Replying to myself: I upgraded slurm on the head node to 14.11.7 and configured the nodes to use cgroup based process tracking on the nodes.
Still, the jobs get canceled when running a higher priority job. Any thoughts on how to debug the issue? Best, Olaf
