Re: How to prevent vacuum and reindex from deadlocking.

On Sun, 10 Aug 2003 14:50:10 -0600
Robert Creager <Robert_Creager(at)LogicalChaos(dot)org> said something like:

>
> I'm running 7.4Beta1 with pg_autovacuum. In one of my operations
> which is executed frequently, a REINDEX is done after a COPY. Well,
> VACUUM's are being executed by pg_autovacuum, and my app is loosing
> the deadlock.
>
> Is there a way I can prevent this from happening? I can obviously
> acquire a LOCK before doing the REINDEX (which one?), but it looks
> like this should be happening anyway by both the VACUUM and REINDEX
> statements.
>

Opps, if it helps, the log of the deadlock:

Aug 10 14:19:36 thunder postgres[18735]: [2-1] ERROR: deadlock detected

Aug 10 14:19:36 thunder postgres[18735]: [2-2] DETAIL: Proc 18735 waits
for AccessExclusiveLock on relation 18028 of database 17140; blocked by
proc 18815.

Aug 10 14:19:36 thunder postgres[18735]: [2-3] Proc 18815 waits for
AccessExclusiveLock on relation 18101 of database 17140; blocked by proc
18735.

--
14:55:41 up 9 days, 7:41, 4 users, load average: 1.56, 1.27, 1.08