Subject: Re: Is disableing nested_loops a bad idea ?
From: Franck Routier <franck.routier@axege.com>
To: pgsql-performance <pgsql-performance@postgresql.org>
In-Reply-To: <4C91DA360200002500035839@gw.wicourts.gov>
References: <1284625427.2527.13.camel@franck-laptop>
	<4C91DA360200002500035839@gw.wicourts.gov>
Content-Type: text/plain; charset="UTF-8"
Organization: =?ISO-8859-1?Q?Ax=E8ge?= sarl
Date: Thu, 16 Sep 2010 16:13:06 +0200
Message-ID: <1284646386.7884.8.camel@franck-laptop>
Mime-Version: 1.0
Content-Transfer-Encoding: 8bit

Thanks Kevin and Samuel for your input.

The point is we already made a lot of tweaking to try to tune postgresql
to behave correctly. I work with Damien, and here is a post he did in
july to explain the kind of problems we have
http://comments.gmane.org/gmane.comp.db.postgresql.performance/25745

The end of the thread was Robert Hass concluding that "Disabling
nestloops altogether, even for one particular query, is
often going to be a sledgehammer where you need a scalpel.   But then
again, a sledgehammer is better than no hammer."

So I wanted to better understand to what extend using a sledgehammer
will impact me :-)

Franck


Le jeudi 16 septembre 2010 à 08:49 -0500, Kevin Grittner a écrit :
> Franck Routier <franck.routier@axege.com> wrote:
>  
> > I come into cases where the planner under-estimates the number of
> > rows in some relations, chooses to go for nested loops, and takes
> > forever to complete the request.
>  
> People can provide more targeted assistance if you pick one of the
> offenders and provide enough information for a thorough analysis. 
> It's at least somewhat likely that some tweaks to your configuration
> or maintenance procedures could help all the queries, but starting
> with just one is likely to highlight what those changes might be.
>  
> For ideas on what information to include, see this page:
>  
> http://wiki.postgresql.org/wiki/SlowQueryQuestions
>  
> -Kevin
>