> Are the tasks migrated between groups? its getting late. the jobacct_gather plugin is disabled here because it is marked as experimental, so the code migrating the task to the root group and setting force_empty is not executed. the only code executed (besides the rmdir from the release_agent) is in https://github.com/SchedMD/slurm/blob/master/src/plugins/task/cgroup/task_cgroup_memory.c the release_agent is found here (in case i missed something): https://github.com/SchedMD/slurm/blob/master/etc/cgroup.release_common.example As far as i can see, this locks the root memory cgroup prior to deleting the corresponding subtree. In the logfile, there are the following error messages, indicating that only the innermost child group for the jobstep could be deleted, but not its parents (hierarchy is /sys/fs/cgroup/memory/slurm/uid_x/job_xxxx/step_xxxxxxxxx). [2013-11-26T21:38:59.583] [62044.0] task/cgroup: not removing job memcg : Device or resource busy [2013-11-26T21:38:59.583] [62044.0] task/cgroup: not removing user memcg : Device or resource busy -- To unsubscribe from this list: send the line "unsubscribe cgroups" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
