X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (neptune.hub.org [200.46.204.2]) by svr1.postgresql.org (Postfix) with ESMTP id 3B103D1B444 for ; Mon, 5 Jan 2004 20:15:01 +0000 (GMT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (neptune.hub.org [200.46.204.2]) (amavisd-new, port 10024) with ESMTP id 49537-08 for ; Mon, 5 Jan 2004 16:14:12 -0400 (AST) Received: from mail.hive.nj2.inquent.com (mc.carriermail.com [205.178.180.9]) by svr1.postgresql.org (Postfix) with SMTP id 58323D1B47F for ; Mon, 5 Jan 2004 16:14:11 -0400 (AST) Received: (qmail 25534 invoked from network); 5 Jan 2004 20:12:48 -0000 Received: from unknown (HELO ?10.0.2.16?) (216.208.117.7) by 205.178.180.9 with SMTP; 5 Jan 2004 20:12:48 -0000 Subject: Re: deferred foreign keys From: Rod Taylor To: Stephan Szabo Cc: Vivek Khera , Postgresql Performance In-Reply-To: <20040105114730.U73612@megazone.bigpanda.com> References: <3FF4FA25.7050802@familyhealth.com.au> <20040105183859.GA28448@wolff.to> <20040105105401.C72490@megazone.bigpanda.com> <20040105114730.U73612@megazone.bigpanda.com> Content-Type: text/plain Message-Id: <1073333664.8958.36.camel@jester> Mime-Version: 1.0 X-Mailer: Ximian Evolution 1.4.5 Date: Mon, 05 Jan 2004 15:14:24 -0500 Content-Transfer-Encoding: 7bit X-Virus-Scanned: by amavisd-new at postgresql.org X-Archive-Number: 200401/52 X-Sequence-Number: 5227 On Mon, 2004-01-05 at 14:48, Stephan Szabo wrote: > On Mon, 5 Jan 2004, Vivek Khera wrote: > > > > > On Jan 5, 2004, at 1:57 PM, Stephan Szabo wrote: > > > > > But, if he's updating the fk table but not the keyed column, it should > > > no > > > longer be doing the check and grabbing the locks. If he's seeing it > > > grab > > > the row locks still a full test case would be handy because it'd > > > probably > > > mean we missed something. > > > > > > > I'm not *sure* it is taking any locks. The transactions appear to be > > running lock step (operating on different parts of the same pair of > > tables) and I was going to see if deferring the locks made the > > difference. It is my feeling now that it will not. However, if there > > is a way to detect if locks are being taken, I'll do that. I'd like to > > avoid dropping and recreating the foreign keys if I can since it takes > > up some bit of time on the table with 20+ million rows. > > The only way I can think of to see the locks is to do just one of the > operations and then manually attempting to select for update the > associated pk row. When a locker runs into a row lock held by another transaction, the locker will show a pending lock on the transaction id in pg_locks.