Received: from localhost (unknown [200.46.204.184]) by postgresql.org (Postfix) with ESMTP id D032E2E00A3 for ; Thu, 21 Feb 2008 18:10:32 -0400 (AST) Received: from postgresql.org ([200.46.204.71]) by localhost (mx1.hub.org [200.46.204.184]) (amavisd-maia, port 10024) with ESMTP id 71810-04 for ; Thu, 21 Feb 2008 18:10:27 -0400 (AST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by postgresql.org (Postfix) with ESMTP id 68A5A2E00B2 for ; Thu, 21 Feb 2008 18:10:32 -0400 (AST) Received: from news.hub.org (news.hub.org [200.46.204.72]) by news.hub.org (8.14.1/8.14.1) with ESMTP id m1LMASFM077219 for ; Thu, 21 Feb 2008 18:10:28 -0400 (AST) (envelope-from news@news.hub.org) Received: (from news@localhost) by news.hub.org (8.14.1/8.14.1/Submit) id m1LLxv48073690 for pgsql-admin@postgresql.org; Thu, 21 Feb 2008 17:59:57 -0400 (AST) (envelope-from news) From: Hannes Dorbath X-Newsgroups: pgsql.admin Subject: Re: Power outage borked things (8.1.10)... Date: Thu, 21 Feb 2008 23:00:08 +0100 Organization: Hub.Org Networking Services Lines: 23 Message-ID: <47BDF468.3090001@theendofthetunnel.de> References: <47BBF298.4010108@fastmail.net> <8625.1203523828@sss.pgh.pa.us> <200802202205.58380@zmi.at> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Complaints-To: usenet@news.hub.org Cc: Tom Lane To: Michael Monnerie User-Agent: Thunderbird 2.0.0.9 (Windows/20071031) In-Reply-To: <200802202205.58380@zmi.at> To: pgsql-admin@postgresql.org X-Virus-Scanned: Maia Mailguard 1.0.1 X-Archive-Number: 200802/253 X-Sequence-Number: 28327 Michael Monnerie wrote: > On Mittwoch, 20. Februar 2008 Tom Lane wrote: >> 5. Figure out why a simple power failure was able to do this to you, >> and fix it. The most likely bet is that your disk drives are lying >> about write completion ... see the PG archives for discussion. > > It could be a caching RAID controller without battery backup unit. > Nothing help you in a power failure then, except a good backup. > Had this with a customer - no more filesystem after reboot... I hardly understand why such things still happen today. It's one of the most important admin tasks to make sure fsync / fua behave as they should. Wouldn't it be a good idea to add a chapter to the installation docs? They cover SHM settings and vm.overcommit -- why not a short introduction to fsync, write caches, write barriers and a few examples of sane and problematic setups as well as links to tools to verify a setup? -- Best regards, Hannes Dorbath