Avoid having vacuum set reltuples to 0 on non-empty relations in the

presence of page pins, which leads to serious estimation errors in the
planner.  This particularly affects small heavily-accessed tables,
especially where locking (e.g. from FK constraints) forces frequent
vacuums for mxid cleanup.

Fix by keeping separate track of pages whose live tuples were actually
counted vs. pages that were only scanned for freezing purposes.  Thus,
reltuples can only be set to 0 if all pages of the relation were
actually counted.

Backpatch to all supported versions.

Per bug #14057 from Nicolas Baccelli, analyzed by me.

Discussion: https://postgr.es/m/20160331103739[email protected]

diff --git a/src/backend/commands/vacuumlazy.c b/src/backend/commands/vacuumlazy.c
index 64626edf702bb2e8c406c91d61236da2438a58d6..2654c9ebef4a498c650b661bdfb0b13a65035c0c 100644 (file)
--- a/src/backend/commands/vacuumlazy.c
+++ b/src/backend/commands/vacuumlazy.c
@@ -106,7 +106,8 @@ typedef struct LVRelStats
        BlockNumber rel_pages;          /* total number of pages */
        BlockNumber scanned_pages;      /* number of pages we examined */
        BlockNumber pinskipped_pages;           /* # of pages we skipped due to a pin */
-       double          scanned_tuples; /* counts only tuples on scanned pages */
+       BlockNumber     tupcount_pages; /* pages whose tuples we counted */
+       double          scanned_tuples; /* counts only tuples on tupcount_pages */
        double          old_rel_tuples; /* previous value of pg_class.reltuples */
        double          new_rel_tuples; /* new estimated total # of tuples */
        double          new_dead_tuples;        /* new estimated total # of dead tuples */
@@ -286,6 +287,10 @@ lazy_vacuum_rel(Relation onerel, int options, VacuumParams *params,
         * density") with nonzero relpages and reltuples=0 (which means "zero
         * tuple density") unless there's some actual evidence for the latter.
         *
+        * It's important that we use tupcount_pages and not scanned_pages for the
+        * check described above; scanned_pages counts pages where we could not get
+        * cleanup lock, and which were processed only for frozenxid purposes.
+        *
         * We do update relallvisible even in the corner case, since if the table
         * is all-visible we'd definitely like to know that.  But clamp the value
         * to be not more than what we're setting relpages to.
@@ -295,7 +300,7 @@ lazy_vacuum_rel(Relation onerel, int options, VacuumParams *params,
         */
        new_rel_pages = vacrelstats->rel_pages;
        new_rel_tuples = vacrelstats->new_rel_tuples;
-       if (vacrelstats->scanned_pages == 0 && new_rel_pages > 0)
+       if (vacrelstats->tupcount_pages == 0 && new_rel_pages > 0)
        {
                new_rel_pages = vacrelstats->old_rel_pages;
                new_rel_tuples = vacrelstats->old_rel_tuples;
@@ -474,6 +479,7 @@ lazy_scan_heap(Relation onerel, LVRelStats *vacrelstats,
        nblocks = RelationGetNumberOfBlocks(onerel);
        vacrelstats->rel_pages = nblocks;
        vacrelstats->scanned_pages = 0;
+       vacrelstats->tupcount_pages = 0;
        vacrelstats->nonempty_pages = 0;
        vacrelstats->latestRemovedXid = InvalidTransactionId;
 
@@ -666,6 +672,7 @@ lazy_scan_heap(Relation onerel, LVRelStats *vacrelstats,
                }
 
                vacrelstats->scanned_pages++;
+               vacrelstats->tupcount_pages++;
 
                page = BufferGetPage(buf);
 
@@ -1074,7 +1081,7 @@ lazy_scan_heap(Relation onerel, LVRelStats *vacrelstats,
        /* now we can compute the new value for pg_class.reltuples */
        vacrelstats->new_rel_tuples = vac_estimate_reltuples(onerel, false,
                                                                                                                 nblocks,
-                                                                                                 vacrelstats->scanned_pages,
+                                                                                                 vacrelstats->tupcount_pages,
                                                                                                                 num_tuples);
 
        /*
@@ -1393,7 +1400,7 @@ lazy_cleanup_index(Relation indrel,
 
        ivinfo.index = indrel;
        ivinfo.analyze_only = false;
-       ivinfo.estimated_count = (vacrelstats->scanned_pages < vacrelstats->rel_pages);
+       ivinfo.estimated_count = (vacrelstats->tupcount_pages < vacrelstats->rel_pages);
        ivinfo.message_level = elevel;
        ivinfo.num_heap_tuples = vacrelstats->new_rel_tuples;
        ivinfo.strategy = vac_strategy;