[Exim] Exim and Journaling Filesystems

We are about to do a complete upgrade of our Linux based Exim mail
servers. Our old system uses good old ext2 which works fine. However
when things go horribly wrong and the machine goes down without a disk
sync we end up with a mail system that takes well over an hour to fsck
(its a 160Gig partition).

Although completely unexpected downtimes don't happen that often it
would be nice to avoid this slow reboot problem!

We have experimented with ext3 and what you gain in fsck time you seem
to lose in mount time resulting in a system which is just as slow as
with ext2 it would seem. We are one of the sites that manage the
national web cache and we have used Reiser FS on those for quite a
while, and it does seem much faster at restarting, but has caused us a
few problems in the past with filesystems so badly broken that they need
to be reformatted from scratch. This isn't a problem on the web cache's
because all the data is transient anyway and our use of LVS avoids
visible downtimes. Also it may well be as a result of the huge disk IO
load that the web caches get and might be fine on a lower IO loaded
machine.

Even so, this wouldn't be a fault that we would want to risk on our Exim
servers. I have XFS on my own personal Linux machine and this seems
pretty good, but then I don't have any 160Gig partitions being written
to by Exim at a high rate so that isn't much of a test!

We have some spare machines at the moment and could do a lot of testing
of the alternative systems, but tests are never as good as production
experience.

The question I really want the answer to is:-

Are people using Exim (Maildir delivery in our case) with a decent level
of incomming messages on Linux with "any" of the Linux Journaling file
systems in a production environment? If so, I we would very much like to
hear about your experiences with this.