Re: Autovacuum firing up during my manual vacuum on same table

Le 02/04/2011 16:06, Sven Haag a écrit :
>
> -------- Original-Nachricht --------
>> Datum: Sat, 2 Apr 2011 14:17:37 +0200
>> Von: Jens Wilke <jens(at)wilke(dot)org>
>> An: pgsql-general(at)postgresql(dot)org
>> Betreff: Re: [GENERAL] Autovacuum firing up during my manual vacuum on same table
>
>> On Samstag, 2. April 2011, Henry C. wrote:
>>
>>> I just noticed something odd: I'm busy with a manual vacuum on a
>>> table and an autovacuum keeps firing up as well.
>>
>> Usually a manual vacuum cancels a running autovacuum task.
>> You should find a notice about the cancelation in th logfile.
>>
>>> current_query | vacuum analyze
>>> age | 11:46:57.245568
>>
>> Where is the age column from?
>> It's not in pg_stat_activity.
>> Is one of the two processes waiting=t in pg_stat_activity?
>> Shure it's the same Table? Do you have one Table named
>> page_citation_text in public and one in anoter sheme?
>>
>>> I'm trying to vacuum the table as quickly as possible so a manual
>>> vacuum seems to be in order as my understanding (and experience)
>>> is that the autovac is a hell of a lot slower to mitigate impact
>>> on general performance.
>>
>> Tune the autovacuum settings, especially the scale factors of the
>> tables in question and the cost limit.
>> Is autovacuum_vacuum_cost_limit != -1 ?
>> This could be one reason, why manual vacuum is faster.
>>
>> Nevertheless since at least 8.4 IMO there's no need to bother with
>> manual vacuum any more.
>>
>> Regards, Jens
>
> a bit off-topic, but:
> i'm using pg 9 and always getting messages to vacuum tables in pgadmin. according to your post this souldn't be the case?
>

pgAdmin's hint doesn't take care of your PostgreSQL release. Its way to
guess that a table needs to be vacuumed is a bit outdated (that's
clearly an understatement :) ).

--
Guillaume
http://www.postgresql.fr
http://dalibo.com