Date: Thu, 26 Oct 2006 18:09:37 -0400
From: Andrew Sullivan <ajs@crankycanuck.ca>
To: pgsql-performance@postgresql.org
Subject: Re: VACUUMs take twice as long across all nodes
Message-ID: <20061026220937.GA6866@phlogiston.dyndns.org>
References: <20061026095827.9e85e1c0.gdh@laterooms.com>
	<20196.1161874041@sss.pgh.pa.us>
	<20061026160609.5e94e581.gdh@laterooms.com>
	<20061026191728.GD26892@nasby.net>
	<20061026213556.174bd7b5.gdh@laterooms.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20061026213556.174bd7b5.gdh@laterooms.com>
User-Agent: Mutt/1.5.9i

On Thu, Oct 26, 2006 at 09:35:56PM +0100, Gavin Hamill wrote:
> 
> I'm absolutely certain. The backups run from only one slave, given that
> it is a full copy of node 1. Our overnight traffic has not increased
> any, and the nightly backups show that the overall size of the DB has
> not increased more than usual growth.

A couple things from your posts:

1.	Don't do VACUUM FULL, please.  It takes longer, and blocks
other things while it's going on, which might mean you're having
table bloat in various slony-related tables.

2.	Are your slony logs showing increased time too?  Are your
targets getting further behind?

3.	Your backups "from the slave" aren't done with pg_dump,
right?

But I suspect Slony has a role here, too.  I'd look carefully at the
slony tables -- especially the sl_log and pg_listen things, which
both are implicated.

A

-- 
Andrew Sullivan  | ajs@crankycanuck.ca
"The year's penultimate month" is not in truth a good way of saying
November.
		--H.W. Fowler