moin moin,
we have a mix of CentOS 5.x boxen. They're currently using ext3 for the OS
and reiserfs for a RAID0 filesystem for fast writing. We have a couple
hundred of them.
We've been experiencing a few kernel panics a month due to reiserfs.
At one point when researching the problem I found
Last I checked XFS was still supported; when tuned correctly it's still the
fastest filesystem for database transaction logs (although some of the
log-structured filesystem may eventually beat it).
EXT4 still has some significant performance issues and has regressed quite a
bit in that regard