X-Original-To: pgsql-performance-postgresql.org@localhost.postgresql.org Received: from localhost (unknown [64.117.224.130]) by svr1.postgresql.org (Postfix) with ESMTP id D6990D1C93B for ; Fri, 8 Aug 2003 20:14:10 +0000 (GMT) Received: from svr1.postgresql.org ([64.117.224.193]) by localhost (neptune.hub.org [64.117.224.130]) (amavisd-new, port 10024) with ESMTP id 74084-03 for ; Fri, 8 Aug 2003 17:12:44 -0300 (ADT) Received: from sss.pgh.pa.us (unknown [192.204.191.242]) by svr1.postgresql.org (Postfix) with ESMTP id C74C6D1C937 for ; Fri, 8 Aug 2003 17:13:59 -0300 (ADT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss.pgh.pa.us (8.12.9/8.12.9) with ESMTP id h78KDlSa029639; Fri, 8 Aug 2003 16:13:47 -0400 (EDT) To: Bruce Momjian Cc: Rod Taylor , Andrew Sullivan , Postgresql Performance Subject: Re: Perfomance Tuning In-reply-to: <200308081934.h78JYi906553@candle.pha.pa.us> References: <200308081934.h78JYi906553@candle.pha.pa.us> Comments: In-reply-to Bruce Momjian message dated "Fri, 08 Aug 2003 15:34:44 -0400" Date: Fri, 08 Aug 2003 16:13:47 -0400 Message-ID: <29638.1060373627@sss.pgh.pa.us> From: Tom Lane X-Virus-Scanned: by amavisd-new at postgresql.org X-Archive-Number: 200308/160 X-Sequence-Number: 2977 Bruce Momjian writes: > Rod Taylor wrote: >> On Fri, 2003-08-08 at 14:53, Andrew Sullivan wrote: >>> I would give you exactly the opposite advice: _never_ use a >>> non-journalling fs for your data and WAL. I suppose if you can >>> afford to lose some transactions, you can do without journalling. >>> Otherwise, you're just borrowing trouble, near as I can tell. >> >> Agreed.. WAL cannot recover something when WAL no longer exists due to a >> filesystem corruption. > It is true that ext2 isn't good because the file system may not recover, > but BSD UFS isn't a journalled file system, but does guarantee file > system recovery after a crash --- it is especially good using soft > updates. The main point here is that the filesystem has to be able to take care of itself; we expect it not to lose any files or forget where the data is. If it wants to use journalling to accomplish that, fine. Journalling file contents updates, as opposed to filesystem metadata, should be redundant with what we do in WAL. So I'd recommend journalling metadata only, if that option is available (and if Postgres stuff is the only stuff on the disk...) regards, tom lane