Message-ID: <4E8B8960.9030903@2ndQuadrant.com>
Date: Tue, 04 Oct 2011 15:32:00 -0700
From: Greg Smith <greg@2ndQuadrant.com>
User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US;
	rv:1.9.1.16) Gecko/20110818 Icedove/3.0.11
MIME-Version: 1.0
To: pgsql-performance@postgresql.org
Subject: Re: : PG9.0 - Checkpoint tuning and pg_stat_bgwriter
References: 
 <CAFrxt0iXn1vzF0wBzva+8CSdK34iqJ+mzcucxGpWqZ5s77LjXQ@mail.gmail.com>
In-Reply-To: 
 <CAFrxt0iXn1vzF0wBzva+8CSdK34iqJ+mzcucxGpWqZ5s77LjXQ@mail.gmail.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

On 10/04/2011 03:50 AM, Venkat Balaji wrote:
> I had a look at the pg_stat_bgwriter as well.

Try saving it like this instead:

select now(),* from pg_stat_bgwriter;

And collect two data points, space a day or more apart.  That gives a 
lot more information about the rate at which things are actually 
happening.  The long-term totals are less interesting than that.

Generally the first round of tuning work here is to increase 
checkpoint_segments until most checkpoints appear in checkpoints_timed 
rather than checkpoints_req.  After that, increasing checkpoint_timeout 
might also be useful.

-- 
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support  www.2ndQuadrant.us