From: "Dave Dutcher" <dave@tridecap.com>
To: "'Christo Du Preez'" <christo@mecola.com>,
	<pgsql-performance@postgresql.org>
References: <466D65CA.6000503@mecola.com>
	<01d001c7ac43$ec686780$2e00a8c0@tridecap.com>
	<466E4D5E.6050103@mecola.com>
Subject: Re: test / live environment, major performance difference
Date: Tue, 12 Jun 2007 09:28:11 -0500
Message-ID: <025001c7acfd$e84d9390$2e00a8c0@tridecap.com>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: 7bit
Thread-Index: Acesw+aN9yP/xS15QfODzRWDrU3ZUgAOD/sg
In-Reply-To: <466E4D5E.6050103@mecola.com>

> From: Christo Du Preez
> Sent: Tuesday, June 12, 2007 2:38 AM
> 
> Where do I set the planner settings or are you reffering to 
> settings in postgres.conf that may affect the planner?
> 

Yes I'm reffering to settings in postgres.conf.  I'm wondering if
enable_indexscan or something got turned off on the server for some reason.
Here is a description of those settings:

http://www.postgresql.org/docs/8.2/interactive/runtime-config-query.html

So when you move data from the laptop to the server, I see that your script
correctly runs an analyze after the load, so have you run analyze on the
fast laptop lately?  Hopefully running analyze wouldn't make the planner
choose a worse plan on the laptop, but if we are trying to get things
consistant between the laptop and server, that is something I would try.

If the consistancy problem really is a problem of the planner not using
index scans on the server, then if you can, please post the table definition
for the table with a million rows and an EXPLAIN ANALYZE of a query which
selects a few rows from the table.

Dave