Question about Cyrus, ext3, and Linux kernel 2.4.18 bug

[Michael Sims]

Is anyone out there running Cyrus on Linux kernel 2.4.18, using ext3 with
"data=journal" on the Cyrus spool volume?  If so, have you had any problems?
I just switched to this mode yesterday and so far it's performing much
better than before but I was wondering if I have to worry about stability
issues.  Let me explain:

I'm currently running Cyrus IMAP 2.1.12 on Red Hat 7.2, using the
2.4.18-27.7.xsmp kernel on an IBM Netfinity 5000 (dual PIII-500).  I have
/var mounted on a separate disk.  I started out with this volume mounted as
ext3 in it's default Red Hat configuration (data=ordered and on my system
~30 MB journal).  I experienced extremely slow disk performance during mail
spikes that would cause tons of LMTPD processes to build up waiting on locks
and my load average would go through the roof.

I've since switched mailboxes.db, all of the seen databases, and deliver.db
to skiplist and that improved things dramatically.  I had seen a lot of
reports on various mailing lists that Cyrus performs better on ext3 when
using "data=journal" mode, so I decided to try it (I also increased the size
of my journal to 128 MB).  There was a noticeable peformance increase,
enough so that I would like to continue using this configuration.

The problem is I have seen vague references to a bug in the 2.4.18 Linux
kernel when using ext3 with "data=journal" turned on:

http://marc.theaimsgroup.com/?l=info-cyrus&m=103185140816795&w=2

Also mentioned in:

http://postfix.state-of-mind.de/ralf.hildebrandt/ext3.shtml

Also:

http://groups.google.com/groups?selm=ahi90k%24194d%241%40FreeBSD.csie.NCTU.e
du.tw&oe=UTF-8

Most of these posts were made months ago, however, and a couple of ext3
related bugs have been fixed in the Red Hat errata kernel that I am running.
The problem is, I can't be sure that the bug mentioned above is one of the
ones that was fixed.

Can anyone shed some light on this?  If you know of the bug, can you tell me
what exactly happens when it is triggered?  Is it a kernel panic?  Does the
filesystem get corrupted?

For right now I'm going to leave the server running and just keep a close
eye on it.  I was just hoping someone on the list would have some experience
with this so I would know to stop worrying about it, or to switch the
journal mode back.

Thanks in advance...

___________________________________________
Michael Sims
Project Analyst - Information Technology
Crye-Leike Realtors
Office: (901)758-5648  Pager: (901)769-3722
___________________________________________