From owner-freebsd-stable@FreeBSD.ORG Fri Dec 9 00:35:10 2005 Return-Path: X-Original-To: freebsd-stable@freebsd.org Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id C479416A420 for ; Fri, 9 Dec 2005 00:35:10 +0000 (GMT) (envelope-from jd@ugcs.caltech.edu) Received: from upchuck.ugcs.caltech.edu (upchuck.ugcs.caltech.edu [131.215.176.100]) by mx1.FreeBSD.org (Postfix) with ESMTP id E174643D91 for ; Fri, 9 Dec 2005 00:34:59 +0000 (GMT) (envelope-from jd@ugcs.caltech.edu) Received: by upchuck.ugcs.caltech.edu (Postfix, from userid 3640) id 119E7AC450; Thu, 8 Dec 2005 16:34:56 -0800 (PST) Received: from localhost (localhost [127.0.0.1]) by upchuck.ugcs.caltech.edu (Postfix) with ESMTP id 03BEAAC43C; Thu, 8 Dec 2005 16:34:56 -0800 (PST) Date: Thu, 8 Dec 2005 16:34:55 -0800 (PST) From: Jon Dama To: secmgr In-Reply-To: <4398CA29.3070602@jim-liesl.org> Message-ID: References: <20051208001857.68ac4fef.torfinn.ingolfsen@broadpark.no> <20051208133100.GC912@pcwin002.win.tue.nl> <20051208194915.GC948@unixpages.org> <4398CA29.3070602@jim-liesl.org> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Cc: freebsd-stable@freebsd.org Subject: Re: FreeBSD 6.0 as storage server with raid5? X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 09 Dec 2005 00:35:10 -0000 > Whatever you do, don't complain about it on this list, or you'll just be > told that if you really wanted raid, you should be running SCSI disks Ah, no please complain so that if s/w raid gives you trouble, there will be something to point to when and if people doubt there are still problems (if indeed there are) Though I think jim is being entirely too harsh: The scary, poorly tested part of software raid is recovery. Thousands might roll out a s/w raid but if the h/w raid wasn't cost justified its unlikely that the HDs in the raid are actually going to be pressed into failing in any reasonable period of time that would reveal trouble in the recovery/degraded operating modes. Second, if you use s/w raid, pay close attention to the way your partitions line up. Third, SATA drives are actually quite good. You're primarily looking at a degraded MTBF versus a server grade SCSI disk. This could well mean just about nothing if your transaction volume is actually pretty low. imo, it would be nice to see MTBF quoted in a few parts: MTBF while seeking regularly (i.e., at some duty cycle) and MTBF in the bearings and other rotational components alone (MTBF while the disk is spun-up), and number of spin-up/spin-down cycles. Fourth, the major limitations on SATA drives right now is that FreeBSD does not support NCQ and therefore has no access to reliable write-completion information wrt SATA drives. > and adapter). I'd strongly suggest anyone using GEOM raid to do some > fault insertion testing of their setup prior to actually relying on it. This is very good advice. -Jon