From owner-freebsd-current@FreeBSD.ORG Sun Nov 4 12:26:27 2007 Return-Path: Delivered-To: freebsd-current@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 1933A16A417 for ; Sun, 4 Nov 2007 12:26:27 +0000 (UTC) (envelope-from kris@FreeBSD.org) Received: from weak.local (pointyhat.freebsd.org [IPv6:2001:4f8:fff6::2b]) by mx1.freebsd.org (Postfix) with ESMTP id 4F9CB13C4A7; Sun, 4 Nov 2007 12:26:26 +0000 (UTC) (envelope-from kris@FreeBSD.org) Message-ID: <472DBA71.9070401@FreeBSD.org> Date: Sun, 04 Nov 2007 13:26:25 +0100 From: Kris Kennaway User-Agent: Thunderbird 2.0.0.6 (Macintosh/20070728) MIME-Version: 1.0 To: Peter Schuller References: <200711021208.25913.Thomas.Sparrevohn@btinternet.com> <200711040948.25732.peter.schuller@infidyne.com> <472D9B2A.5080900@FreeBSD.org> <200711041302.08877.peter.schuller@infidyne.com> In-Reply-To: <200711041302.08877.peter.schuller@infidyne.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: freebsd-current@freebsd.org Subject: Re: ZFS Hangs X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 04 Nov 2007 12:26:27 -0000 Peter Schuller wrote: >> It certainly could be a caching effect but the particular problem you >> were seeing should have been fixed. Are you still seeing it? > > I am not still seeing it; the machine has since been switched to AMD 64, and > RAM increased from 1 GB to 3 GB (and FreeBSD itself upgraded several times > since then). I did not look into it in terrible detail; I just observed by > tracing that the I/O in question was definitely the package database, and > that it was definitely accessing the same files mutiple times. I assumed it > was a caching effect at the time, given all the maxvnode tweaking and sizing > restrictions put on the ARC. I do know I never saw it after the memory > increase. I think the particular issue was fixed by improvements to the package tools. > > Also, this might have been prior to disabling prefetch. If it was, perhaps the > prefetch hosed the cache. (Are the prefetch issues still there, anyone?) Well the readahead "issues" are just the zfs design, AFAIK. It does aggressive prefetching to improve performance on the assumption that you are running with a reasonably fast storage system that can keep up. On crappy low end disk hardware (e.g. single ATA disk) this can easily saturate the disk. Kris