Lustre discuss - Feb 2012 - Troubleshooting high cpu usage on osses

All,

I have seen others have this issue, but I am not seeing any definitive
troubleshooting process for it. Does anyone have or  know of a flowchart that
could help here?

We have 9 OSSes with 2 OSTs on each. There is a particular user who, when he
submits a job, tears the heck out of 2 of the OSSes. Always the same 2.
The striping is across all OSTs for the filesystem. I am trying to figure out
what file(s) he is working with, but he is less than helpful for troubleshooting
info, preferring to claim it is our configuration, not his code or usage.

Unfortunately, his jobs are relatively  quick (a couple minutes usually), so
just to try and see what is going on I have to be fast :)

Any insight would be greatly appreciated.


All the Best,


Brian Andrus
ITACS/Research Computing
Naval Postgraduate School
Monterey, California
voice: 831-656-6238


-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.lustre.org/pipermail/lustre-discuss/attachments/20120206/ea173c63/attachment.html