From: Bruce Momjian <bruce@momjian.us>
Message-Id: <201006291529.o5TFTIn17113@momjian.us>
Subject: Re: cpu bound postgresql setup.
In-Reply-To: <F4E6A2751A2823418A21D4A160B68988614660@fletch.stackdump.local>
To: Benjamin Krajmalnik <kraj@servoyant.com>
Date: Tue, 29 Jun 2010 11:29:18 -0400 (EDT)
CC: Rajesh Kumar Mallah <mallah.rajesh@gmail.com>,
	Kevin Grittner <Kevin.Grittner@wicourts.gov>,
	pgsql-performance@postgresql.org
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset="US-ASCII"

Benjamin Krajmalnik wrote:
> > -----Original Message-----
> > From: Bruce Momjian [mailto:bruce@momjian.us]
> > Sent: Monday, June 28, 2010 3:45 PM
> > To: Benjamin Krajmalnik
> > Cc: Rajesh Kumar Mallah; Kevin Grittner; pgsql-
> > performance@postgresql.org
> > Subject: Re: [PERFORM] cpu bound postgresql setup.
> > 
> > Benjamin Krajmalnik wrote:
> > > Rajesh,
> > >
> > > I had a similar situation a few weeks ago whereby performance all of
> > a
> > > sudden decreased.
> > > The one tunable which resolved the problem in my case was increasing
> > the
> > > number of checkpoint segments.
> > > After increasing them, everything went back to its normal state.
> > 
> > Did you get a db server log message suggesting in increasing that
> > setting? I hope so.

> Unfortunately not.  The behavior I had was ebbs and flows.  On FreeBSD,
> I was seeing a lot of kernel wait states in top.  So every few minutes,
> responsiveness of the db was pretty bad.  8.4.4/amd64 on FreeBSD 7.2

Bummer.  What is supposed to happen is if you are checkpointing too
frequently ( < 30 seconds), you get the suggestion about increasing
checkpoint_segments. I have not heard of cases where you are not
checkpointing too frequently, and increasing checkpoint_segments helps.

-- 
  Bruce Momjian  <bruce@momjian.us>        http://momjian.us
  EnterpriseDB                             http://enterprisedb.com

  + None of us is going to be here forever. +