Basic SLURM commands

Introduction to SLURM commands

You will only be able to see information about your own jobs. Information from other users is currently not available to you.

squeue

See the status of your jobs in the queue

$ squeue
             JOBID PARTITION     NAME     USER ST       TIME  NODES NODELIST(REASON)
              1519   hsw-fdr serial_t USERNAME  R       0:01      1 node001

BASH

A definition of the fields:

JOBID	A unique ID number assigned to your job, this will be stored in the accounting logs and you can use it to query for more information later.
PARTITION	The name of the partition or "queue" the job runs in. Each partition has a set of nodes and billing rate. See the "Choosing your partition" section below.
NAME	The name of your job; you can set this value to anything you'd like.
USER	The username of that is running the job.
ST	The state of the job: PD (pending), R (running), CA (cancelled), CF(configuring), CG (completing), CD (completed), F (failed), TO (timeout), NF (node failure) and SE (special exit state).
TIME	The elapsed run-time of the job.
NODES	How many nodes you are running on.
NODELIST	The list of nodes you are running on.
(REASON)	The reason why your job is pending, or not running.

sbatch

Submit a batch job to run on the cluster

$ sbatch my-job-script.sh

BASH

scancel

Cancel a job running on the cluster. If the job is currently running it will kill it; if the job is pending it will prevent it from running.

$ scancel 1519

BASH

These commands have many options; this is only an introduction. Please check the man pages for each command for more information.