On Tue, 2010-06-08 at 11:53 -0500, Chris Adams wrote:
> I am working on a new server that looks like it is going to go in
> service before RHEL 6 is released.  I'm building it with plans to
> upgrade the OS once RHEL 6 is out and I've tested it (I'm leaving empty
> space on the OS drive to install RHEL 6 to a separate partition when I'm
> ready).
> 
> I'd like to use ext4 for the data filesystem for extent support and
> better allocation patterns.  Is anybody using ext4 on RHEL 5 in
> production?  Any comments, opinions, recommendations, etc.?

We're running them on some internal "production" servers which are
basically large backup/archive servers.  The volumes are 16TB ext4 and
contain a mix of fairly large (100GB-1TB) files and a fairly small
BackupPC repository with roughly 3 million hardlinks.

Performance for the large files is noticeably better than ext3, both
reading and writing, and especially deleting them.  It would regularly
take 10 minutes to delete a 200GB file from the ext3 volumes, and now it
takes seconds.

We've also running been running one 250GB development Oracle system with
ext4.  Haven't really noticed much performance difference in that
environment, but haven't had any problems.

My biggest disappointment with ext4 is that it's still limited to 16TB
with the current e4fsprog versions.  It appears that the filesystem
itself is ready to break the 16TB barrier, but the userspace tools not
so much.  We tried switching to XFS since Redhat is now providing
optional support for that filesystem, but we quickly hit several kernel
crashes and and had to recover data with xfs_repair so we quickly gave
up on paying extra for that feature.

Anyone know if RHEL6 finally provides an out-of-box supported,
single-user filesystem capable of breaking the 16TB barrier?

Later,
Tom


_______________________________________________
rhelv5-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/rhelv5-list

Reply via email to