On Wed, Dec 7, 2011 at 8:34 AM, Anibal David Acosta <aa@devshock.com> wrote:
> Hello, I have a postgres 9.0.2 installation.
>
> Every works fine, but in some hours of day I got several timeout in my
> application (my application wait X seconds before throw a timeout).
>
> Normally hours are not of intensive use, so I think that the autovacuum
> could be the problem.
>
>
>
> Is threre any log where autovacuum write information about it self like
> “duration for each table” or any other relevante information.
>
>
>
> Another inline question, should I exclude bigger tables from autovacuum or
> there are some mechanism to tell autovacuum to not run often on bigger
> tables (tables with more than 400 millions of rows)
More often than not not the problem will be checkpoint segments not
autovacuum. log vacuum and checkpoints, and then run something like
iostat in the background and keep an eye on %util to see if one or the
other is slamming your IO subsystem. Default tuning for autovac is
pretty conservative, to the point that it won't usually hurt your IO,
but may not keep up with vaccuming, leading to table bloating.