From owner-freebsd-geom@FreeBSD.ORG Thu May 19 21:20:35 2011 Return-Path: Delivered-To: freebsd-geom@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 45CF0106566B for ; Thu, 19 May 2011 21:20:35 +0000 (UTC) (envelope-from jwd@SlowBlink.Com) Received: from nmail.slowblink.com (rrcs-24-199-145-34.midsouth.biz.rr.com [24.199.145.34]) by mx1.freebsd.org (Postfix) with ESMTP id E8CA38FC13 for ; Thu, 19 May 2011 21:20:34 +0000 (UTC) Received: from nmail.slowblink.com (localhost [127.0.0.1]) by nmail.slowblink.com (8.14.3/8.14.3) with ESMTP id p4JLKNAh085271; Thu, 19 May 2011 17:20:24 -0400 (EDT) (envelope-from jwd@nmail.slowblink.com) Received: (from jwd@localhost) by nmail.slowblink.com (8.14.3/8.14.3/Submit) id p4JLKMx2085270; Thu, 19 May 2011 17:20:22 -0400 (EDT) (envelope-from jwd) Date: Thu, 19 May 2011 17:20:22 -0400 From: John D To: mj@feral.com Message-ID: <20110519212022.GA84918@slowblink.com> References: <20110519143935.GA83122@slowblink.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.4.2.3i Cc: freebsd-geom@freebsd.org Subject: Re: LSI 9200-8e/gmultipath/ZFS cable pull kernel crash X-BeenThere: freebsd-geom@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: GEOM-specific discussions and implementations List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 19 May 2011 21:20:35 -0000 ----- mjacob@freebsd.org's Original Message ----- > > On Thu, 19 May 2011, John D wrote: > > >.. > > Not a zfs issue I believe, so dropping them -fs the thread. > > When you do the cable pull, is any I/O active? Yes, I have about 600MB/s running to the array: dd if=/dev/zero bs=1m | mbuffer -m 64m -o /lsi/bigfile with mbuffer reporting the following in @ 697 MiB/s, out @ 697 MiB/s, 11.8 GiB total, buffer 100% full when I pull the cable. Playing with it today, I'm wondering if the 1st cable actually has any active providers on it (luck of the draw that I pulled it first), and the 2nd has all the active providers on it. I have not explicity tried to balance the i/o on the two cables with gmultipath rotate. Having a bit of trouble getting a good objdump of g_multipath.o so I may do a new src update. Thx