I did not find any errors pertaining to the autovacuum task.
In both cases we had a data-centric issue which was causing excessive
errors on the database (we have a timestamp field and an external
utility which we use to collect product keys was sending the date as
"dd/mm/yyyy hh:mm:ss" for some entries. We modified our processing
script to catch this condition and rewrite it to "mm/dd/yyyy hh:mm:ss"
on the server script, and I will be modifying the agent to send it in as
"yyyymmdd hh:mm:ss" in our next agent release to bullet proof it.
> -----Original Message-----
> From: Alvaro Herrera [mailto:alvherre@commandprompt.com]
> Sent: Monday, May 03, 2010 7:48 AM
> To: Benjamin Krajmalnik
> Cc: Tom Lane; pgsql-admin@postgresql.org
> Subject: Re: [ADMIN] Autovacuum stopped running
>
> Benjamin Krajmalnik wrote:
> > OK.
> > This happened again on another server with these same settings.
> > I did not see any long running transaction, and the autovacuum
launch
> > process was running.
>
> Did you check the server log for FATAL or ERROR lines? Maybe it's
> dying
> trying to vacuum some table and isn't ever able to get to the next
> ones.
>
>
> --
> Alvaro Herrera
> http://www.CommandPrompt.com/
> PostgreSQL Replication, Consulting, Custom Development, 24x7 support