What is SLURM?

SLRUM is a job-scheduling manager for Unix clusters. It allows easy request for resources and job submission on computing platform such as Nero. A SLURM batch script consists of two parts: resource requests and job steps.

• Resource requests describe the amount of computing resource (CPUs, GPUs, memory, expected run time, etc.) that the job will need to run. These are commands prefixed with #SBATCH, and are interpreted by SLURM as parameters describing resource requests and submissions options.
• Job steps describe tasks/scripts that needs to be executed.

Useful commands for monitoring jobs

• sinfo: reports the partitions and nodes available
• squeue -u <SUNetID>: reports the state of all submitted jobs under the UserID
• scancel <jobID> or scancel -u <SUNetID>: cancel a pending or running job (the latter aborts all jobs under the UserID)
• sbatch <submit_file_name>.sbatch: submit a job script named .sbatch

Sample submit.sbatch script (for Python)

You can create the following job submission script on Nero using a text editor such as vi by simply typing vi submit.sbatch for example. Type i to start inserting/editting and esc :wq to save and exit the text editor.

Most of the specifications should hopefully be self-explanatory. You should, of course, adjust the setting as deemed appropriate for the script you will be running. Here is an exhaustive documentation for reference and here is a shorter cheat-sheet for the essentials.

#!/bin/bash

#SBATCH --job-name=test
#SBATCH --partition=normal
#SBATCH --nodes=2    
#SBATCH --cpus-per-task=1
#SBATCH --ntasks-per-node=2
#SBATCH --mem-per-cpu=2G
#SBATCH --time 00:30:00   # format: d-hh:mm:ss
#SBATCH --output=test.log
#SBATCH --error=test.err

module load anaconda/3

python my_script.py

Note: module avail shows the modules available for loading on the cluster and module load <name_of_module> loads the corresponding module.

After submitting the script to the job queue using sbatch submit.sbatch, you can easily manage its status with the commands from the previous section. After it finishes running the my_script.py file, it will output a log file test.log and a file test.err that catches any error message (if any) during the execution. You can check out the files using cat test.err, for example.