From owner-freebsd-scsi@FreeBSD.ORG Sun Sep 26 18:46:58 2010 Return-Path: Delivered-To: freebsd-scsi@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 3A52E1065672 for ; Sun, 26 Sep 2010 18:46:58 +0000 (UTC) (envelope-from niklas@saers.com) Received: from mail-ew0-f54.google.com (mail-ew0-f54.google.com [209.85.215.54]) by mx1.freebsd.org (Postfix) with ESMTP id C5C4E8FC18 for ; Sun, 26 Sep 2010 18:46:57 +0000 (UTC) Received: by ewy22 with SMTP id 22so1287985ewy.13 for ; Sun, 26 Sep 2010 11:46:56 -0700 (PDT) Received: by 10.213.35.2 with SMTP id n2mr1909583ebd.62.1285526816471; Sun, 26 Sep 2010 11:46:56 -0700 (PDT) Received: from [192.168.0.198] (x1-6-00-24-01-67-20-00.k428.webspeed.dk [83.89.13.33]) by mx.google.com with ESMTPS id a48sm7095644eei.19.2010.09.26.11.46.53 (version=TLSv1/SSLv3 cipher=RC4-MD5); Sun, 26 Sep 2010 11:46:54 -0700 (PDT) Mime-Version: 1.0 (Apple Message framework v1081) Content-Type: text/plain; charset=us-ascii From: Niklas Saers In-Reply-To: Date: Sun, 26 Sep 2010 20:46:51 +0200 Content-Transfer-Encoding: quoted-printable Message-Id: <4EF4B99C-8550-4207-BBDB-391CF4F2D940@saers.com> References: <2EA9CBBC-3F97-4AF2-BFB5-96DF39FDE376@saers.com> <86aan67obp.fsf@ds4.des.no> To: Scott Long X-Mailer: Apple Mail (2.1081) Cc: freebsd-scsi@freebsd.org, Jakob Jensen , =?iso-8859-1?Q?michael_Kj=F8gx?= , =?iso-8859-1?Q?Dag-Erling_Sm=F8rgrav?= Subject: Re: mfi - setting up disks X-BeenThere: freebsd-scsi@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: SCSI subsystem List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 26 Sep 2010 18:46:58 -0000 Hi Scott & Dag-Erling, On Sep 26, 2010, at 1:25 AM, Scott Long wrote: > MFI only uses CAM for passthrough access to component drives, not for = normal I/O. Setting CAM wiring hints will not solve the problem at = hand. And the problem at hand isn't really even numbering, it's that = the MFI firmware freaked out and marked the disks inaccessible. I think = that this happened to us at Yahoo once, and we eventually gave up and = replaced the disks. Putting the disks on a non-LSI, non-RAID controller = and writing 0's to the last 10MB worth of sectors (or just writing 0's = to the entire drive) will likely solve the problem, but YMMV. Indeed, "mfiutil locate" gave us all we needed for identifying the disk, = so numbering is no longer an issue. Thanks for the tip on how to get the = disk back up, I'll make sure we try that this week. I tried the let-the-RAID-controller-make-RAIDs-and-join-them-via-ZFS = model on the mpt-based controller, and that made performance drop from = ~250 mb/sec to ~4 mb/sec. With mfi (the system is otherwise unchanged), = the average speed is ~200 mb/sec, down ~50 mb/sec. Cheers Nik=