[torqueusers] cluster hangs
jwilkinson at stoneeagle.com
Tue Sep 3 14:34:11 MDT 2013
I'm not sure if this is a maui or a torque issue, so I'm being slightly rude and sending this to both lists.
We're running maui 3.3-4 and torque 2.5.7-9 on CentOS 6.3.
Most of the time there's no problem. qsub a set of submit files, they run and we get our output, but every now and then, they submit and get held so that qstat shows something like this...
[eob_merge at srvBatchHead01 ~]$ qstat
Job id Name User Time Use S Queue
------------------------- ---------------- --------------- -------- - -----
48267.srvbatchhead01 FS93130003_000 eob_merge 0 Q batch
48268.srvbatchhead01 FS93130003_001 eob_merge 0 Q batch
48269.srvbatchhead01 FS93130003_002 eob_merge 0 Q batch
48270.srvbatchhead01 FS93130003_003 eob_merge 0 Q batch
48271.srvbatchhead01 FS93130003_004 eob_merge 0 Q batch
48272.srvbatchhead01 FS93130006_000 eob_merge 0 Q batch
48273.srvbatchhead01 FS93130006_001 eob_merge 0 Q batch
48274.srvbatchhead01 FS93130006_002 eob_merge 0 Q batch
48275.srvbatchhead01 FS93130006_003 eob_merge 0 Q batch
48276.srvbatchhead01 FS93130006_004 eob_merge 0 Q batch
and they'll sit there forever like this. We will restart all of the associated services: maui, pbs_server, pbs_mom and munge, yet, it doesn't help. Finally we just reboot all of the boxes in the cluster (fortunately, it's a small number) and everything comes back up and runs.
I've proposed a weekly reboot of everything, but have been told that this can only be a stop gap measure and cannot be the final solution.
Does anyone have any clues?
Services | VPay(r)
jwilkinson at stoneeagle.com<mailto:jwilkinson at stoneeagle.com>
111 W. Spring Valley Rd., #100
Richardson, TX 75081
CONFIDENTIALITY NOTICE: This email, including any attachments, is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure, or distribution is prohibited. If you received this email and are not the intended recipient, please inform the sender by email reply and destroy all copies of the original message.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the torqueusers