From owner-freebsd-stable@FreeBSD.ORG Tue Aug 31 13:37:17 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 DF3FB10656A7 for ; Tue, 31 Aug 2010 13:37:16 +0000 (UTC) (envelope-from auryn@zirakzigil.org) Received: from mail.zirakzigil.org (mail.zirakzigil.org [82.63.178.63]) by mx1.freebsd.org (Postfix) with ESMTP id 367998FC26 for ; Tue, 31 Aug 2010 13:37:16 +0000 (UTC) Received: from localhost (unknown [192.168.1.2]) by mail.zirakzigil.org (Postfix) with ESMTP id CC2779F317 for ; Tue, 31 Aug 2010 15:22:01 +0200 (CEST) X-Virus-Scanned: amavisd-new at zirakzigil.org Received: from mail.zirakzigil.org ([192.168.1.2]) by localhost (ext.zirakzigil.org [192.168.1.2]) (amavisd-new, port 10024) with ESMTP id y5wSV8fvei24 for ; Tue, 31 Aug 2010 15:21:58 +0200 (CEST) Received: from aurynmob2.giulioferro.it (unknown [192.168.1.2]) (Authenticated sender: auryn@zirakzigil.org) by mail.zirakzigil.org (Postfix) with ESMTPA id 1D8A39F300 for ; Tue, 31 Aug 2010 15:21:58 +0200 (CEST) Message-ID: <4C7D01F7.4010003@zirakzigil.org> Date: Tue, 31 Aug 2010 15:21:59 +0200 From: Giulio Ferro User-Agent: Mozilla/5.0 (X11; U; FreeBSD amd64; en-US; rv:1.9.1.10) Gecko/20100621 Thunderbird/3.0.5 MIME-Version: 1.0 To: freebsd-stable@freebsd.org Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Subject: About zfs + nfs stability 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: Tue, 31 Aug 2010 13:37:17 -0000 I have a 8.0 stable (last build around April 2010) which I use as a nfs server : amd64, 8GB RAM, ~7TB storage. I had a lot of grief with the (sadly) well know "kmem map too small" bug, which really compromised the quality of the service that server was deputed to. There wasn't (and there still isn't) any relevant indication in the official freebsd zfs documentation on how to bypass the problem. Only thanks to the effort and goodwill of other users in this list and with hours on end of trying, I could come up with something working: (in /boot/loader.conf) vm.kmem_size="6096M" vfs.zfs.arc_max="3584M" vfs.zfs.prefetch_disable="1" vfs.zfs.txg.timeout="5" The freezes are gone, thankfully, but I often get huge slow-downs: looking in the logs of the nfs clients I get plenty of: ... kernel: nfs server ...:/path/to/dir: lockd not responding ... kernel: nfs server ...:/path/to/dir: lockd is alive again I don't know if this has anything to do with zfs. What I'd like to know is the answer to the following questions by other users and/or developers. I don't need opinions, only punctual facts people have verified for themselves. 1) Is it a good idea to upgrade this production system to the latest 8 stable (8.1 stable I believe)? Is it really stable? 2) Are the zfs aforementioned tuning in /boot/loader.conf still necessary? 3) Is it a good idea to switch to nfsv4? Performance? Stability? and above all: 4) will I get a more stable and performant system by upgrading? Thanks in advance for the answers...