From owner-freebsd-fs@FreeBSD.ORG Wed Feb 6 10:00:19 2008 Return-Path: Delivered-To: freebsd-fs@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 5470816A417; Wed, 6 Feb 2008 10:00:19 +0000 (UTC) (envelope-from des@des.no) Received: from tim.des.no (tim.des.no [194.63.250.121]) by mx1.freebsd.org (Postfix) with ESMTP id 1002613C45A; Wed, 6 Feb 2008 10:00:19 +0000 (UTC) (envelope-from des@des.no) Received: from tim.des.no (localhost [127.0.0.1]) by spam.des.no (Postfix) with ESMTP id 198F0208C; Wed, 6 Feb 2008 11:00:11 +0100 (CET) X-Spam-Tests: AWL X-Spam-Learn: disabled X-Spam-Score: -0.3/3.0 X-Spam-Checker-Version: SpamAssassin 3.2.3 (2007-08-08) on tim.des.no Received: from ds4.des.no (des.no [80.203.243.180]) by smtp.des.no (Postfix) with ESMTP id 74AC5208A; Wed, 6 Feb 2008 11:00:10 +0100 (CET) Received: by ds4.des.no (Postfix, from userid 1001) id 4DD69844A2; Wed, 6 Feb 2008 11:00:10 +0100 (CET) From: =?utf-8?Q?Dag-Erling_Sm=C3=B8rgrav?= To: Brooks Davis References: <47A73C8D.3000107@skyrush.com> <86prvby5o1.fsf@ds4.des.no> <47A864D9.4060504@skyrush.com> <864pcnxz8f.fsf@ds4.des.no> <47A88ADE.7050503@skyrush.com> <86abmfwc6h.fsf@ds4.des.no> <20080205173102.GA85735@lor.one-eyed-alien.net> Date: Wed, 06 Feb 2008 11:00:10 +0100 In-Reply-To: <20080205173102.GA85735@lor.one-eyed-alien.net> (Brooks Davis's message of "Tue\, 5 Feb 2008 11\:31\:02 -0600") Message-ID: <86tzkmv1zp.fsf@ds4.des.no> User-Agent: Gnus/5.110006 (No Gnus v0.6) Emacs/22.1 (berkeley-unix) MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Cc: freebsd-fs@freebsd.org Subject: Re: Forcing full file read in ZFS even when checksum error encountered X-BeenThere: freebsd-fs@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Filesystems List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 06 Feb 2008 10:00:19 -0000 Brooks Davis writes: > We've also experienced several situations were zfs was detecting > corruption caused by bad cabling or bad controller firmware so SMART > had nothing to report. I once worked on a FreeBSD-based turn-key web hosting solution. Our first (and, eventually, only) customer had a four-server rig where two of the servers shared a split rack-mount SCSI cabinet. Either the cables or the cabinet (most likely the latter) were defective, and at some point, all data written to disk became silently corrupted. To add insult to injury, the backups were unusable. I don't recall why; either the customer hadn't put tapes in the backup server, or the backups had been corrupted as well, or maybe the backup solution I'd written simply didn't work and hadn't been properly tested. It was around that time I discovered that FreeBSD's version of GNU tar, when run with certain command-line options (such as those typically used by *shiver* Amanda), would create archives that no tar implementation - itself included - could extract. That rig was a disaster in so many ways... When I first ordered it, I also ordered spares for everything - spare PSUs, spare cooling fans, spare disks etc. Of course, the spares we actually got didn't fit. I can understand delivering the wrong spare fan when the original is factory-installed, but when you ship a SCSI cabinet full of disks along with spare disks that *don't match*, you know you're about to lose a customer. Oh, and the SCSI cables we used between the servers and the cabinet didn't fit or lock properly, so we kept losing touch with the disks until we replaced the cables. Strangely enough, that supplier later lost their exclusive contract with the University of Oslo, and went out of business. Somebody apparently figured that Dell could more easily handle larger orders, was less likely to ship the wrong parts, and more likely to admit it and quickly send a replacement if it did happen (although I've heard of a case where someone complained to Dell about six missing screws, and shortly thereafter received six separate cardboard boxes, each containing one screw) (oh shit, I'm turning into a rambling old fart!) DES --=20 Dag-Erling Sm=C3=B8rgrav - des@des.no