Date: Wed, 10 Jan 2007 21:42:00 -0600
From: "Jim C. Nasby" <jim@nasby.net>
To: Jeremy Haile <jhaile@fastmail.fm>, pgsql-performance@postgresql.org
Subject: Re: High inserts, bulk deletes - autovacuum vs scheduled vacuum
Message-ID: <20070111034159.GJ12217@nasby.net>
References: <1168363601.5177.1168380907@webmail.messagingengine.com>
	<20070110203556.GX12217@nasby.net>
	<1168461780.9408.1168619515@webmail.messagingengine.com>
	<20070110212125.GZ12217@nasby.net>
	<1168465722.17889.1168630713@webmail.messagingengine.com>
	<20070111012709.GG12217@nasby.net>
	<20070111031034.GG15378@alvh.no-ip.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20070111031034.GG15378@alvh.no-ip.org>
User-Agent: Mutt/1.5.13 (2006-08-11)

On Thu, Jan 11, 2007 at 12:10:34AM -0300, Alvaro Herrera wrote:
> Jim C. Nasby wrote:
> 
> > > Is the best way to do that usually to lower the scale factors?  Is it
> > > ever a good approach to lower the scale factor to zero and just set the
> > > thresholds to a pure number of rows? (when setting it for a specific
> > > table)
> > 
> > The problem is what happens if autovac goes off and starts vacuuming
> > some large table? While that's going on your queue table is sitting
> > there bloating. If you have a separate cronjob to handle the queue
> > table, it'll stay small, especially in 8.2.
> 
> You mean "at least in 8.2".  In previous releases, you could vacuum
> that queue table until you were blue on the face, but it would achieve
> nothing because it would consider that the dead tuples were visible to a
> running transaction: that running the vacuum on the large table.  This
> is an annoyance that was fixed in 8.2.

True, but in many environments there are other transactions that run
long enough that additional vacuums while a long vacuum was running
would still help.
-- 
Jim Nasby                                            jim@nasby.net
EnterpriseDB      http://enterprisedb.com      512.569.9461 (cell)