Why are my jobs in the SSUSP state?
To maximize cluster efficiency, the short_ and normal_serial queues target nodes that are owned by individual research groups. These groups have agreed to let their nodes be used by general users on the condition that should a member of the owning research group need to run a job and there are insufficient processors open, then guest jobs using their nodes will be suspended while the owner’s job runs. The guest job will remain suspended until the owning job finishes, even if other nodes open up (i.e., there is no job migration).
If you are unwilling to wait for your job to restart, consider using the ‘rbsub’ command. rbsub is identical to the bsub command, except that jobs submitted with rbsub are automatically requeued (started over from the beginning) when they enter the SSUSP state.