Date: Tue, 5 Jun 2007 08:47:12 -0400
From: Alvaro Herrera <alvherre@commandprompt.com>
To: "Charles.Hou" <ivan.hou@msa.hinet.net>
Cc: pgsql-admin@postgresql.org
Subject: Re: Auto Vacuum question?
Message-ID: <20070605124712.GA3859@alvh.no-ip.org>
References: <1180966368.172224.250500@x35g2000prf.googlegroups.com>
	<20070604144906.GH4779@alvh.no-ip.org>
	<1181002036.120310.83340@r19g2000prf.googlegroups.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <1181002036.120310.83340@r19g2000prf.googlegroups.com>
User-Agent: Mutt/1.5.13 (2006-08-11)

Charles.Hou wrote:
> On 6 4 ,   10 49 , alvhe...@commandprompt.com (Alvaro Herrera) wrote:

> > It is normal that the tables grow a bit to a stationary state of dead
> > space (generated by UPDATE and DELETE).  However it is not normal if it
> > loses track of that dead space.  One thing you can and should do is
> > adjust the autovacuum settings (for example lowering the vacuum scale
> > factor and threshold values).  Another is to correctly set the FSM
> > values.

> autovacuum_vacuum_scale_factor = 1      # fraction of rel size before
> autovacuum_analyze_scale_factor = 1     # fraction of rel size before

These values are rather strange; they mean it would wait until the table
is populated by as many dead tuples as there are live tuples, before
vacuuming it.  This is probably too much.  Try decreasing the vacuum
scale factor to 0.2 or maybe 0.1 and see how it goes.

-- 
Alvaro Herrera                 http://www.amazon.com/gp/registry/DXLWNGRJD34J
"La fuerza no está en los medios físicos
sino que reside en una voluntad indomable" (Gandhi)