Canceling Jobs

From athena

If you want to cancel a job that has been submitted, please do not use the PBS/Torque command qdel. We have noticed that running jobs that are canceled with qdel have a chance of getting stuck in the "Canceling" state. Using the MOAB cancel commands seems to be more stable.

Instead, use canceljob <jobid> or mjobctl -c <jobid>.

Kills job 12345 which had been submitted/created by qsub

 mjobctl -c 12345

or

 canceljob 12345


Kills all of your processes everywhere in the cluster (without even saying please):

 cluster-fork kill -9 -1

If your job gets stuck in the "Canceling" state, execute the following:

 mjobctl -c <jobid>
 qdel <jobid>
 mjobctl -c <jobid>
 qdel <jobid>

If that still does not work, email help@phys.washington.edu, give them your job ID and ask that they run "qdel -p <jobid>" on it to forcibly remove it.

See also Soft_Errors_(busy,_etc) for nodes stuck in the "Busy" state.