5 Postgres DBA Tips

© 2016 EDB. All rights reserved. 1
5 Postgres DBA Tips
• Dave Page | 2016-10-20

• EnterpriseDB:
− Vice President & Chief Architect, Tools & Installers
− EDB Postgres Enterprise Manager
− EDB Failover Manager
− Cloud Products (EDB Ark, OpenShift containers)
− Packaging and distribution (installers, RPMs, StackBuilder etc)
− Configuration Management
• PostgreSQL:
− Core Team Member
− pgAdmin Project Lead
− Secretary of the board, PostgreSQL Europe
− Chairman of the board, PostgreSQL Community Association of Canada
− Webmaster and Sysadmin teams
Introduction – Dave Page

• The leading Open Source management and administration application for
Postgres: 
• Browse the database structure
• Create, edit and drop Postgres objects such as databases, tables and functions
• View current system activity in real time
• View and edit table data
• Execute arbitrary SQL queries and view the results
pgAdmin

• Includes all features of pgAdmin, and adds: 
• A powerful SQL profiler for examining server workload
• 24x7 monitoring and alerting
• Dashboards for visualising monitoring data
• Ad-hoc reporting with trend analysis for capacity management
• Tools for en-masse management:
• Configuration
• Software and upgrade deployment
• Static analysis
• Log analysis
Postgres Enterprise Manager (PEM)

• pgAdmin 4 released with PostgreSQL 9.6
• PEM 7 will include an updated UI based on pgAdmin 4, and is due for
release in Q1 2017 
• This presentation is based on pgAdmin III and PEM 6 for ease of
comparison
• All features shown will also be available in pgAdmin 4 and PEM 7
Roadmap

• Standard functionality for ad-hoc use in pgAdmin
• Integrated into the SQL Profiler in PEM
• Allows quick and easy visualisation of plans when developing SQL:
• Renders each node in the plan as an icon depicting the type of operation (e.g.
sequential scan, hash join, aggregate etc)
• Popup panels show node info (e.g. objects, filter data, row estimates and
estimated costs)
• Variable-thickness arrows show the flow of data from one node into the next,
giving visual cues to the volume of data
Visual Query Plans

Demo #1
Visual Query Plans

• Included with the PEM fat client
• Can optionally be used with a PEM server in order to schedule profiling of
workloads, for example, to examine the execution of a nightly batch job
• Requires a plugin on the database server:
• Lightweight
• No locking
• Profiles only the required combination of databases and/or users
• Safety features prevent storage space exhaustion when profiling busy
systems
SQL Profiler (1)

• Records 20+ statistics as well as the query plan and query text for all
appropriate queries executed during the profiling run
• Browse the captured data, with powerful filtering capabilities to help
discover expensive queries
• View the query text executed, and text and graphical plans
• View aggregate statistics, e.g:
• This query executed 27 times during the run
• This query was responsible for 58% of cache misses during the run
• Integrates with the Index Advisor when used with EDB Postgres Advanced
Server
SQL Profiler (2)

Demo #2
SQL Profiler

• Probes execute on schedules, collecting one or more metrics about the
system
• Each probe operates at a specific level of the logical object hierarchy:
• Host/OS
• Database Server
• Database
• Schema
• Each individual instance of a probe may be tuned:
• Enabled/disabled
• Execution frequency
• Data retention
24x7 Monitoring (1)

• Probes may be:
• Internal (implemented as part of the PEM Agent in C++)
• SQL
• WMI
• Batch/shell
• Custom probes may be implemented in SQL, WMI or Batch/shell to
monitor domain-specific metrics
• Individual metrics may be cumulative or point-in-time values
• Individual metrics may be graph-able numerics, or other datatypes such
as strings
24x7 Monitoring (2)

Demo #3
Probe Configuration
& Dashboards

• Metric data may be used in the Capacity Manager for visualisation
• Save/load report templates
• Re-use graph definitions in custom dashboards
• Plot multiple metrics from anywhere in the logical object hierarchy
• Data aggregation method selected on a per-metric basis
• Specify timeframe based on:
• Start -> End date/time (past, present or future)
• Start date -> Point at which a metric exceeds or falls below a value
• Data that exceeds the recorded available data will be extrapolated using a
linear trend analysis algorithm to allow predictive reporting
Capacity Management

Demo #4
Capacity Manager

• An alerting engine constantly monitors collected metrics for values that fall
outside acceptable minimum or maximum levels
• Templates are used to define alerts on each object type, tailored to the
needs of each instance of the object
• Threshold values may be specified for low, medium and high level alerts,
along with an operator to denote whether low or high values are to be
detected
• Some templates allow additional parameters to be defined, for example,
“Number of ERRORs in the Postgres Log in the last N minutes
• Users can create custom alert templates from any data collected for which
a scalar numeric value can be calculated
Alerting (1)

• Alerts can be sent to users:
• Visually, through the PEM dashboards
• Via SNMP Traps to an upstream monitoring system (a database function is
included to generate a MIB file)
• Via SMTP to email recipients, grouped as needed to cover different alert levels
or hours of the day
• Via Nagios passive service checks to a Nagios installation (database functions
are included to generate host and service definitions)
• Scripts may be executed in response to alert level changes:
• Either on the affected host, for example, to restart a service
• or on the PEM server, for example, to call the RESTful API of an SMS service
provider or chat bot in a sysadmin channel/chatroom
Alerting (2)

Demo #5
Alerting

Questions?

5 Postgres DBA Tips

More Related Content

What's hot (20)

Viewers also liked (6)

Similar to 5 Postgres DBA Tips (20)

More from EDB (20)

Recently uploaded (20)

5 Postgres DBA Tips