Hello CentOS, I''ve had cause to upgrade some 3.x machines that had a few ext2 partitions to ext3 (long story). Anyway, did the tune2fs -j /foo and the journal is created. changed fstab to ext3 Ok, that''s fine. But in a couple cases, the machine was to be rebooted in the 300day uptime range. upon reboot, the machine stated that /foo was due for an fsck. Well, fsck sat there twiddling it''s thumbs indefinitely. I had to reboot to CDROM, and fsck from linux rescue (which had no troubles). Then once fsck''d exit and reboot went fine. Is this a problem with fsck, or ? I''ve never known fsck to hang before. It usually does it''s job or spits out an error. Of course, now that I know that, for most partitions I can unmount and fsck before rebooting. sigh -- Best regards, Mickael mailto:mike@kamloopsbc.com
Is there anything unusual about the disk subsystem? Were the device drivers installed from an alternate source (e.g. 3w-9xxx)? I''ve seen this behavior immediately following updates when a new kernel is installed but the device driver is not carried forward. -geoff On Thu, 10 Mar 2005 13:30:12 -0800 Mickael Maddison <mike@kamloopsbc.com> wrote:> Hello CentOS, > > I''ve had cause to upgrade some 3.x machines that had a few ext2 > partitions to ext3 (long story). Anyway, did the tune2fs -j /foo and > the journal is created. changed fstab to ext3 > > Ok, that''s fine. But in a couple cases, the machine was to be > rebooted in the 300day uptime range. upon reboot, the machine stated > that /foo was due for an fsck. Well, fsck sat there twiddling it''s > thumbs indefinitely. I had to reboot to CDROM, and fsck from linux > rescue (which had no troubles). Then once fsck''d exit and reboot went > fine. > > Is this a problem with fsck, or ? I''ve never known fsck to hang > before. It usually does it''s job or spits out an error. > > Of course, now that I know that, for most partitions I can unmount and > fsck before rebooting. sigh > > -- > Best regards, > Mickael > mailto:mike@kamloopsbc.com > > _______________________________________________ > CentOS mailing list > CentOS@caosity.org > http://lists.caosity.org/mailman/listinfo/centos---------------------------------------------------------- Geoff Galitz Chemistry Research Computing, UC Berkeley galitz@berkeley.edu
Hi, On Thu, Mar 10, 2005 at 01:30:12PM -0800, Mickael Maddison wrote:> Hello CentOS, > > I''ve had cause to upgrade some 3.x machines that had a few ext2 > partitions to ext3 (long story). Anyway, did the tune2fs -j /foo and > the journal is created. changed fstab to ext3 >But you didn''t ''tune2fs -i0 -c0 /foo'', did you? I don''t know how this oot stuff exactly works, but you should reset the -c max-mount-counts -i interval-between-checks[d|m|w] to make it truly ''we do it with journal'' device. Propably you just hit the ''interval-between-cheks'' flag on boot and it started to process as it should start. When one does that tune2fs -j for live system, i do believe that it''s making some temp journal node, which is later on (next mount or so) converted to true journal or something. So it might have been ''in progress of migration'' still ....... But the point being that one should use this tune2fs -i0 -c0 mantra even for making new ext3 fs. That how the installer does it too. HTH -- Pasi Pirhonen - upi@iki.fi - http://iki.fi/upi/
Hello Geoffrey, All the hardware in question are via the standard onboard IDE connections, but that''s a great point, I do have a couple systems with 3ware RAID cards, best keep an eye on those. -- Best regards, Mickael mailto:mike@kamloopsbc.com Thursday, March 10, 2005, 1:59:36 PM, you wrote: GG> Return-Path: <centos-bounces@caosity.org> GG> Received: from c1m14.emaildefenseservice.com GG> (c1m14.emaildefenseservice.com [216.40.36.47]) GG> by phoenix.silverservers.com (8.12.11/8.12.11) with SMTP id j2ANE8Vt032101 GG> for <mike@kamloopsbc.com>; Thu, 10 Mar 2005 15:14:08 -0800 GG> Received: from mail.caosity.org [69.93.111.165] (EHLO mail.caosity.org) GG> by c1m14.emaildefenseservice.com (mxl_mta-1.3.8-10p6) GG> with ESMTP id 8c7d0324.28783.050.c1m14; GG> Thu, 10 Mar 2005 18:27:04 -0500 (EST) GG> Received: from caos1.caosity.org (caos1 [127.0.0.1]) GG> by mail.caosity.org (Postfix) with ESMTP GG> id C1829B3010E; Thu, 10 Mar 2005 21:59:42 +0000 (UTC) GG> X-Original-To: centos@caosity.org GG> Delivered-To: centos@caosity.org GG> Received: from calmail-be3.berkeley.edu GG> (mailfarm.Berkeley.EDU [128.32.61.106]) GG> by mail.caosity.org (Postfix) with ESMTP id 2EF5CB300A3 GG> for <centos@caosity.org>; Thu, 10 Mar 2005 21:59:37 +0000 (UTC) GG> Received: from [128.32.50.7] (account galitz@calmail.berkeley.edu) GG> by calmail-be3.berkeley.edu (CommuniGate Pro WebUser 4.2.6) GG> with HTTP id 5149795 for centos@caosity.org; GG> Thu, 10 Mar 2005 13:59:36 -0800 GG> From: "Geoffrey Galitz" <galitz@uclink.berkeley.edu> GG> Subject: Re: [Centos] fsck -Broke? GG> To: CentOS discussion and information list <centos@caosity.org> GG> X-Mailer: CommuniGate Pro WebUser Interface v.4.2.6 GG> Date: Thu, 10 Mar 2005 13:59:36 -0800 GG> Message-ID: <web-5149795@calmail-be3.berkeley.edu> GG> In-Reply-To: <1242999433.20050310133012@kamloopsbc.com> GG> References: <1242999433.20050310133012@kamloopsbc.com> GG> MIME-Version: 1.0 GG> Content-Type: text/plain; charset="ISO-8859-1"; format="flowed" GG> X-BeenThere: centos@caosity.org GG> X-Mailman-Version: 2.1.5 GG> Precedence: list GG> Reply-To: CentOS discussion and information list <centos@caosity.org> GG> List-Id: CentOS discussion and information list <centos.caosity.org> GG> List-Unsubscribe: GG> <http://lists.caosity.org/mailman/listinfo/centos>, GG> <mailto:centos-request@caosity.org?subject=unsubscribe> GG> List-Archive: <http://lists.caosity.org/pipermail/centos> GG> List-Post: <mailto:centos@caosity.org> GG> List-Help: <mailto:centos-request@caosity.org?subject=help> GG> List-Subscribe: GG> <http://lists.caosity.org/mailman/listinfo/centos>, GG> <mailto:centos-request@caosity.org?subject=subscribe> GG> Sender: centos-bounces@caosity.org GG> Errors-To: centos-bounces@caosity.org GG> X-MX-Spam: final=0.0230263158; heur=0.5000000000(-26400); GG> stat=0.0100000000; spamtraq-heur=0.7000000000(2005031001) GG> X-MX-MAIL-FROM: <centos-bounces@caosity.org> GG> X-MX-SOURCE-IP: [69.93.111.165] GG> Content-Transfer-Encoding: 8bit GG> X-MIME-Autoconverted: from quoted-printable to 8bit by GG> phoenix.silverservers.com id j2ANE8Vt032101 GG> X-UIDL: "g%"!AjU!!jR9"!*>T!! GG> Is there anything unusual about the disk subsystem? Were GG> the device drivers installed from an alternate source (e.g. GG> 3w-9xxx)? I''ve seen this behavior immediately following updates GG> when a new kernel is installed but the device driver is not GG> carried forward. GG> -geoff GG> On Thu, 10 Mar 2005 13:30:12 -0800 GG> Mickael Maddison <mike@kamloopsbc.com> wrote:>> Hello CentOS, >> >> I''ve had cause to upgrade some 3.x machines that had a few ext2 >> partitions to ext3 (long story). Anyway, did the tune2fs -j /foo and >> the journal is created. changed fstab to ext3 >> >> Ok, that''s fine. But in a couple cases, the machine was to be >> rebooted in the 300day uptime range. upon reboot, the machine stated >> that /foo was due for an fsck. Well, fsck sat there twiddling it''s >> thumbs indefinitely. I had to reboot to CDROM, and fsck from linux >> rescue (which had no troubles). Then once fsck''d exit and reboot went >> fine. >> >> Is this a problem with fsck, or ? I''ve never known fsck to hang >> before. It usually does it''s job or spits out an error. >> >> Of course, now that I know that, for most partitions I can unmount and >> fsck before rebooting. sigh >> >> -- >> Best regards, >> Mickael >> mailto:mike@kamloopsbc.com >> >> _______________________________________________ >> CentOS mailing list >> CentOS@caosity.org >> http://lists.caosity.org/mailman/listinfo/centosGG> ---------------------------------------------------------- GG> Geoff Galitz GG> Chemistry Research Computing, UC Berkeley GG> galitz@berkeley.edu GG> _______________________________________________ GG> CentOS mailing list GG> CentOS@caosity.org GG> http://lists.caosity.org/mailman/listinfo/centos GG> __________ NOD32 1.1023 (20050310) Information __________ GG> This message was checked by NOD32 antivirus system. GG> http://www.nod32.com
Hello Pasi, PP> But you didn''t ''tune2fs -i0 -c0 /foo'', did you? PP> I don''t know how this oot stuff exactly works, but you should reset the PP> -c max-mount-counts PP> -i interval-between-checks[d|m|w] PP> to make it truly ''we do it with journal'' device Well shucks. The docs never mentioned that one but it sounds logical. PP> When one does that tune2fs -j for live system, i do believe that it''s PP> making some temp journal node, which is later on (next mount or so) PP> converted to true journal or something. So it might have been ''in PP> progress of migration'' still ....... Hmm... if the ''real'' journal is just being created, I can see that taking a long time indeed. so now I wonder, since I went to rescue mode, fsck''d the partition, then proceeded to reboot and things came up fine, do I have a proper journal now? I''m assuming the kjournald processes would realize there''s a missing journal, and begin building it in the background while the system is in operation. -- Best regards, Mickael mailto:mike@kamloopsbc.com Thursday, March 10, 2005, 4:42:29 PM, you wrote: PP> Hi, PP> On Thu, Mar 10, 2005 at 01:30:12PM -0800, Mickael Maddison wrote:>> Hello CentOS, >> >> I''ve had cause to upgrade some 3.x machines that had a few ext2 >> partitions to ext3 (long story). Anyway, did the tune2fs -j /foo and >> the journal is created. changed fstab to ext3 >>PP> But you didn''t ''tune2fs -i0 -c0 /foo'', did you? PP> I don''t know how this oot stuff exactly works, but you should reset the PP> -c max-mount-counts PP> -i interval-between-checks[d|m|w] PP> to make it truly ''we do it with journal'' device. PP> Propably you just hit the ''interval-between-cheks'' flag on boot and it PP> started to process as it should start. PP> When one does that tune2fs -j for live system, i do believe that it''s PP> making some temp journal node, which is later on (next mount or so) PP> converted to true journal or something. So it might have been ''in PP> progress of migration'' still ....... PP> But the point being that one should use this PP> tune2fs -i0 -c0 PP> mantra even for making new ext3 fs. That how the installer does it too. PP> HTH
Hi, On Fri, Mar 11, 2005 at 09:33:18AM -0800, Mickael Maddison wrote:> > Hmm... if the ''real'' journal is just being created, I can see that > taking a long time indeed. so now I wonder, since I went to rescue > mode, fsck''d the partition, then proceeded to reboot and things came > up fine, do I have a proper journal now? I''m assuming the kjournald > processes would realize there''s a missing journal, and begin building > it in the background while the system is in operation. >I didn''t make myself quite clear on this as i am not quite sure if i do have the facts correct even myself (didn''t bother to look that process too carefully ever). How i see it, is that the boot process does care about those -i and -c times and do ovey those. If you don''t set those, it''ll get back to doing the normal ext2 fsck process. Making mountcount -1 and interval between checks 0, wont'' even trigger anymore this process and the disk i handled as ext3. Something like that? -- Pasi Pirhonen - upi@iki.fi - http://iki.fi/upi/