Date: Thu, 5 Apr 2007 09:33:16 -0400
From: Andrew Sullivan <ajs@crankycanuck.ca>
To: pgsql-sql@postgresql.org
Subject: Re: slow query
Message-ID: <20070405133316.GC6464@phlogiston.dyndns.org>
References: <7539aebb0704042017l70d83d53u9c2e5105620a3fa7@mail.gmail.com>
	<20070405051127.GA24814@a-kretschmer.de>
	<19157.1175750845@sss.pgh.pa.us>
	<20070405053637.GC24814@a-kretschmer.de>
	<19565.1175752023@sss.pgh.pa.us>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <19565.1175752023@sss.pgh.pa.us>
User-Agent: Mutt/1.5.9i

On Thu, Apr 05, 2007 at 01:47:03AM -0400, Tom Lane wrote:
> [ shrug... ]  If the damage is already done, lazy VACUUM won't fix it.
> 

Also, if there are enough open transactions at any one time and
sufficient churn in the table, lazy VACUUM may not be able to keep
up.  (We had that experience with a table that was updated _very
very_ often.  The answer turned out to be to update less often. 
Aggregating queries that could use an index over a large number of
"expired" rows worked better than seqscans over large numbers of dead
tuples.)

A
-- 
Andrew Sullivan  | ajs@crankycanuck.ca
Information security isn't a technological problem.  It's an economics
problem.
		--Bruce Schneier