Message-ID: <4D645702.40203@2ndquadrant.com>
Date: Tue, 22 Feb 2011 19:38:26 -0500
From: Greg Smith <greg@2ndquadrant.com>
User-Agent: Thunderbird 2.0.0.24 (X11/20100317)
MIME-Version: 1.0
To: Sam Nelson <samn@consistentstate.com>
CC: pgsql-admin@postgresql.org
Subject: Re: Checkpoint and Background Writer Statistics
References: <AANLkTikHjPw85v0_5yGy5xNcV3=oR2kX=bXmmH0=3uO-@mail.gmail.com>
In-Reply-To: <AANLkTikHjPw85v0_5yGy5xNcV3=oR2kX=bXmmH0=3uO-@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

Sam Nelson wrote:
> One of our clients has an lru_max_dirty setting of 1000 and a 
> bgwriter_delay of 200, but we're still seeing much higher growth in 
> the buffers_checkpoint than buffers_clean, with buffers_checkpoint 
> increasing at about 20 times the rate that buffers_clean is.  We don't 
> see very much growth in the maxwritten_clean, though - something like 
> one every couple of days.

The background writer can only clean data pages that haven't been used 
in a while.  In your case, it sounds like most of the things that are 
getting dirty are being touched enough that they never meet its 
criteria.  There's nothing wrong with that.  Writing out dirty buffers 
only once per checkpoint is in theory the most efficient way to handle 
regularly changed data.

-- 
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support  www.2ndQuadrant.us
"PostgreSQL 9.0 High Performance": http://www.2ndQuadrant.com/books