Re: We need to log aborted autovacuums

On Sun, Jan 30, 2011 at 10:26 PM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> On Sun, Jan 30, 2011 at 10:03 PM, Alvaro Herrera
> <alvherre(at)commandprompt(dot)com> wrote:
>> Excerpts from Robert Haas's message of dom ene 30 23:37:51 -0300 2011:
>>
>>> Unless I'm missing something, making autovacuum.c call
>>> ConditionalLockRelationOid() is not going to work, because the vacuum
>>> transaction isn't started until we get all the way down to
>>> vacuum_rel().
>>
>> Maybe we need ConditionalLockRelationOidForSession or something like
>> that?
>
> That'd be another way to go, if there are objections to what I've
> implemented here.

Seeing as how there seem to be neither objections nor endorsements,
I'm inclined to commit what I proposed more or less as-is. There
remains the issue of what do about the log spam. Josh Berkus
suggested logging it when log_autovacuum_min_duration != -1, which
seems like a bit of an abuse of that setting, but it's certainly not
worth adding another setting for, and the alternative of logging it
at, say, DEBUG2 seems unappealing because you'll then have to turn on
logging for a lot of unrelated crap to get this information. So on
balance I think that proposal is perhaps the least of evils.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company