[Cookie-users] GPU zombie and reboot on cookie

[Daniel Lowell]

There is a zombie process on cookie's GPU. You can take a look with nvidia-smi -a
Utilization is 99%, meaning no work can be done on the GPU until it is free. Which will be never until cookie is rebooted. nvidia-smi for sdk 4.0 does not have the capability of resetting the GPU on the fly, however the sdk 4.1 version will have it. Until then...

Not sure when is a good time for a reboot. I know Krishna has a pbound job running, so I'd like to get a consensus of cookie users.

Let me know. Thanks.

Daniel Lowell
(630)252-0092



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/cookie-users/attachments/20111110/c021b66d/attachment.htm>