From owner-freebsd-fs@FreeBSD.ORG Fri Jun 9 12:27:55 2006 Return-Path: X-Original-To: freebsd-fs@freebsd.org Delivered-To: freebsd-fs@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id D96D716A41A for ; Fri, 9 Jun 2006 12:27:55 +0000 (UTC) (envelope-from anderson@centtech.com) Received: from mh1.centtech.com (moat3.centtech.com [207.200.51.50]) by mx1.FreeBSD.org (Postfix) with ESMTP id 5CA5643D93 for ; Fri, 9 Jun 2006 12:27:51 +0000 (GMT) (envelope-from anderson@centtech.com) Received: from [10.177.171.220] (neutrino.centtech.com [10.177.171.220]) by mh1.centtech.com (8.13.1/8.13.1) with ESMTP id k59CRog6091369 for ; Fri, 9 Jun 2006 07:27:51 -0500 (CDT) (envelope-from anderson@centtech.com) Message-ID: <4489694F.1050503@centtech.com> Date: Fri, 09 Jun 2006 07:27:59 -0500 From: Eric Anderson User-Agent: Thunderbird 1.5.0.2 (X11/20060506) MIME-Version: 1.0 To: freebsd-fs@freebsd.org References: <200606091053.k59ArYQs029626@lurza.secnetix.de> In-Reply-To: <200606091053.k59ArYQs029626@lurza.secnetix.de> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit X-Virus-Scanned: ClamAV 0.87.1/1523/Fri Jun 9 02:10:10 2006 on mh1.centtech.com X-Virus-Status: Clean Subject: Re: heavy NFS writes lead to corrup summary in superblock 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: Fri, 09 Jun 2006 12:27:55 -0000 Oliver Fromme wrote: > Mikhail Teterin wrote: > > The FS is intended for very few very large files and was created > > with "newfs -b 65536 -O1" (no softupdates). > > Did you also increase the fragment size (-f option)? > The default is 2048 bytes, and I wouldn't expect a b/f > ratio of 32:1 to work very well. In fact I'm surprised > that you have so little problems. :-) > > If you intend to have very few very large files that are > accessed sequentially most of the time, it is probably > better to set both block and fragment size to the same > value (e.g. 16k), essentially disabling fragmentation. > You should also reduce the inode density by specifying > a larger bytes-per-inode value (-i option), a typical > value would be 262144 (2^18). > > Carefully fiddling with the -g and -h options might also > improve performance a bit, see newfs(8). He should also use UFS2, and disable softupdates (if he really doesn't want them). No reason I can think of to use UFS1, but that doesn't mean there isn't a bug lurking in UFS1. Eric -- ------------------------------------------------------------------------ Eric Anderson Sr. Systems Administrator Centaur Technology Anything that works is better than anything that doesn't. ------------------------------------------------------------------------