From owner-freebsd-stable@FreeBSD.ORG Mon Feb 8 19:56:22 2010 Return-Path: Delivered-To: freebsd-stable@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 767E51065692 for ; Mon, 8 Feb 2010 19:56:22 +0000 (UTC) (envelope-from jdc@koitsu.dyndns.org) Received: from qmta15.emeryville.ca.mail.comcast.net (qmta15.emeryville.ca.mail.comcast.net [76.96.27.228]) by mx1.freebsd.org (Postfix) with ESMTP id 5D64A8FC1D for ; Mon, 8 Feb 2010 19:56:21 +0000 (UTC) Received: from omta20.emeryville.ca.mail.comcast.net ([76.96.30.87]) by qmta15.emeryville.ca.mail.comcast.net with comcast id fXj61d00E1smiN4AFXwNSi; Mon, 08 Feb 2010 19:56:22 +0000 Received: from koitsu.dyndns.org ([98.248.46.159]) by omta20.emeryville.ca.mail.comcast.net with comcast id fXwN1d0043S48mS8gXwNCT; Mon, 08 Feb 2010 19:56:22 +0000 Received: by icarus.home.lan (Postfix, from userid 1000) id 092C11E3033; Mon, 8 Feb 2010 11:56:21 -0800 (PST) Date: Mon, 8 Feb 2010 11:56:21 -0800 From: Jeremy Chadwick To: freebsd-stable@freebsd.org Message-ID: <20100208195620.GA10209@icarus.home.lan> References: <20100208143329.GA12057@megatron.madpilot.net> <20100208145147.GA3733@icarus.home.lan> <20100208171822.12773278@r500.local> <20100208165853.GA43617@megatron.madpilot.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20100208165853.GA43617@megatron.madpilot.net> User-Agent: Mutt/1.5.20 (2009-06-14) Subject: Re: ATA_CAM + ZFS gives short 1-2 seconds system freeze on disk load 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: Mon, 08 Feb 2010 19:56:22 -0000 On Mon, Feb 08, 2010 at 05:58:53PM +0100, Guido Falsi wrote: > On Mon, Feb 08, 2010 at 05:18:22PM +0100, Fabian Keil wrote: > > > > I experienced what I think is the same problem. ZFS's bulk disk flushes > > caused vlc to occasionally stutter when viewing a DVD rip from disk while > > ripping a DVD at the same time. > > > > My workaround is to put "vfs.zfs.txg.timeout=3" in /boot/loader.conf. > > I think I read about this on zfs-discuss@. I assume on faster systems > > one can use a higher value. > > This was a very good idea. I'm trying this and the system freezes much > less if at all. > > Are there any drawbacks I should be aware of? > > I can imagine that forcing more disk writes somewhat slows down normal > disk activity, but I can bear that on a desktop system. I'd like a technical explanation of exactly what this loader.conf tunable does. The sysctl -d explanation is useful if one has insight to what purpose it serves. I can find mention on Solaris lists about "txg timeout", but the context is over my head (intended for those very familiar with the inner workings of ZFS). I'd also like to know what vfs.zfs.txg.synctime does. I've been playing with vfs.zfs.txg.timeout (decreasing it from 30 to 5) and testing on a 3-disk raidz1 pool on a test machine I have, but I haven't finished my testing yet. A person said that decreasing the value increased throughput[1] for them, but in my brief test (using dd) it doesn't appear to improve I/O speed at all. Anyone know of a good (non-X-related) repro test case which can induce the ZFS "bursty sluggishness" problem so I can try it with different loader.conf tunables? Thanks. -- | Jeremy Chadwick jdc@parodius.com | | Parodius Networking http://www.parodius.com/ | | UNIX Systems Administrator Mountain View, CA, USA | | Making life hard for others since 1977. PGP: 4BD6C0CB |