From owner-freebsd-fs@FreeBSD.ORG Mon Oct 8 16:39:52 2012 Return-Path: Delivered-To: freebsd-fs@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [69.147.83.52]) by hub.freebsd.org (Postfix) with ESMTP id 1D7901065670 for ; Mon, 8 Oct 2012 16:39:52 +0000 (UTC) (envelope-from dustinwenz@ebureau.com) Received: from internet02.ebureau.com (internet02.tru-signal.biz [65.127.24.21]) by mx1.freebsd.org (Postfix) with ESMTP id D56588FC0A for ; Mon, 8 Oct 2012 16:39:51 +0000 (UTC) Received: from service02.office.ebureau.com (internet06.ebureau.com [65.127.24.25]) by internet02.ebureau.com (Postfix) with ESMTP id 567F6DDD4BD for ; Mon, 8 Oct 2012 11:32:15 -0500 (CDT) Received: from localhost (localhost [127.0.0.1]) by service02.office.ebureau.com (Postfix) with ESMTP id 521368D3EB6 for ; Mon, 8 Oct 2012 11:32:15 -0500 (CDT) X-Virus-Scanned: amavisd-new at ebureau.com Received: from service02.office.ebureau.com ([127.0.0.1]) by localhost (internet06.ebureau.com [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id RsU8UBf8w_Yr for ; Mon, 8 Oct 2012 11:32:15 -0500 (CDT) Received: from square.office.ebureau.com (square.office.ebureau.com [10.10.20.22]) by service02.office.ebureau.com (Postfix) with ESMTPSA id 0ABEF8D3EA9 for ; Mon, 8 Oct 2012 11:32:15 -0500 (CDT) From: Dustin Wenz Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: quoted-printable Message-Id: <6116A56E-4565-4485-887E-46E3ED231606@ebureau.com> Date: Mon, 8 Oct 2012 11:32:14 -0500 To: freebsd-fs@freebsd.org Mime-Version: 1.0 (Mac OS X Mail 6.1 \(1498\)) X-Mailer: Apple Mail (2.1498) Subject: Imposing ZFS latency limits 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: Mon, 08 Oct 2012 16:39:52 -0000 I'm trying to find a way to get lower read latency in ZFS when there is = a failing disk in a mirror. The specific failure mode that I'm trying to = handle is when a disk takes significantly longer than usual to respond = to read requests, but is not so far gone that the OS drops it from the = bus. In that case, the whole pool can become much less responsive than I = think it needs to be. Since ZFS has access to the data I'm reading through other members in a = mirror, I would like ZFS to defer to the redundant disks when individual = disk latency exceeds some threshold that I would define. Is there any = provision for this in the FreeBSD ZFS implementation? - .Dustin