From owner-freebsd-amd64@FreeBSD.ORG Thu Nov 1 04:09:16 2007 Return-Path: Delivered-To: freebsd-amd64@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id B31BE16A417 for ; Thu, 1 Nov 2007 04:09:16 +0000 (UTC) (envelope-from rhurlin@gwdg.de) Received: from amailer.gwdg.de (amailer.gwdg.de [134.76.10.18]) by mx1.freebsd.org (Postfix) with ESMTP id 4F2E413C480 for ; Thu, 1 Nov 2007 04:09:16 +0000 (UTC) (envelope-from rhurlin@gwdg.de) Received: from p578b68b8.dip0.t-ipconnect.de ([87.139.104.184] helo=krabat.raven.hur) by mailer.gwdg.de with esmtpsa (TLSv1:AES256-SHA:256) (Exim 4.67) (envelope-from ) id 1InHQp-0005YU-KQ; Wed, 31 Oct 2007 18:32:43 +0100 Message-ID: <4728BC3A.7040100@gwdg.de> Date: Wed, 31 Oct 2007 18:32:42 +0100 From: Rainer Hurling User-Agent: Thunderbird 2.0.0.6 (X11/20070803) MIME-Version: 1.0 To: Marek Blaszkowski References: <4728906E.8020804@kis.p.lodz.pl> In-Reply-To: <4728906E.8020804@kis.p.lodz.pl> Content-Type: text/plain; charset=ISO-8859-2; format=flowed Content-Transfer-Encoding: 7bit X-Authenticated: Id:rhurlin X-Spam-Level: - X-Virus-Scanned: (clean) by exiscan+sophie Cc: freebsd-amd64@freebsd.org Subject: Re: strange i/o and tty lockups/delays on 6.1, strange sync() issue X-BeenThere: freebsd-amd64@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Porting FreeBSD to the AMD64 platform List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 01 Nov 2007 04:09:16 -0000 Hi Marek, a thread on stable@ is dealing with this kind of hangings: http://www.nabble.com/forum/ViewPost.jtp?post=13296788 Perhaps there is some relationship between them? Regards, Rainer Marek Blaszkowski schrieb: > Hi, > my config: > thunder k8sr, dual opteron 244, 8GB RAM, FreeBSD 6.1, sata disk > > > for time to time (every 10 or 20 secs) any process which try write or read any, even small file, > is locked for a couple seconds (about 7 or 9 secs), very iritating when I try save > small files edited in vim (size about 2-3 kbytes), it's looks like this: > #vim /path/to/file > ...editing.. > :w > and now (afetr w command) vim freeze for 7-10 secs until it displays "file written". > > Delays accours on tty, too. In vim I can't move around edited file (up/down in rows), > vim just got hang, strange is that I can switch (durning vim lockup) beetwen virtual screen > consoles. > > System is not swapping (4GB ram free), iostat says: > > #iostat 1 > tty ad4 ad6 cpu > tin tout KB/t tps MB/s KB/t tps MB/s us ni sy in id > 1 353 17.94 45 0.78 0.64 0 0.00 11 1 6 1 82 > 0 360 16.00 72 1.12 0.00 0 0.00 17 0 2 1 81 > 0 120 16.00 77 1.20 0.00 0 0.00 6 0 3 1 90 > 0 120 14.66 73 1.04 0.00 0 0.00 41 0 11 1 47 > 0 120 16.00 78 1.22 0.00 0 0.00 27 0 1 0 72 > 0 120 16.22 74 1.17 0.00 0 0.00 0 0 0 0 100 > 0 122 17.22 79 1.33 0.00 0 0.00 8 1 5 0 85 > 0 120 16.46 79 1.27 0.00 0 0.00 14 0 3 2 81 > 0 120 13.85 48 0.65 0.00 0 0.00 7 0 4 0 88 > 0 120 15.13 79 1.17 0.00 0 0.00 12 0 4 1 83 > 0 120 11.11 83 0.90 0.00 0 0.00 8 0 3 1 89 > > as You see, there is no heavy i/o load. > > Strange things is I have two identical setup of the same system configuration with the same > apps running (web server mirror) but other motherboard (K8W), on the second system there no such delay/lockup > issue. > It's looks like some other process calls sync() on every 7 or 10 seconds and it couse such i/o delay. > Other question is why sync takes so much time, when i/o disk performance its not bad (tested with dd is about 50MB/sec) > and iostat (durning sync calls) doesn't show any high disk transfers. > Example: > #time sync > real 2m0.853s > user 0m0.000s > sys 0m0.145s > > 2 mins was the longest time amount I saw durning tests, in most cases sync takies about 10 secs, when sync was called on the > other tty iostat doesn't print any high transfer, never it exceed 1MB/s. > Qestion is: if disks transfers (using cp or dd) are high, why sync waits so long ? > > I played with vfs.write_behind=0 but it doesn't change anything. > > PS. > I checkced motherboards manuals: K8W and K8SR have the same sata chip (Silicon Image SIL3114), so I think > above problems are not ata driver issues. > > regards,