Message-ID: <44F2B737.7070301@logix-tt.com>
Date: Mon, 28 Aug 2006 11:28:23 +0200
From: Markus Schaber <schabi@logix-tt.com>
Organization: Logical Tracking and Tracing International AG, Switzerland
User-Agent: Thunderbird 1.5.0.4 (X11/20060713)
MIME-Version: 1.0
To: Guillaume Cottenceau <gc@mnc.ch>
Cc: pgsql-performance@postgresql.org
Subject: Re: perf pb solved only after pg_dump and restore
References: <8764gd1dkz.fsf@meuh.mnc.lan>
In-Reply-To: <8764gd1dkz.fsf@meuh.mnc.lan>
Content-Type: text/plain; charset=ISO-8859-15
Content-Transfer-Encoding: 7bit

Hi, Guillaume

Guillaume Cottenceau wrote:

> We noticed a slowdown on our application while traffic was kinda
> heavy. The logics after reading the docs commanded us to trim the
> enlarged tables, run VACUUM ANALYZE and then expect fast
> performance again; but it wasn't the case[1].

What exactly do you mean with "trim the enlarged tables"?

> Out of the blue, we dumped the database, removed it, recreated
> from the restore, and now the performance is lightning fast
> again.
> 
> Does it look familiar to anyone? I thought running VACUUM ANALYZE
> after a trim should be enough so that pg has assembled the data
> and has good statistical knowledge of the tables contents..

This looks like either your free_space_map setting is way to low, or you
have index bloat.

Maybe a VACUUM FULL fullowed by a REINDEX will have solved your problem.

It also might make sense to issue a CLUSTER instead (which combines the
effects of VACUUM FULL, REINDEX and physically reordering the data).

When the free_space_map is to low, VACUUM ANALYZE should have told you
via a warning (at least, if your logging is set appropriately).


HTH,
Markus
-- 
Markus Schaber | Logical Tracking&Tracing International AG
Dipl. Inf.     | Software Development GIS

Fight against software patents in EU! www.ffii.org www.nosoftwarepatents.org