I was under the impression that most issues with XFS and stacks came up
when stacking it up with MD and LVM.  There is a patch to serialize call
chains such as XFS->LVM->MD, but I don't know if Redhat as added it
to their distributed kernel.

SGI does have a test source rpm in ftp://oss.sgi.com/projects/xfs/testing/RHEL4/
for 2.6.9-22.EL.  This does also build properly against 2.6.9-22.0.1.EL.  It
looks like it put together sometime in Nov2005 so should be pretty representitive
of the current state of XFS development.

Tony Hoffmann | Tel/Tél: (250) 493-2277 | Fax: (250) 493-7767
[log in to unmask] <mailto:[log in to unmask]>
National Research Council Canada | P.O. Box 248, Penticton BC V2A 6J9
Conseil national de recherches Canada | B.P. 248, Penticton (C.-B.) V2A 6J9
Government of Canada | Gouvernement du Canada



-----Original Message-----
From: [log in to unmask]
[mailto:[log in to unmask]]On Behalf Of
Michael Mansour
Sent: Monday, February 06, 2006 12:56 PM
To: Bonnie Alcorn; [log in to unmask]
Subject: Re: XFS and SL4.2


Hi,

> We are currently using SL3.05 with XFS and are considering going to 
> SL 
> 4.2  We use XFS for our data areas.  Does anyone have any experience 
> using XFS with 4.2?  Are there any known problems?  I would be happy 
> to hear about any experiences, good or bad.

I used XFS on 4.2 for a couple of months in test, but found an issue with the
4k and 8k buffers? (I think that's what they were called) when using them. My
severs would hard crash when putting any legitimate load on them.

From memory (you can check the list archives for my emails related to this),
it turned out that RH compiled only 4k into the kernel, and XFS needed 8k.

I begrungingly removed XFS from the data volume and reformatted to ext3.

Regards,

Michael.