Hi,
On 2023-08-07 12:57:40 +0200, Christoph Berg wrote:
> Re: Thomas Munro
> > Thanks for testing! Would you mind trying v8 from that thread? V7
> > had a silly bug (I accidentally deleted a 'case' label while cleaning
> > some stuff up, resulting in the above error...)
>
> v8 worked better. It succeeded a few times (at least 12, my screen
> scrollback didn't catch more) before erroring like this:
> [10:21:58.410](0.151s) ok 15 - startup deadlock: logfile contains terminated
> connection due to recovery conflict
> [10:21:58.463](0.053s) not ok 16 - startup deadlock: stats show conflict on
> standby
> [10:21:58.463](0.000s)
> [10:21:58.463](0.000s) # Failed test 'startup deadlock: stats show conflict
> on standby'
> # at t/031_recovery_conflict.pl line 332.
> [10:21:58.463](0.000s) # got: '0'
> # expected: '1'
Hm, that could just be a "harmless" race. Does it still happen if you apply
the attached patch in addition?
Greetings,
Andres Freund
diff --git i/src/backend/utils/activity/pgstat_database.c w/src/backend/utils/activity/pgstat_database.c
index 7149f22f729..bb36d73ec04 100644
--- i/src/backend/utils/activity/pgstat_database.c
+++ w/src/backend/utils/activity/pgstat_database.c
@@ -81,12 +81,22 @@ void
pgstat_report_recovery_conflict(int reason)
{
PgStat_StatDBEntry *dbentry;
+ PgStat_EntryRef *entry_ref;
+ PgStatShared_Database *sharedent;
Assert(IsUnderPostmaster);
if (!pgstat_track_counts)
return;
- dbentry = pgstat_prep_database_pending(MyDatabaseId);
+ /*
+ * Update the shared stats directly - recovery conflicts should never be
+ * common enough for that to be a problem.
+ */
+ entry_ref =
+ pgstat_get_entry_ref_locked(PGSTAT_KIND_DATABASE, MyDatabaseId, InvalidOid, false);
+
+ sharedent = (PgStatShared_Database *) entry_ref->shared_stats;
+ dbentry = &sharedent->stats;
switch (reason)
{
@@ -116,6 +126,8 @@ pgstat_report_recovery_conflict(int reason)
dbentry->conflict_startup_deadlock++;
break;
}
+
+ pgstat_unlock_entry(entry_ref);
}
/*
