Frequently Asked Questions

« Previous
How do I launch parallel jobs using MPI?
Next »
Backing up via Rsync to ZFS or Btrfs snapshots

windows ubuntu macos xibo network zenworks android storage


9. What are some nice commands for Slurm?

See also: How do I launch parallel jobs using MPI?

User commands

View the current queue for all users

squeue

View the jobs belionging to me in the queue.

squeue -u jny25782

Or a specific job.

squeue 12345

Stop all jobs for a specific user.

scancel -u jnhy25782

Run just a single command via Slurm

srun hostname

Run a job called my.job.sh

sbatch my.job.sh

Administrator commands

List all users on the cluster in the specified group and the max number of jobs they can run.

sacctmgr show user withassoc where cluster=cluster account=pmgroup format=user,maxjobs

Set the max number of jobs a specific user can run.

sacctmgr -i modify user where name=jny25782 cluster=cluster account=pmgroup set maxjobs=120

List the max number of jobx for a specific user.

sacctmgr show user name=jny25782 withassoc where cluster=cluster account=pmgroup format=user,maxjobs

This will make all the nodes named n41 to n60 to stop accepting new jobs.

for i in {41..60} ; do echo $i ; scontrol update NodeName=fbv-n$i state=DRAIN reason="kernel upgrade" ; done

This will make a node start accepting jobs. This is good either when a node has been drained or when it is down, maybe due to a reboot.

scontrol update NodeName=fbv-n62 state=RESUME



 

This entry slurm was last modified 2019-08-22

   

This documentation is covered by GNU Free Documentation License. 143 ms