Re: postgres deadlocks - Mailing list pgsql-general

From Stephen Robert Norris
Subject Re: postgres deadlocks
Date
Msg-id 20010726081913.A21242@sunhill.commsecure.com.au
Whole thread Raw
In response to postgres deadlocks  (Joseph Shraibman <jks@selectacast.net>)
List pgsql-general
On Wed, Jul 25, 2001 at 05:38:06PM -0400, Joseph Shraibman wrote:
> My postgres deadlocked again today.  Logs indicated that before the
> nightly vacuum -z everything was fine, but after it was locked.  The
> postgres.log contained the end of the vacuum which appeared to be
> consistant with the other vacuums in the log.  A ps showed no vacuuming
> processes, but it did show one 'idle in transaction'.  This problem last
> happened in Februrary when I was using 7.0.3.
>
>  select version();
>                                version
> ---------------------------------------------------------------------
>  PostgreSQL 7.1.2 on i686-pc-linux-gnu, compiled by GCC egcs-2.91.66
> (1 row)
>
>
> [root@p1 /var/log] ps ax | grep postgres
> 16973 ?        S      0:36 postgres: user2 owl 127.0.0.1 idle
> 17105 ?        S      0:01 postgres: user2 owl 127.0.0.1 idle
> 17122 ?        S      1:51 postgres: user2 owl 127.0.0.1 idle
> 17302 ?        S      1:19 postgres: user2 owl 127.0.0.1 UPDATE waiting
> 17453 ?        S      0:32 postgres: user2 owl 127.0.0.1 idle
> 18202 ?        S      0:03 postgres: user2 owl 127.0.0.1 idle
> 18351 ?        S      0:12 postgres: user2 owl 127.0.0.1 idle in
> transaction
> 18352 ?        S      0:00 postgres: user2 owl 127.0.0.1 idle
>  7137 ?        S      0:00 postgres: user2 owl 127.0.0.1 idle
>
> --
> Joseph Shraibman
> jks@selectacast.net
> Increase signal to noise ratio.  http://www.targabot.com
>
> ---------------------------(end of broadcast)---------------------------
> TIP 4: Don't 'kill -9' the postmaster

Whenever I've seen something like this it's turned out to be a deadlock
between the process in the transaction and the transaction (UPDATE in this
case) waiting.

I think it's generally bad for a client to be idle in a transaction?

    Stephen

Attachment

pgsql-general by date:

Previous
From: "Mitch Vincent"
Date:
Subject: Re: Hardware Recommendations
Next
From: Joseph Shraibman
Date:
Subject: Re: postgres deadlocks