Message-ID: <50D2A77D.10809@richardneill.org>
Date: Thu, 20 Dec 2012 05:51:57 +0000
From: Richard Neill <rn214@richardneill.org>
User-Agent: Mozilla/5.0 (X11; Linux x86_64;
	rv:16.0) Gecko/20121010 Thunderbird/16.0.1
MIME-Version: 1.0
CC: pgsql-performance@postgresql.org
Subject: Re: Why does the query planner use two full indexes, when
	a dedicated partial index exists?
References: <50D22DE2.4090301@richardneill.org>
	<28778.1355972790@sss.pgh.pa.us>
In-Reply-To: <28778.1355972790@sss.pgh.pa.us>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

Dear Tom,

Thanks very much for your advice.

>> A psql session is below.  This shows that, if I force the planner to use
>> the partial index, by dropping the others, then it's fast. But as soon
>> as I put the full indexes back (which I need for other queries), the
>> query planner chooses them instead, and is slow.
>
> [ experiments with a similar test case ... ]  I think the reason why the
> planner is overestimating the cost of using the partial index is that
> 9.1 and earlier fail to account for the partial-index predicate when
> estimating the number of index rows that will be visited.  Because the
> partial-index predicate is so highly selective in this case, that
> results in a significant overestimate of how much of the index will be
> traversed.

I think that seems likely to me.

I'll try out 9.2 and see if it helps. As it's a production server, I 
have to wait for some downtime, probably Friday night before I can find 
out - will report back.

Best wishes,

Richard


-- 
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance