Preemption mode

On premium/flix partitions, we use a preemption mode, in which jobs with special high priority will requeue jobs with normal priority.

Preempted jobs will be requeued by default, hence:

• If requeued job handle restart (via anything like ‘’starting point’’, ‘’restart files’’, etc.), it will be restarted at the last written output, when resources become available.

• If requeued job does not handle restarting, it will be restarted from it’s beginning. If you whish to avoid this behavior, use --norequeue sbatch option, and the said job will be cancelled, not requeued, if preempted.