To: <rafael@akyasociados.com.ar>
Cc: <mpartio@gmail.com>,  <pgsql-performance@postgresql.org>
Subject: Re: Vacum Analyze problem
In-Reply-To: 
 <1415.192.168.1.229.1188927997.squirrel@webmail.akyasociados.com.ar>
	(rafael@akyasociados.com.ar's message of "Tue\,
	4 Sep 2007 14\:46\:37 -0300 \(ART\)")
User-Agent: Gnus/5.11 (Gnus v5.11) Emacs/22.1 (gnu/linux)
References: <2ca799770709041033n664cb0ep3a5679548dac3a2e@mail.gmail.com>
	<1415.192.168.1.229.1188927997.squirrel@webmail.akyasociados.com.ar>
From: Gregory Stark <stark@enterprisedb.com>
Organization: EnterpriseDB
Date: Wed, 05 Sep 2007 00:24:32 +0100
Message-ID: <87bqci580f.fsf@oxford.xeocode.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii

<rafael@akyasociados.com.ar> writes:

> I knew that in the long run the VACUUM ANALYZE comand has to be executed
> again. My question is if something can happen over night and cause the need
> of a new VACUUM ANALYZE (regenerating indexes or other thing related with
> performance).

The answer to your question is possibly yes for two reasons:

1) If you're running an autovacuum daemon it might decide it's time to vacuum
the table and kick off a vacuum. That sounds most like what you're describing.

2) If the size of the table changes substantially becoming much larger (or
smaller but that wouldn't happen just due to deletes unless you run vacuum)
then recent versions of Postgres will notice even if you don't run analyze and
take that into account.

-- 
  Gregory Stark
  EnterpriseDB          http://www.enterprisedb.com