Re: Vacuum v/s Autovacuum

You will need to properly tune the thresholds for VACUUM and ANALYZE in case
of autovacuuming process, so that you do not need to increase the
max_fsm_pages oftenly...

-------------
Shoaib Mir
EnterpriseDB (www.enterprisedb.com)

On 1/18/07, Bill Moran <wmoran(at)collaborativefusion(dot)com> wrote:
>
> In response to "Gauri Kanekar" <meetgaurikanekar(at)gmail(dot)com>:
> > On 1/18/07, Michael Glaesemann <grzm(at)seespotcode(dot)net> wrote:
> > >
> > > On Jan 18, 2007, at 22:24 , Gauri Kanekar wrote:
> > >
> > > > is autovacuum similar to vacuum full analyse verbose.
> > >
> > > http://www.postgresql.org/docs/8.2/interactive/routine-
> > > vacuuming.html#AUTOVACUUM
> > >
> > > Apparently, no FULL, no VERBOSE (which is only really useful if you
> > > want to see the results, not for routine maintenance).
>
> [please don't top-post]
>
> Actually, you can raise the debugging level in PostgreSQL and get
> something
> similar to VERBOSE. The only problem is that it also increases the amount
> of logging that occurs with everything.
>
> > We have autovacuum ON , but still postgres server warns to
> > increas max_fsm_pages value.
> >
> > Do autovacuum release space after it is over?
>
> Yes.
>
> If you're still getting warnings about max_fsm_pages while autovac is
> running, you need to do one of two things:
> 1) Increase max_fsm_pages
> 2) Adjust autovacuum's settings so it vacuums more often.
>
> Depending on this, you may also need to temporarily adjust max_fsm_pages,
> then manually vacuum -- you may then find that autovacuum can keep
> everything
> clean with lower settings of max_fsm_pages.
>
> Overall, the best settings for 1 and 2 depend on the nature of your
> workload, and simulation and monitoring will be required to find the
> best values. I feel that the docs on this are very good. If the
> amount of data that changes between runs of autovacuum is greater
> than max_fsm_pages, then vacuum will be unable to reclaim all the
> space.
>
> --
> Bill Moran
> Collaborative Fusion Inc.
>
> ---------------------------(end of broadcast)---------------------------
> TIP 2: Don't 'kill -9' the postmaster
>