btw. Had this going on and working well after compiled my own 2.6.18 kernel.

XFS still hangs sometimes, sometimes only gives error. I'm stuck with this, don't know how to get rid of the problem. 2.6.17 or newer kernel might do it but it is said not beeing clever thing to do. But it isn't working with this anyway so I have to do something.

Hold on, kernels installed:
rpm -qa |grep kernel
kernel-devel-2.6.9-34.0.2.EL
kernel-2.6.9-42.0.2.plus.c4
kernel-module-xfs-2.6.9-34.0.2.EL-0.1-3
kernel-module-xfs-2.6.9-42.0.2.EL-0.1-3
kernel-devel-2.6.9-42.0.2.plus.c4
kernel-utils-2.4-13.1.83
kernel-2.6.9-34.0.2.EL
kernel-module-xfs-2.6.9-42.plus.c4-0.1-3
kernel-module-xfs-2.6.9-42.0.2.plus.c4-0.1-3

uname -a
Linux r.xyz.fi 2.6.9-42.0.2.plus.c4 #1 Fri Aug 25 17:33:49 CDT 2006 i686 athlon i386 GNU/Linux

How to tell which xfs-module is used?
/sbin/lsmod |grep xfs
xfs_quota 53600 0
xfs 527600 2 xfs_quota

I have /etc/mdadm.conf ARRAY /dev/md0 devices=/dev/sda2,/dev/sdb2,/dev/hdc1
and this causes hangs abt. once a week, if stressed from machine itself + 2 clients via smb.

I'll gladly give any information, if there is any hint how to get rid of the hang-ups.
Sometimes it only gives this warning abt. 50 lines or so and then recovers, but abt. 1/3 of the cases it panics.
(edited, addition) I run abt. a year the same system with ext3 without no problems and then changed to xfs for performance.

--
Veli-Pekka Lehtosaari
inetvpl@ww (dot) iki (dot) fi

vplehto, what kernel and what xfs code are you using ?

Isn't there any working solution on Centos.

I'm having xfs: possible memory allocation deadlock in kmem_alloc (mode:0x2d0) problems.
Foud out searching google that 2.6.17 should correct it http://oss.sgi.com/bugzilla/show_bug.cgi?id=410.

Once thought that 2.6.9-42.0.2.plus.c4 corrected this but nope.

Having this on high disk / net io.

Do I have to compile my own kernel for this?

-
vpl

While having the separate module package for both kernels is cool, I hope it doesn't cause the problem I had with the nvidia modules in Fedora. Basically they packaged the kernel version into the package name, so the individual modules never knew to update when the kernel was updated, because each module package was considered separate from the others. It was nice from the perspective where you could have the modules installed for each kernel installed, but in the case where you just wanted the updated kernel to have the updated driver without having to do anything extra it was a pain.

There was a post in the mailing lists today. The XFS stuff has been moved to module form that can be used in both the standard kernel and the plus kernel. I don't recall the details. If you go to the CentOS home, mailing lists and use the search box in the upper left corner to search for today's date and XFS, or maybe just XFS will be enough, you'll see it.

HTH

Bill

I don't think the project folks are monitoring here right now - getting 4.4 ready. But the last thing I heard from Johhny was a call for things to enable in the Plus kernel. It may be that he hasn't finished that update yet. I would post to the general support lists. I see several of the project folks there a lot.

I'm not pissing and moaning about an un-supported kernel, but when Johnny 1st mentioned the availability of an un-supported but additional kernel option, he specifically said XFS support was enabled. The recent upgrade from 2.6.9-34.108 to 2.6.9-42.0.2 breaks XFS.

Should this be enabled and didn't get in by accident or should we no longer expect XFS in the 'Plus' kernels?