DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma;
	h=mime-version:in-reply-to:references:date:message-id:subject:from:to
	:cc:content-type:content-transfer-encoding;
	b=mwBYRHX/D5QFnkfS7azKR5C45ASz1dqp6RX6fMoYyonvtOLCFI9JmFGXSYEanwvDwM
	jO2HKC4Lf4lXUWjkcXCAdO8y/KbPt7CHckFEBojgDEoY5YWZ/i0hSAOGpmwkCriLWYQU
	dsr+VOKD+vRUiDNmFfwPf4WHKObFz7e9pTQaE=
MIME-Version: 1.0
In-Reply-To: <49A9F834.8060100@ca.com>
References: <49A9F834.8060100@ca.com>
Date: Sat, 28 Feb 2009 22:15:17 -0500
Message-ID: <603c8f070902281915u4d402589wef143ff433a8a3d2@mail.gmail.com>
Subject: Re: "slow" queries
From: Robert Haas <robertmhaas@gmail.com>
To: Brian Cox <brian.cox@ca.com>
Cc: "pgsql-performance@postgresql.org" <pgsql-performance@postgresql.org>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

On Sat, Feb 28, 2009 at 9:51 PM, Brian Cox <brian.cox@ca.com> wrote:
> Actually, they're all deadlocked. The question is why?
>
> Here's a brief background. The ts_defects table is partitioned by occurrence
> date; each partition contains the rows for 1 day. When the data gets old
> enough, the partition is dropped. Since the correct partition can be
> determined from the occurrence date, there is no trigger: inserts are done
> directly into the correct partition. Multiple threads may be inserting into
> a partition at the same time. The thread that checks for old data to be
> dropped runs at 00:30 each night. It also creates the partition for the next
> day.
>
> Below is the output from:
> select xact_start,query_start,substring(current_query from 0 for 40) from
> pg_stat_activity order by xact_start;

Can you post this again with procpid added to the column list and
without truncating current_query?  And then also post the results of
"select * from pg_locks"?

Is there anything interesting in the postmaster log?

...Robert