From: Tom Lane <tgl@sss.pgh.pa.us>
To: Jonatan Evald Buus <jonatan.buus@cellpointmobile.com>
cc: pgsql-performance@postgresql.org
Subject: Re: Poor performance for delete query
In-reply-to: 
 <CABnF-8Lyps8ribpp=BOFnqGEeWp2K5iuLukMo3txQ5UkgHjF7g@mail.gmail.com>
References: 
 <CABnF-8Lyps8ribpp=BOFnqGEeWp2K5iuLukMo3txQ5UkgHjF7g@mail.gmail.com>
Comments: In-reply-to Jonatan Evald Buus <jonatan.buus@cellpointmobile.com>
 message dated "Thu, 24 Apr 2014 21:42:41 +0200"
Date: Thu, 24 Apr 2014 16:29:00 -0400
Message-ID: <28604.1398371340@sss.pgh.pa.us>
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

Jonatan Evald Buus <jonatan.buus@cellpointmobile.com> writes:
> We're currently having very poor performance for the following delete query.
> DELETE FROM TopTable WHERE id IN (xx, yy, zz);

> We've observed that it takes around 7 seconds under normal load to for each
> row that's being from TopTable and several minutes pr deleted row under
> heavy load.

I'd really have to bet that you forgot to index one of the referencing
tables.  Are any of the foreign keys multi-column?  If so you probably
need a matching multi-column index, not just indexes on the individual
referencing columns.

> How do we track down the cause of the poorly performing delete query?

EXPLAIN ANALYZE on a DELETE, for starters.  That would isolate whether
it's the DELETE itself or one of the foreign-key updates, and if the
latter which one.  It's a little bit difficult to see the exact plan being
used for a foreign-key update query, but I think one way you could do it
is to enable auto_explain with auto_explain.log_nested_statements turned
on.

			regards, tom lane


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