Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Commit456bf26

Browse files
committed
Ignore old stats file timestamps when starting the stats collector.
The stats collector disregards inquiry messages that bear a cutoff_timebefore when it last wrote the relevant stats file. That's fine, but atstartup when it reads the "permanent" stats files, it absorbed theirtimestamps as if they were the times at which the corresponding temporarystats files had been written. In reality, of course, there's no dataout there at all. This led to disregarding inquiry messages soon afterstartup if the postmaster had been shut down and restarted within lessthan PGSTAT_STAT_INTERVAL; which is a pretty common scenario, both fortesting and in the field. Requesting backends would hang for 10 secondsand then report failure to read statistics, unless they got bailed outby some other backend coming along and making a newer request withinthat interval.I came across this through investigating unexpected delays in thesrc/test/recovery TAP tests: it manifests there because the autovacuumlauncher hangs for 10 seconds when it can't get statistics at startup,thus preventing a second shutdown from occurring promptly. We mightwant to do some things in the autovac code to make it less prone togetting stuck that way, but this change is a good bug fix regardless.In passing, also fix pgstat_read_statsfiles() to ensure that itre-zeroes its global stats variables if they are corrupted by ashort read from the stats file. (Other reads in that functiongo into temp variables, so that the issue doesn't arise.)This has been broken since we created the separation between permanentand temporary stats files in 8.4, so back-patch to all supported branches.Discussion:https://postgr.es/m/16860.1498442626@sss.pgh.pa.us
1 parent2563f07 commit456bf26

File tree

1 file changed

+20
-0
lines changed

1 file changed

+20
-0
lines changed

‎src/backend/postmaster/pgstat.c

Lines changed: 20 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3883,9 +3883,20 @@ pgstat_read_statsfiles(Oid onlydb, bool permanent, bool deep)
38833883
{
38843884
ereport(pgStatRunningInCollector ?LOG :WARNING,
38853885
(errmsg("corrupted statistics file \"%s\"",statfile)));
3886+
memset(&globalStats,0,sizeof(globalStats));
38863887
gotodone;
38873888
}
38883889

3890+
/*
3891+
* In the collector, disregard the timestamp we read from the permanent
3892+
* stats file; we should be willing to write a temp stats file immediately
3893+
* upon the first request from any backend. This only matters if the old
3894+
* file's timestamp is less than PGSTAT_STAT_INTERVAL ago, but that's not
3895+
* an unusual scenario.
3896+
*/
3897+
if (pgStatRunningInCollector)
3898+
globalStats.stats_timestamp=0;
3899+
38893900
/*
38903901
* We found an existing collector stats file. Read it and put all the
38913902
* hashtable entries into place.
@@ -3927,6 +3938,15 @@ pgstat_read_statsfiles(Oid onlydb, bool permanent, bool deep)
39273938
dbentry->tables=NULL;
39283939
dbentry->functions=NULL;
39293940

3941+
/*
3942+
* In the collector, disregard the timestamp we read from the
3943+
* permanent stats file; we should be willing to write a temp
3944+
* stats file immediately upon the first request from any
3945+
* backend.
3946+
*/
3947+
if (pgStatRunningInCollector)
3948+
dbentry->stats_timestamp=0;
3949+
39303950
/*
39313951
* Don't create tables/functions hashtables for uninteresting
39323952
* databases.

0 commit comments

Comments
 (0)
[8]ページ先頭
©2009-2025 Movatter.jp