From owner-freebsd-fs@FreeBSD.ORG Tue Nov 24 12:47:35 2009 Return-Path: Delivered-To: freebsd-fs@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 3224C106568F for ; Tue, 24 Nov 2009 12:47:35 +0000 (UTC) (envelope-from se@freebsd.org) Received: from smtp143.mail.ukl.yahoo.com (smtp143.mail.ukl.yahoo.com [77.238.184.74]) by mx1.freebsd.org (Postfix) with SMTP id 9D91B8FC17 for ; Tue, 24 Nov 2009 12:47:34 +0000 (UTC) Received: (qmail 37761 invoked from network); 24 Nov 2009 12:47:33 -0000 Received: from xdsl-81-173-156-74.netcologne.de (se@81.173.156.74 with plain) by smtp143.mail.ukl.yahoo.com with SMTP; 24 Nov 2009 12:47:33 +0000 GMT X-Yahoo-SMTP: iDf2N9.swBDAhYEh7VHfpgq0lnq. X-YMail-OSG: 7rNlRx0VM1nIDgNOPuAQF8XeAc9KlEyB0lAyh8_vNbR.n6T7z3Nvlm5DzhUcwzNcstjisNO6Ctu9aIwL6DjOt_tfMYOAcHyKmUwkoT2oQ_fh9.eggigmkkO8nTB8dN_ztsp6p0fl.6jWxrQOdo8S2_9PK.TYmLPCMbbKet1u6DPR3KfkfXfBWejB7Qm_IivYMclnXJ1eyKCLm_Pim37KRG41eAfrbx8Zs9rd743ProKMJj_u_k.3Uxt7WOEKyyiu X-Yahoo-Newman-Property: ymail-3 Message-ID: <4B0BD5E7.3050604@freebsd.org> Date: Tue, 24 Nov 2009 13:47:35 +0100 From: Stefan Esser User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; de; rv:1.9.1.4pre) Gecko/20090915 Lightning/1.0pre Thunderbird/3.0b4 MIME-Version: 1.0 To: Ollivier Robert References: <20091122154228.GB55532@rron.freenix.org> In-Reply-To: <20091122154228.GB55532@rron.freenix.org> X-Enigmail-Version: 0.97b Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: freebsd-fs@freebsd.org Subject: Re: 7.2 dies in zfs 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: Tue, 24 Nov 2009 12:47:35 -0000 On 22.11.2009 16:42, Ollivier Robert wrote: > According to Randy Bush: >> i think the issue is how to tune for zfs >> >> i386 with 4G of RAM > > I've given up on ZFS on i386. Whatever tuning you could do is only > delaying the inevitable. Even with lots of RAM, it will panic. I'd > love being proven wrong as I also hav a 4 GB i386 with ZFS and it panics > regularely. If your i386 based system has much RAM (2GB or more), than you should definitely increase KVA_PAGES. Not doing so will lead to panics, not in spite of but exactly because of the large RAM. I have been using ZFS on i386 since it became available, first for testing and soon as only file-system (with UFS boot, initially, now switching over to gptzfsboot). Systems range from Pentium-3 to AMD64x2 and I see no problems even under significant load. The following is the complete contents of /boot/loader.conf on my home server with 512MB RAM (the maximum supported on this P3/733 based SFF box) and a single 320MB IDE drive: zfs_load="YES" vfs.root.mountfrom="zfs:gk" vfs.zfs.arc_max="80000000" vm.kmem_size="350000000" The box is a mail gateway with spam-filter, IMAP server, web-server, SMB and NFS server for media applications (incl. storage backend for a networked digital TV receiver). It has only 280 days of uptime, since it took a reboot to upgrade kernel and world when ZFS version 13 had been committed to 8-current (previous uptime was at least as long). With 4GB of RAM you need to raise KVA_PAGES or you'll run into a panic. Perhaps, the default of 256 should be raised to 512? The cost of KVA_PAGES=512 is 1MB of RAM allocated to the kernel page table and 1GB less maximum user process size ... Sun specifically mentions, that ZFS makes assumptions that are easily valid on 64bit architectures, but not so easy to meet on 32bit systems. But for moderate load, ZFS can run on a 512MB P3 with good reliability and the known advantages from an admin POV. Regards, STefan