database/postgres
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2025-07-03 20:02:46 +03:00
Code Activity

Fix VACUUM so that it always updates pg_class.reltuples/relpages.

Browse Source 
When we added the ability for vacuum to skip heap pages by consulting the
visibility map, we made it just not update the reltuples/relpages
statistics if it skipped any pages.  But this could leave us with extremely
out-of-date stats for a table that contains any unchanging areas,
especially for TOAST tables which never get processed by ANALYZE.  In
particular this could result in autovacuum making poor decisions about when
to process the table, as in recent report from Florian Helmberger.  And in
general it's a bad idea to not update the stats at all.  Instead, use the
previous values of reltuples/relpages as an estimate of the tuple density
in unvisited pages.  This approach results in a "moving average" estimate
of reltuples, which should converge to the correct value over multiple
VACUUM and ANALYZE cycles even when individual measurements aren't very
good.

This new method for updating reltuples is used by both VACUUM and ANALYZE,
with the result that we no longer need the grotty interconnections that
caused ANALYZE to not update the stats depending on what had happened
in the parent VACUUM command.

Also, fix the logic for skipping all-visible pages during VACUUM so that it
looks ahead rather than behind to decide what to do, as per a suggestion
from Greg Stark.  This eliminates useless scanning of all-visible pages at
the start of the relation or just after a not-all-visible page.  In
particular, the first few pages of the relation will not be invariably
included in the scanned pages, which seems to help in not overweighting
them in the reltuples estimate.

Back-patch to 8.4, where the visibility map was introduced.
This commit is contained in:
Tom Lane
2011-05-30 17:05:26 -04:00
parent 3001b76308
commit b4b6923e03
6 changed files with 228 additions and 129 deletions
Download Patch File Download Diff File Expand all files Collapse all files

59
src/backend/commands/analyze.c
View File

@ -84,8 +84,7 @@ static MemoryContext anl_context = NULL;
static BufferAccessStrategy vac_strategy;
static void do_analyze_rel(Relation onerel, VacuumStmt *vacstmt,
bool update_reltuples, bool inh);
static void do_analyze_rel(Relation onerel, VacuumStmt *vacstmt, bool inh);
static void BlockSampler_Init(BlockSampler bs, BlockNumber nblocks,
int samplesize);
static bool BlockSampler_HasMore(BlockSampler bs);
@ -115,18 +114,9 @@ static bool std_typanalyze(VacAttrStats *stats);
/*
* analyze_rel() -- analyze one relation
*
* If update_reltuples is true, we update reltuples and relpages columns
* in pg_class. Caller should pass false if we're part of VACUUM ANALYZE,
* and the VACUUM didn't skip any pages. We only have an approximate count,
* so we don't want to overwrite the accurate values already inserted by the
* VACUUM in that case. VACUUM always scans all indexes, however, so the
* pg_class entries for indexes are never updated if we're part of VACUUM
* ANALYZE.
*/
void
analyze_rel(Oid relid, VacuumStmt *vacstmt,
BufferAccessStrategy bstrategy, bool update_reltuples)
analyze_rel(Oid relid, VacuumStmt *vacstmt, BufferAccessStrategy bstrategy)
{
Relation onerel;
@ -238,13 +228,13 @@ analyze_rel(Oid relid, VacuumStmt *vacstmt,
/*
* Do the normal non-recursive ANALYZE.
*/
do_analyze_rel(onerel, vacstmt, update_reltuples, false);
do_analyze_rel(onerel, vacstmt, false);
/*
* If there are child tables, do recursive ANALYZE.
*/
if (onerel->rd_rel->relhassubclass)
do_analyze_rel(onerel, vacstmt, false, true);
do_analyze_rel(onerel, vacstmt, true);
/*
* Close source relation now, but keep lock so that no one deletes it
@ -267,8 +257,7 @@ analyze_rel(Oid relid, VacuumStmt *vacstmt,
* do_analyze_rel() -- analyze one relation, recursively or not
*/
static void
do_analyze_rel(Relation onerel, VacuumStmt *vacstmt,
bool update_reltuples, bool inh)
do_analyze_rel(Relation onerel, VacuumStmt *vacstmt, bool inh)
{
int attr_cnt,
tcnt,
@ -437,9 +426,9 @@ do_analyze_rel(Relation onerel, VacuumStmt *vacstmt,
}
/*
* Quit if no analyzable columns and no pg_class update needed.
* Quit if no analyzable columns.
*/
if (attr_cnt <= 0 && !analyzableindex && !update_reltuples)
if (attr_cnt <= 0 && !analyzableindex)
goto cleanup;
/*
@ -549,10 +538,10 @@ do_analyze_rel(Relation onerel, VacuumStmt *vacstmt,
}
/*
* Update pages/tuples stats in pg_class, but not if we're inside a VACUUM
* that got a more precise number.
* Update pages/tuples stats in pg_class ... but not if we're doing
* inherited stats.
*/
if (update_reltuples)
if (!inh)
vac_update_relstats(onerel,
RelationGetNumberOfBlocks(onerel),
totalrows, hasindex, InvalidTransactionId);
@ -562,7 +551,7 @@ do_analyze_rel(Relation onerel, VacuumStmt *vacstmt,
* VACUUM ANALYZE, don't overwrite the accurate count already inserted by
* VACUUM.
*/
if (!(vacstmt->options & VACOPT_VACUUM))
if (!inh && !(vacstmt->options & VACOPT_VACUUM))
{
for (ind = 0; ind < nindexes; ind++)
{
@ -577,13 +566,12 @@ do_analyze_rel(Relation onerel, VacuumStmt *vacstmt,
}
/*
* Report ANALYZE to the stats collector, too; likewise, tell it to adopt
* these numbers only if we're not inside a VACUUM that got a better
* number. However, a call with inh = true shouldn't reset the stats.
* Report ANALYZE to the stats collector, too. However, if doing
* inherited stats we shouldn't report, because the stats collector only
* tracks per-table stats.
*/
if (!inh)
pgstat_report_analyze(onerel, update_reltuples,
totalrows, totaldeadrows);
pgstat_report_analyze(onerel, totalrows, totaldeadrows);
/* We skip to here if there were no analyzable columns */
cleanup:
@ -1243,18 +1231,19 @@ acquire_sample_rows(Relation onerel, HeapTuple *rows, int targrows,
qsort((void *) rows, numrows, sizeof(HeapTuple), compare_rows);
/*
* Estimate total numbers of rows in relation.
* Estimate total numbers of rows in relation. For live rows, use
* vac_estimate_reltuples; for dead rows, we have no source of old
* information, so we have to assume the density is the same in unseen
* pages as in the pages we scanned.
*/
*totalrows = vac_estimate_reltuples(onerel, true,
totalblocks,
bs.m,
liverows);
if (bs.m > 0)
{
*totalrows = floor((liverows * totalblocks) / bs.m + 0.5);
*totaldeadrows = floor((deadrows * totalblocks) / bs.m + 0.5);
}
*totaldeadrows = floor((deadrows / bs.m) * totalblocks + 0.5);
else
{
*totalrows = 0.0;
*totaldeadrows = 0.0;
}
/*
* Emit some interesting relation info