From owner-freebsd-questions@FreeBSD.ORG Fri Jan 16 14:35:16 2009 Return-Path: Delivered-To: freebsd-questions@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 12474106564A for ; Fri, 16 Jan 2009 14:35:16 +0000 (UTC) (envelope-from vince@unsane.co.uk) Received: from unsane.co.uk (unsane-pt.tunnel.tserv5.lon1.ipv6.he.net [IPv6:2001:470:1f08:110::2]) by mx1.freebsd.org (Postfix) with ESMTP id 72BE98FC16 for ; Fri, 16 Jan 2009 14:35:15 +0000 (UTC) (envelope-from vince@unsane.co.uk) Received: from vhoffman.lon.namesco.net (150.117-84-212.staticip.namesco.net [212.84.117.150]) (authenticated bits=0) by unsane.co.uk (8.14.0/8.14.0) with ESMTP id n0GEaQJP039215 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO); Fri, 16 Jan 2009 14:36:28 GMT (envelope-from vince@unsane.co.uk) Message-ID: <49709B1F.4080902@unsane.co.uk> Date: Fri, 16 Jan 2009 14:35:11 +0000 From: Vincent Hoffman User-Agent: Thunderbird 2.0.0.19 (Macintosh/20081209) MIME-Version: 1.0 To: "B. Cook" References: <9c7e8a84da1d469df4d8155b88f59c39.squirrel@mail.poughkeepsieschools.org> In-Reply-To: <9c7e8a84da1d469df4d8155b88f59c39.squirrel@mail.poughkeepsieschools.org> X-Enigmail-Version: 0.95.7 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: freebsd-questions@freebsd.org Subject: Re: zfs on i386 trauma :P X-BeenThere: freebsd-questions@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: User questions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 16 Jan 2009 14:35:16 -0000 B. Cook wrote: > Hello all, > > We had a raid card die in a dell box and reinstalled FreeBSD 7 and > restored from backups.. the problem is amd64 didn't boot on the box and > the person doing the restore wanted to 'help' and changed from ufs2 to > zfs.. > > while being a noble effort, he was testing on amd64 and "didn't think" > that amd64 to i386 was that big a deal for zfs.. *sigh* > > So here we are.. if I can't figure out how to make this 'stable' I will > dump and restore and reinstall i386 ufs2/gmirror.. > > So we have a box w/ 3G of ram running samba and squid for a few hundred > people. > > I have tried to fix /boot/loader.conf: > # cat /boot/loader.conf > autoboot_delay="4" > vm.kmem_size_max="1024M" > vm.kmem_size="1024M" > > zfs_load="YES" > vfs.zfs.prefetch_disable=1 > vfs.root.mountfrom="zfs:tank/root" > Try adding vfs.zfs.arc_max="100M" (or less possibly, tuning as you require it.) I gather you've looked at the tuning guide on the wiki so cant really suggest anything further. The gains from zfs are nice and there are people running it in production, but its definitely still experimental. Vince > accf_http_load="YES" > accf_data_load="YES" > > and rebuilt the kernel with many device drivers disabled.. > > I found a script on the zfs tuning page on the wiki, but I do not > understand what it is telling me.. > > TEXT=7031512, 6.70577 MB > DATA=514754560, 490.908 MB > TOTAL=521786072, 497.614 MB > > kldstat yeilds this output: > Id Refs Address Size Name > 1 9 0x80400000 55f3e8 kernel > 2 1 0x80960000 b23b0 zfs.ko > 3 2 0x80a13000 23bc opensolaris.ko > 4 1 0x80a16000 191c accf_data.ko > 5 1 0x80a18000 22d4 accf_http.ko > 6 1 0x80a1b000 68394 acpi.ko > 7 1 0x87ecb000 3000 pflog.ko > 8 1 0x87ed7000 32000 pf.ko > > We have updated to 7.1-p2 with the hopes of fixing somethings.. > > So the problems that we are having is that the machine stops responding.. > ping never stops, it just stops answering tcp requests. > > and we only have a few minor tweaks in sysctl.conf: > > security.bsd.see_other_uids=0 > > net.inet.ip.random_id=1 > kern.randompid=32768 > kern.coredump=0 > > net.inet.tcp.delayed_ack=0 > net.inet.tcp.inflight.enable=0 > net.inet.tcp.path_mtu_discovery=0 > net.inet.tcp.recvbuf_auto=1 > net.inet.tcp.rfc1323=1 > net.inet.tcp.sendbuf_auto=1 > > and in rc.conf we have set some things on the bge0 interface: > > ifconfig_bge0="inet 10.20.0.230 netmask 255.255.254.0 rxcsum txcsum media > 1000baseTX mediaopt full-duplex polling" > > Is there something that is set, that is making the system stop responding? > > is there something else that could be done? > > If stability and performace are the goals (in that order) would it just > make sense to reinstall and go back to gmirror on ufs2? > > Thanks in advance. > > _______________________________________________ > freebsd-questions@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-questions > To unsubscribe, send any mail to "freebsd-questions-unsubscribe@freebsd.org" >