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 23AA85E46C0 for ; Mon, 16 Aug 2004 12:41:19 -0300 (ADT) 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 40230-04 for ; Mon, 16 Aug 2004 15:41:17 +0000 (GMT) Received: from sss.pgh.pa.us (sss.pgh.pa.us [66.207.139.130]) by svr1.postgresql.org (Postfix) with ESMTP id 3DFE55E46C4 for ; Mon, 16 Aug 2004 12:41:10 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.12.11/8.12.11) with ESMTP id i7GFfGdY014010; Mon, 16 Aug 2004 11:41:16 -0400 (EDT) To: Stef Cc: pgsql-performance@postgresql.org Subject: Re: Strange problems with more memory. In-reply-to: <20040816171234.5a9bac8f@svb.ucs.co.za> References: <20040816171234.5a9bac8f@svb.ucs.co.za> Comments: In-reply-to Stef message dated "Mon, 16 Aug 2004 17:12:34 +0200" Date: Mon, 16 Aug 2004 11:41:16 -0400 Message-ID: <14009.1092670876@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at hub.org X-Spam-Status: No, hits=0.0 tagged_above=0.0 required=5.0 tests= X-Spam-Level: X-Archive-Number: 200408/247 X-Sequence-Number: 7883 Stef writes: > It seems I was mistaking, as I started getting these kind of errors in dmesg : > VM: killing process postmaster > __alloc_pages: 0-order allocation failed (gfp=0x1d2/0) > __alloc_pages: 0-order allocation failed (gfp=0x1d2/0) > VM: killing process postmaster This looks like the infamous OOM-kill kernel bug^H^H^Hfeature. Turn off memory overallocation in your kernel to get more stable behavior when pushing the limits of available memory. > But the database is just as slow as before. (By slow I mean not as > fast as it should be on such a powerful machine compared to much worse > machines running the same processes) If your concern is with a single nightly process, then that quad Xeon is doing squat for you, because only one of the processors will be working. See if you can divide up the processing into several jobs that can run in parallel. (Of course, if the real problem is that you are disk I/O bound, nothing will help except better disk hardware. Way too many people think they should buy a super-fast CPU and attach it to consumer-grade IDE disks. For database work you're usually better off spending your money on good disks...) regards, tom lane