogid, xrecoff) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "--inclusive=false", + "-j", "4", '--lsn={0}'.format(target_lsn), + "--recovery-target-action=promote"] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + self.assertEqual( + len(node.execute("postgres", "SELECT * FROM tbl0005")), 1) + + # @unittest.skip("skip") + def test_restore_full_ptrack_archive(self): + """recovery to latest from archive full+ptrack backups""" + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + ptrack_enable=True) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + node.pgbench_init(scale=2) + + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type="ptrack") + + before = node.table_checksum("pgbench_branches") + + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=["-j", "4"]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + + # @unittest.skip("skip") + def test_restore_ptrack(self): + """recovery to latest from archive full+ptrack+ptrack backups""" + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + ptrack_enable=True) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + node.pgbench_init(scale=2) + + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + self.backup_node(backup_dir, 'node', node, backup_type="ptrack") + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type="ptrack") + + before = node.table_checksum("pgbench_branches") + + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=["-j", "4"]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + + # @unittest.skip("skip") + def test_restore_full_ptrack_stream(self): + """recovery in stream mode to latest from full + ptrack backups""" + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + node.pgbench_init(scale=2) + + self.backup_node(backup_dir, 'node', node, options=["--stream"]) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + backup_id = self.backup_node( + backup_dir, 'node', node, + backup_type="ptrack", options=["--stream"]) + + before = node.table_checksum("pgbench_branches") + + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + after = node.table_checksum("pgbench_branches") + self.assertEqual(before, after) + + # @unittest.skip("skip") + def test_restore_full_ptrack_under_load(self): + """ + recovery to latest from full + ptrack backups + with loads when ptrack backup do + """ + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + node.pgbench_init(scale=2) + + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "8"] + ) + + backup_id = self.backup_node( + backup_dir, 'node', node, + backup_type="ptrack", options=["--stream"]) + + pgbench.wait() + pgbench.stdout.close() + + bbalance = node.execute( + "postgres", "SELECT sum(bbalance) FROM pgbench_branches") + delta = node.execute( + "postgres", "SELECT sum(delta) FROM pgbench_history") + + self.assertEqual(bbalance, delta) + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + bbalance = node.execute( + "postgres", "SELECT sum(bbalance) FROM pgbench_branches") + delta = node.execute( + "postgres", "SELECT sum(delta) FROM pgbench_history") + self.assertEqual(bbalance, delta) + + # @unittest.skip("skip") + def test_restore_full_under_load_ptrack(self): + """ + recovery to latest from full + page backups + with loads when full backup do + """ + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "CREATE EXTENSION ptrack") + + # wal_segment_size = self.guc_wal_segment_size(node) + node.pgbench_init(scale=2) + + pgbench = node.pgbench( + stdout=subprocess.PIPE, + stderr=subprocess.STDOUT, + options=["-c", "4", "-T", "8"] + ) + + self.backup_node(backup_dir, 'node', node) + + pgbench.wait() + pgbench.stdout.close() + + backup_id = self.backup_node( + backup_dir, 'node', node, + backup_type="ptrack", options=["--stream"]) + + bbalance = node.execute( + "postgres", "SELECT sum(bbalance) FROM pgbench_branches") + delta = node.execute( + "postgres", "SELECT sum(delta) FROM pgbench_history") + + self.assertEqual(bbalance, delta) + + node.stop() + node.cleanup() + # self.wrong_wal_clean(node, wal_segment_size) + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + node.slow_start() + bbalance = node.execute( + "postgres", "SELECT sum(bbalance) FROM pgbench_branches") + delta = node.execute( + "postgres", "SELECT sum(delta) FROM pgbench_history") + self.assertEqual(bbalance, delta) + + # @unittest.skip("skip") + def test_restore_with_tablespace_mapping_1(self): + """recovery using tablespace-mapping option""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Create tablespace + tblspc_path = os.path.join(node.base_dir, "tblspc") + os.makedirs(tblspc_path) + with node.connect("postgres") as con: + con.connection.autocommit = True + con.execute("CREATE TABLESPACE tblspc LOCATION '%s'" % tblspc_path) + con.connection.autocommit = False + con.execute("CREATE TABLE test (id int) TABLESPACE tblspc") + con.execute("INSERT INTO test VALUES (1)") + con.commit() + + backup_id = self.backup_node(backup_dir, 'node', node) + self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") + + # 1 - Try to restore to existing directory + node.stop() + try: + self.restore_node(backup_dir, 'node', node) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because restore destination is not empty.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Restore destination is not empty:', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # 2 - Try to restore to existing tablespace directory + tblspc_path_tmp = os.path.join(node.base_dir, "tblspc_tmp") + os.rename(tblspc_path, tblspc_path_tmp) + node.cleanup() + os.rename(tblspc_path_tmp, tblspc_path) + try: + self.restore_node(backup_dir, 'node', node) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because restore tablespace destination is " + "not empty.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Restore tablespace destination is not empty:', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # 3 - Restore using tablespace-mapping to not empty directory + tblspc_path_temp = os.path.join(node.base_dir, "tblspc_temp") + os.mkdir(tblspc_path_temp) + with open(os.path.join(tblspc_path_temp, 'file'), 'w+') as f: + f.close() + + try: + self.restore_node( + backup_dir, 'node', node, + options=["-T", "%s=%s" % (tblspc_path, tblspc_path_temp)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because restore tablespace destination is " + "not empty.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Restore tablespace destination is not empty:', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # 4 - Restore using tablespace-mapping + tblspc_path_new = os.path.join(node.base_dir, "tblspc_new") + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-T", "%s=%s" % (tblspc_path, tblspc_path_new)] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + + result = node.execute("postgres", "SELECT id FROM test") + self.assertEqual(result[0][0], 1) + + # 4 - Restore using tablespace-mapping using page backup + self.backup_node(backup_dir, 'node', node) + with node.connect("postgres") as con: + con.execute("INSERT INTO test VALUES (2)") + con.commit() + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type="page") + + show_pb = self.show_pb(backup_dir, 'node') + self.assertEqual(show_pb[1]['status'], "OK") + self.assertEqual(show_pb[2]['status'], "OK") + + node.stop() + node.cleanup() + tblspc_path_page = os.path.join(node.base_dir, "tblspc_page") + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-T", "%s=%s" % (tblspc_path_new, tblspc_path_page)]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + result = node.execute("postgres", "SELECT id FROM test OFFSET 1") + self.assertEqual(result[0][0], 2) + + # @unittest.skip("skip") + def test_restore_with_tablespace_mapping_2(self): + """recovery using tablespace-mapping option and page backup""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Full backup + self.backup_node(backup_dir, 'node', node) + self.assertEqual(self.show_pb(backup_dir, 'node')[0]['status'], "OK") + + # Create tablespace + tblspc_path = os.path.join(node.base_dir, "tblspc") + os.makedirs(tblspc_path) + with node.connect("postgres") as con: + con.connection.autocommit = True + con.execute("CREATE TABLESPACE tblspc LOCATION '%s'" % tblspc_path) + con.connection.autocommit = False + con.execute( + "CREATE TABLE tbl AS SELECT * " + "FROM generate_series(0,3) AS integer") + con.commit() + + # First page backup + self.backup_node(backup_dir, 'node', node, backup_type="page") + self.assertEqual(self.show_pb(backup_dir, 'node')[1]['status'], "OK") + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['backup-mode'], "PAGE") + + # Create tablespace table + with node.connect("postgres") as con: +# con.connection.autocommit = True +# con.execute("CHECKPOINT") +# con.connection.autocommit = False + con.execute("CREATE TABLE tbl1 (a int) TABLESPACE tblspc") + con.execute( + "INSERT INTO tbl1 SELECT * " + "FROM generate_series(0,3) AS integer") + con.commit() + + # Second page backup + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type="page") + self.assertEqual(self.show_pb(backup_dir, 'node')[2]['status'], "OK") + self.assertEqual( + self.show_pb(backup_dir, 'node')[2]['backup-mode'], "PAGE") + + node.stop() + node.cleanup() + + tblspc_path_new = os.path.join(node.base_dir, "tblspc_new") + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-T", "%s=%s" % (tblspc_path, tblspc_path_new)]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + node.slow_start() + + count = node.execute("postgres", "SELECT count(*) FROM tbl") + self.assertEqual(count[0][0], 4) + count = node.execute("postgres", "SELECT count(*) FROM tbl1") + self.assertEqual(count[0][0], 4) + + # @unittest.skip("skip") + def test_restore_with_missing_or_corrupted_tablespace_map(self): + """restore backup with missing or corrupted tablespace_map""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Create tablespace + self.create_tblspace_in_node(node, 'tblspace') + node.pgbench_init(scale=1, tablespace='tblspace') + + # Full backup + self.backup_node(backup_dir, 'node', node) + + # Change some data + pgbench = node.pgbench(options=['-T', '10', '-c', '1', '--no-vacuum']) + pgbench.wait() + + # Page backup + page_id = self.backup_node(backup_dir, 'node', node, backup_type="page") + + pgdata = self.pgdata_content(node.data_dir) + + node2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node2')) + node2.cleanup() + + olddir = self.get_tblspace_path(node, 'tblspace') + newdir = self.get_tblspace_path(node2, 'tblspace') + + # drop tablespace_map + tablespace_map = os.path.join( + backup_dir, 'backups', 'node', + page_id, 'database', 'tablespace_map') + + tablespace_map_tmp = os.path.join( + backup_dir, 'backups', 'node', + page_id, 'database', 'tablespace_map_tmp') + + os.rename(tablespace_map, tablespace_map_tmp) + + try: + self.restore_node( + backup_dir, 'node', node2, + options=["-T", "{0}={1}".format(olddir, newdir)]) + self.assertEqual( + 1, 0, + "Expecting Error because tablespace_map is missing.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Tablespace map is missing: "{0}", ' + 'probably backup {1} is corrupt, validate it'.format( + tablespace_map, page_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + try: + self.restore_node(backup_dir, 'node', node2) + self.assertEqual( + 1, 0, + "Expecting Error because tablespace_map is missing.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Tablespace map is missing: "{0}", ' + 'probably backup {1} is corrupt, validate it'.format( + tablespace_map, page_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + copyfile(tablespace_map_tmp, tablespace_map) + + with open(tablespace_map, "a") as f: + f.write("HELLO\n") + + try: + self.restore_node( + backup_dir, 'node', node2, + options=["-T", "{0}={1}".format(olddir, newdir)]) + self.assertEqual( + 1, 0, + "Expecting Error because tablespace_map is corupted.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Invalid CRC of tablespace map file "{0}"'.format(tablespace_map), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + try: + self.restore_node(backup_dir, 'node', node2) + self.assertEqual( + 1, 0, + "Expecting Error because tablespace_map is corupted.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Invalid CRC of tablespace map file "{0}"'.format(tablespace_map), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # rename it back + os.rename(tablespace_map_tmp, tablespace_map) + + print(self.restore_node( + backup_dir, 'node', node2, + options=["-T", "{0}={1}".format(olddir, newdir)])) + + pgdata_restored = self.pgdata_content(node2.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_archive_node_backup_stream_restore_to_recovery_time(self): + """ + make node with archiving, make stream backup, + make PITR to Recovery Time + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node( + backup_dir, 'node', node, options=["--stream"]) + node.safe_psql("postgres", "create table t_heap(a int)") + + node.stop() + node.cleanup() + + recovery_time = self.show_pb( + backup_dir, 'node', backup_id)['recovery-time'] + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", '--time={0}'.format(recovery_time), + "--recovery-target-action=promote" + ] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + + result = node.psql("postgres", 'select * from t_heap') + self.assertTrue('does not exist' in result[2].decode("utf-8")) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_archive_node_backup_stream_restore_to_recovery_time(self): + """ + make node with archiving, make stream backup, + make PITR to Recovery Time + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node( + backup_dir, 'node', node, options=["--stream"]) + node.safe_psql("postgres", "create table t_heap(a int)") + node.stop() + node.cleanup() + + recovery_time = self.show_pb( + backup_dir, 'node', backup_id)['recovery-time'] + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", '--time={0}'.format(recovery_time), + "--recovery-target-action=promote" + ] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + result = node.psql("postgres", 'select * from t_heap') + self.assertTrue('does not exist' in result[2].decode("utf-8")) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_archive_node_backup_stream_pitr(self): + """ + make node with archiving, make stream backup, + create table t_heap, make pitr to Recovery Time, + check that t_heap do not exists + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node( + backup_dir, 'node', node, options=["--stream"]) + node.safe_psql("postgres", "create table t_heap(a int)") + node.cleanup() + + recovery_time = self.show_pb( + backup_dir, 'node', backup_id)['recovery-time'] + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-j", "4", '--time={0}'.format(recovery_time), + "--recovery-target-action=promote" + ] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + + result = node.psql("postgres", 'select * from t_heap') + self.assertEqual(True, 'does not exist' in result[2].decode("utf-8")) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_archive_node_backup_archive_pitr_2(self): + """ + make node with archiving, make archive backup, + create table t_heap, make pitr to Recovery Time, + check that t_heap do not exists + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node(backup_dir, 'node', node) + if self.paranoia: + pgdata = self.pgdata_content(node.data_dir) + + node.safe_psql("postgres", "create table t_heap(a int)") + node.stop() + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + recovery_time = self.show_pb( + backup_dir, 'node', backup_id)['recovery-time'] + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node_restored, + options=[ + "-j", "4", '--time={0}'.format(recovery_time), + "--recovery-target-action=promote"] + ), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + if self.paranoia: + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + self.set_auto_conf(node_restored, {'port': node_restored.port}) + + node_restored.slow_start() + + result = node_restored.psql("postgres", 'select * from t_heap') + self.assertTrue('does not exist' in result[2].decode("utf-8")) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_archive_restore_to_restore_point(self): + """ + make node with archiving, make archive backup, + create table t_heap, make pitr to Recovery Time, + check that t_heap do not exists + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + "postgres", + "create table t_heap as select generate_series(0,10000)") + result = node.table_checksum("t_heap") + node.safe_psql( + "postgres", "select pg_create_restore_point('savepoint')") + node.safe_psql( + "postgres", + "create table t_heap_1 as select generate_series(0,10000)") + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, + options=[ + "--recovery-target-name=savepoint", + "--recovery-target-action=promote"]) + + node.slow_start() + + result_new = node.table_checksum("t_heap") + res = node.psql("postgres", "select * from t_heap_1") + self.assertEqual( + res[0], 1, + "Table t_heap_1 should not exist in restored instance") + + self.assertEqual(result, result_new) + + @unittest.skip("skip") + # @unittest.expectedFailure + def test_zags_block_corrupt(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + + conn = node.connect() + with node.connect("postgres") as conn: + + conn.execute( + "create table tbl(i int)") + conn.commit() + conn.execute( + "create index idx ON tbl (i)") + conn.commit() + conn.execute( + "insert into tbl select i from generate_series(0,400) as i") + conn.commit() + conn.execute( + "select pg_relation_size('idx')") + conn.commit() + conn.execute( + "delete from tbl where i < 100") + conn.commit() + conn.execute( + "explain analyze select i from tbl order by i") + conn.commit() + conn.execute( + "select i from tbl order by i") + conn.commit() + conn.execute( + "create extension pageinspect") + conn.commit() + print(conn.execute( + "select * from bt_page_stats('idx',1)")) + conn.commit() + conn.execute( + "insert into tbl select i from generate_series(0,100) as i") + conn.commit() + conn.execute( + "insert into tbl select i from generate_series(0,100) as i") + conn.commit() + conn.execute( + "insert into tbl select i from generate_series(0,100) as i") + conn.commit() + conn.execute( + "insert into tbl select i from generate_series(0,100) as i") + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored'), + initdb_params=['--data-checksums']) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored) + + self.set_auto_conf( + node_restored, + {'archive_mode': 'off', 'hot_standby': 'on', 'port': node_restored.port}) + + node_restored.slow_start() + + @unittest.skip("skip") + # @unittest.expectedFailure + def test_zags_block_corrupt_1(self): + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + pg_options={ + 'full_page_writes': 'on'} + ) + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.backup_node(backup_dir, 'node', node) + + node.safe_psql('postgres', 'create table tbl(i int)') + + node.safe_psql('postgres', 'create index idx ON tbl (i)') + + node.safe_psql( + 'postgres', + 'insert into tbl select i from generate_series(0,100000) as i') + + node.safe_psql( + 'postgres', + 'delete from tbl where i%2 = 0') + + node.safe_psql( + 'postgres', + 'explain analyze select i from tbl order by i') + + node.safe_psql( + 'postgres', + 'select i from tbl order by i') + + node.safe_psql( + 'postgres', + 'create extension pageinspect') + + node.safe_psql( + 'postgres', + 'insert into tbl select i from generate_series(0,100) as i') + + node.safe_psql( + 'postgres', + 'insert into tbl select i from generate_series(0,100) as i') + + node.safe_psql( + 'postgres', + 'insert into tbl select i from generate_series(0,100) as i') + + node.safe_psql( + 'postgres', + 'insert into tbl select i from generate_series(0,100) as i') + + self.switch_wal_segment(node) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored'), + initdb_params=['--data-checksums']) + + pgdata = self.pgdata_content(node.data_dir) + + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored) + + self.set_auto_conf( + node_restored, + {'archive_mode': 'off', 'hot_standby': 'on', 'port': node_restored.port}) + + node_restored.slow_start() + + while True: + with open(node_restored.pg_log_file, 'r') as f: + if 'selected new timeline ID' in f.read(): + break + + # with open(node_restored.pg_log_file, 'r') as f: + # print(f.read()) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + + self.compare_pgdata(pgdata, pgdata_restored) + +# pg_xlogdump_path = self.get_bin_path('pg_xlogdump') + +# pg_xlogdump = self.run_binary( +# [ +# pg_xlogdump_path, '-b', +# os.path.join(backup_dir, 'wal', 'node', '000000010000000000000003'), +# ' | ', 'grep', 'Btree', '' +# ], async=False) + + if pg_xlogdump.returncode: + self.assertFalse( + True, + 'Failed to start pg_wal_dump: {0}'.format( + pg_receivexlog.communicate()[1])) + + # @unittest.skip("skip") + def test_restore_chain(self): + """ + make node, take full backup, take several + ERROR delta backups, take valid delta backup, + restore must be successfull + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL + self.backup_node( + backup_dir, 'node', node) + + # Take DELTA + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # Take ERROR DELTA + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['-U', 'wrong_name']) + except ProbackupException as e: + pass + + # Take ERROR DELTA + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['-U', 'wrong_name']) + except ProbackupException as e: + pass + + # Take DELTA + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # Take ERROR DELTA + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['-U', 'wrong_name']) + except ProbackupException as e: + pass + + self.assertEqual( + 'OK', + self.show_pb(backup_dir, 'node')[0]['status'], + 'Backup STATUS should be "OK"') + + self.assertEqual( + 'OK', + self.show_pb(backup_dir, 'node')[1]['status'], + 'Backup STATUS should be "OK"') + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node')[2]['status'], + 'Backup STATUS should be "ERROR"') + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node')[3]['status'], + 'Backup STATUS should be "ERROR"') + + self.assertEqual( + 'OK', + self.show_pb(backup_dir, 'node')[4]['status'], + 'Backup STATUS should be "OK"') + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node')[5]['status'], + 'Backup STATUS should be "ERROR"') + + node.cleanup() + + self.restore_node(backup_dir, 'node', node) + + # @unittest.skip("skip") + def test_restore_chain_with_corrupted_backup(self): + """more complex test_restore_chain()""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL + self.backup_node( + backup_dir, 'node', node) + + # Take DELTA + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Take ERROR DELTA + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=['-U', 'wrong_name']) + except ProbackupException as e: + pass + + # Take 1 DELTA + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # Take ERROR DELTA + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['-U', 'wrong_name']) + except ProbackupException as e: + pass + + # Take 2 DELTA + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # Take ERROR DELTA + try: + self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['-U', 'wrong_name']) + except ProbackupException as e: + pass + + # Take 3 DELTA + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # Corrupted 4 DELTA + corrupt_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # ORPHAN 5 DELTA + restore_target_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # ORPHAN 6 DELTA + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # NEXT FULL BACKUP + self.backup_node( + backup_dir, 'node', node, backup_type='full') + + # Next Delta + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # do corrupt 6 DELTA backup + file = os.path.join( + backup_dir, 'backups', 'node', + corrupt_id, 'database', 'global', 'pg_control') + + file_new = os.path.join(backup_dir, 'pg_control') + os.rename(file, file_new) + + # RESTORE BACKUP + node.cleanup() + + try: + self.restore_node( + backup_dir, 'node', node, backup_id=restore_target_id) + self.assertEqual( + 1, 0, + "Expecting Error because restore backup is corrupted.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Backup {0} is orphan'.format(restore_target_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.assertEqual( + 'OK', + self.show_pb(backup_dir, 'node')[0]['status'], + 'Backup STATUS should be "OK"') + + self.assertEqual( + 'OK', + self.show_pb(backup_dir, 'node')[1]['status'], + 'Backup STATUS should be "OK"') + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node')[2]['status'], + 'Backup STATUS should be "ERROR"') + + self.assertEqual( + 'OK', + self.show_pb(backup_dir, 'node')[3]['status'], + 'Backup STATUS should be "OK"') + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node')[4]['status'], + 'Backup STATUS should be "ERROR"') + + self.assertEqual( + 'OK', + self.show_pb(backup_dir, 'node')[5]['status'], + 'Backup STATUS should be "OK"') + + self.assertEqual( + 'ERROR', + self.show_pb(backup_dir, 'node')[6]['status'], + 'Backup STATUS should be "ERROR"') + + self.assertEqual( + 'OK', + self.show_pb(backup_dir, 'node')[7]['status'], + 'Backup STATUS should be "OK"') + + # corruption victim + self.assertEqual( + 'CORRUPT', + self.show_pb(backup_dir, 'node')[8]['status'], + 'Backup STATUS should be "CORRUPT"') + + # orphaned child + self.assertEqual( + 'ORPHAN', + self.show_pb(backup_dir, 'node')[9]['status'], + 'Backup STATUS should be "ORPHAN"') + + # orphaned child + self.assertEqual( + 'ORPHAN', + self.show_pb(backup_dir, 'node')[10]['status'], + 'Backup STATUS should be "ORPHAN"') + + # next FULL + self.assertEqual( + 'OK', + self.show_pb(backup_dir, 'node')[11]['status'], + 'Backup STATUS should be "OK"') + + # next DELTA + self.assertEqual( + 'OK', + self.show_pb(backup_dir, 'node')[12]['status'], + 'Backup STATUS should be "OK"') + + node.cleanup() + + # Skipped, because backups from the future are invalid. + # This cause a "ERROR: Can't assign backup_id, there is already a backup in future" + # now (PBCKP-259). We can conduct such a test again when we + # untie 'backup_id' from 'start_time' + @unittest.skip("skip") + def test_restore_backup_from_future(self): + """more complex test_restore_chain()""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL + self.backup_node(backup_dir, 'node', node) + + node.pgbench_init(scale=5) + # pgbench = node.pgbench(options=['-T', '20', '-c', '2']) + # pgbench.wait() + + # Take PAGE from future + backup_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + with open( + os.path.join( + backup_dir, 'backups', 'node', + backup_id, "backup.control"), "a") as conf: + conf.write("start-time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() + timedelta(days=3))) + + # rename directory + new_id = self.show_pb(backup_dir, 'node')[1]['id'] + + os.rename( + os.path.join(backup_dir, 'backups', 'node', backup_id), + os.path.join(backup_dir, 'backups', 'node', new_id)) + + pgbench = node.pgbench(options=['-T', '7', '-c', '1', '--no-vacuum']) + pgbench.wait() + + backup_id = self.backup_node(backup_dir, 'node', node, backup_type='page') + pgdata = self.pgdata_content(node.data_dir) + + node.cleanup() + self.restore_node(backup_dir, 'node', node, backup_id=backup_id) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_restore_target_immediate_stream(self): + """ + correct handling of immediate recovery target + for STREAM backups + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # Take FULL + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + # Take delta + backup_id = self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + # TODO update test + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_conf = os.path.join(node.data_dir, 'postgresql.auto.conf') + with open(recovery_conf, 'r') as f: + print(f.read()) + else: + recovery_conf = os.path.join(node.data_dir, 'recovery.conf') + + # restore delta backup + node.cleanup() + self.restore_node( + backup_dir, 'node', node, options=['--immediate']) + + self.assertTrue( + os.path.isfile(recovery_conf), + "File {0} do not exists".format(recovery_conf)) + + # restore delta backup + node.cleanup() + self.restore_node( + backup_dir, 'node', node, options=['--recovery-target=immediate']) + + self.assertTrue( + os.path.isfile(recovery_conf), + "File {0} do not exists".format(recovery_conf)) + + # @unittest.skip("skip") + def test_restore_target_immediate_archive(self): + """ + correct handling of immediate recovery target + for ARCHIVE backups + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL + self.backup_node( + backup_dir, 'node', node) + + # Take delta + backup_id = self.backup_node( + backup_dir, 'node', node, + backup_type='delta') + + pgdata = self.pgdata_content(node.data_dir) + + # TODO update test + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_conf = os.path.join(node.data_dir, 'postgresql.auto.conf') + with open(recovery_conf, 'r') as f: + print(f.read()) + else: + recovery_conf = os.path.join(node.data_dir, 'recovery.conf') + + # restore page backup + node.cleanup() + self.restore_node( + backup_dir, 'node', node, options=['--immediate']) + + # For archive backup with immediate recovery target + # recovery.conf is mandatory + with open(recovery_conf, 'r') as f: + self.assertIn("recovery_target = 'immediate'", f.read()) + + # restore page backup + node.cleanup() + self.restore_node( + backup_dir, 'node', node, options=['--recovery-target=immediate']) + + # For archive backup with immediate recovery target + # recovery.conf is mandatory + with open(recovery_conf, 'r') as f: + self.assertIn("recovery_target = 'immediate'", f.read()) + + # Skipped, because default recovery_target_timeline is 'current' + # Before PBCKP-598 the --recovery-target=latest' option did not work and this test allways passed + @unittest.skip("skip") + def test_restore_target_latest_archive(self): + """ + make sure that recovery_target 'latest' + is default recovery target + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL + self.backup_node(backup_dir, 'node', node) + + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_conf = os.path.join(node.data_dir, 'postgresql.auto.conf') + else: + recovery_conf = os.path.join(node.data_dir, 'recovery.conf') + + # restore + node.cleanup() + self.restore_node(backup_dir, 'node', node) + + # hash_1 = hashlib.md5( + # open(recovery_conf, 'rb').read()).hexdigest() + + with open(recovery_conf, 'r') as f: + content_1 = '' + while True: + line = f.readline() + + if not line: + break + if line.startswith("#"): + continue + content_1 += line + + node.cleanup() + self.restore_node(backup_dir, 'node', node, options=['--recovery-target=latest']) + + # hash_2 = hashlib.md5( + # open(recovery_conf, 'rb').read()).hexdigest() + + with open(recovery_conf, 'r') as f: + content_2 = '' + while True: + line = f.readline() + + if not line: + break + if line.startswith("#"): + continue + content_2 += line + + self.assertEqual(content_1, content_2) + + # @unittest.skip("skip") + def test_restore_target_new_options(self): + """ + check that new --recovery-target-* + options are working correctly + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL + self.backup_node(backup_dir, 'node', node) + + # TODO update test + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_conf = os.path.join(node.data_dir, 'postgresql.auto.conf') + with open(recovery_conf, 'r') as f: + print(f.read()) + else: + recovery_conf = os.path.join(node.data_dir, 'recovery.conf') + + node.pgbench_init(scale=2) + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT) + pgbench.wait() + pgbench.stdout.close() + + node.safe_psql( + "postgres", + "CREATE TABLE tbl0005 (a text)") + + node.safe_psql( + "postgres", "select pg_create_restore_point('savepoint')") + + target_name = 'savepoint' + + # in python-3.6+ it can be ...now()..astimezone()... + target_time = datetime.utcnow().replace(tzinfo=timezone.utc).astimezone().strftime("%Y-%m-%d %H:%M:%S %z") + with node.connect("postgres") as con: + res = con.execute( + "INSERT INTO tbl0005 VALUES ('inserted') RETURNING (xmin)") + con.commit() + target_xid = res[0][0] + + with node.connect("postgres") as con: + con.execute("INSERT INTO tbl0005 VALUES (1)") + con.commit() + if self.get_version(node) > self.version_to_num('10.0'): + res = con.execute("SELECT pg_current_wal_lsn()") + else: + res = con.execute("SELECT pg_current_xlog_location()") + + con.commit() + con.execute("INSERT INTO tbl0005 VALUES (2)") + con.commit() + xlogid, xrecoff = res[0][0].split('/') + xrecoff = hex(int(xrecoff, 16) + 1)[2:] + target_lsn = "{0}/{1}".format(xlogid, xrecoff) + + # Restore with recovery target time + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target-time={0}'.format(target_time), + "--recovery-target-action=promote", + '--recovery-target-timeline=1', + ]) + + with open(recovery_conf, 'r') as f: + recovery_conf_content = f.read() + + self.assertIn( + "recovery_target_time = '{0}'".format(target_time), + recovery_conf_content) + + self.assertIn( + "recovery_target_action = 'promote'", + recovery_conf_content) + + self.assertIn( + "recovery_target_timeline = '1'", + recovery_conf_content) + + node.slow_start() + + # Restore with recovery target xid + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + "--recovery-target-action=promote", + '--recovery-target-timeline=1', + ]) + + with open(recovery_conf, 'r') as f: + recovery_conf_content = f.read() + + self.assertIn( + "recovery_target_xid = '{0}'".format(target_xid), + recovery_conf_content) + + self.assertIn( + "recovery_target_action = 'promote'", + recovery_conf_content) + + self.assertIn( + "recovery_target_timeline = '1'", + recovery_conf_content) + + node.slow_start() + + # Restore with recovery target name + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target-name={0}'.format(target_name), + "--recovery-target-action=promote", + '--recovery-target-timeline=1', + ]) + + with open(recovery_conf, 'r') as f: + recovery_conf_content = f.read() + + self.assertIn( + "recovery_target_name = '{0}'".format(target_name), + recovery_conf_content) + + self.assertIn( + "recovery_target_action = 'promote'", + recovery_conf_content) + + self.assertIn( + "recovery_target_timeline = '1'", + recovery_conf_content) + + node.slow_start() + + # Restore with recovery target lsn + if self.get_version(node) >= 100000: + + node.cleanup() + self.restore_node( + backup_dir, 'node', node, + options=[ + '--recovery-target-lsn={0}'.format(target_lsn), + "--recovery-target-action=promote", + '--recovery-target-timeline=1', + ]) + + with open(recovery_conf, 'r') as f: + recovery_conf_content = f.read() + + self.assertIn( + "recovery_target_lsn = '{0}'".format(target_lsn), + recovery_conf_content) + + self.assertIn( + "recovery_target_action = 'promote'", + recovery_conf_content) + + self.assertIn( + "recovery_target_timeline = '1'", + recovery_conf_content) + + node.slow_start() + + # @unittest.skip("skip") + def test_smart_restore(self): + """ + make node, create database, take full backup, drop database, + take incremental backup and restore it, + make sure that files from dropped database are not + copied during restore + https://github.com/postgrespro/pg_probackup/issues/63 + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # create database + node.safe_psql( + "postgres", + "CREATE DATABASE testdb") + + # take FULL backup + full_id = self.backup_node(backup_dir, 'node', node) + + # drop database + node.safe_psql( + "postgres", + "DROP DATABASE testdb") + + # take PAGE backup + page_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # restore PAGE backup + node.cleanup() + self.restore_node( + backup_dir, 'node', node, backup_id=page_id, + options=['--no-validate', '--log-level-file=VERBOSE']) + + logfile = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(logfile, 'r') as f: + logfile_content = f.read() + + # get delta between FULL and PAGE filelists + filelist_full = self.get_backup_filelist( + backup_dir, 'node', full_id) + + filelist_page = self.get_backup_filelist( + backup_dir, 'node', page_id) + + filelist_diff = self.get_backup_filelist_diff( + filelist_full, filelist_page) + + for file in filelist_diff: + self.assertNotIn(file, logfile_content) + + # @unittest.skip("skip") + def test_pg_11_group_access(self): + """ + test group access for PG >= 11 + """ + if self.pg_config_version < self.version_to_num('11.0'): + self.skipTest('You need PostgreSQL >= 11 for this test') + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=[ + '--data-checksums', + '--allow-group-access']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # take FULL backup + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + # restore backup + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node( + backup_dir, 'node', node_restored) + + # compare pgdata permissions + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_restore_concurrent_drop_table(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=1) + + # FULL backup + self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--compress']) + + # DELTA backup + gdb = self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--stream', '--compress', '--no-validate'], + gdb=True) + + gdb.set_breakpoint('backup_data_file') + gdb.run_until_break() + + node.safe_psql( + 'postgres', + 'DROP TABLE pgbench_accounts') + + # do checkpoint to guarantee filenode removal + node.safe_psql( + 'postgres', + 'CHECKPOINT') + + gdb.remove_all_breakpoints() + gdb.continue_execution_until_exit() + + pgdata = self.pgdata_content(node.data_dir) + node.cleanup() + + self.restore_node( + backup_dir, 'node', node, options=['--no-validate']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_lost_non_data_file(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + backup_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + file = os.path.join( + backup_dir, 'backups', 'node', + backup_id, 'database', 'postgresql.auto.conf') + + os.remove(file) + + node.cleanup() + + try: + self.restore_node( + backup_dir, 'node', node, options=['--no-validate']) + self.assertEqual( + 1, 0, + "Expecting Error because of non-data file dissapearance.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + 'No such file or directory', e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + self.assertIn( + 'ERROR: Backup files restoring failed', e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + def test_partial_restore_exclude(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + db_list_raw = node.safe_psql( + 'postgres', + 'SELECT to_json(a) ' + 'FROM (SELECT oid, datname FROM pg_database) a').decode('utf-8').rstrip() + + db_list_splitted = db_list_raw.splitlines() + + db_list = {} + for line in db_list_splitted: + line = json.loads(line) + db_list[line['datname']] = line['oid'] + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + pgdata = self.pgdata_content(node.data_dir) + + # restore FULL backup + node_restored_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored_1')) + node_restored_1.cleanup() + + try: + self.restore_node( + backup_dir, 'node', + node_restored_1, options=[ + "--db-include=db1", + "--db-exclude=db2"]) + self.assertEqual( + 1, 0, + "Expecting Error because of 'db-exclude' and 'db-include'.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: You cannot specify '--db-include' " + "and '--db-exclude' together", e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.restore_node( + backup_dir, 'node', node_restored_1) + + pgdata_restored_1 = self.pgdata_content(node_restored_1.data_dir) + self.compare_pgdata(pgdata, pgdata_restored_1) + + db1_path = os.path.join( + node_restored_1.data_dir, 'base', db_list['db1']) + db5_path = os.path.join( + node_restored_1.data_dir, 'base', db_list['db5']) + + self.truncate_every_file_in_dir(db1_path) + self.truncate_every_file_in_dir(db5_path) + pgdata_restored_1 = self.pgdata_content(node_restored_1.data_dir) + + node_restored_2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored_2')) + node_restored_2.cleanup() + + self.restore_node( + backup_dir, 'node', + node_restored_2, options=[ + "--db-exclude=db1", + "--db-exclude=db5"]) + + pgdata_restored_2 = self.pgdata_content(node_restored_2.data_dir) + self.compare_pgdata(pgdata_restored_1, pgdata_restored_2) + + self.set_auto_conf(node_restored_2, {'port': node_restored_2.port}) + + node_restored_2.slow_start() + + node_restored_2.safe_psql( + 'postgres', + 'select 1') + + try: + node_restored_2.safe_psql( + 'db1', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + try: + node_restored_2.safe_psql( + 'db5', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + with open(node_restored_2.pg_log_file, 'r') as f: + output = f.read() + + self.assertNotIn('PANIC', output) + + def test_partial_restore_exclude_tablespace(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + cat_version = node.get_control_data()["Catalog version number"] + version_specific_dir = 'PG_' + node.major_version_str + '_' + cat_version + + # PG_10_201707211 + # pg_tblspc/33172/PG_9.5_201510051/16386/ + + self.create_tblspace_in_node(node, 'somedata') + + node_tablespace = self.get_tblspace_path(node, 'somedata') + + tbl_oid = node.safe_psql( + 'postgres', + "SELECT oid " + "FROM pg_tablespace " + "WHERE spcname = 'somedata'").decode('utf-8').rstrip() + + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0} tablespace somedata'.format(i)) + + db_list_raw = node.safe_psql( + 'postgres', + 'SELECT to_json(a) ' + 'FROM (SELECT oid, datname FROM pg_database) a').decode('utf-8').rstrip() + + db_list_splitted = db_list_raw.splitlines() + + db_list = {} + for line in db_list_splitted: + line = json.loads(line) + db_list[line['datname']] = line['oid'] + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + pgdata = self.pgdata_content(node.data_dir) + + # restore FULL backup + node_restored_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored_1')) + node_restored_1.cleanup() + + node1_tablespace = self.get_tblspace_path(node_restored_1, 'somedata') + + self.restore_node( + backup_dir, 'node', + node_restored_1, options=[ + "-T", "{0}={1}".format( + node_tablespace, node1_tablespace)]) + + pgdata_restored_1 = self.pgdata_content(node_restored_1.data_dir) + self.compare_pgdata(pgdata, pgdata_restored_1) + + # truncate every db + for db in db_list: + # with exception below + if db in ['db1', 'db5']: + self.truncate_every_file_in_dir( + os.path.join( + node_restored_1.data_dir, 'pg_tblspc', + tbl_oid, version_specific_dir, db_list[db])) + + pgdata_restored_1 = self.pgdata_content(node_restored_1.data_dir) + + node_restored_2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored_2')) + node_restored_2.cleanup() + node2_tablespace = self.get_tblspace_path(node_restored_2, 'somedata') + + self.restore_node( + backup_dir, 'node', + node_restored_2, options=[ + "--db-exclude=db1", + "--db-exclude=db5", + "-T", "{0}={1}".format( + node_tablespace, node2_tablespace)]) + + pgdata_restored_2 = self.pgdata_content(node_restored_2.data_dir) + self.compare_pgdata(pgdata_restored_1, pgdata_restored_2) + + self.set_auto_conf(node_restored_2, {'port': node_restored_2.port}) + + node_restored_2.slow_start() + + node_restored_2.safe_psql( + 'postgres', + 'select 1') + + try: + node_restored_2.safe_psql( + 'db1', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + try: + node_restored_2.safe_psql( + 'db5', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + with open(node_restored_2.pg_log_file, 'r') as f: + output = f.read() + + self.assertNotIn('PANIC', output) + + def test_partial_restore_include(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + db_list_raw = node.safe_psql( + 'postgres', + 'SELECT to_json(a) ' + 'FROM (SELECT oid, datname FROM pg_database) a').decode('utf-8').rstrip() + + db_list_splitted = db_list_raw.splitlines() + + db_list = {} + for line in db_list_splitted: + line = json.loads(line) + db_list[line['datname']] = line['oid'] + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + pgdata = self.pgdata_content(node.data_dir) + + # restore FULL backup + node_restored_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored_1')) + node_restored_1.cleanup() + + try: + self.restore_node( + backup_dir, 'node', + node_restored_1, options=[ + "--db-include=db1", + "--db-exclude=db2"]) + self.assertEqual( + 1, 0, + "Expecting Error because of 'db-exclude' and 'db-include'.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: You cannot specify '--db-include' " + "and '--db-exclude' together", e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.restore_node( + backup_dir, 'node', node_restored_1) + + pgdata_restored_1 = self.pgdata_content(node_restored_1.data_dir) + self.compare_pgdata(pgdata, pgdata_restored_1) + + # truncate every db + for db in db_list: + # with exception below + if db in ['template0', 'template1', 'postgres', 'db1', 'db5']: + continue + self.truncate_every_file_in_dir( + os.path.join( + node_restored_1.data_dir, 'base', db_list[db])) + + pgdata_restored_1 = self.pgdata_content(node_restored_1.data_dir) + + node_restored_2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored_2')) + node_restored_2.cleanup() + + self.restore_node( + backup_dir, 'node', + node_restored_2, options=[ + "--db-include=db1", + "--db-include=db5", + "--db-include=postgres"]) + + pgdata_restored_2 = self.pgdata_content(node_restored_2.data_dir) + self.compare_pgdata(pgdata_restored_1, pgdata_restored_2) + + self.set_auto_conf(node_restored_2, {'port': node_restored_2.port}) + node_restored_2.slow_start() + + node_restored_2.safe_psql( + 'db1', + 'select 1') + + node_restored_2.safe_psql( + 'db5', + 'select 1') + + node_restored_2.safe_psql( + 'template1', + 'select 1') + + try: + node_restored_2.safe_psql( + 'db2', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + try: + node_restored_2.safe_psql( + 'db10', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + with open(node_restored_2.pg_log_file, 'r') as f: + output = f.read() + + self.assertNotIn('PANIC', output) + + def test_partial_restore_backward_compatibility_1(self): + """ + old binary should be of version < 2.2.0 + """ + if not self.probackup_old_path: + self.skipTest("You must specify PGPROBACKUPBIN_OLD" + " for run this test") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + + node.slow_start() + + # create databases + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + # FULL backup with old binary, without partial restore support + backup_id = self.backup_node( + backup_dir, 'node', node, + old_binary=True, options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + try: + self.restore_node( + backup_dir, 'node', + node_restored, options=[ + "--db-exclude=db5"]) + self.assertEqual( + 1, 0, + "Expecting Error because backup do not support partial restore.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Backup {0} doesn't contain a database_map, " + "partial restore is impossible".format(backup_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.restore_node(backup_dir, 'node', node_restored) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # incremental backup with partial restore support + for i in range(11, 15, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + # get db list + db_list_raw = node.safe_psql( + 'postgres', + 'SELECT to_json(a) ' + 'FROM (SELECT oid, datname FROM pg_database) a').rstrip() + db_list_splitted = db_list_raw.splitlines() + db_list = {} + for line in db_list_splitted: + line = json.loads(line) + db_list[line['datname']] = line['oid'] + + backup_id = self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['--stream']) + + # get etalon + node_restored.cleanup() + self.restore_node(backup_dir, 'node', node_restored) + self.truncate_every_file_in_dir( + os.path.join( + node_restored.data_dir, 'base', db_list['db5'])) + self.truncate_every_file_in_dir( + os.path.join( + node_restored.data_dir, 'base', db_list['db14'])) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + + # get new node + node_restored_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored_1')) + node_restored_1.cleanup() + + self.restore_node( + backup_dir, 'node', + node_restored_1, options=[ + "--db-exclude=db5", + "--db-exclude=db14"]) + + pgdata_restored_1 = self.pgdata_content(node_restored_1.data_dir) + + self.compare_pgdata(pgdata_restored, pgdata_restored_1) + + def test_partial_restore_backward_compatibility_merge(self): + """ + old binary should be of version < 2.2.0 + """ + if not self.probackup_old_path: + self.skipTest("You must specify PGPROBACKUPBIN_OLD" + " for run this test") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir, old_binary=True) + self.add_instance(backup_dir, 'node', node, old_binary=True) + + node.slow_start() + + # create databases + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + # FULL backup with old binary, without partial restore support + backup_id = self.backup_node( + backup_dir, 'node', node, + old_binary=True, options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + try: + self.restore_node( + backup_dir, 'node', + node_restored, options=[ + "--db-exclude=db5"]) + self.assertEqual( + 1, 0, + "Expecting Error because backup do not support partial restore.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Backup {0} doesn't contain a database_map, " + "partial restore is impossible.".format(backup_id), + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.restore_node(backup_dir, 'node', node_restored) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # incremental backup with partial restore support + for i in range(11, 15, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + # get db list + db_list_raw = node.safe_psql( + 'postgres', + 'SELECT to_json(a) ' + 'FROM (SELECT oid, datname FROM pg_database) a').rstrip() + db_list_splitted = db_list_raw.splitlines() + db_list = {} + for line in db_list_splitted: + line = json.loads(line) + db_list[line['datname']] = line['oid'] + + backup_id = self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['--stream']) + + # get etalon + node_restored.cleanup() + self.restore_node(backup_dir, 'node', node_restored) + self.truncate_every_file_in_dir( + os.path.join( + node_restored.data_dir, 'base', db_list['db5'])) + self.truncate_every_file_in_dir( + os.path.join( + node_restored.data_dir, 'base', db_list['db14'])) + pgdata_restored = self.pgdata_content(node_restored.data_dir) + + # get new node + node_restored_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored_1')) + node_restored_1.cleanup() + + # merge + self.merge_backup(backup_dir, 'node', backup_id=backup_id) + + self.restore_node( + backup_dir, 'node', + node_restored_1, options=[ + "--db-exclude=db5", + "--db-exclude=db14"]) + pgdata_restored_1 = self.pgdata_content(node_restored_1.data_dir) + + self.compare_pgdata(pgdata_restored, pgdata_restored_1) + + def test_empty_and_mangled_database_map(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + node.slow_start() + + # create databases + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + # FULL backup with database_map + backup_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + pgdata = self.pgdata_content(node.data_dir) + + # truncate database_map + path = os.path.join( + backup_dir, 'backups', 'node', + backup_id, 'database', 'database_map') + with open(path, "w") as f: + f.close() + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + try: + self.restore_node( + backup_dir, 'node', node_restored, + options=["--db-include=db1", '--no-validate']) + self.assertEqual( + 1, 0, + "Expecting Error because database_map is empty.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Backup {0} has empty or mangled database_map, " + "partial restore is impossible".format(backup_id), e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + try: + self.restore_node( + backup_dir, 'node', node_restored, + options=["--db-exclude=db1", '--no-validate']) + self.assertEqual( + 1, 0, + "Expecting Error because database_map is empty.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Backup {0} has empty or mangled database_map, " + "partial restore is impossible".format(backup_id), e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # mangle database_map + with open(path, "w") as f: + f.write("42") + f.close() + + try: + self.restore_node( + backup_dir, 'node', node_restored, + options=["--db-include=db1", '--no-validate']) + self.assertEqual( + 1, 0, + "Expecting Error because database_map is empty.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Field "dbOid" is not found in the line 42 of ' + 'the file backup_content.control', e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + try: + self.restore_node( + backup_dir, 'node', node_restored, + options=["--db-exclude=db1", '--no-validate']) + self.assertEqual( + 1, 0, + "Expecting Error because database_map is empty.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Field "dbOid" is not found in the line 42 of ' + 'the file backup_content.control', e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # check that simple restore is still possible + self.restore_node( + backup_dir, 'node', node_restored, options=['--no-validate']) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + def test_missing_database_map(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + node.slow_start() + + # create databases + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + node.safe_psql( + "postgres", + "CREATE DATABASE backupdb") + + # PG 9.5 + if self.get_version(node) < 90600: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;") + # PG 9.6 + elif self.get_version(node) > 90600 and self.get_version(node) < 100000: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.textout(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.timestamptz(timestamp with time zone, integer) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_xlog() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_xlog_replay_location() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + # >= 10 && < 15 + elif self.get_version(node) >= 100000 and self.get_version(node) < 150000: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_start_backup(text, boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_stop_backup(boolean, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + # >= 15 + else: + node.safe_psql( + 'backupdb', + "REVOKE ALL ON DATABASE backupdb from PUBLIC; " + "REVOKE ALL ON SCHEMA public from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA public FROM PUBLIC; " + "REVOKE ALL ON SCHEMA pg_catalog from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA pg_catalog FROM PUBLIC; " + "REVOKE ALL ON SCHEMA information_schema from PUBLIC; " + "REVOKE ALL ON ALL TABLES IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL FUNCTIONS IN SCHEMA information_schema FROM PUBLIC; " + "REVOKE ALL ON ALL SEQUENCES IN SCHEMA information_schema FROM PUBLIC; " + "CREATE ROLE backup WITH LOGIN REPLICATION; " + "GRANT CONNECT ON DATABASE backupdb to backup; " + "GRANT USAGE ON SCHEMA pg_catalog TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_proc TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_extension TO backup; " + "GRANT SELECT ON TABLE pg_catalog.pg_database TO backup; " # for partial restore, checkdb and ptrack + "GRANT EXECUTE ON FUNCTION pg_catalog.oideq(oid, oid) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.nameeq(name, name) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.current_setting(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.set_config(text, text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_is_in_recovery() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_control_system() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_start(text, boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_backup_stop(boolean) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_create_restore_point(text) TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_switch_wal() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pg_last_wal_replay_lsn() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_current_snapshot() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.txid_snapshot_xmax(txid_snapshot) TO backup;" + ) + + if self.ptrack: + # TODO why backup works without these grants ? + # 'pg_ptrack_get_pagemapset(pg_lsn)', + # 'pg_ptrack_control_lsn()', + # because PUBLIC + node.safe_psql( + "backupdb", + "CREATE SCHEMA ptrack; " + "GRANT USAGE ON SCHEMA ptrack TO backup; " + "CREATE EXTENSION ptrack WITH SCHEMA ptrack") + + if ProbackupTest.enterprise: + + node.safe_psql( + "backupdb", + "GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_version() TO backup; " + "GRANT EXECUTE ON FUNCTION pg_catalog.pgpro_edition() TO backup;") + + # FULL backup without database_map + backup_id = self.backup_node( + backup_dir, 'node', node, datname='backupdb', + options=['--stream', "-U", "backup", '--log-level-file=verbose']) + + pgdata = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + # backup has missing database_map and that is legal + try: + self.restore_node( + backup_dir, 'node', node_restored, + options=["--db-exclude=db5", "--db-exclude=db9"]) + self.assertEqual( + 1, 0, + "Expecting Error because user do not have pg_database access.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Backup {0} doesn't contain a database_map, " + "partial restore is impossible.".format( + backup_id), e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + try: + self.restore_node( + backup_dir, 'node', node_restored, + options=["--db-include=db1"]) + self.assertEqual( + 1, 0, + "Expecting Error because user do not have pg_database access.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Backup {0} doesn't contain a database_map, " + "partial restore is impossible.".format( + backup_id), e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # check that simple restore is still possible + self.restore_node(backup_dir, 'node', node_restored) + + pgdata_restored = self.pgdata_content(node_restored.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_stream_restore_command_option(self): + """ + correct handling of restore command options + when restoring STREAM backup + + 1. Restore STREAM backup with --restore-command only + parameter, check that PostgreSQL recovery uses + restore_command to obtain WAL from archive. + + 2. Restore STREAM backup wuth --restore-command + as replica, check that PostgreSQL recovery uses + restore_command to obtain WAL from archive. + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={'max_wal_size': '32MB'}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # TODO update test + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_conf = os.path.join(node.data_dir, 'postgresql.auto.conf') + with open(recovery_conf, 'r') as f: + print(f.read()) + else: + recovery_conf = os.path.join(node.data_dir, 'recovery.conf') + + # Take FULL + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + node.pgbench_init(scale=5) + + node.safe_psql( + 'postgres', + 'create table t1()') + + # restore backup + node.cleanup() + shutil.rmtree(os.path.join(node.logs_dir)) + + restore_cmd = self.get_restore_command(backup_dir, 'node', node) + + self.restore_node( + backup_dir, 'node', node, + options=[ + '--restore-command={0}'.format(restore_cmd)]) + + self.assertTrue( + os.path.isfile(recovery_conf), + "File '{0}' do not exists".format(recovery_conf)) + + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_signal = os.path.join(node.data_dir, 'recovery.signal') + self.assertTrue( + os.path.isfile(recovery_signal), + "File '{0}' do not exists".format(recovery_signal)) + + node.slow_start() + + node.safe_psql( + 'postgres', + 'select * from t1') + + timeline_id = node.safe_psql( + 'postgres', + 'select timeline_id from pg_control_checkpoint()').decode('utf-8').rstrip() + + self.assertEqual('2', timeline_id) + + # @unittest.skip("skip") + def test_restore_primary_conninfo(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # Take FULL + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + node.pgbench_init(scale=1) + + #primary_conninfo = 'host=192.168.1.50 port=5432 user=foo password=foopass' + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + str_conninfo='host=192.168.1.50 port=5432 user=foo password=foopass' + + self.restore_node( + backup_dir, 'node', replica, + options=['-R', '--primary-conninfo={0}'.format(str_conninfo)]) + + if self.get_version(node) >= self.version_to_num('12.0'): + standby_signal = os.path.join(replica.data_dir, 'standby.signal') + self.assertTrue( + os.path.isfile(standby_signal), + "File '{0}' do not exists".format(standby_signal)) + + # TODO update test + if self.get_version(node) >= self.version_to_num('12.0'): + recovery_conf = os.path.join(replica.data_dir, 'postgresql.auto.conf') + with open(recovery_conf, 'r') as f: + print(f.read()) + else: + recovery_conf = os.path.join(replica.data_dir, 'recovery.conf') + + with open(os.path.join(replica.data_dir, recovery_conf), 'r') as f: + recovery_conf_content = f.read() + + self.assertIn(str_conninfo, recovery_conf_content) + + # @unittest.skip("skip") + def test_restore_primary_slot_info(self): + """ + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # Take FULL + self.backup_node(backup_dir, 'node', node, options=['--stream']) + + node.pgbench_init(scale=1) + + replica = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'replica')) + replica.cleanup() + + node.safe_psql( + "SELECT pg_create_physical_replication_slot('master_slot')") + + self.restore_node( + backup_dir, 'node', replica, + options=['-R', '--primary-slot-name=master_slot']) + + self.set_auto_conf(replica, {'port': replica.port}) + self.set_auto_conf(replica, {'hot_standby': 'on'}) + + if self.get_version(node) >= self.version_to_num('12.0'): + standby_signal = os.path.join(replica.data_dir, 'standby.signal') + self.assertTrue( + os.path.isfile(standby_signal), + "File '{0}' do not exists".format(standby_signal)) + + replica.slow_start(replica=True) + + def test_issue_249(self): + """ + https://github.com/postgrespro/pg_probackup/issues/249 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + 'postgres', + 'CREATE database db1') + + node.pgbench_init(scale=5) + + node.safe_psql( + 'postgres', + 'CREATE TABLE t1 as SELECT * from pgbench_accounts where aid > 200000 and aid < 450000') + + node.safe_psql( + 'postgres', + 'DELETE from pgbench_accounts where aid > 200000 and aid < 450000') + + node.safe_psql( + 'postgres', + 'select * from pgbench_accounts') + + # FULL backup + self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + 'postgres', + 'INSERT INTO pgbench_accounts SELECT * FROM t1') + + # restore FULL backup + node_restored_1 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored_1')) + node_restored_1.cleanup() + + self.restore_node( + backup_dir, 'node', + node_restored_1, options=["--db-include=db1"]) + + self.set_auto_conf( + node_restored_1, + {'port': node_restored_1.port, 'hot_standby': 'off'}) + + node_restored_1.slow_start() + + node_restored_1.safe_psql( + 'db1', + 'select 1') + + try: + node_restored_1.safe_psql( + 'postgres', + 'select 1') + except QueryException as e: + self.assertIn('FATAL', e.message) + + def test_pg_12_probackup_recovery_conf_compatibility(self): + """ + https://github.com/postgrespro/pg_probackup/issues/249 + + pg_probackup version must be 12 or greater + """ + if not self.probackup_old_path: + self.skipTest("You must specify PGPROBACKUPBIN_OLD" + " for run this test") + if self.pg_config_version < self.version_to_num('12.0'): + self.skipTest('You need PostgreSQL >= 12 for this test') + + if self.version_to_num(self.old_probackup_version) >= self.version_to_num('2.4.5'): + self.assertTrue(False, 'You need pg_probackup < 2.4.5 for this test') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + self.backup_node(backup_dir, 'node', node, old_binary=True) + + node.pgbench_init(scale=5) + + node.safe_psql( + 'postgres', + 'CREATE TABLE t1 as SELECT * from pgbench_accounts where aid > 200000 and aid < 450000') + + time = node.safe_psql( + 'SELECT current_timestamp(0)::timestamptz;').decode('utf-8').rstrip() + + node.safe_psql( + 'postgres', + 'DELETE from pgbench_accounts where aid > 200000 and aid < 450000') + + node.cleanup() + + self.restore_node( + backup_dir, 'node',node, + options=[ + "--recovery-target-time={0}".format(time), + "--recovery-target-action=promote"], + old_binary=True) + + node.slow_start() + + self.backup_node(backup_dir, 'node', node, old_binary=True) + + node.pgbench_init(scale=5) + + xid = node.safe_psql( + 'SELECT txid_current()').decode('utf-8').rstrip() + node.pgbench_init(scale=1) + + node.cleanup() + + self.restore_node( + backup_dir, 'node',node, + options=[ + "--recovery-target-xid={0}".format(xid), + "--recovery-target-action=promote"]) + + node.slow_start() + + def test_drop_postgresql_auto_conf(self): + """ + https://github.com/postgrespro/pg_probackup/issues/249 + + pg_probackup version must be 12 or greater + """ + + if self.pg_config_version < self.version_to_num('12.0'): + self.skipTest('You need PostgreSQL >= 12 for this test') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + self.backup_node(backup_dir, 'node', node) + + # drop postgresql.auto.conf + auto_path = os.path.join(node.data_dir, "postgresql.auto.conf") + os.remove(auto_path) + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + node.cleanup() + + self.restore_node( + backup_dir, 'node',node, + options=[ + "--recovery-target=latest", + "--recovery-target-action=promote"]) + + node.slow_start() + + self.assertTrue(os.path.exists(auto_path)) + + def test_truncate_postgresql_auto_conf(self): + """ + https://github.com/postgrespro/pg_probackup/issues/249 + + pg_probackup version must be 12 or greater + """ + + if self.pg_config_version < self.version_to_num('12.0'): + self.skipTest('You need PostgreSQL >= 12 for this test') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + self.backup_node(backup_dir, 'node', node) + + # truncate postgresql.auto.conf + auto_path = os.path.join(node.data_dir, "postgresql.auto.conf") + with open(auto_path, "w+") as f: + f.truncate() + + self.backup_node(backup_dir, 'node', node, backup_type='page') + + node.cleanup() + + self.restore_node( + backup_dir, 'node',node, + options=[ + "--recovery-target=latest", + "--recovery-target-action=promote"]) + node.slow_start() + + self.assertTrue(os.path.exists(auto_path)) + + # @unittest.skip("skip") + def test_concurrent_restore(self): + """""" + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=1) + + # FULL backup + self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--compress']) + + pgbench = node.pgbench(options=['-T', '7', '-c', '1', '--no-vacuum']) + pgbench.wait() + + # DELTA backup + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--stream', '--compress', '--no-validate']) + + pgdata1 = self.pgdata_content(node.data_dir) + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + node.cleanup() + node_restored.cleanup() + + gdb = self.restore_node( + backup_dir, 'node', node, options=['--no-validate'], gdb=True) + + gdb.set_breakpoint('restore_data_file') + gdb.run_until_break() + + self.restore_node( + backup_dir, 'node', node_restored, options=['--no-validate']) + + gdb.remove_all_breakpoints() + gdb.continue_execution_until_exit() + + pgdata2 = self.pgdata_content(node.data_dir) + pgdata3 = self.pgdata_content(node_restored.data_dir) + + self.compare_pgdata(pgdata1, pgdata2) + self.compare_pgdata(pgdata2, pgdata3) + + # @unittest.skip("skip") + def test_restore_with_waldir(self): + """recovery using tablespace-mapping option and page backup""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + + with node.connect("postgres") as con: + con.execute( + "CREATE TABLE tbl AS SELECT * " + "FROM generate_series(0,3) AS integer") + con.commit() + + # Full backup + backup_id = self.backup_node(backup_dir, 'node', node) + + node.stop() + node.cleanup() + + # Create waldir + waldir_path = os.path.join(node.base_dir, "waldir") + os.makedirs(waldir_path) + + # Test recovery from latest + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, + options=[ + "-X", "%s" % (waldir_path)]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + node.slow_start() + + count = node.execute("postgres", "SELECT count(*) FROM tbl") + self.assertEqual(count[0][0], 4) + + # check pg_wal is symlink + if node.major_version >= 10: + wal_path=os.path.join(node.data_dir, "pg_wal") + else: + wal_path=os.path.join(node.data_dir, "pg_xlog") + + self.assertEqual(os.path.islink(wal_path), True) + + # @unittest.skip("skip") + def test_restore_to_latest_timeline(self): + """recovery to latest timeline""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + node.pgbench_init(scale=2) + + before1 = node.table_checksum("pgbench_branches") + backup_id = self.backup_node(backup_dir, 'node', node) + + node.stop() + node.cleanup() + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), + self.restore_node( + backup_dir, 'node', node, options=["-j", "4"]), + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + node.slow_start() + pgbench = node.pgbench( + stdout=subprocess.PIPE, stderr=subprocess.STDOUT, + options=['-T', '10', '-c', '2', '--no-vacuum']) + pgbench.wait() + pgbench.stdout.close() + + before2 = node.table_checksum("pgbench_branches") + self.backup_node(backup_dir, 'node', node) + + node.stop() + node.cleanup() + # restore from first backup + restore_result = self.restore_node(backup_dir, 'node', node, + options=[ + "-j", "4", "--recovery-target-timeline=latest", "-i", backup_id] + ) + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), restore_result, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + # check recovery_target_timeline option in the recovery_conf + recovery_target_timeline = self.get_recovery_conf(node)["recovery_target_timeline"] + self.assertEqual(recovery_target_timeline, "latest") + # check recovery-target=latest option for compatibility with previous versions + node.cleanup() + restore_result = self.restore_node(backup_dir, 'node', node, + options=[ + "-j", "4", "--recovery-target=latest", "-i", backup_id] + ) + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), restore_result, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + # check recovery_target_timeline option in the recovery_conf + recovery_target_timeline = self.get_recovery_conf(node)["recovery_target_timeline"] + self.assertEqual(recovery_target_timeline, "latest") + + # start postgres and promote wal files to latest timeline + node.slow_start() + + # check for the latest updates + after = node.table_checksum("pgbench_branches") + self.assertEqual(before2, after) + + # checking recovery_target_timeline=current is the default option + if self.pg_config_version >= self.version_to_num('12.0'): + node.stop() + node.cleanup() + + # restore from first backup + restore_result = self.restore_node(backup_dir, 'node', node, + options=[ + "-j", "4", "-i", backup_id] + ) + + self.assertIn( + "INFO: Restore of backup {0} completed.".format(backup_id), restore_result, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(self.output), self.cmd)) + + # check recovery_target_timeline option in the recovery_conf + recovery_target_timeline = self.get_recovery_conf(node)["recovery_target_timeline"] + self.assertEqual(recovery_target_timeline, "current") + + # start postgres with current timeline + node.slow_start() + + # check for the current updates + after = node.table_checksum("pgbench_branches") + self.assertEqual(before1, after) + + def test_restore_issue_313(self): + """ + Check that partially restored PostgreSQL instance cannot be started + """ + self._check_gdb_flag_or_skip_test + node = self.make_simple_node('node', + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + node.cleanup() + + count = 0 + filelist = self.get_backup_filelist(backup_dir, 'node', backup_id) + for file in filelist: + # count only nondata files + if int(filelist[file]['is_datafile']) == 0 and \ + not stat.S_ISDIR(int(filelist[file]['mode'])) and \ + not filelist[file]['size'] == '0' and \ + file != 'database_map': + count += 1 + + node_restored = self.make_simple_node('node_restored') + node_restored.cleanup() + self.restore_node(backup_dir, 'node', node_restored) + + gdb = self.restore_node(backup_dir, 'node', node, gdb=True, options=['--progress']) + gdb.verbose = False + gdb.set_breakpoint('restore_non_data_file') + gdb.run_until_break() + gdb.continue_execution_until_break(count - 1) + gdb.quit() + + # emulate the user or HA taking care of PG configuration + for fname in os.listdir(node_restored.data_dir): + if fname.endswith('.conf'): + os.rename( + os.path.join(node_restored.data_dir, fname), + os.path.join(node.data_dir, fname)) + + try: + node.slow_start() + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because backup is not fully restored") + except StartNodeException as e: + self.assertIn( + 'Cannot start node', + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + with open(os.path.join(node.logs_dir, 'postgresql.log'), 'r') as f: + if self.pg_config_version >= 120000: + self.assertIn( + "PANIC: could not read file \"global/pg_control\"", + f.read()) + else: + self.assertIn( + "PANIC: could not read from control file", + f.read()) diff --git a/tests/retention.py b/tests/retention.py deleted file mode 100644 index 9eb5b0206..000000000 --- a/tests/retention.py +++ /dev/null @@ -1,1321 +0,0 @@ -import os -import unittest -from datetime import datetime, timedelta -from .helpers.ptrack_helpers import ProbackupTest -from time import sleep - - -module_name = 'retention' - - -class RetentionTest(ProbackupTest, unittest.TestCase): - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_retention_redundancy_1(self): - """purge backups using redundancy-based retention policy""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - with open(os.path.join( - backup_dir, 'backups', 'node', - "pg_probackup.conf"), "a") as conf: - conf.write("retention-redundancy = 1\n") - - # Make backups to be purged - self.backup_node(backup_dir, 'node', node) - self.backup_node(backup_dir, 'node', node, backup_type="page") - # Make backups to be keeped - self.backup_node(backup_dir, 'node', node) - self.backup_node(backup_dir, 'node', node, backup_type="page") - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) - - # Purge backups - log = self.delete_expired( - backup_dir, 'node', options=['--expired', '--wal']) - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) - - # Check that WAL segments were deleted - min_wal = None - max_wal = None - for line in log.splitlines(): - if line.startswith("INFO: removed min WAL segment"): - min_wal = line[31:-1] - elif line.startswith("INFO: removed max WAL segment"): - max_wal = line[31:-1] - - if not min_wal: - self.assertTrue(False, "min_wal is empty") - - if not max_wal: - self.assertTrue(False, "max_wal is not set") - - for wal_name in os.listdir(os.path.join(backup_dir, 'wal', 'node')): - if not wal_name.endswith(".backup"): - # wal_name_b = wal_name.encode('ascii') - self.assertEqual(wal_name[8:] > min_wal[8:], True) - self.assertEqual(wal_name[8:] > max_wal[8:], True) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_retention_window_2(self): - """purge backups using window-based retention policy""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - with open( - os.path.join( - backup_dir, - 'backups', - 'node', - "pg_probackup.conf"), "a") as conf: - conf.write("retention-redundancy = 1\n") - conf.write("retention-window = 1\n") - - # Make backups to be purged - self.backup_node(backup_dir, 'node', node) - self.backup_node(backup_dir, 'node', node, backup_type="page") - # Make backup to be keeped - self.backup_node(backup_dir, 'node', node) - - backups = os.path.join(backup_dir, 'backups', 'node') - days_delta = 5 - for backup in os.listdir(backups): - if backup == 'pg_probackup.conf': - continue - with open( - os.path.join( - backups, backup, "backup.control"), "a") as conf: - conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() - timedelta(days=days_delta))) - days_delta -= 1 - - # Make backup to be keeped - self.backup_node(backup_dir, 'node', node, backup_type="page") - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) - - # Purge backups - self.delete_expired(backup_dir, 'node', options=['--expired']) - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_retention_window_3(self): - """purge all backups using window-based retention policy""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # take FULL BACKUP - backup_id_1 = self.backup_node(backup_dir, 'node', node) - - # Take second FULL BACKUP - backup_id_2 = self.backup_node(backup_dir, 'node', node) - - # Take third FULL BACKUP - backup_id_3 = self.backup_node(backup_dir, 'node', node) - - backups = os.path.join(backup_dir, 'backups', 'node') - for backup in os.listdir(backups): - if backup == 'pg_probackup.conf': - continue - with open( - os.path.join( - backups, backup, "backup.control"), "a") as conf: - conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() - timedelta(days=3))) - - # Purge backups - self.delete_expired( - backup_dir, 'node', options=['--retention-window=1', '--expired']) - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 0) - - print(self.show_pb( - backup_dir, 'node', as_json=False, as_text=True)) - - # count wal files in ARCHIVE - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_retention_window_4(self): - """purge all backups using window-based retention policy""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # take FULL BACKUPs - backup_id_1 = self.backup_node(backup_dir, 'node', node) - - backup_id_2 = self.backup_node(backup_dir, 'node', node) - - backup_id_3 = self.backup_node(backup_dir, 'node', node) - - backups = os.path.join(backup_dir, 'backups', 'node') - for backup in os.listdir(backups): - if backup == 'pg_probackup.conf': - continue - with open( - os.path.join( - backups, backup, "backup.control"), "a") as conf: - conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() - timedelta(days=3))) - - self.delete_pb(backup_dir, 'node', backup_id_2) - self.delete_pb(backup_dir, 'node', backup_id_3) - - # Purge backups - self.delete_expired( - backup_dir, 'node', - options=['--retention-window=1', '--expired', '--wal']) - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 0) - - print(self.show_pb( - backup_dir, 'node', as_json=False, as_text=True)) - - # count wal files in ARCHIVE - wals_dir = os.path.join(backup_dir, 'wal', 'node') - # n_wals = len(os.listdir(wals_dir)) - - # self.assertTrue(n_wals > 0) - - # self.delete_expired( - # backup_dir, 'node', - # options=['--retention-window=1', '--expired', '--wal']) - - # count again - n_wals = len(os.listdir(wals_dir)) - self.assertTrue(n_wals == 0) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_window_expire_interleaved_incremental_chains(self): - """complicated case of interleaved backup chains""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # take FULL BACKUPs - backup_id_a = self.backup_node(backup_dir, 'node', node) - backup_id_b = self.backup_node(backup_dir, 'node', node) - - # Change FULLb backup status to ERROR - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - # FULLb ERROR - # FULLa OK - - # Take PAGEa1 backup - page_id_a1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - # Change FULLb backup status to OK - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') - - # Change PAGEa1 and FULLa to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') - - # PAGEa1 ERROR - # FULLb OK - # FULLa ERROR - - page_id_b1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEb1 OK - # PAGEa1 ERROR - # FULLb OK - # FULLa ERROR - - # Now we start to play with first generation of PAGE backups - # Change PAGEb1 and FULLb to ERROR - self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - # Change PAGEa1 and FULLa to OK - self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') - - # PAGEb1 ERROR - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - page_id_a2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEa2 OK - # PAGEb1 ERROR - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - # Change PAGEa2 and FULLa to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a2, 'ERROR') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') - - # Change PAGEb1 and FULLb to OK - self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') - - # PAGEa2 ERROR - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa ERROR - - page_id_b2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # Change PAGEa2 and FULla to OK - self.change_backup_status(backup_dir, 'node', page_id_a2, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') - - # PAGEb2 OK - # PAGEa2 OK - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa OK - - # Purge backups - backups = os.path.join(backup_dir, 'backups', 'node') - for backup in os.listdir(backups): - if backup not in [page_id_a2, page_id_b2, 'pg_probackup.conf']: - with open( - os.path.join( - backups, backup, "backup.control"), "a") as conf: - conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() - timedelta(days=3))) - - self.delete_expired( - backup_dir, 'node', - options=['--retention-window=1', '--expired']) - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 6) - - print(self.show_pb( - backup_dir, 'node', as_json=False, as_text=True)) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_redundancy_expire_interleaved_incremental_chains(self): - """complicated case of interleaved backup chains""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # take FULL BACKUPs - backup_id_a = self.backup_node(backup_dir, 'node', node) - backup_id_b = self.backup_node(backup_dir, 'node', node) - - # Change FULL B backup status to ERROR - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - # FULLb ERROR - # FULLa OK - # Take PAGEa1 backup - page_id_a1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - # Change FULLb backup status to OK - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') - - # Change PAGEa1 and FULLa backup status to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') - - # PAGEa1 ERROR - # FULLb OK - # FULLa ERROR - - page_id_b1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEb1 OK - # PAGEa1 ERROR - # FULLb OK - # FULLa ERROR - - # Now we start to play with first generation of PAGE backups - # Change PAGEb1 status to ERROR - self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - # Change PAGEa1 status to OK - self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') - - # PAGEb1 ERROR - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - page_id_a2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEa2 OK - # PAGEb1 ERROR - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - # Change PAGEa2 and FULLa status to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a2, 'ERROR') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') - - # Change PAGEb1 and FULLb status to OK - self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') - - # PAGEa2 ERROR - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa ERROR - page_id_b2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # Change PAGEa2 and FULLa status to OK - self.change_backup_status(backup_dir, 'node', page_id_a2, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') - - # PAGEb2 OK - # PAGEa2 OK - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa OK - - self.delete_expired( - backup_dir, 'node', - options=['--retention-redundancy=1', '--expired']) - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 3) - - print(self.show_pb( - backup_dir, 'node', as_json=False, as_text=True)) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_window_merge_interleaved_incremental_chains(self): - """complicated case of interleaved backup chains""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - # Take FULL BACKUPs - backup_id_a = self.backup_node(backup_dir, 'node', node) - backup_id_b = self.backup_node(backup_dir, 'node', node) - - # Change FULLb backup status to ERROR - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - # FULLb ERROR - # FULLa OK - - # Take PAGEa1 backup - page_id_a1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - # Change FULLb to OK - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') - - # Change PAGEa1 backup status to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') - - # PAGEa1 ERROR - # FULLb OK - # FULLa OK - - page_id_b1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEb1 OK - # PAGEa1 ERROR - # FULLb OK - # FULLa OK - - # Now we start to play with first generation of PAGE backups - # Change PAGEb1 and FULLb to ERROR - self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - # Change PAGEa1 to OK - self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') - - # PAGEb1 ERROR - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - page_id_a2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEa2 OK - # PAGEb1 ERROR - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - # Change PAGEa2 and FULLa to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a2, 'ERROR') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') - - # Change PAGEb1 and FULLb to OK - self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') - - # PAGEa2 ERROR - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa ERROR - - page_id_b2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # Change PAGEa2 and FULLa to OK - self.change_backup_status(backup_dir, 'node', page_id_a2, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') - - # PAGEb2 OK - # PAGEa2 OK - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa OK - - # Purge backups - backups = os.path.join(backup_dir, 'backups', 'node') - for backup in os.listdir(backups): - if backup not in [page_id_a2, page_id_b2, 'pg_probackup.conf']: - with open( - os.path.join( - backups, backup, "backup.control"), "a") as conf: - conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() - timedelta(days=3))) - - output = self.delete_expired( - backup_dir, 'node', - options=['--retention-window=1', '--expired', '--merge-expired']) - - self.assertIn( - "Merge incremental chain between FULL backup {0} and backup {1}".format( - backup_id_a, page_id_a2), - output) - - self.assertIn( - "Merging backup {0} with backup {1}".format( - page_id_a1, backup_id_a), output) - - self.assertIn( - "Rename {0} to {1}".format( - backup_id_a, page_id_a1), output) - - self.assertIn( - "Merging backup {0} with backup {1}".format( - page_id_a2, page_id_a1), output) - - self.assertIn( - "Rename {0} to {1}".format( - page_id_a1, page_id_a2), output) - - self.assertIn( - "Merge incremental chain between FULL backup {0} and backup {1}".format( - backup_id_b, page_id_b2), - output) - - self.assertIn( - "Merging backup {0} with backup {1}".format( - page_id_b1, backup_id_b), output) - - self.assertIn( - "Rename {0} to {1}".format( - backup_id_b, page_id_b1), output) - - self.assertIn( - "Merging backup {0} with backup {1}".format( - page_id_b2, page_id_b1), output) - - self.assertIn( - "Rename {0} to {1}".format( - page_id_b1, page_id_b2), output) - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_window_merge_interleaved_incremental_chains_1(self): - """ - PAGEb3 - PAGEb2 - PAGEb1 - PAGEa1 - FULLb - FULLa - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums'], - pg_options={'autovacuum':'off'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=3) - - # Take FULL BACKUPs - backup_id_a = self.backup_node(backup_dir, 'node', node) - pgbench = node.pgbench(options=['-t', '10', '-c', '2']) - pgbench.wait() - - backup_id_b = self.backup_node(backup_dir, 'node', node) - pgbench = node.pgbench(options=['-t', '10', '-c', '2']) - pgbench.wait() - - # Change FULL B backup status to ERROR - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - page_id_a1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - pgdata_a1 = self.pgdata_content(node.data_dir) - - pgbench = node.pgbench(options=['-t', '10', '-c', '2']) - pgbench.wait() - - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - # Change FULL B backup status to OK - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') - - # Change PAGEa1 backup status to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') - - # PAGEa1 ERROR - # FULLb OK - # FULLa OK - page_id_b1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - pgbench = node.pgbench(options=['-t', '10', '-c', '2']) - pgbench.wait() - - page_id_b2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - pgbench = node.pgbench(options=['-t', '10', '-c', '2']) - pgbench.wait() - - page_id_b3 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - pgdata_b3 = self.pgdata_content(node.data_dir) - - pgbench = node.pgbench(options=['-t', '10', '-c', '2']) - pgbench.wait() - - # PAGEb3 OK - # PAGEb2 OK - # PAGEb1 OK - # PAGEa1 ERROR - # FULLb OK - # FULLa OK - - # Change PAGEa1 backup status to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') - - # PAGEb3 OK - # PAGEb2 OK - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa OK - - # Purge backups - backups = os.path.join(backup_dir, 'backups', 'node') - for backup in os.listdir(backups): - if backup in [page_id_a1, page_id_b3, 'pg_probackup.conf']: - continue - - with open( - os.path.join( - backups, backup, "backup.control"), "a") as conf: - conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() - timedelta(days=3))) - - output = self.delete_expired( - backup_dir, 'node', - options=['--retention-window=1', '--expired', '--merge-expired']) - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) - - self.assertEqual( - self.show_pb(backup_dir, 'node')[1]['id'], - page_id_b3) - - self.assertEqual( - self.show_pb(backup_dir, 'node')[0]['id'], - page_id_a1) - - self.assertEqual( - self.show_pb(backup_dir, 'node')[1]['backup-mode'], - 'FULL') - - self.assertEqual( - self.show_pb(backup_dir, 'node')[0]['backup-mode'], - 'FULL') - - node.cleanup() - - # Data correctness of PAGEa3 - self.restore_node(backup_dir, 'node', node, backup_id=page_id_a1) - pgdata_restored_a1 = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata_a1, pgdata_restored_a1) - - node.cleanup() - - # Data correctness of PAGEb3 - self.restore_node(backup_dir, 'node', node, backup_id=page_id_b3) - pgdata_restored_b3 = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata_b3, pgdata_restored_b3) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_basic_window_merge_multiple_descendants(self): - """ - PAGEb3 - | PAGEa3 - -----------------------------retention window - PAGEb2 / - | PAGEa2 / should be deleted - PAGEb1 \ / - | PAGEa1 - FULLb | - FULLa - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=3) - - # Take FULL BACKUPs - backup_id_a = self.backup_node(backup_dir, 'node', node) - # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) - # pgbench.wait() - - backup_id_b = self.backup_node(backup_dir, 'node', node) - # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) - # pgbench.wait() - - # Change FULLb backup status to ERROR - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - page_id_a1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) - # pgbench.wait() - - # Change FULLb to OK - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') - - # Change PAGEa1 to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') - - # PAGEa1 ERROR - # FULLb OK - # FULLa OK - - page_id_b1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEb1 OK - # PAGEa1 ERROR - # FULLb OK - # FULLa OK - - # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) - # pgbench.wait() - - # Change PAGEa1 to OK - self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') - - # Change PAGEb1 and FULLb to ERROR - self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - # PAGEb1 ERROR - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - page_id_a2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) - # pgbench.wait() - - # PAGEa2 OK - # PAGEb1 ERROR - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - # Change PAGEb1 and FULLb to OK - self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') - - # Change PAGEa2 and FULLa to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a2, 'ERROR') - self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') - - # PAGEa2 ERROR - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa ERROR - - page_id_b2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) - # pgbench.wait() - - # PAGEb2 OK - # PAGEa2 ERROR - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa ERROR - - # Change PAGEb2 and PAGEb1 to ERROR - self.change_backup_status(backup_dir, 'node', page_id_b2, 'ERROR') - self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') - - # and FULL stuff - self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - # PAGEb2 ERROR - # PAGEa2 ERROR - # PAGEb1 ERROR - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - page_id_a3 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) - # pgbench.wait() - - # PAGEa3 OK - # PAGEb2 ERROR - # PAGEa2 ERROR - # PAGEb1 ERROR - # PAGEa1 OK - # FULLb ERROR - # FULLa OK - - # Change PAGEa3 to ERROR - self.change_backup_status(backup_dir, 'node', page_id_a3, 'ERROR') - - # Change PAGEb2, PAGEb1 and FULLb to OK - self.change_backup_status(backup_dir, 'node', page_id_b2, 'OK') - self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') - self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') - - page_id_b3 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # PAGEb3 OK - # PAGEa3 ERROR - # PAGEb2 OK - # PAGEa2 ERROR - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa OK - - # Change PAGEa3, PAGEa2 and PAGEb1 status to OK - self.change_backup_status(backup_dir, 'node', page_id_a3, 'OK') - self.change_backup_status(backup_dir, 'node', page_id_a2, 'OK') - self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') - - # PAGEb3 OK - # PAGEa3 OK - # PAGEb2 OK - # PAGEa2 OK - # PAGEb1 OK - # PAGEa1 OK - # FULLb OK - # FULLa OK - - # Check that page_id_a3 and page_id_a2 are both direct descendants of page_id_a1 - self.assertEqual( - self.show_pb( - backup_dir, 'node', backup_id=page_id_a3)['parent-backup-id'], - page_id_a1) - - self.assertEqual( - self.show_pb( - backup_dir, 'node', backup_id=page_id_a2)['parent-backup-id'], - page_id_a1) - - # Purge backups - backups = os.path.join(backup_dir, 'backups', 'node') - for backup in os.listdir(backups): - if backup in [page_id_a3, page_id_b3, 'pg_probackup.conf']: - continue - - with open( - os.path.join( - backups, backup, "backup.control"), "a") as conf: - conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() - timedelta(days=3))) - - output = self.delete_expired( - backup_dir, 'node', - options=[ - '--retention-window=1', '--expired', - '--merge-expired', '--log-level-console=log']) - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 3) - - # Merging chain A - self.assertIn( - "Merge incremental chain between FULL backup {0} and backup {1}".format( - backup_id_a, page_id_a3), - output) - - self.assertIn( - "Merging backup {0} with backup {1}".format( - page_id_a1, backup_id_a), output) - - self.assertIn( - "INFO: Rename {0} to {1}".format( - backup_id_a, page_id_a1), output) - - self.assertIn( - "WARNING: Backup {0} has multiple valid descendants. " - "Automatic merge is not possible.".format( - page_id_a1), output) - - # Merge chain B - self.assertIn( - "Merge incremental chain between FULL backup {0} and backup {1}".format( - backup_id_b, page_id_b3), - output) - - self.assertIn( - "Merging backup {0} with backup {1}".format( - page_id_b1, backup_id_b), output) - - self.assertIn( - "INFO: Rename {0} to {1}".format( - backup_id_b, page_id_b1), output) - - self.assertIn( - "Merging backup {0} with backup {1}".format( - page_id_b2, page_id_b1), output) - - self.assertIn( - "INFO: Rename {0} to {1}".format( - page_id_b1, page_id_b2), output) - - self.assertIn( - "Merging backup {0} with backup {1}".format( - page_id_b3, page_id_b2), output) - - self.assertIn( - "INFO: Rename {0} to {1}".format( - page_id_b2, page_id_b3), output) - - # this backup deleted because it is not guarded by retention - self.assertIn( - "INFO: Delete: {0}".format( - page_id_a1), output) - - self.assertEqual( - self.show_pb(backup_dir, 'node')[2]['id'], - page_id_b3) - - self.assertEqual( - self.show_pb(backup_dir, 'node')[1]['id'], - page_id_a3) - - self.assertEqual( - self.show_pb(backup_dir, 'node')[0]['id'], - page_id_a1) - - self.assertEqual( - self.show_pb(backup_dir, 'node')[2]['backup-mode'], - 'FULL') - - self.assertEqual( - self.show_pb(backup_dir, 'node')[1]['backup-mode'], - 'PAGE') - - self.assertEqual( - self.show_pb(backup_dir, 'node')[0]['backup-mode'], - 'FULL') - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_window_chains(self): - """ - PAGE - -------window - PAGE - PAGE - FULL - PAGE - PAGE - FULL - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums'], - pg_options={'autovacuum': 'off'}) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=3) - - # Chain A - backup_id_a = self.backup_node(backup_dir, 'node', node) - page_id_a1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - page_id_a2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # Chain B - backup_id_b = self.backup_node(backup_dir, 'node', node) - - pgbench = node.pgbench(options=['-T', '10', '-c', '2']) - pgbench.wait() - - page_id_b1 = self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - pgbench = node.pgbench(options=['-T', '10', '-c', '2']) - pgbench.wait() - - page_id_b2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - pgbench = node.pgbench(options=['-T', '10', '-c', '2']) - pgbench.wait() - - page_id_b3 = self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - pgdata = self.pgdata_content(node.data_dir) - - # Purge backups - backups = os.path.join(backup_dir, 'backups', 'node') - for backup in os.listdir(backups): - if backup in [page_id_b3, 'pg_probackup.conf']: - continue - - with open( - os.path.join( - backups, backup, "backup.control"), "a") as conf: - conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() - timedelta(days=3))) - - output = self.delete_expired( - backup_dir, 'node', - options=[ - '--retention-window=1', '--expired', - '--merge-expired', '--log-level-console=log']) - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 1) - - node.cleanup() - - self.restore_node(backup_dir, 'node', node) - - pgdata_restored = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_window_chains_1(self): - """ - PAGE - -------window - PAGE - PAGE - FULL - PAGE - PAGE - FULL - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=3) - - # Chain A - backup_id_a = self.backup_node(backup_dir, 'node', node) - page_id_a1 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - page_id_a2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # Chain B - backup_id_b = self.backup_node(backup_dir, 'node', node) - - page_id_b1 = self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - page_id_b2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - page_id_b3 = self.backup_node( - backup_dir, 'node', node, backup_type='delta') - - pgdata = self.pgdata_content(node.data_dir) - - # Purge backups - backups = os.path.join(backup_dir, 'backups', 'node') - for backup in os.listdir(backups): - if backup in [page_id_b3, 'pg_probackup.conf']: - continue - - with open( - os.path.join( - backups, backup, "backup.control"), "a") as conf: - conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( - datetime.now() - timedelta(days=3))) - - output = self.delete_expired( - backup_dir, 'node', - options=[ - '--retention-window=1', - '--merge-expired', '--log-level-console=log']) - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) - - self.assertIn( - "There are no backups to delete by retention policy", - output) - - self.assertIn( - "Retention merging finished", - output) - - output = self.delete_expired( - backup_dir, 'node', - options=[ - '--retention-window=1', - '--expired', '--log-level-console=log']) - - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 1) - - self.assertIn( - "There are no backups to merge by retention policy", - output) - - self.assertIn( - "Purging finished", - output) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - @unittest.skip("skip") - def test_window_error_backups(self): - """ - PAGE ERROR - -------window - PAGE ERROR - PAGE ERROR - PAGE ERROR - FULL ERROR - FULL - -------redundancy - """ - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - node.pgbench_init(scale=3) - - # Take FULL BACKUPs - backup_id_a1 = self.backup_node(backup_dir, 'node', node) - page_id_a2 = self.backup_node( - backup_dir, 'node', node, backup_type='page') - - # Change FULLb backup status to ERROR - self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') - - def test_retention_redundancy_overlapping_chains(self): - """""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - if self.get_version(node) < 90600: - self.del_test_dir(module_name, fname) - return unittest.skip('Skipped because ptrack support is disabled') - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.set_config( - backup_dir, 'node', options=['--retention-redundancy=1']) - - # Make backups to be purged - self.backup_node(backup_dir, 'node', node) - self.backup_node(backup_dir, 'node', node, backup_type="page") - - # Make backups to be keeped - gdb = self.backup_node(backup_dir, 'node', node, gdb=True) - gdb.set_breakpoint('backup_files') - gdb.run_until_break() - - sleep(1) - - self.backup_node(backup_dir, 'node', node, backup_type="page") - - gdb.remove_all_breakpoints() - gdb.continue_execution_until_exit() - - self.backup_node(backup_dir, 'node', node, backup_type="page") - - # Purge backups - log = self.delete_expired( - backup_dir, 'node', options=['--expired', '--wal']) - self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/retention_test.py b/tests/retention_test.py new file mode 100644 index 000000000..88432a00f --- /dev/null +++ b/tests/retention_test.py @@ -0,0 +1,2529 @@ +import os +import unittest +from datetime import datetime, timedelta +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +from time import sleep +from distutils.dir_util import copy_tree + + +class RetentionTest(ProbackupTest, unittest.TestCase): + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_retention_redundancy_1(self): + """purge backups using redundancy-based retention policy""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.set_config( + backup_dir, 'node', options=['--retention-redundancy=1']) + + # Make backups to be purged + self.backup_node(backup_dir, 'node', node) + self.backup_node(backup_dir, 'node', node, backup_type="page") + # Make backups to be keeped + self.backup_node(backup_dir, 'node', node) + self.backup_node(backup_dir, 'node', node, backup_type="page") + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) + + output_before = self.show_archive(backup_dir, 'node', tli=1) + + # Purge backups + self.delete_expired( + backup_dir, 'node', options=['--expired', '--wal']) + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) + + output_after = self.show_archive(backup_dir, 'node', tli=1) + + self.assertEqual( + output_before['max-segno'], + output_after['max-segno']) + + self.assertNotEqual( + output_before['min-segno'], + output_after['min-segno']) + + # Check that WAL segments were deleted + min_wal = output_after['min-segno'] + max_wal = output_after['max-segno'] + + for wal_name in os.listdir(os.path.join(backup_dir, 'wal', 'node')): + if not wal_name.endswith(".backup"): + + if self.archive_compress: + wal_name = wal_name[-27:] + wal_name = wal_name[:-3] + else: + wal_name = wal_name[-24:] + + self.assertTrue(wal_name >= min_wal) + self.assertTrue(wal_name <= max_wal) + + # @unittest.skip("skip") + def test_retention_window_2(self): + """purge backups using window-based retention policy""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + with open( + os.path.join( + backup_dir, + 'backups', + 'node', + "pg_probackup.conf"), "a") as conf: + conf.write("retention-redundancy = 1\n") + conf.write("retention-window = 1\n") + + # Make backups to be purged + self.backup_node(backup_dir, 'node', node) + self.backup_node(backup_dir, 'node', node, backup_type="page") + # Make backup to be keeped + self.backup_node(backup_dir, 'node', node) + + backups = os.path.join(backup_dir, 'backups', 'node') + days_delta = 5 + for backup in os.listdir(backups): + if backup == 'pg_probackup.conf': + continue + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=days_delta))) + days_delta -= 1 + + # Make backup to be keeped + self.backup_node(backup_dir, 'node', node, backup_type="page") + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) + + # Purge backups + self.delete_expired(backup_dir, 'node', options=['--expired']) + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) + + # @unittest.skip("skip") + def test_retention_window_3(self): + """purge all backups using window-based retention policy""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # take FULL BACKUP + self.backup_node(backup_dir, 'node', node) + + # Take second FULL BACKUP + self.backup_node(backup_dir, 'node', node) + + # Take third FULL BACKUP + self.backup_node(backup_dir, 'node', node) + + backups = os.path.join(backup_dir, 'backups', 'node') + for backup in os.listdir(backups): + if backup == 'pg_probackup.conf': + continue + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + # Purge backups + self.delete_expired( + backup_dir, 'node', options=['--retention-window=1', '--expired']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 0) + + print(self.show_pb( + backup_dir, 'node', as_json=False, as_text=True)) + + # count wal files in ARCHIVE + + # @unittest.skip("skip") + def test_retention_window_4(self): + """purge all backups using window-based retention policy""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # take FULL BACKUPs + self.backup_node(backup_dir, 'node', node) + + backup_id_2 = self.backup_node(backup_dir, 'node', node) + + backup_id_3 = self.backup_node(backup_dir, 'node', node) + + backups = os.path.join(backup_dir, 'backups', 'node') + for backup in os.listdir(backups): + if backup == 'pg_probackup.conf': + continue + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + self.delete_pb(backup_dir, 'node', backup_id_2) + self.delete_pb(backup_dir, 'node', backup_id_3) + + # Purge backups + self.delete_expired( + backup_dir, 'node', + options=['--retention-window=1', '--expired', '--wal']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 0) + + print(self.show_pb( + backup_dir, 'node', as_json=False, as_text=True)) + + # count wal files in ARCHIVE + wals_dir = os.path.join(backup_dir, 'wal', 'node') + # n_wals = len(os.listdir(wals_dir)) + + # self.assertTrue(n_wals > 0) + + # self.delete_expired( + # backup_dir, 'node', + # options=['--retention-window=1', '--expired', '--wal']) + + # count again + n_wals = len(os.listdir(wals_dir)) + self.assertTrue(n_wals == 0) + + # @unittest.skip("skip") + def test_window_expire_interleaved_incremental_chains(self): + """complicated case of interleaved backup chains""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # take FULL BACKUPs + backup_id_a = self.backup_node(backup_dir, 'node', node) + backup_id_b = self.backup_node(backup_dir, 'node', node) + + # Change FULLb backup status to ERROR + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # FULLb ERROR + # FULLa OK + + # Take PAGEa1 backup + page_id_a1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + # Change FULLb backup status to OK + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # Change PAGEa1 and FULLa to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') + + # PAGEa1 ERROR + # FULLb OK + # FULLa ERROR + + page_id_b1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEb1 OK + # PAGEa1 ERROR + # FULLb OK + # FULLa ERROR + + # Now we start to play with first generation of PAGE backups + # Change PAGEb1 and FULLb to ERROR + self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # Change PAGEa1 and FULLa to OK + self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') + + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + page_id_a2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEa2 OK + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + # Change PAGEa2 and FULLa to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a2, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') + + # Change PAGEb1 and FULLb to OK + self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # PAGEa2 ERROR + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa ERROR + + page_id_b2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Change PAGEa2 and FULla to OK + self.change_backup_status(backup_dir, 'node', page_id_a2, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') + + # PAGEb2 OK + # PAGEa2 OK + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa OK + + # Purge backups + backups = os.path.join(backup_dir, 'backups', 'node') + for backup in os.listdir(backups): + if backup not in [page_id_a2, page_id_b2, 'pg_probackup.conf']: + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + self.delete_expired( + backup_dir, 'node', + options=['--retention-window=1', '--expired']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 6) + + print(self.show_pb( + backup_dir, 'node', as_json=False, as_text=True)) + + # @unittest.skip("skip") + def test_redundancy_expire_interleaved_incremental_chains(self): + """complicated case of interleaved backup chains""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # take FULL BACKUPs + backup_id_a = self.backup_node(backup_dir, 'node', node) + backup_id_b = self.backup_node(backup_dir, 'node', node) + + # Change FULL B backup status to ERROR + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # FULLb ERROR + # FULLa OK + # Take PAGEa1 backup + page_id_a1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + # Change FULLb backup status to OK + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # Change PAGEa1 and FULLa backup status to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') + + # PAGEa1 ERROR + # FULLb OK + # FULLa ERROR + + page_id_b1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEb1 OK + # PAGEa1 ERROR + # FULLb OK + # FULLa ERROR + + # Now we start to play with first generation of PAGE backups + # Change PAGEb1 status to ERROR + self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # Change PAGEa1 status to OK + self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') + + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + page_id_a2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEa2 OK + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + # Change PAGEa2 and FULLa status to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a2, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') + + # Change PAGEb1 and FULLb status to OK + self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # PAGEa2 ERROR + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa ERROR + self.backup_node(backup_dir, 'node', node, backup_type='page') + + # Change PAGEa2 and FULLa status to OK + self.change_backup_status(backup_dir, 'node', page_id_a2, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') + + # PAGEb2 OK + # PAGEa2 OK + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa OK + + self.delete_expired( + backup_dir, 'node', + options=['--retention-redundancy=1', '--expired']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 3) + + print(self.show_pb( + backup_dir, 'node', as_json=False, as_text=True)) + + # @unittest.skip("skip") + def test_window_merge_interleaved_incremental_chains(self): + """complicated case of interleaved backup chains""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL BACKUPs + backup_id_a = self.backup_node(backup_dir, 'node', node) + backup_id_b = self.backup_node(backup_dir, 'node', node) + + # Change FULLb backup status to ERROR + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # FULLb ERROR + # FULLa OK + + # Take PAGEa1 backup + page_id_a1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + # Change FULLb to OK + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # Change PAGEa1 backup status to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') + + # PAGEa1 ERROR + # FULLb OK + # FULLa OK + + page_id_b1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEb1 OK + # PAGEa1 ERROR + # FULLb OK + # FULLa OK + + # Now we start to play with first generation of PAGE backups + # Change PAGEb1 and FULLb to ERROR + self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # Change PAGEa1 to OK + self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') + + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + page_id_a2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEa2 OK + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + # Change PAGEa2 and FULLa to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a2, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') + + # Change PAGEb1 and FULLb to OK + self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # PAGEa2 ERROR + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa ERROR + + page_id_b2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Change PAGEa2 and FULLa to OK + self.change_backup_status(backup_dir, 'node', page_id_a2, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') + + # PAGEb2 OK + # PAGEa2 OK + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa OK + + # Purge backups + backups = os.path.join(backup_dir, 'backups', 'node') + for backup in os.listdir(backups): + if backup not in [page_id_a2, page_id_b2, 'pg_probackup.conf']: + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + output = self.delete_expired( + backup_dir, 'node', + options=['--retention-window=1', '--expired', '--merge-expired']) + + self.assertIn( + "Merge incremental chain between full backup {0} and backup {1}".format( + backup_id_a, page_id_a2), + output) + + self.assertIn( + "Rename merged full backup {0} to {1}".format( + backup_id_a, page_id_a2), output) + + self.assertIn( + "Merge incremental chain between full backup {0} and backup {1}".format( + backup_id_b, page_id_b2), + output) + + self.assertIn( + "Rename merged full backup {0} to {1}".format( + backup_id_b, page_id_b2), output) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) + + # @unittest.skip("skip") + def test_window_merge_interleaved_incremental_chains_1(self): + """ + PAGEb3 + PAGEb2 + PAGEb1 + PAGEa1 + FULLb + FULLa + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=5) + + # Take FULL BACKUPs + self.backup_node(backup_dir, 'node', node) + pgbench = node.pgbench(options=['-t', '20', '-c', '1']) + pgbench.wait() + + backup_id_b = self.backup_node(backup_dir, 'node', node) + pgbench = node.pgbench(options=['-t', '20', '-c', '1']) + pgbench.wait() + + # Change FULL B backup status to ERROR + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + page_id_a1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + pgdata_a1 = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench(options=['-t', '20', '-c', '1']) + pgbench.wait() + + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + # Change FULL B backup status to OK + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # Change PAGEa1 backup status to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') + + # PAGEa1 ERROR + # FULLb OK + # FULLa OK + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench(options=['-t', '20', '-c', '1']) + pgbench.wait() + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench(options=['-t', '20', '-c', '1']) + pgbench.wait() + + page_id_b3 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + pgdata_b3 = self.pgdata_content(node.data_dir) + + pgbench = node.pgbench(options=['-t', '20', '-c', '1']) + pgbench.wait() + + # PAGEb3 OK + # PAGEb2 OK + # PAGEb1 OK + # PAGEa1 ERROR + # FULLb OK + # FULLa OK + + # Change PAGEa1 backup status to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') + + # PAGEb3 OK + # PAGEb2 OK + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa OK + + # Purge backups + backups = os.path.join(backup_dir, 'backups', 'node') + for backup in os.listdir(backups): + if backup in [page_id_a1, page_id_b3, 'pg_probackup.conf']: + continue + + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + self.delete_expired( + backup_dir, 'node', + options=['--retention-window=1', '--expired', '--merge-expired']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['id'], + page_id_b3) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['id'], + page_id_a1) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['backup-mode'], + 'FULL') + + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['backup-mode'], + 'FULL') + + node.cleanup() + + # Data correctness of PAGEa3 + self.restore_node(backup_dir, 'node', node, backup_id=page_id_a1) + pgdata_restored_a1 = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata_a1, pgdata_restored_a1) + + node.cleanup() + + # Data correctness of PAGEb3 + self.restore_node(backup_dir, 'node', node, backup_id=page_id_b3) + pgdata_restored_b3 = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata_b3, pgdata_restored_b3) + + # @unittest.skip("skip") + def test_basic_window_merge_multiple_descendants(self): + """ + PAGEb3 + | PAGEa3 + -----------------------------retention window + PAGEb2 / + | PAGEa2 / should be deleted + PAGEb1 \ / + | PAGEa1 + FULLb | + FULLa + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=3) + + # Take FULL BACKUPs + backup_id_a = self.backup_node(backup_dir, 'node', node) + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + backup_id_b = self.backup_node(backup_dir, 'node', node) + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # Change FULLb backup status to ERROR + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + page_id_a1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # Change FULLb to OK + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # Change PAGEa1 to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') + + # PAGEa1 ERROR + # FULLb OK + # FULLa OK + + page_id_b1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEb1 OK + # PAGEa1 ERROR + # FULLb OK + # FULLa OK + + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # Change PAGEa1 to OK + self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') + + # Change PAGEb1 and FULLb to ERROR + self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + page_id_a2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # PAGEa2 OK + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + # Change PAGEb1 and FULLb to OK + self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # Change PAGEa2 and FULLa to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a2, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') + + # PAGEa2 ERROR + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa ERROR + + page_id_b2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # PAGEb2 OK + # PAGEa2 ERROR + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa ERROR + + # Change PAGEb2 and PAGEb1 to ERROR + self.change_backup_status(backup_dir, 'node', page_id_b2, 'ERROR') + self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') + + # and FULL stuff + self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # PAGEb2 ERROR + # PAGEa2 ERROR + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + page_id_a3 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # PAGEa3 OK + # PAGEb2 ERROR + # PAGEa2 ERROR + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + # Change PAGEa3 to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a3, 'ERROR') + + # Change PAGEb2, PAGEb1 and FULLb to OK + self.change_backup_status(backup_dir, 'node', page_id_b2, 'OK') + self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + page_id_b3 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEb3 OK + # PAGEa3 ERROR + # PAGEb2 OK + # PAGEa2 ERROR + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa OK + + # Change PAGEa3, PAGEa2 and PAGEb1 status to OK + self.change_backup_status(backup_dir, 'node', page_id_a3, 'OK') + self.change_backup_status(backup_dir, 'node', page_id_a2, 'OK') + self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') + + # PAGEb3 OK + # PAGEa3 OK + # PAGEb2 OK + # PAGEa2 OK + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa OK + + # Check that page_id_a3 and page_id_a2 are both direct descendants of page_id_a1 + self.assertEqual( + self.show_pb( + backup_dir, 'node', backup_id=page_id_a3)['parent-backup-id'], + page_id_a1) + + self.assertEqual( + self.show_pb( + backup_dir, 'node', backup_id=page_id_a2)['parent-backup-id'], + page_id_a1) + + # Purge backups + backups = os.path.join(backup_dir, 'backups', 'node') + for backup in os.listdir(backups): + if backup in [page_id_a3, page_id_b3, 'pg_probackup.conf']: + continue + + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + output = self.delete_expired( + backup_dir, 'node', + options=[ + '--retention-window=1', '--delete-expired', + '--merge-expired', '--log-level-console=log']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) + + # Merging chain A + self.assertIn( + "Merge incremental chain between full backup {0} and backup {1}".format( + backup_id_a, page_id_a3), + output) + + self.assertIn( + "INFO: Rename merged full backup {0} to {1}".format( + backup_id_a, page_id_a3), output) + +# self.assertIn( +# "WARNING: Backup {0} has multiple valid descendants. " +# "Automatic merge is not possible.".format( +# page_id_a1), output) + + self.assertIn( + "LOG: Consider backup {0} for purge".format( + page_id_a2), output) + + # Merge chain B + self.assertIn( + "Merge incremental chain between full backup {0} and backup {1}".format( + backup_id_b, page_id_b3), + output) + + self.assertIn( + "INFO: Rename merged full backup {0} to {1}".format( + backup_id_b, page_id_b3), output) + + self.assertIn( + "Delete: {0}".format(page_id_a2), output) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['id'], + page_id_b3) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['id'], + page_id_a3) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['backup-mode'], + 'FULL') + + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['backup-mode'], + 'FULL') + + # @unittest.skip("skip") + def test_basic_window_merge_multiple_descendants_1(self): + """ + PAGEb3 + | PAGEa3 + -----------------------------retention window + PAGEb2 / + | PAGEa2 / + PAGEb1 \ / + | PAGEa1 + FULLb | + FULLa + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=3) + + # Take FULL BACKUPs + backup_id_a = self.backup_node(backup_dir, 'node', node) + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + backup_id_b = self.backup_node(backup_dir, 'node', node) + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # Change FULLb backup status to ERROR + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + page_id_a1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # Change FULLb to OK + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # Change PAGEa1 to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a1, 'ERROR') + + # PAGEa1 ERROR + # FULLb OK + # FULLa OK + + page_id_b1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEb1 OK + # PAGEa1 ERROR + # FULLb OK + # FULLa OK + + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # Change PAGEa1 to OK + self.change_backup_status(backup_dir, 'node', page_id_a1, 'OK') + + # Change PAGEb1 and FULLb to ERROR + self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + page_id_a2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # PAGEa2 OK + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + # Change PAGEb1 and FULLb to OK + self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + # Change PAGEa2 and FULLa to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a2, 'ERROR') + self.change_backup_status(backup_dir, 'node', backup_id_a, 'ERROR') + + # PAGEa2 ERROR + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa ERROR + + page_id_b2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # PAGEb2 OK + # PAGEa2 ERROR + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa ERROR + + # Change PAGEb2 and PAGEb1 to ERROR + self.change_backup_status(backup_dir, 'node', page_id_b2, 'ERROR') + self.change_backup_status(backup_dir, 'node', page_id_b1, 'ERROR') + + # and FULL stuff + self.change_backup_status(backup_dir, 'node', backup_id_a, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # PAGEb2 ERROR + # PAGEa2 ERROR + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + page_id_a3 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + # pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + # pgbench.wait() + + # PAGEa3 OK + # PAGEb2 ERROR + # PAGEa2 ERROR + # PAGEb1 ERROR + # PAGEa1 OK + # FULLb ERROR + # FULLa OK + + # Change PAGEa3 to ERROR + self.change_backup_status(backup_dir, 'node', page_id_a3, 'ERROR') + + # Change PAGEb2, PAGEb1 and FULLb to OK + self.change_backup_status(backup_dir, 'node', page_id_b2, 'OK') + self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') + self.change_backup_status(backup_dir, 'node', backup_id_b, 'OK') + + page_id_b3 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # PAGEb3 OK + # PAGEa3 ERROR + # PAGEb2 OK + # PAGEa2 ERROR + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa OK + + # Change PAGEa3, PAGEa2 and PAGEb1 status to OK + self.change_backup_status(backup_dir, 'node', page_id_a3, 'OK') + self.change_backup_status(backup_dir, 'node', page_id_a2, 'OK') + self.change_backup_status(backup_dir, 'node', page_id_b1, 'OK') + + # PAGEb3 OK + # PAGEa3 OK + # PAGEb2 OK + # PAGEa2 OK + # PAGEb1 OK + # PAGEa1 OK + # FULLb OK + # FULLa OK + + # Check that page_id_a3 and page_id_a2 are both direct descendants of page_id_a1 + self.assertEqual( + self.show_pb( + backup_dir, 'node', backup_id=page_id_a3)['parent-backup-id'], + page_id_a1) + + self.assertEqual( + self.show_pb( + backup_dir, 'node', backup_id=page_id_a2)['parent-backup-id'], + page_id_a1) + + # Purge backups + backups = os.path.join(backup_dir, 'backups', 'node') + for backup in os.listdir(backups): + if backup in [page_id_a3, page_id_b3, 'pg_probackup.conf']: + continue + + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + output = self.delete_expired( + backup_dir, 'node', + options=[ + '--retention-window=1', + '--merge-expired', '--log-level-console=log']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 3) + + # Merging chain A + self.assertIn( + "Merge incremental chain between full backup {0} and backup {1}".format( + backup_id_a, page_id_a3), + output) + + self.assertIn( + "INFO: Rename merged full backup {0} to {1}".format( + backup_id_a, page_id_a3), output) + +# self.assertIn( +# "WARNING: Backup {0} has multiple valid descendants. " +# "Automatic merge is not possible.".format( +# page_id_a1), output) + + # Merge chain B + self.assertIn( + "Merge incremental chain between full backup {0} and backup {1}".format( + backup_id_b, page_id_b3), output) + + self.assertIn( + "INFO: Rename merged full backup {0} to {1}".format( + backup_id_b, page_id_b3), output) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[2]['id'], + page_id_b3) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['id'], + page_id_a3) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['id'], + page_id_a2) + + self.assertEqual( + self.show_pb(backup_dir, 'node')[2]['backup-mode'], + 'FULL') + + self.assertEqual( + self.show_pb(backup_dir, 'node')[1]['backup-mode'], + 'FULL') + + self.assertEqual( + self.show_pb(backup_dir, 'node')[0]['backup-mode'], + 'PAGE') + + output = self.delete_expired( + backup_dir, 'node', + options=[ + '--retention-window=1', + '--delete-expired', '--log-level-console=log']) + + # @unittest.skip("skip") + def test_window_chains(self): + """ + PAGE + -------window + PAGE + PAGE + FULL + PAGE + PAGE + FULL + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=3) + + # Chain A + self.backup_node(backup_dir, 'node', node) + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Chain B + self.backup_node(backup_dir, 'node', node) + + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + + page_id_b3 = self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + pgdata = self.pgdata_content(node.data_dir) + + # Purge backups + backups = os.path.join(backup_dir, 'backups', 'node') + for backup in os.listdir(backups): + if backup in [page_id_b3, 'pg_probackup.conf']: + continue + + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + self.delete_expired( + backup_dir, 'node', + options=[ + '--retention-window=1', '--expired', + '--merge-expired', '--log-level-console=log']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 1) + + node.cleanup() + + self.restore_node(backup_dir, 'node', node) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # @unittest.skip("skip") + def test_window_chains_1(self): + """ + PAGE + -------window + PAGE + PAGE + FULL + PAGE + PAGE + FULL + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=3) + + # Chain A + self.backup_node(backup_dir, 'node', node) + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Chain B + self.backup_node(backup_dir, 'node', node) + + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + page_id_b3 = self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + self.pgdata_content(node.data_dir) + + # Purge backups + backups = os.path.join(backup_dir, 'backups', 'node') + for backup in os.listdir(backups): + if backup in [page_id_b3, 'pg_probackup.conf']: + continue + + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + output = self.delete_expired( + backup_dir, 'node', + options=[ + '--retention-window=1', + '--merge-expired', '--log-level-console=log']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) + + self.assertIn( + "There are no backups to delete by retention policy", + output) + + self.assertIn( + "Retention merging finished", + output) + + output = self.delete_expired( + backup_dir, 'node', + options=[ + '--retention-window=1', + '--expired', '--log-level-console=log']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 1) + + self.assertIn( + "There are no backups to merge by retention policy", + output) + + self.assertIn( + "Purging finished", + output) + + @unittest.skip("skip") + def test_window_error_backups(self): + """ + PAGE ERROR + -------window + PAGE ERROR + PAGE ERROR + PAGE ERROR + FULL ERROR + FULL + -------redundancy + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL BACKUPs + self.backup_node(backup_dir, 'node', node) + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Change FULLb backup status to ERROR + # self.change_backup_status(backup_dir, 'node', backup_id_b, 'ERROR') + + # @unittest.skip("skip") + def test_window_error_backups_1(self): + """ + DELTA + PAGE ERROR + FULL + -------window + """ + self._check_gdb_flag_or_skip_test() + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL BACKUP + self.backup_node(backup_dir, 'node', node) + + # Take PAGE BACKUP + gdb = self.backup_node( + backup_dir, 'node', node, backup_type='page', gdb=True) + + # Attention! this breakpoint has been set on internal probackup function, not on a postgres core one + gdb.set_breakpoint('pg_stop_backup') + gdb.run_until_break() + gdb.remove_all_breakpoints() + gdb._execute('signal SIGINT') + gdb.continue_execution_until_error() + + self.show_pb(backup_dir, 'node')[1]['id'] + + # Take DELTA backup + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--retention-window=2', '--delete-expired']) + + # Take FULL BACKUP + self.backup_node(backup_dir, 'node', node) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) + + # @unittest.skip("skip") + def test_window_error_backups_2(self): + """ + DELTA + PAGE ERROR + FULL + -------window + """ + self._check_gdb_flag_or_skip_test() + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Take FULL BACKUP + self.backup_node(backup_dir, 'node', node) + + # Take PAGE BACKUP + gdb = self.backup_node( + backup_dir, 'node', node, backup_type='page', gdb=True) + + # Attention! this breakpoint has been set on internal probackup function, not on a postgres core one + gdb.set_breakpoint('pg_stop_backup') + gdb.run_until_break() + gdb._execute('signal SIGKILL') + gdb.continue_execution_until_error() + + self.show_pb(backup_dir, 'node')[1]['id'] + + if self.get_version(node) < 90600: + node.safe_psql( + 'postgres', + 'SELECT pg_catalog.pg_stop_backup()') + + # Take DELTA backup + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--retention-window=2', '--delete-expired']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 3) + + def test_retention_redundancy_overlapping_chains(self): + """""" + self._check_gdb_flag_or_skip_test() + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + if self.get_version(node) < 90600: + self.skipTest('Skipped because ptrack support is disabled') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.set_config( + backup_dir, 'node', options=['--retention-redundancy=1']) + + # Make backups to be purged + self.backup_node(backup_dir, 'node', node) + self.backup_node(backup_dir, 'node', node, backup_type="page") + + # Make backups to be keeped + gdb = self.backup_node(backup_dir, 'node', node, gdb=True) + gdb.set_breakpoint('backup_files') + gdb.run_until_break() + + sleep(1) + + self.backup_node(backup_dir, 'node', node, backup_type="page") + + gdb.remove_all_breakpoints() + gdb.continue_execution_until_exit() + + self.backup_node(backup_dir, 'node', node, backup_type="page") + + # Purge backups + self.delete_expired( + backup_dir, 'node', options=['--expired', '--wal']) + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) + + self.validate_pb(backup_dir, 'node') + + def test_retention_redundancy_overlapping_chains_1(self): + """""" + self._check_gdb_flag_or_skip_test() + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + if self.get_version(node) < 90600: + self.skipTest('Skipped because ptrack support is disabled') + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.set_config( + backup_dir, 'node', options=['--retention-redundancy=1']) + + # Make backups to be purged + self.backup_node(backup_dir, 'node', node) + self.backup_node(backup_dir, 'node', node, backup_type="page") + + # Make backups to be keeped + gdb = self.backup_node(backup_dir, 'node', node, gdb=True) + gdb.set_breakpoint('backup_files') + gdb.run_until_break() + + sleep(1) + + self.backup_node(backup_dir, 'node', node, backup_type="page") + + gdb.remove_all_breakpoints() + gdb.continue_execution_until_exit() + + self.backup_node(backup_dir, 'node', node, backup_type="page") + + # Purge backups + self.delete_expired( + backup_dir, 'node', options=['--expired', '--wal']) + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) + + self.validate_pb(backup_dir, 'node') + + def test_wal_purge_victim(self): + """ + https://github.com/postgrespro/pg_probackup/issues/103 + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Make ERROR incremental backup + try: + self.backup_node(backup_dir, 'node', node, backup_type='page') + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because page backup should not be possible " + "without valid full backup.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertTrue( + "WARNING: Valid full backup on current timeline 1 is not found" in e.message and + "ERROR: Create new full backup before an incremental one" in e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + page_id = self.show_pb(backup_dir, 'node')[0]['id'] + + sleep(1) + + # Make FULL backup + full_id = self.backup_node(backup_dir, 'node', node, options=['--delete-wal']) + + try: + self.validate_pb(backup_dir, 'node') + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because page backup should not be possible " + "without valid full backup.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "INFO: Backup {0} WAL segments are valid".format(full_id), + e.message) + self.assertIn( + "WARNING: Backup {0} has missing parent 0".format(page_id), + e.message) + + # @unittest.skip("skip") + def test_failed_merge_redundancy_retention(self): + """ + Check that retention purge works correctly with MERGING backups + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join( + self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL1 backup + full_id = self.backup_node(backup_dir, 'node', node) + + # DELTA BACKUP + delta_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # DELTA BACKUP + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # DELTA BACKUP + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # FULL2 backup + self.backup_node(backup_dir, 'node', node) + + # DELTA BACKUP + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # DELTA BACKUP + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # FULL3 backup + self.backup_node(backup_dir, 'node', node) + + # DELTA BACKUP + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + # DELTA BACKUP + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + self.set_config( + backup_dir, 'node', options=['--retention-redundancy=2']) + + self.set_config( + backup_dir, 'node', options=['--retention-window=2']) + + # create pair of MERGING backup as a result of failed merge + gdb = self.merge_backup( + backup_dir, 'node', delta_id, gdb=True) + gdb.set_breakpoint('backup_non_data_file') + gdb.run_until_break() + gdb.continue_execution_until_break(2) + gdb._execute('signal SIGKILL') + + # "expire" first full backup + backups = os.path.join(backup_dir, 'backups', 'node') + with open( + os.path.join( + backups, full_id, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + # run retention merge + self.delete_expired( + backup_dir, 'node', options=['--delete-expired']) + + self.assertEqual( + 'MERGING', + self.show_pb(backup_dir, 'node', full_id)['status'], + 'Backup STATUS should be "MERGING"') + + self.assertEqual( + 'MERGING', + self.show_pb(backup_dir, 'node', delta_id)['status'], + 'Backup STATUS should be "MERGING"') + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 10) + + def test_wal_depth_1(self): + """ + |-------------B5----------> WAL timeline3 + |-----|-------------------------> WAL timeline2 + B1 B2---| B3 B4-------B6-----> WAL timeline1 + + wal-depth=2 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={ + 'archive_timeout': '30s', + 'checkpoint_timeout': '30s'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + + self.set_config(backup_dir, 'node', options=['--archive-timeout=60s']) + + node.slow_start() + + # FULL + node.pgbench_init(scale=1) + self.backup_node(backup_dir, 'node', node) + + # PAGE + node.pgbench_init(scale=1) + B2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # generate_some more data + node.pgbench_init(scale=1) + + target_xid = node.safe_psql( + "postgres", + "select txid_current()").decode('utf-8').rstrip() + + node.pgbench_init(scale=1) + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + node.pgbench_init(scale=1) + + self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Timeline 2 + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + + node_restored.cleanup() + + output = self.restore_node( + backup_dir, 'node', node_restored, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-action=promote']) + + self.assertIn( + 'Restore of backup {0} completed'.format(B2), + output) + + self.set_auto_conf(node_restored, options={'port': node_restored.port}) + + node_restored.slow_start() + + node_restored.pgbench_init(scale=1) + + target_xid = node_restored.safe_psql( + "postgres", + "select txid_current()").decode('utf-8').rstrip() + + node_restored.pgbench_init(scale=2) + + # Timeline 3 + node_restored.cleanup() + + output = self.restore_node( + backup_dir, 'node', node_restored, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-timeline=2', + '--recovery-target-action=promote']) + + self.assertIn( + 'Restore of backup {0} completed'.format(B2), + output) + + self.set_auto_conf(node_restored, options={'port': node_restored.port}) + + node_restored.slow_start() + + node_restored.pgbench_init(scale=1) + self.backup_node( + backup_dir, 'node', node_restored, data_dir=node_restored.data_dir) + + node.pgbench_init(scale=1) + self.backup_node(backup_dir, 'node', node) + + lsn = self.show_archive(backup_dir, 'node', tli=2)['switchpoint'] + + self.validate_pb( + backup_dir, 'node', backup_id=B2, + options=['--recovery-target-lsn={0}'.format(lsn)]) + + self.validate_pb(backup_dir, 'node') + + def test_wal_purge(self): + """ + -------------------------------------> tli5 + ---------------------------B6--------> tli4 + S2`---------------> tli3 + S1`------------S2---B4-------B5--> tli2 + B1---S1-------------B2--------B3------> tli1 + + B* - backups + S* - switchpoints + + Expected result: + TLI5 will be purged entirely + B6--------> tli4 + S2`---------------> tli3 + S1`------------S2---B4-------B5--> tli2 + B1---S1-------------B2--------B3------> tli1 + + wal-depth=2 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_config(backup_dir, 'node', options=['--archive-timeout=60s']) + + node.slow_start() + + # STREAM FULL + stream_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + node.stop() + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL + B1 = self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=1) + + target_xid = node.safe_psql( + "postgres", + "select txid_current()").decode('utf-8').rstrip() + node.pgbench_init(scale=5) + + # B2 FULL on TLI1 + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=4) + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=4) + + self.delete_pb(backup_dir, 'node', options=['--delete-wal']) + + # TLI 2 + node_tli2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_tli2')) + node_tli2.cleanup() + + output = self.restore_node( + backup_dir, 'node', node_tli2, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-timeline=1', + '--recovery-target-action=promote']) + + self.assertIn( + 'INFO: Restore of backup {0} completed'.format(B1), + output) + + self.set_auto_conf(node_tli2, options={'port': node_tli2.port}) + node_tli2.slow_start() + node_tli2.pgbench_init(scale=4) + + target_xid = node_tli2.safe_psql( + "postgres", + "select txid_current()").decode('utf-8').rstrip() + node_tli2.pgbench_init(scale=1) + + self.backup_node( + backup_dir, 'node', node_tli2, data_dir=node_tli2.data_dir) + node_tli2.pgbench_init(scale=3) + + self.backup_node( + backup_dir, 'node', node_tli2, data_dir=node_tli2.data_dir) + node_tli2.pgbench_init(scale=1) + node_tli2.cleanup() + + # TLI3 + node_tli3 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_tli3')) + node_tli3.cleanup() + + # Note, that successful validation here is a happy coincidence + output = self.restore_node( + backup_dir, 'node', node_tli3, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-timeline=2', + '--recovery-target-action=promote']) + + self.assertIn( + 'INFO: Restore of backup {0} completed'.format(B1), + output) + self.set_auto_conf(node_tli3, options={'port': node_tli3.port}) + node_tli3.slow_start() + node_tli3.pgbench_init(scale=5) + node_tli3.cleanup() + + # TLI4 + node_tli4 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_tli4')) + node_tli4.cleanup() + + self.restore_node( + backup_dir, 'node', node_tli4, backup_id=stream_id, + options=[ + '--recovery-target=immediate', + '--recovery-target-action=promote']) + + self.set_auto_conf(node_tli4, options={'port': node_tli4.port}) + self.set_archiving(backup_dir, 'node', node_tli4) + node_tli4.slow_start() + + node_tli4.pgbench_init(scale=5) + + self.backup_node( + backup_dir, 'node', node_tli4, data_dir=node_tli4.data_dir) + node_tli4.pgbench_init(scale=5) + node_tli4.cleanup() + + # TLI5 + node_tli5 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_tli5')) + node_tli5.cleanup() + + self.restore_node( + backup_dir, 'node', node_tli5, backup_id=stream_id, + options=[ + '--recovery-target=immediate', + '--recovery-target-action=promote']) + + self.set_auto_conf(node_tli5, options={'port': node_tli5.port}) + self.set_archiving(backup_dir, 'node', node_tli5) + node_tli5.slow_start() + node_tli5.pgbench_init(scale=10) + + # delete '.history' file of TLI4 + os.remove(os.path.join(backup_dir, 'wal', 'node', '00000004.history')) + # delete '.history' file of TLI5 + os.remove(os.path.join(backup_dir, 'wal', 'node', '00000005.history')) + + output = self.delete_pb( + backup_dir, 'node', + options=[ + '--delete-wal', '--dry-run', + '--log-level-console=verbose']) + + self.assertIn( + 'INFO: On timeline 4 WAL segments between 000000040000000000000002 ' + 'and 000000040000000000000006 can be removed', + output) + + self.assertIn( + 'INFO: On timeline 5 all files can be removed', + output) + + show_tli1_before = self.show_archive(backup_dir, 'node', tli=1) + show_tli2_before = self.show_archive(backup_dir, 'node', tli=2) + show_tli3_before = self.show_archive(backup_dir, 'node', tli=3) + show_tli4_before = self.show_archive(backup_dir, 'node', tli=4) + show_tli5_before = self.show_archive(backup_dir, 'node', tli=5) + + self.assertTrue(show_tli1_before) + self.assertTrue(show_tli2_before) + self.assertTrue(show_tli3_before) + self.assertTrue(show_tli4_before) + self.assertTrue(show_tli5_before) + + output = self.delete_pb( + backup_dir, 'node', + options=['--delete-wal', '--log-level-console=verbose']) + + self.assertIn( + 'INFO: On timeline 4 WAL segments between 000000040000000000000002 ' + 'and 000000040000000000000006 will be removed', + output) + + self.assertIn( + 'INFO: On timeline 5 all files will be removed', + output) + + show_tli1_after = self.show_archive(backup_dir, 'node', tli=1) + show_tli2_after = self.show_archive(backup_dir, 'node', tli=2) + show_tli3_after = self.show_archive(backup_dir, 'node', tli=3) + show_tli4_after = self.show_archive(backup_dir, 'node', tli=4) + show_tli5_after = self.show_archive(backup_dir, 'node', tli=5) + + self.assertEqual(show_tli1_before, show_tli1_after) + self.assertEqual(show_tli2_before, show_tli2_after) + self.assertEqual(show_tli3_before, show_tli3_after) + self.assertNotEqual(show_tli4_before, show_tli4_after) + self.assertNotEqual(show_tli5_before, show_tli5_after) + + self.assertEqual( + show_tli4_before['min-segno'], + '000000040000000000000002') + + self.assertEqual( + show_tli4_after['min-segno'], + '000000040000000000000006') + + self.assertFalse(show_tli5_after) + + self.validate_pb(backup_dir, 'node') + + def test_wal_depth_2(self): + """ + -------------------------------------> tli5 + ---------------------------B6--------> tli4 + S2`---------------> tli3 + S1`------------S2---B4-------B5--> tli2 + B1---S1-------------B2--------B3------> tli1 + + B* - backups + S* - switchpoints + wal-depth=2 + + Expected result: + TLI5 will be purged entirely + B6--------> tli4 + S2`---------------> tli3 + S1`------------S2 B4-------B5--> tli2 + B1---S1 B2--------B3------> tli1 + + wal-depth=2 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_config(backup_dir, 'node', options=['--archive-timeout=60s']) + + node.slow_start() + + # STREAM FULL + stream_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + node.stop() + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL + B1 = self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=1) + + target_xid = node.safe_psql( + "postgres", + "select txid_current()").decode('utf-8').rstrip() + node.pgbench_init(scale=5) + + # B2 FULL on TLI1 + B2 = self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=4) + self.backup_node(backup_dir, 'node', node) + node.pgbench_init(scale=4) + + # TLI 2 + node_tli2 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_tli2')) + node_tli2.cleanup() + + output = self.restore_node( + backup_dir, 'node', node_tli2, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-timeline=1', + '--recovery-target-action=promote']) + + self.assertIn( + 'INFO: Restore of backup {0} completed'.format(B1), + output) + + self.set_auto_conf(node_tli2, options={'port': node_tli2.port}) + node_tli2.slow_start() + node_tli2.pgbench_init(scale=4) + + target_xid = node_tli2.safe_psql( + "postgres", + "select txid_current()").decode('utf-8').rstrip() + node_tli2.pgbench_init(scale=1) + + B4 = self.backup_node( + backup_dir, 'node', node_tli2, data_dir=node_tli2.data_dir) + node_tli2.pgbench_init(scale=3) + + self.backup_node( + backup_dir, 'node', node_tli2, data_dir=node_tli2.data_dir) + node_tli2.pgbench_init(scale=1) + node_tli2.cleanup() + + # TLI3 + node_tli3 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_tli3')) + node_tli3.cleanup() + + # Note, that successful validation here is a happy coincidence + output = self.restore_node( + backup_dir, 'node', node_tli3, + options=[ + '--recovery-target-xid={0}'.format(target_xid), + '--recovery-target-timeline=2', + '--recovery-target-action=promote']) + + self.assertIn( + 'INFO: Restore of backup {0} completed'.format(B1), + output) + self.set_auto_conf(node_tli3, options={'port': node_tli3.port}) + node_tli3.slow_start() + node_tli3.pgbench_init(scale=5) + node_tli3.cleanup() + + # TLI4 + node_tli4 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_tli4')) + node_tli4.cleanup() + + self.restore_node( + backup_dir, 'node', node_tli4, backup_id=stream_id, + options=[ + '--recovery-target=immediate', + '--recovery-target-action=promote']) + + self.set_auto_conf(node_tli4, options={'port': node_tli4.port}) + self.set_archiving(backup_dir, 'node', node_tli4) + node_tli4.slow_start() + + node_tli4.pgbench_init(scale=5) + + self.backup_node( + backup_dir, 'node', node_tli4, data_dir=node_tli4.data_dir) + node_tli4.pgbench_init(scale=5) + node_tli4.cleanup() + + # TLI5 + node_tli5 = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_tli5')) + node_tli5.cleanup() + + self.restore_node( + backup_dir, 'node', node_tli5, backup_id=stream_id, + options=[ + '--recovery-target=immediate', + '--recovery-target-action=promote']) + + self.set_auto_conf(node_tli5, options={'port': node_tli5.port}) + self.set_archiving(backup_dir, 'node', node_tli5) + node_tli5.slow_start() + node_tli5.pgbench_init(scale=10) + + # delete '.history' file of TLI4 + os.remove(os.path.join(backup_dir, 'wal', 'node', '00000004.history')) + # delete '.history' file of TLI5 + os.remove(os.path.join(backup_dir, 'wal', 'node', '00000005.history')) + + output = self.delete_pb( + backup_dir, 'node', + options=[ + '--delete-wal', '--dry-run', + '--wal-depth=2', '--log-level-console=verbose']) + + start_lsn_B2 = self.show_pb(backup_dir, 'node', B2)['start-lsn'] + self.assertIn( + 'On timeline 1 WAL is protected from purge at {0}'.format(start_lsn_B2), + output) + + self.assertIn( + 'LOG: Archive backup {0} to stay consistent protect from ' + 'purge WAL interval between 000000010000000000000004 ' + 'and 000000010000000000000005 on timeline 1'.format(B1), output) + + start_lsn_B4 = self.show_pb(backup_dir, 'node', B4)['start-lsn'] + self.assertIn( + 'On timeline 2 WAL is protected from purge at {0}'.format(start_lsn_B4), + output) + + self.assertIn( + 'LOG: Timeline 3 to stay reachable from timeline 1 protect ' + 'from purge WAL interval between 000000020000000000000006 and ' + '000000020000000000000009 on timeline 2', output) + + self.assertIn( + 'LOG: Timeline 3 to stay reachable from timeline 1 protect ' + 'from purge WAL interval between 000000010000000000000004 and ' + '000000010000000000000006 on timeline 1', output) + + show_tli1_before = self.show_archive(backup_dir, 'node', tli=1) + show_tli2_before = self.show_archive(backup_dir, 'node', tli=2) + show_tli3_before = self.show_archive(backup_dir, 'node', tli=3) + show_tli4_before = self.show_archive(backup_dir, 'node', tli=4) + show_tli5_before = self.show_archive(backup_dir, 'node', tli=5) + + self.assertTrue(show_tli1_before) + self.assertTrue(show_tli2_before) + self.assertTrue(show_tli3_before) + self.assertTrue(show_tli4_before) + self.assertTrue(show_tli5_before) + + sleep(5) + + output = self.delete_pb( + backup_dir, 'node', + options=['--delete-wal', '--wal-depth=2', '--log-level-console=verbose']) + +# print(output) + + show_tli1_after = self.show_archive(backup_dir, 'node', tli=1) + show_tli2_after = self.show_archive(backup_dir, 'node', tli=2) + show_tli3_after = self.show_archive(backup_dir, 'node', tli=3) + show_tli4_after = self.show_archive(backup_dir, 'node', tli=4) + show_tli5_after = self.show_archive(backup_dir, 'node', tli=5) + + self.assertNotEqual(show_tli1_before, show_tli1_after) + self.assertNotEqual(show_tli2_before, show_tli2_after) + self.assertEqual(show_tli3_before, show_tli3_after) + self.assertNotEqual(show_tli4_before, show_tli4_after) + self.assertNotEqual(show_tli5_before, show_tli5_after) + + self.assertEqual( + show_tli4_before['min-segno'], + '000000040000000000000002') + + self.assertEqual( + show_tli4_after['min-segno'], + '000000040000000000000006') + + self.assertFalse(show_tli5_after) + + self.assertTrue(show_tli1_after['lost-segments']) + self.assertTrue(show_tli2_after['lost-segments']) + self.assertFalse(show_tli3_after['lost-segments']) + self.assertFalse(show_tli4_after['lost-segments']) + self.assertFalse(show_tli5_after) + + self.assertEqual(len(show_tli1_after['lost-segments']), 1) + self.assertEqual(len(show_tli2_after['lost-segments']), 1) + + self.assertEqual( + show_tli1_after['lost-segments'][0]['begin-segno'], + '000000010000000000000007') + + self.assertEqual( + show_tli1_after['lost-segments'][0]['end-segno'], + '00000001000000000000000A') + + self.assertEqual( + show_tli2_after['lost-segments'][0]['begin-segno'], + '00000002000000000000000A') + + self.assertEqual( + show_tli2_after['lost-segments'][0]['end-segno'], + '00000002000000000000000A') + + self.validate_pb(backup_dir, 'node') + + def test_basic_wal_depth(self): + """ + B1---B1----B3-----B4----B5------> tli1 + + Expected result with wal-depth=1: + B1 B1 B3 B4 B5------> tli1 + + wal-depth=1 + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_config(backup_dir, 'node', options=['--archive-timeout=60s']) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL + node.pgbench_init(scale=1) + B1 = self.backup_node(backup_dir, 'node', node) + + + # B2 + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + B2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # B3 + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + B3 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # B4 + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + B4 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # B5 + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + B5 = self.backup_node( + backup_dir, 'node', node, backup_type='page', + options=['--wal-depth=1', '--delete-wal']) + + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + + target_xid = node.safe_psql( + "postgres", + "select txid_current()").decode('utf-8').rstrip() + + self.switch_wal_segment(node) + + pgbench = node.pgbench(options=['-T', '10', '-c', '2']) + pgbench.wait() + + tli1 = self.show_archive(backup_dir, 'node', tli=1) + + # check that there are 4 lost_segments intervals + self.assertEqual(len(tli1['lost-segments']), 4) + + output = self.validate_pb( + backup_dir, 'node', B5, + options=['--recovery-target-xid={0}'.format(target_xid)]) + + print(output) + + self.assertIn( + 'INFO: Backup validation completed successfully on time', + output) + + self.assertIn( + 'xid {0} and LSN'.format(target_xid), + output) + + for backup_id in [B1, B2, B3, B4]: + try: + self.validate_pb( + backup_dir, 'node', backup_id, + options=['--recovery-target-xid={0}'.format(target_xid)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because page backup should not be possible " + "without valid full backup.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Not enough WAL records to xid {0}".format(target_xid), + e.message) + + self.validate_pb(backup_dir, 'node') + + def test_concurrent_running_full_backup(self): + """ + https://github.com/postgrespro/pg_probackup/issues/328 + """ + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL + self.backup_node(backup_dir, 'node', node) + + gdb = self.backup_node(backup_dir, 'node', node, gdb=True) + gdb.set_breakpoint('backup_data_file') + gdb.run_until_break() + gdb.kill() + + self.assertTrue( + self.show_pb(backup_dir, 'node')[0]['status'], + 'RUNNING') + + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--retention-redundancy=2', '--delete-expired']) + + self.assertTrue( + self.show_pb(backup_dir, 'node')[1]['status'], + 'RUNNING') + + self.backup_node(backup_dir, 'node', node) + + gdb = self.backup_node(backup_dir, 'node', node, gdb=True) + gdb.set_breakpoint('backup_data_file') + gdb.run_until_break() + gdb.kill() + + gdb = self.backup_node(backup_dir, 'node', node, gdb=True) + gdb.set_breakpoint('backup_data_file') + gdb.run_until_break() + gdb.kill() + + self.backup_node(backup_dir, 'node', node) + + gdb = self.backup_node(backup_dir, 'node', node, gdb=True) + gdb.set_breakpoint('backup_data_file') + gdb.run_until_break() + gdb.kill() + + self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--retention-redundancy=2', '--delete-expired'], + return_id=False) + + self.assertTrue( + self.show_pb(backup_dir, 'node')[0]['status'], + 'OK') + + self.assertTrue( + self.show_pb(backup_dir, 'node')[1]['status'], + 'RUNNING') + + self.assertTrue( + self.show_pb(backup_dir, 'node')[2]['status'], + 'OK') + + self.assertEqual( + len(self.show_pb(backup_dir, 'node')), + 6) diff --git a/tests/set_backup_test.py b/tests/set_backup_test.py new file mode 100644 index 000000000..31334cfba --- /dev/null +++ b/tests/set_backup_test.py @@ -0,0 +1,476 @@ +import unittest +import subprocess +import os +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +from sys import exit +from datetime import datetime, timedelta + + +class SetBackupTest(ProbackupTest, unittest.TestCase): + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_set_backup_sanity(self): + """general sanity for set-backup command""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + recovery_time = self.show_pb( + backup_dir, 'node', backup_id=backup_id)['recovery-time'] + + expire_time_1 = "{:%Y-%m-%d %H:%M:%S}".format( + datetime.now() + timedelta(days=5)) + + try: + self.set_backup(backup_dir, False, options=['--ttl=30d']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of missing instance. " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Required parameter not specified: --instance', + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + try: + self.set_backup( + backup_dir, 'node', + options=[ + "--ttl=30d", + "--expire-time='{0}'".format(expire_time_1)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because options cannot be mixed. " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: You cannot specify '--expire-time' " + "and '--ttl' options together", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + try: + self.set_backup(backup_dir, 'node', options=["--ttl=30d"]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because of missing backup_id. " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: You must specify parameter (-i, --backup-id) " + "for 'set-backup' command", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + self.set_backup( + backup_dir, 'node', backup_id, options=["--ttl=30d"]) + + actual_expire_time = self.show_pb( + backup_dir, 'node', backup_id=backup_id)['expire-time'] + + self.assertNotEqual(expire_time_1, actual_expire_time) + + expire_time_2 = "{:%Y-%m-%d %H:%M:%S}".format( + datetime.now() + timedelta(days=6)) + + self.set_backup( + backup_dir, 'node', backup_id, + options=["--expire-time={0}".format(expire_time_2)]) + + actual_expire_time = self.show_pb( + backup_dir, 'node', backup_id=backup_id)['expire-time'] + + self.assertIn(expire_time_2, actual_expire_time) + + # unpin backup + self.set_backup( + backup_dir, 'node', backup_id, options=["--ttl=0"]) + + attr_list = self.show_pb( + backup_dir, 'node', backup_id=backup_id) + + self.assertNotIn('expire-time', attr_list) + + self.set_backup( + backup_dir, 'node', backup_id, options=["--expire-time={0}".format(recovery_time)]) + + # parse string to datetime object + #new_expire_time = datetime.strptime(new_expire_time, '%Y-%m-%d %H:%M:%S%z') + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_retention_redundancy_pinning(self): + """""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + with open(os.path.join( + backup_dir, 'backups', 'node', + "pg_probackup.conf"), "a") as conf: + conf.write("retention-redundancy = 1\n") + + self.set_config( + backup_dir, 'node', options=['--retention-redundancy=1']) + + # Make backups to be purged + full_id = self.backup_node(backup_dir, 'node', node) + page_id = self.backup_node( + backup_dir, 'node', node, backup_type="page") + # Make backups to be keeped + self.backup_node(backup_dir, 'node', node) + self.backup_node(backup_dir, 'node', node, backup_type="page") + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) + + self.set_backup( + backup_dir, 'node', page_id, options=['--ttl=5d']) + + # Purge backups + log = self.delete_expired( + backup_dir, 'node', + options=['--delete-expired', '--log-level-console=LOG']) + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 4) + + self.assertIn('Time Window: 0d/5d', log) + self.assertIn( + 'LOG: Backup {0} is pinned until'.format(page_id), + log) + self.assertIn( + 'LOG: Retain backup {0} because his descendant ' + '{1} is guarded by retention'.format(full_id, page_id), + log) + + # @unittest.skip("skip") + def test_retention_window_pinning(self): + """purge all backups using window-based retention policy""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # take FULL BACKUP + backup_id_1 = self.backup_node(backup_dir, 'node', node) + page1 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Take second FULL BACKUP + backup_id_2 = self.backup_node(backup_dir, 'node', node) + page2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # Take third FULL BACKUP + backup_id_3 = self.backup_node(backup_dir, 'node', node) + page2 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + backups = os.path.join(backup_dir, 'backups', 'node') + for backup in os.listdir(backups): + if backup == 'pg_probackup.conf': + continue + with open( + os.path.join( + backups, backup, "backup.control"), "a") as conf: + conf.write("recovery_time='{:%Y-%m-%d %H:%M:%S}'\n".format( + datetime.now() - timedelta(days=3))) + + self.set_backup( + backup_dir, 'node', page1, options=['--ttl=30d']) + + # Purge backups + out = self.delete_expired( + backup_dir, 'node', + options=[ + '--log-level-console=LOG', + '--retention-window=1', + '--delete-expired']) + + self.assertEqual(len(self.show_pb(backup_dir, 'node')), 2) + + self.assertIn( + 'LOG: Backup {0} is pinned until'.format(page1), out) + + self.assertIn( + 'LOG: Retain backup {0} because his descendant ' + '{1} is guarded by retention'.format(backup_id_1, page1), + out) + + # @unittest.skip("skip") + def test_wal_retention_and_pinning(self): + """ + B1---B2---P---B3---> + wal-depth=2 + P - pinned backup + + expected result after WAL purge: + B1 B2---P---B3---> + + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # take FULL BACKUP + self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + node.pgbench_init(scale=1) + + # Take PAGE BACKUP + self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=['--stream']) + + node.pgbench_init(scale=1) + + # Take DELTA BACKUP and pin it + expire_time = "{:%Y-%m-%d %H:%M:%S}".format( + datetime.now() + timedelta(days=6)) + backup_id_pinned = self.backup_node( + backup_dir, 'node', node, + backup_type='delta', + options=[ + '--stream', + '--expire-time={0}'.format(expire_time)]) + + node.pgbench_init(scale=1) + + # Take second PAGE BACKUP + self.backup_node( + backup_dir, 'node', node, backup_type='delta', options=['--stream']) + + node.pgbench_init(scale=1) + + # Purge backups + out = self.delete_expired( + backup_dir, 'node', + options=[ + '--log-level-console=LOG', + '--delete-wal', '--wal-depth=2']) + + # print(out) + self.assertIn( + 'Pinned backup {0} is ignored for the ' + 'purpose of WAL retention'.format(backup_id_pinned), + out) + + for instance in self.show_archive(backup_dir): + timelines = instance['timelines'] + + # sanity + for timeline in timelines: + self.assertEqual( + timeline['min-segno'], + '000000010000000000000004') + self.assertEqual(timeline['status'], 'OK') + + # @unittest.skip("skip") + def test_wal_retention_and_pinning_1(self): + """ + P---B1---> + wal-depth=2 + P - pinned backup + + expected result after WAL purge: + P---B1---> + + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + expire_time = "{:%Y-%m-%d %H:%M:%S}".format( + datetime.now() + timedelta(days=6)) + + # take FULL BACKUP + backup_id_pinned = self.backup_node( + backup_dir, 'node', node, + options=['--expire-time={0}'.format(expire_time)]) + + node.pgbench_init(scale=2) + + # Take second PAGE BACKUP + self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + node.pgbench_init(scale=2) + + # Purge backups + out = self.delete_expired( + backup_dir, 'node', + options=[ + '--log-level-console=verbose', + '--delete-wal', '--wal-depth=2']) + + print(out) + self.assertIn( + 'Pinned backup {0} is ignored for the ' + 'purpose of WAL retention'.format(backup_id_pinned), + out) + + for instance in self.show_archive(backup_dir): + timelines = instance['timelines'] + + # sanity + for timeline in timelines: + self.assertEqual( + timeline['min-segno'], + '000000010000000000000002') + self.assertEqual(timeline['status'], 'OK') + + self.validate_pb(backup_dir) + + # @unittest.skip("skip") + def test_add_note_newlines(self): + """""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + # FULL + backup_id = self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--note={0}'.format('hello\nhello')]) + + backup_meta = self.show_pb(backup_dir, 'node', backup_id) + self.assertEqual(backup_meta['note'], "hello") + + self.set_backup(backup_dir, 'node', backup_id, options=['--note=hello\nhello']) + + backup_meta = self.show_pb(backup_dir, 'node', backup_id) + self.assertEqual(backup_meta['note'], "hello") + + self.set_backup(backup_dir, 'node', backup_id, options=['--note=none']) + + backup_meta = self.show_pb(backup_dir, 'node', backup_id) + self.assertNotIn('note', backup_meta) + + # @unittest.skip("skip") + def test_add_big_note(self): + """""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + +# note = node.safe_psql( +# "postgres", +# "SELECT repeat('hello', 400)").rstrip() # TODO: investigate + + note = node.safe_psql( + "postgres", + "SELECT repeat('hello', 210)").rstrip() + + # FULL + try: + self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--note={0}'.format(note)]) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because note is too large " + "\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Backup note cannot exceed 1024 bytes", + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + note = node.safe_psql( + "postgres", + "SELECT repeat('hello', 200)").decode('utf-8').rstrip() + + backup_id = self.backup_node( + backup_dir, 'node', node, + options=['--stream', '--note={0}'.format(note)]) + + backup_meta = self.show_pb(backup_dir, 'node', backup_id) + self.assertEqual(backup_meta['note'], note) + + + # @unittest.skip("skip") + def test_add_big_note_1(self): + """""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + note = node.safe_psql( + "postgres", + "SELECT repeat('q', 1024)").decode('utf-8').rstrip() + + # FULL + backup_id = self.backup_node(backup_dir, 'node', node, options=['--stream']) + + self.set_backup( + backup_dir, 'node', backup_id, + options=['--note={0}'.format(note)]) + + backup_meta = self.show_pb(backup_dir, 'node', backup_id) + + print(backup_meta) + self.assertEqual(backup_meta['note'], note) diff --git a/tests/show.py b/tests/show.py deleted file mode 100644 index 6044e15b8..000000000 --- a/tests/show.py +++ /dev/null @@ -1,209 +0,0 @@ -import os -import unittest -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException - - -module_name = 'show' - - -class OptionTest(ProbackupTest, unittest.TestCase): - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_show_1(self): - """Status DONE and OK""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.assertEqual( - self.backup_node( - backup_dir, 'node', node, - options=["--log-level-console=off"]), - None - ) - self.assertIn("OK", self.show_pb(backup_dir, 'node', as_text=True)) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_show_json(self): - """Status DONE and OK""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.assertEqual( - self.backup_node( - backup_dir, 'node', node, - options=["--log-level-console=off"]), - None - ) - self.backup_node(backup_dir, 'node', node) - self.assertIn("OK", self.show_pb(backup_dir, 'node', as_text=True)) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_corrupt_2(self): - """Status CORRUPT""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node(backup_dir, 'node', node) - - # delete file which belong to backup - file = os.path.join( - backup_dir, "backups", "node", - backup_id, "database", "postgresql.conf") - os.remove(file) - - try: - self.validate_pb(backup_dir, 'node', backup_id) - # we should die here because exception is what we expect to happen - self.assertEqual( - 1, 0, - "Expecting Error because backup corrupted." - " Output: {0} \n CMD: {1}".format( - repr(self.output), self.cmd - ) - ) - except ProbackupException as e: - self.assertIn( - 'data files are corrupted', - e.message, - '\n Unexpected Error Message: {0}\n' - ' CMD: {1}'.format(repr(e.message), self.cmd) - ) - self.assertIn("CORRUPT", self.show_pb(backup_dir, as_text=True)) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_no_control_file(self): - """backup.control doesn't exist""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node(backup_dir, 'node', node) - - # delete backup.control file - file = os.path.join( - backup_dir, "backups", "node", - backup_id, "backup.control") - os.remove(file) - - output = self.show_pb(backup_dir, 'node', as_text=True, as_json=False) - - self.assertIn( - 'Control file', - output) - - self.assertIn( - 'doesn\'t exist', - output) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - def test_empty_control_file(self): - """backup.control is empty""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node(backup_dir, 'node', node) - - # truncate backup.control file - file = os.path.join( - backup_dir, "backups", "node", - backup_id, "backup.control") - fd = open(file, 'w') - fd.close() - - output = self.show_pb(backup_dir, 'node', as_text=True, as_json=False) - - self.assertIn( - 'Control file', - output) - - self.assertIn( - 'is empty', - output) - - # Clean after yourself - self.del_test_dir(module_name, fname) - - # @unittest.skip("skip") - # @unittest.expectedFailure - def test_corrupt_control_file(self): - """backup.control contains invalid option""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - backup_id = self.backup_node(backup_dir, 'node', node) - - # corrupt backup.control file - file = os.path.join( - backup_dir, "backups", "node", - backup_id, "backup.control") - fd = open(file, 'a') - fd.write("statuss = OK") - fd.close() - - self.assertIn( - 'WARNING: Invalid option "statuss" in file', - self.show_pb(backup_dir, 'node', as_json=False, as_text=True)) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/show_test.py b/tests/show_test.py new file mode 100644 index 000000000..27b6fab96 --- /dev/null +++ b/tests/show_test.py @@ -0,0 +1,545 @@ +import os +import unittest +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException + + +class ShowTest(ProbackupTest, unittest.TestCase): + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_show_1(self): + """Status DONE and OK""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.assertEqual( + self.backup_node( + backup_dir, 'node', node, + options=["--log-level-console=off"]), + None + ) + self.assertIn("OK", self.show_pb(backup_dir, 'node', as_text=True)) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_show_json(self): + """Status DONE and OK""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + self.assertEqual( + self.backup_node( + backup_dir, 'node', node, + options=["--log-level-console=off"]), + None + ) + self.backup_node(backup_dir, 'node', node) + self.assertIn("OK", self.show_pb(backup_dir, 'node', as_text=True)) + + # @unittest.skip("skip") + def test_corrupt_2(self): + """Status CORRUPT""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node(backup_dir, 'node', node) + + # delete file which belong to backup + file = os.path.join( + backup_dir, "backups", "node", + backup_id, "database", "postgresql.conf") + os.remove(file) + + try: + self.validate_pb(backup_dir, 'node', backup_id) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because backup corrupted." + " Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd + ) + ) + except ProbackupException as e: + self.assertIn( + 'data files are corrupted', + e.message, + '\n Unexpected Error Message: {0}\n' + ' CMD: {1}'.format(repr(e.message), self.cmd) + ) + self.assertIn("CORRUPT", self.show_pb(backup_dir, as_text=True)) + + # @unittest.skip("skip") + def test_no_control_file(self): + """backup.control doesn't exist""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node(backup_dir, 'node', node) + + # delete backup.control file + file = os.path.join( + backup_dir, "backups", "node", + backup_id, "backup.control") + os.remove(file) + + output = self.show_pb(backup_dir, 'node', as_text=True, as_json=False) + + self.assertIn( + 'Control file', + output) + + self.assertIn( + 'doesn\'t exist', + output) + + # @unittest.skip("skip") + def test_empty_control_file(self): + """backup.control is empty""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node(backup_dir, 'node', node) + + # truncate backup.control file + file = os.path.join( + backup_dir, "backups", "node", + backup_id, "backup.control") + fd = open(file, 'w') + fd.close() + + output = self.show_pb(backup_dir, 'node', as_text=True, as_json=False) + + self.assertIn( + 'Control file', + output) + + self.assertIn( + 'is empty', + output) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_corrupt_control_file(self): + """backup.control contains invalid option""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node(backup_dir, 'node', node) + + # corrupt backup.control file + file = os.path.join( + backup_dir, "backups", "node", + backup_id, "backup.control") + fd = open(file, 'a') + fd.write("statuss = OK") + fd.close() + + self.assertIn( + 'WARNING: Invalid option "statuss" in file', + self.show_pb(backup_dir, 'node', as_json=False, as_text=True)) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_corrupt_correctness(self): + """backup.control contains invalid option""" + if not self.remote: + self.skipTest("You must enable PGPROBACKUP_SSH_REMOTE" + " for run this test") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=1) + + # FULL + backup_local_id = self.backup_node( + backup_dir, 'node', node, no_remote=True) + + output_local = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_local_id) + + backup_remote_id = self.backup_node(backup_dir, 'node', node) + + output_remote = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_remote_id) + + # check correctness + self.assertEqual( + output_local['data-bytes'], + output_remote['data-bytes']) + + self.assertEqual( + output_local['uncompressed-bytes'], + output_remote['uncompressed-bytes']) + + # DELTA + backup_local_id = self.backup_node( + backup_dir, 'node', node, + backup_type='delta', no_remote=True) + + output_local = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_local_id) + self.delete_pb(backup_dir, 'node', backup_local_id) + + backup_remote_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + output_remote = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_remote_id) + self.delete_pb(backup_dir, 'node', backup_remote_id) + + # check correctness + self.assertEqual( + output_local['data-bytes'], + output_remote['data-bytes']) + + self.assertEqual( + output_local['uncompressed-bytes'], + output_remote['uncompressed-bytes']) + + # PAGE + backup_local_id = self.backup_node( + backup_dir, 'node', node, + backup_type='page', no_remote=True) + + output_local = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_local_id) + self.delete_pb(backup_dir, 'node', backup_local_id) + + backup_remote_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + output_remote = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_remote_id) + self.delete_pb(backup_dir, 'node', backup_remote_id) + + # check correctness + self.assertEqual( + output_local['data-bytes'], + output_remote['data-bytes']) + + self.assertEqual( + output_local['uncompressed-bytes'], + output_remote['uncompressed-bytes']) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_corrupt_correctness_1(self): + """backup.control contains invalid option""" + if not self.remote: + self.skipTest("You must enable PGPROBACKUP_SSH_REMOTE" + " for run this test") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=1) + + # FULL + backup_local_id = self.backup_node( + backup_dir, 'node', node, no_remote=True) + + output_local = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_local_id) + + backup_remote_id = self.backup_node(backup_dir, 'node', node) + + output_remote = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_remote_id) + + # check correctness + self.assertEqual( + output_local['data-bytes'], + output_remote['data-bytes']) + + self.assertEqual( + output_local['uncompressed-bytes'], + output_remote['uncompressed-bytes']) + + # change data + pgbench = node.pgbench(options=['-T', '10', '--no-vacuum']) + pgbench.wait() + + # DELTA + backup_local_id = self.backup_node( + backup_dir, 'node', node, + backup_type='delta', no_remote=True) + + output_local = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_local_id) + self.delete_pb(backup_dir, 'node', backup_local_id) + + backup_remote_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta') + + output_remote = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_remote_id) + self.delete_pb(backup_dir, 'node', backup_remote_id) + + # check correctness + self.assertEqual( + output_local['data-bytes'], + output_remote['data-bytes']) + + self.assertEqual( + output_local['uncompressed-bytes'], + output_remote['uncompressed-bytes']) + + # PAGE + backup_local_id = self.backup_node( + backup_dir, 'node', node, + backup_type='page', no_remote=True) + + output_local = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_local_id) + self.delete_pb(backup_dir, 'node', backup_local_id) + + backup_remote_id = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + output_remote = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_remote_id) + self.delete_pb(backup_dir, 'node', backup_remote_id) + + # check correctness + self.assertEqual( + output_local['data-bytes'], + output_remote['data-bytes']) + + self.assertEqual( + output_local['uncompressed-bytes'], + output_remote['uncompressed-bytes']) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_corrupt_correctness_2(self): + """backup.control contains invalid option""" + if not self.remote: + self.skipTest("You must enable PGPROBACKUP_SSH_REMOTE" + " for run this test") + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=1) + + # FULL + backup_local_id = self.backup_node( + backup_dir, 'node', node, + options=['--compress'], no_remote=True) + + output_local = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_local_id) + + if self.remote: + backup_remote_id = self.backup_node( + backup_dir, 'node', node, options=['--compress']) + else: + backup_remote_id = self.backup_node( + backup_dir, 'node', node, + options=['--remote-proto=ssh', '--remote-host=localhost', '--compress']) + + output_remote = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_remote_id) + + # check correctness + self.assertEqual( + output_local['data-bytes'], + output_remote['data-bytes']) + + self.assertEqual( + output_local['uncompressed-bytes'], + output_remote['uncompressed-bytes']) + + # change data + pgbench = node.pgbench(options=['-T', '10', '--no-vacuum']) + pgbench.wait() + + # DELTA + backup_local_id = self.backup_node( + backup_dir, 'node', node, + backup_type='delta', options=['--compress'], no_remote=True) + + output_local = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_local_id) + self.delete_pb(backup_dir, 'node', backup_local_id) + + if self.remote: + backup_remote_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta', options=['--compress']) + else: + backup_remote_id = self.backup_node( + backup_dir, 'node', node, backup_type='delta', + options=['--remote-proto=ssh', '--remote-host=localhost', '--compress']) + + output_remote = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_remote_id) + self.delete_pb(backup_dir, 'node', backup_remote_id) + + # check correctness + self.assertEqual( + output_local['data-bytes'], + output_remote['data-bytes']) + + self.assertEqual( + output_local['uncompressed-bytes'], + output_remote['uncompressed-bytes']) + + # PAGE + backup_local_id = self.backup_node( + backup_dir, 'node', node, + backup_type='page', options=['--compress'], no_remote=True) + + output_local = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_local_id) + self.delete_pb(backup_dir, 'node', backup_local_id) + + if self.remote: + backup_remote_id = self.backup_node( + backup_dir, 'node', node, backup_type='page', options=['--compress']) + else: + backup_remote_id = self.backup_node( + backup_dir, 'node', node, backup_type='page', + options=['--remote-proto=ssh', '--remote-host=localhost', '--compress']) + + output_remote = self.show_pb( + backup_dir, 'node', as_json=False, backup_id=backup_remote_id) + self.delete_pb(backup_dir, 'node', backup_remote_id) + + # check correctness + self.assertEqual( + output_local['data-bytes'], + output_remote['data-bytes']) + + self.assertEqual( + output_local['uncompressed-bytes'], + output_remote['uncompressed-bytes']) + + # @unittest.skip("skip") + # @unittest.expectedFailure + def test_color_with_no_terminal(self): + """backup.control contains invalid option""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums'], + pg_options={'autovacuum': 'off'}) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + node.pgbench_init(scale=1) + + # FULL + try: + self.backup_node( + backup_dir, 'node', node, options=['--archive-timeout=1s']) + # we should die here because exception is what we expect to happen + self.assertEqual( + 1, 0, + "Expecting Error because archiving is disabled\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertNotIn( + '[0m', e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # @unittest.skip("skip") + def test_tablespace_print_issue_431(self): + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # Create tablespace + tblspc_path = os.path.join(node.base_dir, "tblspc") + os.makedirs(tblspc_path) + with node.connect("postgres") as con: + con.connection.autocommit = True + con.execute("CREATE TABLESPACE tblspc LOCATION '%s'" % tblspc_path) + con.connection.autocommit = False + con.execute("CREATE TABLE test (id int) TABLESPACE tblspc") + con.execute("INSERT INTO test VALUES (1)") + con.commit() + + full_backup_id = self.backup_node(backup_dir, 'node', node) + self.assertIn("OK", self.show_pb(backup_dir,'node', as_text=True)) + # Check that tablespace info exists. JSON + self.assertIn("tablespace_map", self.show_pb(backup_dir, 'node', as_text=True)) + self.assertIn("oid", self.show_pb(backup_dir, 'node', as_text=True)) + self.assertIn("path", self.show_pb(backup_dir, 'node', as_text=True)) + self.assertIn(tblspc_path, self.show_pb(backup_dir, 'node', as_text=True)) + # Check that tablespace info exists. PLAIN + self.assertIn("tablespace_map", self.show_pb(backup_dir, 'node', backup_id=full_backup_id, as_text=True, as_json=False)) + self.assertIn(tblspc_path, self.show_pb(backup_dir, 'node', backup_id=full_backup_id, as_text=True, as_json=False)) + # Check that tablespace info NOT exists if backup id not provided. PLAIN + self.assertNotIn("tablespace_map", self.show_pb(backup_dir, 'node', as_text=True, as_json=False)) diff --git a/tests/snapfs.py b/tests/snapfs.py deleted file mode 100644 index a7f926c4c..000000000 --- a/tests/snapfs.py +++ /dev/null @@ -1,60 +0,0 @@ -import unittest -import os -from time import sleep -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException - - -module_name = 'snapfs' - - -class SnapFSTest(ProbackupTest, unittest.TestCase): - - # @unittest.expectedFailure - @unittest.skipUnless(ProbackupTest.enterprise, 'skip') - def test_snapfs_simple(self): - """standart backup modes with ARCHIVE WAL method""" - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), - initdb_params=['--data-checksums']) - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.slow_start() - - self.backup_node(backup_dir, 'node', node) - - node.safe_psql( - 'postgres', - 'select pg_make_snapshot()') - - node.pgbench_init(scale=10) - - pgbench = node.pgbench(options=['-T', '50', '-c', '2', '--no-vacuum']) - pgbench.wait() - - self.backup_node( - backup_dir, 'node', node, backup_type='page') - - node.safe_psql( - 'postgres', - 'select pg_remove_snapshot(1)') - - self.backup_node( - backup_dir, 'node', node, backup_type='page') - - pgdata = self.pgdata_content(node.data_dir) - - node.cleanup() - - self.restore_node( - backup_dir, 'node', - node, options=["-j", "4"]) - - pgdata_restored = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/time_consuming_test.py b/tests/time_consuming_test.py new file mode 100644 index 000000000..c0038c085 --- /dev/null +++ b/tests/time_consuming_test.py @@ -0,0 +1,77 @@ +import os +import unittest +from .helpers.ptrack_helpers import ProbackupTest +import subprocess +from time import sleep + + +class TimeConsumingTests(ProbackupTest, unittest.TestCase): + def test_pbckp150(self): + """ + https://jira.postgrespro.ru/browse/PBCKP-150 + create a node filled with pgbench + create FULL backup followed by PTRACK backup + run pgbench, vacuum VERBOSE FULL and ptrack backups in parallel + """ + # init node + if self.pg_config_version < self.version_to_num('11.0'): + self.skipTest('You need PostgreSQL >= 11 for this test') + if not self.ptrack: + self.skipTest('Skipped because ptrack support is disabled') + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + ptrack_enable=self.ptrack, + initdb_params=['--data-checksums'], + pg_options={ + 'max_connections': 100, + 'log_statement': 'none', + 'log_checkpoints': 'on', + 'autovacuum': 'off', + 'ptrack.map_size': 1}) + + if node.major_version >= 13: + self.set_auto_conf(node, {'wal_keep_size': '16000MB'}) + else: + self.set_auto_conf(node, {'wal_keep_segments': '1000'}) + + # init probackup and add an instance + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + # run the node and init ptrack + node.slow_start() + node.safe_psql("postgres", "CREATE EXTENSION ptrack") + # populate it with pgbench + node.pgbench_init(scale=5) + + # FULL backup followed by PTRACK backup + self.backup_node(backup_dir, 'node', node, options=['--stream']) + self.backup_node(backup_dir, 'node', node, backup_type='ptrack', options=['--stream']) + + # run ordinary pgbench scenario to imitate some activity and another pgbench for vacuuming in parallel + nBenchDuration = 30 + pgbench = node.pgbench(options=['-c', '20', '-j', '8', '-T', str(nBenchDuration)]) + with open('/tmp/pbckp150vacuum.sql', 'w') as f: + f.write('VACUUM (FULL) pgbench_accounts, pgbench_tellers, pgbench_history; SELECT pg_sleep(1);\n') + pgbenchval = node.pgbench(options=['-c', '1', '-f', '/tmp/pbckp150vacuum.sql', '-T', str(nBenchDuration)]) + + # several PTRACK backups + for i in range(nBenchDuration): + print("[{}] backing up PTRACK diff...".format(i+1)) + self.backup_node(backup_dir, 'node', node, backup_type='ptrack', options=['--stream', '--log-level-console', 'VERBOSE']) + sleep(0.1) + # if the activity pgbench has finished, stop backing up + if pgbench.poll() is not None: + break + + pgbench.kill() + pgbenchval.kill() + pgbench.wait() + pgbenchval.wait() + + backups = self.show_pb(backup_dir, 'node') + for b in backups: + self.assertEqual("OK", b['status']) diff --git a/tests/time_stamp.py b/tests/time_stamp.py deleted file mode 100644 index adc6cd3e6..000000000 --- a/tests/time_stamp.py +++ /dev/null @@ -1,54 +0,0 @@ -import os -import unittest -from .helpers.ptrack_helpers import ProbackupTest, ProbackupException - - -module_name = 'time_stamp' - -class CheckTimeStamp(ProbackupTest, unittest.TestCase): - - def test_start_time_format(self): - """Test backup ID changing after start-time editing in backup.control. - We should convert local time in UTC format""" - # Create simple node - fname = self.id().split('.')[3] - node = self.make_simple_node( - base_dir="{0}/{1}/node".format(module_name, fname), - set_replication=True, - initdb_params=['--data-checksums']) - - - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') - self.init_pb(backup_dir) - self.add_instance(backup_dir, 'node', node) - self.set_archiving(backup_dir, 'node', node) - node.start() - - backup_id = self.backup_node(backup_dir, 'node', node, options=['--stream', '-j 2']) - show_backup = self.show_pb(backup_dir, 'node') - - i = 0 - while i < 2: - with open(os.path.join(backup_dir, "backups", "node", backup_id, "backup.control"), "r+") as f: - output = "" - for line in f: - if line.startswith('start-time') is True: - if i == 0: - output = output + str(line[:-5])+'+00\''+'\n' - else: - output = output + str(line[:-5]) + '\'' + '\n' - else: - output = output + str(line) - f.close() - - with open(os.path.join(backup_dir, "backups", "node", backup_id, "backup.control"), "w") as fw: - fw.write(output) - fw.flush() - show_backup = show_backup + self.show_pb(backup_dir, 'node') - i += 1 - - self.assertTrue(show_backup[1]['id'] == show_backup[2]['id'], "ERROR: Localtime format using instead of UTC") - - node.stop() - # Clean after yourself - self.del_test_dir(module_name, fname) diff --git a/tests/time_stamp_test.py b/tests/time_stamp_test.py new file mode 100644 index 000000000..170c62cd4 --- /dev/null +++ b/tests/time_stamp_test.py @@ -0,0 +1,236 @@ +import os +import unittest +from .helpers.ptrack_helpers import ProbackupTest, ProbackupException +import subprocess +from time import sleep + + +class TimeStamp(ProbackupTest, unittest.TestCase): + + def test_start_time_format(self): + """Test backup ID changing after start-time editing in backup.control. + We should convert local time in UTC format""" + # Create simple node + node = self.make_simple_node( + base_dir="{0}/{1}/node".format(self.module_name, self.fname), + set_replication=True, + initdb_params=['--data-checksums']) + + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.start() + + backup_id = self.backup_node(backup_dir, 'node', node, options=['--stream', '-j 2']) + show_backup = self.show_pb(backup_dir, 'node') + + i = 0 + while i < 2: + with open(os.path.join(backup_dir, "backups", "node", backup_id, "backup.control"), "r+") as f: + output = "" + for line in f: + if line.startswith('start-time') is True: + if i == 0: + output = output + str(line[:-5])+'+00\''+'\n' + else: + output = output + str(line[:-5]) + '\'' + '\n' + else: + output = output + str(line) + f.close() + + with open(os.path.join(backup_dir, "backups", "node", backup_id, "backup.control"), "w") as fw: + fw.write(output) + fw.flush() + show_backup = show_backup + self.show_pb(backup_dir, 'node') + i += 1 + + print(show_backup[1]['id']) + print(show_backup[2]['id']) + + self.assertTrue(show_backup[1]['id'] == show_backup[2]['id'], "ERROR: Localtime format using instead of UTC") + + output = self.show_pb(backup_dir, as_json=False, as_text=True) + self.assertNotIn("backup ID in control file", output) + + node.stop() + + def test_server_date_style(self): + """Issue #112""" + node = self.make_simple_node( + base_dir="{0}/{1}/node".format(self.module_name, self.fname), + set_replication=True, + initdb_params=['--data-checksums'], + pg_options={"datestyle": "GERMAN, DMY"}) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.start() + + self.backup_node( + backup_dir, 'node', node, options=['--stream', '-j 2']) + + def test_handling_of_TZ_env_variable(self): + """Issue #284""" + node = self.make_simple_node( + base_dir="{0}/{1}/node".format(self.module_name, self.fname), + set_replication=True, + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.start() + + my_env = os.environ.copy() + my_env["TZ"] = "America/Detroit" + + self.backup_node( + backup_dir, 'node', node, options=['--stream', '-j 2'], env=my_env) + + output = self.show_pb(backup_dir, 'node', as_json=False, as_text=True, env=my_env) + + self.assertNotIn("backup ID in control file", output) + + @unittest.skip("skip") + # @unittest.expectedFailure + def test_dst_timezone_handling(self): + """for manual testing""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + print(subprocess.Popen( + ['sudo', 'timedatectl', 'set-timezone', 'America/Detroit'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate()) + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-ntp', 'false'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-time', '2020-05-25 12:00:00'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + # FULL + output = self.backup_node(backup_dir, 'node', node, return_id=False) + self.assertNotIn("backup ID in control file", output) + + # move to dst + subprocess.Popen( + ['sudo', 'timedatectl', 'set-time', '2020-10-25 12:00:00'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + # DELTA + output = self.backup_node( + backup_dir, 'node', node, backup_type='delta', return_id=False) + self.assertNotIn("backup ID in control file", output) + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-time', '2020-12-01 12:00:00'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + # DELTA + self.backup_node(backup_dir, 'node', node, backup_type='delta') + + output = self.show_pb(backup_dir, as_json=False, as_text=True) + self.assertNotIn("backup ID in control file", output) + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-ntp', 'true'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + sleep(10) + + self.backup_node(backup_dir, 'node', node, backup_type='delta') + + output = self.show_pb(backup_dir, as_json=False, as_text=True) + self.assertNotIn("backup ID in control file", output) + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-timezone', 'US/Moscow'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + @unittest.skip("skip") + def test_dst_timezone_handling_backward_compatibilty(self): + """for manual testing""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-timezone', 'America/Detroit'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-ntp', 'false'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-time', '2020-05-25 12:00:00'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + # FULL + self.backup_node(backup_dir, 'node', node, old_binary=True, return_id=False) + + # move to dst + subprocess.Popen( + ['sudo', 'timedatectl', 'set-time', '2020-10-25 12:00:00'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + # DELTA + output = self.backup_node( + backup_dir, 'node', node, backup_type='delta', old_binary=True, return_id=False) + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-time', '2020-12-01 12:00:00'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + # DELTA + self.backup_node(backup_dir, 'node', node, backup_type='delta') + + output = self.show_pb(backup_dir, as_json=False, as_text=True) + self.assertNotIn("backup ID in control file", output) + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-ntp', 'true'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() + + sleep(10) + + self.backup_node(backup_dir, 'node', node, backup_type='delta') + + output = self.show_pb(backup_dir, as_json=False, as_text=True) + self.assertNotIn("backup ID in control file", output) + + subprocess.Popen( + ['sudo', 'timedatectl', 'set-timezone', 'US/Moscow'], + stdout=subprocess.PIPE, + stderr=subprocess.PIPE).communicate() diff --git a/tests/validate.py b/tests/validate_test.py similarity index 80% rename from tests/validate.py rename to tests/validate_test.py index 1c919907f..4ff44941f 100644 --- a/tests/validate.py +++ b/tests/validate_test.py @@ -2,15 +2,13 @@ import unittest from .helpers.ptrack_helpers import ProbackupTest, ProbackupException from datetime import datetime, timedelta +from pathlib import Path import subprocess from sys import exit import time import hashlib -module_name = 'validate' - - class ValidateTest(ProbackupTest, unittest.TestCase): # @unittest.skip("skip") @@ -19,12 +17,11 @@ def test_basic_validate_nullified_heap_page_backup(self): """ make node with nullified heap block """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -34,7 +31,7 @@ def test_basic_validate_nullified_heap_page_backup(self): file_path = node.safe_psql( "postgres", - "select pg_relation_filepath('pgbench_accounts')").rstrip() + "select pg_relation_filepath('pgbench_accounts')").decode('utf-8').rstrip() node.safe_psql( "postgres", @@ -51,14 +48,15 @@ def test_basic_validate_nullified_heap_page_backup(self): self.backup_node( backup_dir, 'node', node, options=['--log-level-file=verbose']) - if self.paranoia: - pgdata = self.pgdata_content(node.data_dir) + pgdata = self.pgdata_content(node.data_dir) if not self.remote: log_file_path = os.path.join(backup_dir, "log", "pg_probackup.log") with open(log_file_path) as f: - self.assertTrue( - '{0} blknum 1, empty page'.format(file_path) in f.read(), + log_content = f.read() + self.assertIn( + 'File: "{0}" blknum 1, empty page'.format(Path(file).as_posix()), + log_content, 'Failed to detect nullified block') self.validate_pb(backup_dir, options=["-j", "4"]) @@ -66,12 +64,8 @@ def test_basic_validate_nullified_heap_page_backup(self): self.restore_node(backup_dir, 'node', node) - if self.paranoia: - pgdata_restored = self.pgdata_content(node.data_dir) - self.compare_pgdata(pgdata, pgdata_restored) - - # Clean after yourself - self.del_test_dir(module_name, fname) + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) # @unittest.skip("skip") # @unittest.expectedFailure @@ -80,12 +74,11 @@ def test_validate_wal_unreal_values(self): make node with archiving, make archive backup validate to both real and unreal values """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -212,9 +205,6 @@ def test_validate_wal_unreal_values(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(self.output), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_basic_validate_corrupted_intermediate_backup(self): """ @@ -223,12 +213,11 @@ def test_basic_validate_corrupted_intermediate_backup(self): run validate on PAGE1, expect PAGE1 to gain status CORRUPT and PAGE2 gain status ORPHAN """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -244,7 +233,7 @@ def test_basic_validate_corrupted_intermediate_backup(self): "from generate_series(0,10000) i") file_path = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap')").rstrip() + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() # PAGE1 backup_id_2 = self.backup_node( backup_dir, 'node', node, backup_type='page') @@ -297,9 +286,6 @@ def test_basic_validate_corrupted_intermediate_backup(self): self.show_pb(backup_dir, 'node', backup_id_3)['status'], 'Backup STATUS should be "ORPHAN"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_corrupted_intermediate_backups(self): """ @@ -308,12 +294,11 @@ def test_validate_corrupted_intermediate_backups(self): expect FULL and PAGE1 to gain status CORRUPT and PAGE2 gain status ORPHAN """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -326,7 +311,7 @@ def test_validate_corrupted_intermediate_backups(self): "from generate_series(0,10000) i") file_path_t_heap = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap')").rstrip() + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() # FULL backup_id_1 = self.backup_node(backup_dir, 'node', node) @@ -337,7 +322,7 @@ def test_validate_corrupted_intermediate_backups(self): "from generate_series(0,10000) i") file_path_t_heap_1 = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap_1')").rstrip() + "select pg_relation_filepath('t_heap_1')").decode('utf-8').rstrip() # PAGE1 backup_id_2 = self.backup_node( backup_dir, 'node', node, backup_type='page') @@ -418,9 +403,6 @@ def test_validate_corrupted_intermediate_backups(self): self.show_pb(backup_dir, 'node', backup_id_3)['status'], 'Backup STATUS should be "ORPHAN"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_specific_error_intermediate_backups(self): """ @@ -430,12 +412,11 @@ def test_validate_specific_error_intermediate_backups(self): purpose of this test is to be sure that not only CORRUPT backup descendants can be orphanized """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -506,9 +487,6 @@ def test_validate_specific_error_intermediate_backups(self): self.show_pb(backup_dir, 'node', backup_id_3)['status'], 'Backup STATUS should be "ORPHAN"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_error_intermediate_backups(self): """ @@ -518,12 +496,11 @@ def test_validate_error_intermediate_backups(self): purpose of this test is to be sure that not only CORRUPT backup descendants can be orphanized """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -590,9 +567,6 @@ def test_validate_error_intermediate_backups(self): self.show_pb(backup_dir, 'node', backup_id_3)['status'], 'Backup STATUS should be "ORPHAN"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_corrupted_intermediate_backups_1(self): """ @@ -601,12 +575,11 @@ def test_validate_corrupted_intermediate_backups_1(self): expect PAGE1 to gain status CORRUPT, PAGE2, PAGE3, PAGE4 and PAGE5 to gain status ORPHAN """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -632,7 +605,7 @@ def test_validate_corrupted_intermediate_backups_1(self): "from generate_series(0,10000) i") file_page_2 = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap')").rstrip() + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() backup_id_3 = self.backup_node( backup_dir, 'node', node, backup_type='page') @@ -662,7 +635,7 @@ def test_validate_corrupted_intermediate_backups_1(self): "from generate_series(0,10000) i") file_page_5 = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap1')").rstrip() + "select pg_relation_filepath('t_heap1')").decode('utf-8').rstrip() backup_id_6 = self.backup_node( backup_dir, 'node', node, backup_type='page') @@ -787,9 +760,6 @@ def test_validate_corrupted_intermediate_backups_1(self): 'OK', self.show_pb(backup_dir, 'node', backup_id_8)['status'], 'Backup STATUS should be "OK"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_specific_target_corrupted_intermediate_backups(self): """ @@ -798,12 +768,11 @@ def test_validate_specific_target_corrupted_intermediate_backups(self): expect PAGE1 to gain status CORRUPT, PAGE2, PAGE3, PAGE4 and PAGE5 to gain status ORPHAN """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -829,7 +798,7 @@ def test_validate_specific_target_corrupted_intermediate_backups(self): "from generate_series(0,10000) i") file_page_2 = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap')").rstrip() + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() backup_id_3 = self.backup_node( backup_dir, 'node', node, backup_type='page') @@ -843,11 +812,18 @@ def test_validate_specific_target_corrupted_intermediate_backups(self): backup_dir, 'node', node, backup_type='page') # PAGE4 + node.safe_psql( + "postgres", + "insert into t_heap select i as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(20000,30000) i") + target_xid = node.safe_psql( "postgres", "insert into t_heap select i as id, md5(i::text) as text, " "md5(repeat(i::text,10))::tsvector as tsvector " - "from generate_series(20000,30000) i RETURNING (xmin)")[0][0] + "from generate_series(30001, 30001) i RETURNING (xmin)").decode('utf-8').rstrip() + backup_id_5 = self.backup_node( backup_dir, 'node', node, backup_type='page') @@ -859,7 +835,7 @@ def test_validate_specific_target_corrupted_intermediate_backups(self): "from generate_series(0,10000) i") file_page_5 = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap1')").rstrip() + "select pg_relation_filepath('t_heap1')").decode('utf-8').rstrip() backup_id_6 = self.backup_node( backup_dir, 'node', node, backup_type='page') @@ -899,8 +875,7 @@ def test_validate_specific_target_corrupted_intermediate_backups(self): self.validate_pb( backup_dir, 'node', options=[ - '-i', backup_id_4, '--xid={0}'.format(target_xid), - "-j", "4"]) + '-i', backup_id_4, '--xid={0}'.format(target_xid), "-j", "4"]) self.assertEqual( 1, 0, "Expecting Error because of data files corruption.\n " @@ -973,8 +948,197 @@ def test_validate_specific_target_corrupted_intermediate_backups(self): self.assertEqual('ORPHAN', self.show_pb(backup_dir, 'node', backup_id_7)['status'], 'Backup STATUS should be "ORPHAN"') self.assertEqual('OK', self.show_pb(backup_dir, 'node', backup_id_8)['status'], 'Backup STATUS should be "OK"') - # Clean after yourself - self.del_test_dir(module_name, fname) + # @unittest.skip("skip") + def test_validate_instance_with_several_corrupt_backups(self): + """ + make archive node, take FULL1, PAGE1_1, FULL2, PAGE2_1 backups, FULL3 + corrupt file in FULL and FULL2 and run validate on instance, + expect FULL1 to gain status CORRUPT, PAGE1_1 to gain status ORPHAN + FULL2 to gain status CORRUPT, PAGE2_1 to gain status ORPHAN + """ + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "create table t_heap as select generate_series(0,1) i") + # FULL1 + backup_id_1 = self.backup_node( + backup_dir, 'node', node, options=['--no-validate']) + + # FULL2 + backup_id_2 = self.backup_node(backup_dir, 'node', node) + rel_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + node.safe_psql( + "postgres", + "insert into t_heap values(2)") + + backup_id_3 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # FULL3 + backup_id_4 = self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + "postgres", + "insert into t_heap values(3)") + + backup_id_5 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # FULL4 + backup_id_6 = self.backup_node( + backup_dir, 'node', node, options=['--no-validate']) + + # Corrupt some files in FULL2 and FULL3 backup + os.remove(os.path.join( + backup_dir, 'backups', 'node', backup_id_2, + 'database', rel_path)) + os.remove(os.path.join( + backup_dir, 'backups', 'node', backup_id_4, + 'database', rel_path)) + + # Validate Instance + try: + self.validate_pb(backup_dir, 'node', options=["-j", "4", "--log-level-file=LOG"]) + self.assertEqual( + 1, 0, + "Expecting Error because of data files corruption.\n " + "Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertTrue( + "INFO: Validate backups of the instance 'node'" in e.message, + "\n Unexpected Error Message: {0}\n " + "CMD: {1}".format(repr(e.message), self.cmd)) + self.assertTrue( + 'WARNING: Some backups are not valid' in e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'node', backup_id_1)['status'], + 'Backup STATUS should be "OK"') + self.assertEqual( + 'CORRUPT', self.show_pb(backup_dir, 'node', backup_id_2)['status'], + 'Backup STATUS should be "CORRUPT"') + self.assertEqual( + 'ORPHAN', self.show_pb(backup_dir, 'node', backup_id_3)['status'], + 'Backup STATUS should be "ORPHAN"') + self.assertEqual( + 'CORRUPT', self.show_pb(backup_dir, 'node', backup_id_4)['status'], + 'Backup STATUS should be "CORRUPT"') + self.assertEqual( + 'ORPHAN', self.show_pb(backup_dir, 'node', backup_id_5)['status'], + 'Backup STATUS should be "ORPHAN"') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'node', backup_id_6)['status'], + 'Backup STATUS should be "OK"') + + # @unittest.skip("skip") + def test_validate_instance_with_several_corrupt_backups_interrupt(self): + """ + check that interrupt during validation is handled correctly + """ + self._check_gdb_flag_or_skip_test() + + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + node.safe_psql( + "postgres", + "create table t_heap as select generate_series(0,1) i") + # FULL1 + backup_id_1 = self.backup_node( + backup_dir, 'node', node, options=['--no-validate']) + + # FULL2 + backup_id_2 = self.backup_node(backup_dir, 'node', node) + rel_path = node.safe_psql( + "postgres", + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() + + node.safe_psql( + "postgres", + "insert into t_heap values(2)") + + backup_id_3 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # FULL3 + backup_id_4 = self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + "postgres", + "insert into t_heap values(3)") + + backup_id_5 = self.backup_node( + backup_dir, 'node', node, backup_type='page') + + # FULL4 + backup_id_6 = self.backup_node( + backup_dir, 'node', node, options=['--no-validate']) + + # Corrupt some files in FULL2 and FULL3 backup + os.remove(os.path.join( + backup_dir, 'backups', 'node', backup_id_1, + 'database', rel_path)) + os.remove(os.path.join( + backup_dir, 'backups', 'node', backup_id_3, + 'database', rel_path)) + + # Validate Instance + gdb = self.validate_pb( + backup_dir, 'node', options=["-j", "4", "--log-level-file=LOG"], gdb=True) + + gdb.set_breakpoint('validate_file_pages') + gdb.run_until_break() + gdb.continue_execution_until_break() + gdb.remove_all_breakpoints() + gdb._execute('signal SIGINT') + gdb.continue_execution_until_error() + + self.assertEqual( + 'DONE', self.show_pb(backup_dir, 'node', backup_id_1)['status'], + 'Backup STATUS should be "OK"') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'node', backup_id_2)['status'], + 'Backup STATUS should be "OK"') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'node', backup_id_3)['status'], + 'Backup STATUS should be "CORRUPT"') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'node', backup_id_4)['status'], + 'Backup STATUS should be "ORPHAN"') + self.assertEqual( + 'OK', self.show_pb(backup_dir, 'node', backup_id_5)['status'], + 'Backup STATUS should be "OK"') + self.assertEqual( + 'DONE', self.show_pb(backup_dir, 'node', backup_id_6)['status'], + 'Backup STATUS should be "OK"') + + log_file = os.path.join(backup_dir, 'log', 'pg_probackup.log') + with open(log_file, 'r') as f: + log_content = f.read() + self.assertNotIn( + 'Interrupted while locking backup', log_content) # @unittest.skip("skip") def test_validate_instance_with_corrupted_page(self): @@ -983,12 +1147,11 @@ def test_validate_instance_with_corrupted_page(self): corrupt file in PAGE1 backup and run validate on instance, expect PAGE1 to gain status CORRUPT, PAGE2 to gain status ORPHAN """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1009,7 +1172,7 @@ def test_validate_instance_with_corrupted_page(self): "from generate_series(0,10000) i") file_path_t_heap1 = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap1')").rstrip() + "select pg_relation_filepath('t_heap1')").decode('utf-8').rstrip() # PAGE1 backup_id_2 = self.backup_node( backup_dir, 'node', node, backup_type='page') @@ -1120,20 +1283,16 @@ def test_validate_instance_with_corrupted_page(self): 'OK', self.show_pb(backup_dir, 'node', backup_id_5)['status'], 'Backup STATUS should be "OK"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_instance_with_corrupted_full_and_try_restore(self): """make archive node, take FULL, PAGE1, PAGE2, FULL2, PAGE3 backups, corrupt file in FULL backup and run validate on instance, expect FULL to gain status CORRUPT, PAGE1 and PAGE2 to gain status ORPHAN, try to restore backup with --no-validation option""" - fname = self.id().split('.')[3] - node = self.make_simple_node(base_dir=os.path.join(module_name, fname, 'node'), + node = self.make_simple_node(base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1146,7 +1305,7 @@ def test_validate_instance_with_corrupted_full_and_try_restore(self): "from generate_series(0,10000) i") file_path_t_heap = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap')").rstrip() + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() # FULL1 backup_id_1 = self.backup_node(backup_dir, 'node', node) @@ -1216,19 +1375,15 @@ def test_validate_instance_with_corrupted_full_and_try_restore(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(self.output), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_instance_with_corrupted_full(self): """make archive node, take FULL, PAGE1, PAGE2, FULL2, PAGE3 backups, corrupt file in FULL backup and run validate on instance, expect FULL to gain status CORRUPT, PAGE1 and PAGE2 to gain status ORPHAN""" - fname = self.id().split('.')[3] - node = self.make_simple_node(base_dir=os.path.join(module_name, fname, 'node'), + node = self.make_simple_node(base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1242,7 +1397,7 @@ def test_validate_instance_with_corrupted_full(self): file_path_t_heap = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap')").rstrip() + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() # FULL1 backup_id_1 = self.backup_node(backup_dir, 'node', node) @@ -1310,18 +1465,14 @@ def test_validate_instance_with_corrupted_full(self): self.assertEqual('OK', self.show_pb(backup_dir, 'node', backup_id_4)['status'], 'Backup STATUS should be "OK"') self.assertEqual('OK', self.show_pb(backup_dir, 'node', backup_id_5)['status'], 'Backup STATUS should be "OK"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_corrupt_wal_1(self): """make archive node, take FULL1, PAGE1,PAGE2,FULL2,PAGE3,PAGE4 backups, corrupt all wal files, run validate, expect errors""" - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1372,17 +1523,13 @@ def test_validate_corrupt_wal_1(self): self.show_pb(backup_dir, 'node', backup_id_2)['status'], 'Backup STATUS should be "CORRUPT"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_corrupt_wal_2(self): """make archive node, make full backup, corrupt all wal files, run validate to real xid, expect errors""" - fname = self.id().split('.')[3] - node = self.make_simple_node(base_dir=os.path.join(module_name, fname, 'node'), + node = self.make_simple_node(base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1438,9 +1585,6 @@ def test_validate_corrupt_wal_2(self): self.show_pb(backup_dir, 'node', backup_id)['status'], 'Backup STATUS should be "CORRUPT"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_wal_lost_segment_1(self): """make archive node, make archive full backup, @@ -1448,12 +1592,11 @@ def test_validate_wal_lost_segment_1(self): run validate, expecting error because of missing wal segment make sure that backup status is 'CORRUPT' """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1515,21 +1658,17 @@ def test_validate_wal_lost_segment_1(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_corrupt_wal_between_backups(self): """ make archive node, make full backup, corrupt all wal files, run validate to real xid, expect errors """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1553,11 +1692,11 @@ def test_validate_corrupt_wal_between_backups(self): if self.get_version(node) < self.version_to_num('10.0'): walfile = node.safe_psql( 'postgres', - 'select pg_xlogfile_name(pg_current_xlog_location())').rstrip() + 'select pg_xlogfile_name(pg_current_xlog_location())').decode('utf-8').rstrip() else: walfile = node.safe_psql( 'postgres', - 'select pg_walfile_name(pg_current_wal_lsn())').rstrip() + 'select pg_walfile_name(pg_current_wal_lsn())').decode('utf-8').rstrip() if self.archive_compress: walfile = walfile + '.gz' @@ -1608,22 +1747,18 @@ def test_validate_corrupt_wal_between_backups(self): self.show_pb(backup_dir, 'node')[1]['status'], 'Backup STATUS should be "OK"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_pgpro702_688(self): """ make node without archiving, make stream backup, get Recovery Time, validate to Recovery Time """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) node.slow_start() @@ -1649,22 +1784,18 @@ def test_pgpro702_688(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_pgpro688(self): """ make node with archiving, make backup, get Recovery Time, validate to Recovery Time. Waiting PGPRO-688. RESOLVED """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1678,9 +1809,6 @@ def test_pgpro688(self): backup_dir, 'node', options=["--time={0}".format(recovery_time), "-j", "4"]) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") # @unittest.expectedFailure def test_pgpro561(self): @@ -1688,13 +1816,12 @@ def test_pgpro561(self): make node with archiving, make stream backup, restore it to node1, check that archiving is not successful on node1 """ - fname = self.id().split('.')[3] node1 = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node1'), + base_dir=os.path.join(self.module_name, self.fname, 'node1'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node1', node1) self.set_archiving(backup_dir, 'node1', node1) @@ -1704,7 +1831,7 @@ def test_pgpro561(self): backup_dir, 'node1', node1, options=["--stream"]) node2 = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node2')) + base_dir=os.path.join(self.module_name, self.fname, 'node2')) node2.cleanup() node1.psql( @@ -1718,14 +1845,14 @@ def test_pgpro561(self): backup_type='page', options=["--stream"]) self.restore_node(backup_dir, 'node1', data_dir=node2.data_dir) - node2.append_conf( - 'postgresql.auto.conf', 'port = {0}'.format(node2.port)) - node2.append_conf( - 'postgresql.auto.conf', 'archive_mode = off') + self.set_auto_conf( + node2, {'port': node2.port, 'archive_mode': 'off'}) + node2.slow_start() - node2.append_conf( - 'postgresql.auto.conf', 'archive_mode = on') + self.set_auto_conf( + node2, {'archive_mode': 'on'}) + node2.stop() node2.slow_start() @@ -1760,7 +1887,7 @@ def test_pgpro561(self): # wals_dir = os.path.join(backup_dir, 'wal', 'node1') # wals = [f for f in os.listdir(wals_dir) if os.path.isfile(os.path.join( -# wals_dir, f)) and not f.endswith('.backup') and not f.endswith('.partial')] +# wals_dir, f)) and not f.endswith('.backup') and not f.endswith('.part')] # wals = map(str, wals) # print(wals) @@ -1768,7 +1895,7 @@ def test_pgpro561(self): # wals_dir = os.path.join(backup_dir, 'wal', 'node1') # wals = [f for f in os.listdir(wals_dir) if os.path.isfile(os.path.join( -# wals_dir, f)) and not f.endswith('.backup') and not f.endswith('.partial')] +# wals_dir, f)) and not f.endswith('.backup') and not f.endswith('.part')] # wals = map(str, wals) # print(wals) @@ -1780,15 +1907,12 @@ def test_pgpro561(self): self.assertTrue( 'LOG: archive command failed with exit code 1' in log_content and 'DETAIL: The failed archive command was:' in log_content and - 'INFO: pg_probackup archive-push from' in log_content, + 'WAL file already exists in archive with different checksum' in log_content, 'Expecting error messages about failed archive_command' ) self.assertFalse( 'pg_probackup archive-push completed successfully' in log_content) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_corrupted_full(self): """ @@ -1799,15 +1923,14 @@ def test_validate_corrupted_full(self): remove corruption and run valudate again, check that second full backup and his page backups are OK """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums'], pg_options={ 'checkpoint_timeout': '30'}) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -1907,9 +2030,6 @@ def test_validate_corrupted_full(self): self.show_pb(backup_dir, 'node')[6]['status'] == 'ERROR') self.assertTrue(self.show_pb(backup_dir, 'node')[7]['status'] == 'OK') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_corrupted_full_1(self): """ @@ -1923,13 +2043,12 @@ def test_validate_corrupted_full_1(self): second page should be CORRUPT third page should be ORPHAN """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -2009,9 +2128,6 @@ def test_validate_corrupted_full_1(self): self.assertTrue(self.show_pb(backup_dir, 'node')[5]['status'] == 'CORRUPT') self.assertTrue(self.show_pb(backup_dir, 'node')[6]['status'] == 'ORPHAN') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_corrupted_full_2(self): """ @@ -2034,13 +2150,12 @@ def test_validate_corrupted_full_2(self): remove corruption from PAGE2_2 and run validate on PAGE2_4 """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -2371,9 +2486,6 @@ def test_validate_corrupted_full_2(self): self.assertTrue(self.show_pb(backup_dir, 'node')[1]['status'] == 'OK') self.assertTrue(self.show_pb(backup_dir, 'node')[0]['status'] == 'OK') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_corrupted_full_missing(self): """ @@ -2386,13 +2498,12 @@ def test_validate_corrupted_full_missing(self): second full backup and his firts page backups are OK, third page should be ORPHAN """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -2604,18 +2715,14 @@ def test_validate_corrupted_full_missing(self): self.assertTrue(self.show_pb(backup_dir, 'node')[1]['status'] == 'OK') self.assertTrue(self.show_pb(backup_dir, 'node')[0]['status'] == 'OK') - # Clean after yourself - self.del_test_dir(module_name, fname) - def test_file_size_corruption_no_validate(self): - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), # initdb_params=['--data-checksums'], ) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) @@ -2634,7 +2741,7 @@ def test_file_size_corruption_no_validate(self): heap_path = node.safe_psql( "postgres", - "select pg_relation_filepath('t_heap')").rstrip() + "select pg_relation_filepath('t_heap')").decode('utf-8').rstrip() heap_size = node.safe_psql( "postgres", "select pg_relation_size('t_heap')") @@ -2663,12 +2770,8 @@ def test_file_size_corruption_no_validate(self): options=["--no-validate"]) except ProbackupException as e: self.assertTrue( - "ERROR: Data files restoring failed" in e.message, + "ERROR: Backup files restoring failed" in e.message, repr(e.message)) - # print "\nExpected error: \n" + e.message - - # Clean after yourself - self.del_test_dir(module_name, fname) # @unittest.skip("skip") def test_validate_specific_backup_with_missing_backup(self): @@ -2686,13 +2789,12 @@ def test_validate_specific_backup_with_missing_backup(self): PAGE1_1 FULL1 """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -2810,9 +2912,6 @@ def test_validate_specific_backup_with_missing_backup(self): self.assertTrue(self.show_pb(backup_dir, 'node')[1]['status'] == 'OK') self.assertTrue(self.show_pb(backup_dir, 'node')[0]['status'] == 'OK') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_specific_backup_with_missing_backup_1(self): """ @@ -2829,13 +2928,12 @@ def test_validate_specific_backup_with_missing_backup_1(self): PAGE1_1 FULL1 """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -2931,9 +3029,6 @@ def test_validate_specific_backup_with_missing_backup_1(self): self.assertTrue(self.show_pb(backup_dir, 'node')[1]['status'] == 'OK') self.assertTrue(self.show_pb(backup_dir, 'node')[0]['status'] == 'OK') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_with_missing_backup_1(self): """ @@ -2950,13 +3045,12 @@ def test_validate_with_missing_backup_1(self): PAGE1_1 FULL1 """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -3120,9 +3214,6 @@ def test_validate_with_missing_backup_1(self): self.assertTrue(self.show_pb(backup_dir, 'node')[1]['status'] == 'OK') self.assertTrue(self.show_pb(backup_dir, 'node')[0]['status'] == 'OK') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validate_with_missing_backup_2(self): """ @@ -3139,13 +3230,12 @@ def test_validate_with_missing_backup_2(self): PAGE1_1 FULL1 """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -3280,19 +3370,15 @@ def test_validate_with_missing_backup_2(self): self.assertTrue(self.show_pb(backup_dir, 'node')[1]['status'] == 'OK') self.assertTrue(self.show_pb(backup_dir, 'node')[0]['status'] == 'OK') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_corrupt_pg_control_via_resetxlog(self): """ PGPRO-2096 """ - fname = self.id().split('.')[3] node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') self.init_pb(backup_dir) self.add_instance(backup_dir, 'node', node) self.set_archiving(backup_dir, 'node', node) @@ -3350,16 +3436,14 @@ def test_corrupt_pg_control_via_resetxlog(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.skip("skip") def test_validation_after_backup(self): """""" - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + self._check_gdb_flag_or_skip_test() + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -3388,19 +3472,15 @@ def test_validation_after_backup(self): self.show_pb(backup_dir, 'node', backup_id)['status'], 'Backup STATUS should be "ERROR"') - # Clean after yourself - self.del_test_dir(module_name, fname) - # @unittest.expectedFailure # @unittest.skip("skip") def test_validate_corrupt_tablespace_map(self): """ Check that corruption in tablespace_map is detected """ - fname = self.id().split('.')[3] - backup_dir = os.path.join(self.tmp_path, module_name, fname, 'backup') + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') node = self.make_simple_node( - base_dir=os.path.join(module_name, fname, 'node'), + base_dir=os.path.join(self.module_name, self.fname, 'node'), set_replication=True, initdb_params=['--data-checksums']) @@ -3443,6 +3523,543 @@ def test_validate_corrupt_tablespace_map(self): '\n Unexpected Error Message: {0}\n CMD: {1}'.format( repr(e.message), self.cmd)) + #TODO fix the test + @unittest.expectedFailure + # @unittest.skip("skip") + def test_validate_target_lsn(self): + """ + Check validation to specific LSN + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + # FULL backup + self.backup_node(backup_dir, 'node', node) + + node.safe_psql( + "postgres", + "create table t_heap as select 1 as id, md5(i::text) as text, " + "md5(repeat(i::text,10))::tsvector as tsvector " + "from generate_series(0,10000) i") + + node_restored = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node_restored')) + node_restored.cleanup() + + self.restore_node(backup_dir, 'node', node_restored) + + self.set_auto_conf( + node_restored, {'port': node_restored.port}) + + node_restored.slow_start() + + self.switch_wal_segment(node) + + backup_id = self.backup_node( + backup_dir, 'node', node_restored, + data_dir=node_restored.data_dir) + + target_lsn = self.show_pb(backup_dir, 'node')[1]['stop-lsn'] + + self.delete_pb(backup_dir, 'node', backup_id) + + self.validate_pb( + backup_dir, 'node', + options=[ + '--recovery-target-timeline=2', + '--recovery-target-lsn={0}'.format(target_lsn)]) + + @unittest.skip("skip") + def test_partial_validate_empty_and_mangled_database_map(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + + node.slow_start() + + # create databases + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + # FULL backup with database_map + backup_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + pgdata = self.pgdata_content(node.data_dir) + + # truncate database_map + path = os.path.join( + backup_dir, 'backups', 'node', + backup_id, 'database', 'database_map') + with open(path, "w") as f: + f.close() + + try: + self.validate_pb( + backup_dir, 'node', + options=["--db-include=db1"]) + self.assertEqual( + 1, 0, + "Expecting Error because database_map is empty.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "WARNING: Backup {0} data files are corrupted".format( + backup_id), e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + # mangle database_map + with open(path, "w") as f: + f.write("42") + f.close() + + try: + self.validate_pb( + backup_dir, 'node', + options=["--db-include=db1"]) + self.assertEqual( + 1, 0, + "Expecting Error because database_map is empty.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "WARNING: Backup {0} data files are corrupted".format( + backup_id), e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + @unittest.skip("skip") + def test_partial_validate_exclude(self): + """""" + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + + try: + self.validate_pb( + backup_dir, 'node', + options=[ + "--db-include=db1", + "--db-exclude=db2"]) + self.assertEqual( + 1, 0, + "Expecting Error because of 'db-exclude' and 'db-include'.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: You cannot specify '--db-include' " + "and '--db-exclude' together", e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + try: + self.validate_pb( + backup_dir, 'node', + options=[ + "--db-exclude=db1", + "--db-exclude=db5", + "--log-level-console=verbose"]) + self.assertEqual( + 1, 0, + "Expecting Error because of missing backup ID.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: You must specify parameter (-i, --backup-id) for partial validation", + e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + output = self.validate_pb( + backup_dir, 'node', backup_id, + options=[ + "--db-exclude=db1", + "--db-exclude=db5", + "--log-level-console=verbose"]) + + self.assertIn( + "VERBOSE: Skip file validation due to partial restore", output) + + @unittest.skip("skip") + def test_partial_validate_include(self): + """ + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + for i in range(1, 10, 1): + node.safe_psql( + 'postgres', + 'CREATE database db{0}'.format(i)) + + # FULL backup + backup_id = self.backup_node(backup_dir, 'node', node) + + try: + self.validate_pb( + backup_dir, 'node', + options=[ + "--db-include=db1", + "--db-exclude=db2"]) + self.assertEqual( + 1, 0, + "Expecting Error because of 'db-exclude' and 'db-include'.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: You cannot specify '--db-include' " + "and '--db-exclude' together", e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + output = self.validate_pb( + backup_dir, 'node', backup_id, + options=[ + "--db-include=db1", + "--db-include=db5", + "--db-include=postgres", + "--log-level-console=verbose"]) + + self.assertIn( + "VERBOSE: Skip file validation due to partial restore", output) + + output = self.validate_pb( + backup_dir, 'node', backup_id, + options=["--log-level-console=verbose"]) + + self.assertNotIn( + "VERBOSE: Skip file validation due to partial restore", output) + + # @unittest.skip("skip") + def test_not_validate_diffenent_pg_version(self): + """Do not validate backup, if binary is compiled with different PG version""" + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + initdb_params=['--data-checksums']) + + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + self.set_archiving(backup_dir, 'node', node) + node.slow_start() + + backup_id = self.backup_node(backup_dir, 'node', node) + + control_file = os.path.join( + backup_dir, "backups", "node", backup_id, + "backup.control") + + pg_version = node.major_version + + if pg_version.is_integer(): + pg_version = int(pg_version) + + fake_new_pg_version = pg_version + 1 + + with open(control_file, 'r') as f: + data = f.read(); + + data = data.replace( + "server-version = {0}".format(str(pg_version)), + "server-version = {0}".format(str(fake_new_pg_version))) + + with open(control_file, 'w') as f: + f.write(data); + + try: + self.validate_pb(backup_dir) + self.assertEqual( + 1, 0, + "Expecting Error because validation is forbidden if server version of backup " + "is different from the server version of pg_probackup.\n Output: {0} \n CMD: {1}".format( + repr(self.output), self.cmd)) + except ProbackupException as e: + self.assertIn( + "ERROR: Backup {0} has server version".format(backup_id), + e.message, + "\n Unexpected Error Message: {0}\n CMD: {1}".format( + repr(e.message), self.cmd)) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_validate_corrupt_page_header_map(self): + """ + Check that corruption in page_header_map is detected + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + ok_1 = self.backup_node(backup_dir, 'node', node, options=['--stream']) + + # FULL backup + backup_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + ok_2 = self.backup_node(backup_dir, 'node', node, options=['--stream']) + + page_header_map = os.path.join( + backup_dir, 'backups', 'node', backup_id, 'page_header_map') + + # Corrupt tablespace_map file in FULL backup + with open(page_header_map, "rb+", 0) as f: + f.seek(42) + f.write(b"blah") + f.flush() + + with self.assertRaises(ProbackupException) as cm: + self.validate_pb(backup_dir, 'node', backup_id=backup_id) + + e = cm.exception + self.assertRegex( + cm.exception.message, + r'WARNING: An error occured during metadata decompression for file "[\w/]+": (data|buffer) error', + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.assertIn("Backup {0} is corrupt".format(backup_id), e.message) + + with self.assertRaises(ProbackupException) as cm: + self.validate_pb(backup_dir) + + e = cm.exception + self.assertRegex( + e.message, + r'WARNING: An error occured during metadata decompression for file "[\w/]+": (data|buffer) error', + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + self.assertIn("INFO: Backup {0} data files are valid".format(ok_1), e.message) + self.assertIn("WARNING: Backup {0} data files are corrupted".format(backup_id), e.message) + self.assertIn("INFO: Backup {0} data files are valid".format(ok_2), e.message) + + self.assertIn("WARNING: Some backups are not valid", e.message) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_validate_truncated_page_header_map(self): + """ + Check that corruption in page_header_map is detected + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + ok_1 = self.backup_node(backup_dir, 'node', node, options=['--stream']) + + # FULL backup + backup_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + ok_2 = self.backup_node(backup_dir, 'node', node, options=['--stream']) + + page_header_map = os.path.join( + backup_dir, 'backups', 'node', backup_id, 'page_header_map') + + # truncate page_header_map file + with open(page_header_map, "rb+", 0) as f: + f.truncate(121) + f.flush() + f.close + + try: + self.validate_pb(backup_dir, 'node', backup_id=backup_id) + self.assertEqual( + 1, 0, + "Expecting Error because page_header is corrupted.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Backup {0} is corrupt'.format(backup_id), e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + try: + self.validate_pb(backup_dir) + self.assertEqual( + 1, 0, + "Expecting Error because page_header is corrupted.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn("INFO: Backup {0} data files are valid".format(ok_1), e.message) + self.assertIn("WARNING: Backup {0} data files are corrupted".format(backup_id), e.message) + self.assertIn("INFO: Backup {0} data files are valid".format(ok_2), e.message) + self.assertIn("WARNING: Some backups are not valid", e.message) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_validate_missing_page_header_map(self): + """ + Check that corruption in page_header_map is detected + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + ok_1 = self.backup_node(backup_dir, 'node', node, options=['--stream']) + + # FULL backup + backup_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + ok_2 = self.backup_node(backup_dir, 'node', node, options=['--stream']) + + page_header_map = os.path.join( + backup_dir, 'backups', 'node', backup_id, 'page_header_map') + + # unlink page_header_map file + os.remove(page_header_map) + + try: + self.validate_pb(backup_dir, 'node', backup_id=backup_id) + self.assertEqual( + 1, 0, + "Expecting Error because page_header is corrupted.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn( + 'ERROR: Backup {0} is corrupt'.format(backup_id), e.message, + '\n Unexpected Error Message: {0}\n CMD: {1}'.format( + repr(e.message), self.cmd)) + + try: + self.validate_pb(backup_dir) + self.assertEqual( + 1, 0, + "Expecting Error because page_header is corrupted.\n " + "Output: {0} \n CMD: {1}".format( + self.output, self.cmd)) + except ProbackupException as e: + self.assertIn("INFO: Backup {0} data files are valid".format(ok_1), e.message) + self.assertIn("WARNING: Backup {0} data files are corrupted".format(backup_id), e.message) + self.assertIn("INFO: Backup {0} data files are valid".format(ok_2), e.message) + self.assertIn("WARNING: Some backups are not valid", e.message) + + # @unittest.expectedFailure + # @unittest.skip("skip") + def test_no_validate_tablespace_map(self): + """ + Check that --no-validate is propagated to tablespace_map + """ + backup_dir = os.path.join(self.tmp_path, self.module_name, self.fname, 'backup') + node = self.make_simple_node( + base_dir=os.path.join(self.module_name, self.fname, 'node'), + set_replication=True, + initdb_params=['--data-checksums']) + + self.init_pb(backup_dir) + self.add_instance(backup_dir, 'node', node) + node.slow_start() + + self.create_tblspace_in_node(node, 'external_dir') + + node.safe_psql( + 'postgres', + 'CREATE TABLE t_heap(a int) TABLESPACE "external_dir"') + + tblspace_new = self.get_tblspace_path(node, 'external_dir_new') + + oid = node.safe_psql( + 'postgres', + "select oid from pg_tablespace where spcname = 'external_dir'").decode('utf-8').rstrip() + + # FULL backup + backup_id = self.backup_node( + backup_dir, 'node', node, options=['--stream']) + + pgdata = self.pgdata_content(node.data_dir) + + tablespace_map = os.path.join( + backup_dir, 'backups', 'node', + backup_id, 'database', 'tablespace_map') + + # overwrite tablespace_map file + with open(tablespace_map, "w") as f: + f.write("{0} {1}".format(oid, tblspace_new)) + f.close + + node.cleanup() + + self.restore_node(backup_dir, 'node', node, options=['--no-validate']) + + pgdata_restored = self.pgdata_content(node.data_dir) + self.compare_pgdata(pgdata, pgdata_restored) + + # check that tablespace restore as symlink + tablespace_link = os.path.join(node.data_dir, 'pg_tblspc', oid) + self.assertTrue( + os.path.islink(tablespace_link), + "'%s' is not a symlink" % tablespace_link) + + self.assertEqual( + os.readlink(tablespace_link), + tblspace_new, + "Symlink '{0}' do not points to '{1}'".format(tablespace_link, tblspace_new)) + # validate empty backup list # page from future during validate # page from future during backup diff --git a/travis/backup_restore.sh b/travis/backup_restore.sh deleted file mode 100644 index b3c9df1ed..000000000 --- a/travis/backup_restore.sh +++ /dev/null @@ -1,66 +0,0 @@ -#!/bin/sh -ex - -# vars -export PGVERSION=9.5.4 -export PATH=$PATH:/usr/pgsql-9.5/bin -export PGUSER=pgbench -export PGDATABASE=pgbench -export PGDATA=/var/lib/pgsql/9.5/data -export BACKUP_PATH=/backups -export ARCLOG_PATH=$BACKUP_PATH/backup/pg_xlog -export PGDATA2=/var/lib/pgsql/9.5/data2 -export PGBENCH_SCALE=100 -export PGBENCH_TIME=60 - -# prepare directory -cp -a /tests /build -pushd /build - -# download postgresql -yum install -y wget -wget -k https://ftp.postgresql.org/pub/source/v$PGVERSION/postgresql-$PGVERSION.tar.gz -O postgresql.tar.gz -tar xf postgresql.tar.gz - -# install pg_probackup -yum install -y https://download.postgresql.org/pub/repos/yum/9.5/redhat/rhel-7-x86_64/pgdg-centos95-9.5-2.noarch.rpm -yum install -y postgresql95-devel make gcc readline-devel openssl-devel pam-devel libxml2-devel libxslt-devel -make top_srcdir=postgresql-$PGVERSION -make install top_srcdir=postgresql-$PGVERSION - -# initialize cluster and database -yum install -y postgresql95-server -su postgres -c "/usr/pgsql-9.5/bin/initdb -D $PGDATA -k" -cat < $PGDATA/pg_hba.conf -local all all trust -host all all 127.0.0.1/32 trust -local replication pgbench trust -host replication pgbench 127.0.0.1/32 trust -EOF -cat < $PGDATA/postgresql.auto.conf -max_wal_senders = 2 -wal_level = logical -wal_log_hints = on -EOF -su postgres -c "/usr/pgsql-9.5/bin/pg_ctl start -w -D $PGDATA" -su postgres -c "createdb -U postgres $PGUSER" -su postgres -c "createuser -U postgres -a -d -E $PGUSER" -pgbench -i -s $PGBENCH_SCALE - -# Count current -COUNT=$(psql -Atc "select count(*) from pgbench_accounts") -pgbench -s $PGBENCH_SCALE -T $PGBENCH_TIME -j 2 -c 10 & - -# create backup -pg_probackup init -pg_probackup backup -b full --disable-ptrack-clear --stream -v -pg_probackup show -sleep $PGBENCH_TIME - -# restore from backup -chown -R postgres:postgres $BACKUP_PATH -su postgres -c "pg_probackup restore -D $PGDATA2" - -# start backup server -su postgres -c "/usr/pgsql-9.5/bin/pg_ctl stop -w -D $PGDATA" -su postgres -c "/usr/pgsql-9.5/bin/pg_ctl start -w -D $PGDATA2" -( psql -Atc "select count(*) from pgbench_accounts" | grep $COUNT ) || (cat $PGDATA2/pg_log/*.log ; exit 1) diff --git a/travis/before-install.sh b/travis/before-install.sh new file mode 100755 index 000000000..376de5e6e --- /dev/null +++ b/travis/before-install.sh @@ -0,0 +1,6 @@ +#!/usr/bin/env bash + +set -xe + +mkdir /pg +chown travis /pg \ No newline at end of file diff --git a/travis/before-script-user.sh b/travis/before-script-user.sh new file mode 100755 index 000000000..d9c07f1e4 --- /dev/null +++ b/travis/before-script-user.sh @@ -0,0 +1,7 @@ +#!/usr/bin/env bash + +set -xe + +ssh-keygen -t rsa -f ~/.ssh/id_rsa -q -N "" +cat ~/.ssh/id_rsa.pub >> ~/.ssh/authorized_keys +ssh-keyscan -H localhost >> ~/.ssh/known_hosts diff --git a/travis/before-script.sh b/travis/before-script.sh new file mode 100755 index 000000000..ca59bcf23 --- /dev/null +++ b/travis/before-script.sh @@ -0,0 +1,19 @@ +#!/usr/bin/env bash + +set -xe + +/etc/init.d/ssh start + +# Show pg_config path (just in case) +echo "############### pg_config path:" +which pg_config + +# Show pg_config just in case +echo "############### pg_config:" +pg_config + +# Show kernel parameters +echo "############### kernel params:" +cat /proc/sys/kernel/yama/ptrace_scope +sudo sysctl kernel.yama.ptrace_scope=0 +cat /proc/sys/kernel/yama/ptrace_scope diff --git a/travis/install.sh b/travis/install.sh new file mode 100755 index 000000000..43ada47b7 --- /dev/null +++ b/travis/install.sh @@ -0,0 +1,66 @@ +#!/usr/bin/env bash + +set -xe + +if [ -z ${PG_VERSION+x} ]; then + echo PG_VERSION is not set! + exit 1 +fi + +if [ -z ${PG_BRANCH+x} ]; then + echo PG_BRANCH is not set! + exit 1 +fi + +if [ -z ${PTRACK_PATCH_PG_BRANCH+x} ]; then + PTRACK_PATCH_PG_BRANCH=OFF +fi + +# fix +sudo chown -R travis /home/travis/.ccache + +export PGHOME=/pg + +# Clone Postgres +echo "############### Getting Postgres sources:" +git clone https://github.com/postgres/postgres.git -b $PG_BRANCH --depth=1 + +# Clone ptrack +if [ "$PTRACK_PATCH_PG_BRANCH" != "OFF" ]; then + git clone https://github.com/postgrespro/ptrack.git -b master --depth=1 postgres/contrib/ptrack + export PG_PROBACKUP_PTRACK=ON +else + export PG_PROBACKUP_PTRACK=OFF +fi + +# Compile and install Postgres +echo "############### Compiling Postgres:" +cd postgres # Go to postgres dir +if [ "$PG_PROBACKUP_PTRACK" = "ON" ]; then + git apply -3 contrib/ptrack/patches/${PTRACK_PATCH_PG_BRANCH}-ptrack-core.diff +fi +CC='ccache gcc' CFLAGS="-Og" ./configure --prefix=$PGHOME \ + --cache-file=~/.ccache/configure-cache \ + --enable-debug --enable-cassert --enable-depend \ + --enable-tap-tests --enable-nls +make -s -j$(nproc) install +make -s -j$(nproc) -C contrib/ install + +# Override default Postgres instance +export PATH=$PGHOME/bin:$PATH +export LD_LIBRARY_PATH=$PGHOME/lib +export PG_CONFIG=$(which pg_config) + +if [ "$PG_PROBACKUP_PTRACK" = "ON" ]; then + echo "############### Compiling Ptrack:" + make -C contrib/ptrack install +fi + +# Get amcheck if missing +if [ ! -d "contrib/amcheck" ]; then + echo "############### Getting missing amcheck:" + git clone https://github.com/petergeoghegan/amcheck.git --depth=1 contrib/amcheck + make -C contrib/amcheck install +fi + +pip3 install testgres \ No newline at end of file diff --git a/travis/script.sh b/travis/script.sh new file mode 100755 index 000000000..31ef09726 --- /dev/null +++ b/travis/script.sh @@ -0,0 +1,41 @@ +#!/usr/bin/env bash + +set -xe + +export PGHOME=/pg +export PG_SRC=$PWD/postgres +export PATH=$PGHOME/bin:$PATH +export LD_LIBRARY_PATH=$PGHOME/lib +export PG_CONFIG=$(which pg_config) + +# Build and install pg_probackup (using PG_CPPFLAGS and SHLIB_LINK for gcov) +echo "############### Compiling and installing pg_probackup:" +# make USE_PGXS=1 PG_CPPFLAGS="-coverage" SHLIB_LINK="-coverage" top_srcdir=$CUSTOM_PG_SRC install +make USE_PGXS=1 top_srcdir=$PG_SRC install + +if [ -z ${MODE+x} ]; then + MODE=basic +fi + +if [ -z ${PGPROBACKUP_GDB+x} ]; then + PGPROBACKUP_GDB=ON +fi + +echo "############### Testing:" +echo PG_PROBACKUP_PARANOIA=${PG_PROBACKUP_PARANOIA} +echo ARCHIVE_COMPRESSION=${ARCHIVE_COMPRESSION} +echo PGPROBACKUPBIN_OLD=${PGPROBACKUPBIN_OLD} +echo PGPROBACKUPBIN=${PGPROBACKUPBIN} +echo PGPROBACKUP_SSH_REMOTE=${PGPROBACKUP_SSH_REMOTE} +echo PGPROBACKUP_GDB=${PGPROBACKUP_GDB} +echo PG_PROBACKUP_PTRACK=${PG_PROBACKUP_PTRACK} + +if [ "$MODE" = "basic" ]; then + export PG_PROBACKUP_TEST_BASIC=ON + echo PG_PROBACKUP_TEST_BASIC=${PG_PROBACKUP_TEST_BASIC} + python3 -m unittest -v tests + python3 -m unittest -v tests.init_test +else + echo PG_PROBACKUP_TEST_BASIC=${PG_PROBACKUP_TEST_BASIC} + python3 -m unittest -v tests.$MODE +fi