X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [200.46.204.144]) by svr1.postgresql.org (Postfix) with ESMTP id D98E15348A for ; Fri, 25 Mar 2005 23:21:28 +0000 (GMT) Received: from svr1.postgresql.org ([200.46.204.71]) by localhost (av.hub.org [200.46.204.144]) (amavisd-new, port 10024) with ESMTP id 99847-07 for ; Fri, 25 Mar 2005 23:21:22 +0000 (GMT) Received: from candle.pha.pa.us (candle.pha.pa.us [64.139.89.126]) by svr1.postgresql.org (Postfix) with ESMTP id 40C28534DE for ; Fri, 25 Mar 2005 23:21:21 +0000 (GMT) Received: (from pgman@localhost) by candle.pha.pa.us (8.11.6/8.11.6) id j2PNLO826150; Fri, 25 Mar 2005 18:21:24 -0500 (EST) From: Bruce Momjian Message-Id: <200503252321.j2PNLO826150@candle.pha.pa.us> Subject: Re: pg_autovacuum not having enough suction ? In-Reply-To: <16180.1111792693@sss.pgh.pa.us> To: Tom Lane Date: Fri, 25 Mar 2005 18:21:24 -0500 (EST) Cc: "Matthew T. O'Connor" , Otto Blomqvist , "pgsql-performance@postgresql.org" X-Mailer: ELM [version 2.4ME+ PL121 (25)] MIME-Version: 1.0 Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=US-ASCII X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.011 tagged_above=0 required=5 tests=AWL X-Spam-Level: X-Archive-Number: 200503/573 X-Sequence-Number: 11278 Tom Lane wrote: > Bruce Momjian writes: > > Tom Lane wrote: > >> I'm not sure if autovacuum could be taught to do that --- it could > >> perhaps launch a vacuum as soon as it notices a large fraction of the > >> table got deleted, but do we really want to authorize it to launch > >> VACUUM FULL? > > > One problem with VACUUM FULL would be autovacuum waiting for an > > exclusive lock on the table. Anyway, it is documented now as a possible > > issue. > > I don't care too much about autovacuum waiting awhile to get a lock. > I do care about other processes getting queued up behind it, though. > > Perhaps it would be possible to alter the normal lock queuing semantics > for this case, so that autovacuum's request doesn't block later > arrivals, and it can only get the lock when no one is interested in the > table. Of course, that might never happen, or by the time it does > there's no point in VACUUM FULL anymore :-( Can we issue a LOCK TABLE with a statement_timeout, and only do the VACUUM FULL if we can get a lock quickly? That seems like a plan. The only problem is that you can't VACUUM FULL in a transaction: test=> create table test (x int); CREATE TABLE test=> insert into test values (1); INSERT 0 1 test=> begin; BEGIN test=> lock table test; LOCK TABLE test=> vacuum full; ERROR: VACUUM cannot run inside a transaction block -- Bruce Momjian | http://candle.pha.pa.us pgman@candle.pha.pa.us | (610) 359-1001 + If your life is a hard drive, | 13 Roberts Road + Christ can be your backup. | Newtown Square, Pennsylvania 19073