[Ocfs2-users] ocfs2 on mailserver lock up
Sunil Mushran
Sunil.Mushran at oracle.com
Tue Sep 26 13:13:00 PDT 2006
I will need atleast "strace -tt -T" and /proc/meminfo, /proc/slabinfo to
proceed.
write() could be slow for a lot of reasons.
augustasg at gmail.com wrote:
>
>
> On 9/26/06, *Sunil Mushran* <Sunil.Mushran at oracle.com
> <mailto:Sunil.Mushran at oracle.com>> wrote:
>
> You may want to ping Novell to get the 1.2.3 drop of OCFS2. That's
> because it is the latest and greatest.
>
> Having said that we'll need more information.
>
> As in, what syscall did strace show as taking time.
>
>
> write() is extremely slow.
>
> What is the memory usage like? cat /proc/meminfo, cat /proc/slabinfo
> That is, under production load.
>
>
> unfortunately i have no information about the exact contents of the
> files you request, but I should say that memory and CPU usage was
> normal when the lockups happened.
>
> augustasg at gmail.com <mailto:augustasg at gmail.com> wrote:
> > Hello,
> >
> > we are running SLES9 SP3 with OCFS2 for mailsystem. There are three
> > nodes in the cluster with shared storage and OCFS2 filesystem on it.
> > The filesystem is used for mailbox storage and is accessed by smtpd
> > ,pop3 and imap processes. The system works fine for a few hours but
> > locks up so that the OCFS2 filesystem is accessed by mail system
> > extremely slowly. It is possible to list the contents of the
> > filesystem or to change the directories, but mailsystem is working
> > slowly. Tracing smtp processes with strace showed that mail is
> > delivered to mailboxes but it is done in a speed of few
> kilobytes per
> > second.
> > This lock up happens under heavier load when the cluster is in
> > production use and when it happens only restart fixes the
> problem. The
> > umounting of the OCFS2 filesystem is not possible even when all
> > process accessing it are killed. Currently mail system cluster
> is not
> > used in production so we tried to do some tests, but we were not
> able
> > to replicate the problem while sending several thousands of
> messages
> > to a single account.
> > The only thing that comes in mind, that the filesystem locks in the
> > case when a file (message) is access by smtp process and pop3 or
> imap
> > process simultaniously when the processes are running on distinct
> > cluster nodes. This might happen because of lack of global knowledge
> > of localy locked files among OCFS2 cluster nodes.
> > Could this be true?
> > And what other reasons might be there for the problem?
> > Any suggestions on solving the problem?
> >
> > The system information follows:
> > OS: Suse Linux Enterprise Server 9 Service Pack 3
> > Mailbox format: Maildir
> > SMTP daemon: Postfix (postfix-2.2.6-0.1 )
> > OCFS2 version: 1.2.1
> > Kernel Version: 2.6.5-7.276-smp
> > Pop3 and imap daemon: Courier IMAP (courier-imap-4.1.0-1.suse910)
> >
> > --
> > Augustas Gutautas
> >
> ------------------------------------------------------------------------
>
> >
> > _______________________________________________
> > Ocfs2-users mailing list
> > Ocfs2-users at oss.oracle.com <mailto:Ocfs2-users at oss.oracle.com>
> > http://oss.oracle.com/mailman/listinfo/ocfs2-users
> >
>
>
>
>
> --
> Augustas Gutautas
> ------------------------------------------------------------------------
>
> _______________________________________________
> Ocfs2-users mailing list
> Ocfs2-users at oss.oracle.com
> http://oss.oracle.com/mailman/listinfo/ocfs2-users
>
More information about the Ocfs2-users
mailing list