Lustre discuss - Feb 2011 - Kernel panic on lustre 1.8.5 clients due to statahead?

I am seeing the attached kernel panics on the lustre client on a regular basis
whenever robinhood is running on this client. The following are the
lustre and kernel version on the client and the server.

Patchless Client Version
Lustre version: 1.8.5
Kernel: 2.6.18-194.32.1.el5

Lustre server version:
Lustre version: 1.8.5
kernel: 2.6.18-194.17.1.el5_lustre.1.8.5

Could someone please confirm if statahead is still a problem when running
Lustre-1.8.5 on the clients and the servers and does statahead have to
disabled on this client.

Thanks
Nirmal
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: cc1KernelPanic
Url:
http://lists.lustre.org/pipermail/lustre-discuss/attachments/20110201/650b7780/attachment.pl

Nirmal Seenu <nirmal at fnal.gov> writes:
> I am seeing the attached kernel panics on the lustre client on a
> regular basis whenever robinhood is running on this client. The
> following are the lustre and kernel version on the client and the
> server.
>
> Patchless Client Version
> Lustre version: 1.8.5
> Kernel: 2.6.18-194.32.1.el5
>
> Lustre server version:
> Lustre version: 1.8.5
> kernel: 2.6.18-194.17.1.el5_lustre.1.8.5
>
> Could someone please confirm if statahead is still a problem when
> running Lustre-1.8.5 on the clients and the servers and does statahead
> have to disabled on this client.
We have been seeing similar panics, which we think have been caused by
users doing "ls" (aliased to "ls --color=yes", causing
stat:ing) in a
directory with ~ 350000 files).

Patchless client 1.8.5 on 2.6.18-194.32.1.el5 (CentOS), which we have since
upgraded to 2.6.18-238.1.1.el5 (recompiled RHEL5.6 kernel). After
the upgrade we have not seen any crash, but that might just be because
of our users are now being more careful with their ls behaviour.  (Server
2.6.18-92.1.17.el5_lustre.1.6.7.1smp.)

-- 
Kent Engstr?m, National Supercomputer Centre
kent at nsc.liu.se, +46 13 28 4444