SLURM Quick Reference Guide

This page summarises the most relevant information to work with the HPC, to be used as a quick-reference guide.

This is used in the examples that follow:

SLURM Commands

Command Description
sbatch simulation.sh submit script to scheduler
squeue -u xyz123 jobs currently in the queue
scancel JOBID cancel the job with the specified ID (get the ID from the command above)
scancel -u xyz123 cancel all your jobs at once
seff JOBID basic information about the job
sacct -o jobname,account,state,reqmem,maxrss,averss,elapsed -j JOBID custom information about your job

Submission Script Template

At the top of the submission shell script, you should have your #SBATCH options. Use this as a general template for your scripts:

#!/bin/bash
#SBATCH -A TRAINING-SL3-CPU        # account name
#SBATCH -J my_simulation           # a job name for convenience
#SBATCH -D /home/xyz123/rds/hpc-work/simulations  # your working directory
#SBATCH -o logs/simulation.log     # standard output and standard error will be saved in this file
#SBATCH -p skylake                 # partition
#SBATCH -c 2                       # number of CPUs
#SBATCH --mem=1GB                  # RAM memory
#SBATCH -t 00:02:00                # Time for the job in HH:MM:SS