Message-ID: <1114121722.426825fa4474f@webmail.telus.net>
Date: Thu, 21 Apr 2005 15:15:22 -0700
From: Mischa Sandberg <mischa.sandberg@telus.net>
To: Bill Chandler <billybobc1210@yahoo.com>
Cc: a3a18850@telus.net,
	pgsql-perform <pgsql-performance@postgresql.org>
Subject: Re: Index bloat problem?
References: <20050421190318.26022.qmail@web51404.mail.yahoo.com>
In-Reply-To: <20050421190318.26022.qmail@web51404.mail.yahoo.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 8bit
User-Agent: Internet Messaging Program (IMP) 3.1-cvs

Quoting Bill Chandler <billybobc1210@yahoo.com>:

> ... The normal activity is to delete 3-5% of the rows per day,
> followed by a VACUUM ANALYZE. 
...
> However, on occasion, deleting 75% of rows is a 
> legitimate action for the client to take.  

> > In case nobody else has asked: is your max_fsm_pages
> > big enough to handle all the deleted pages, 
> > across ALL tables hit by the purge?

> This parameter is most likely set incorrectly.  So
> that could be causing problems.  Could that be a
> culprit for the index bloat, though?

Look at the last few lines of vacuum verbose output.
It will say something like:

free space map: 55 relations, 88416 pages stored; 89184 total pages needed
  Allocated FSM size: 1000 relations + 1000000 pages = 5920 kB shared memory.

"1000000" here is [max_fsm_pages] from my postgresql.conf.
If the "total pages needed" is bigger than the pages 
fsm is allocated for, then you are bleeding.
-- 
"Dreams come true, not free." -- S.Sondheim, ITW