Batch scripts

As seen earlier (Job script), the typical way of creating a job is to write a job submission script. Which is a shell script (e.g. a Bash script) whose first comments, if they are prefixed with #SBATCH, are interpreted by Slurm as parameters describing resource requests and submissions options [1].

See sbatch manual page [1] for all available directives,
See our repository of examples scripts.

For instance, the following script would request one task with one CPU for 10 minutes (default allowed time), along with 2 GiB of RAM memory, in the default partition (E5):

#!/bin/bash
#
#SBATCH --job-name=test
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=1
#SBATCH --mem-per-cpu=2G

hostname -s
sleep 60s

When started, this job will launch the command hostname on the node on which the requested CPU was allocated. Then, a second job step will run the sleep command.

You can create this job submission script on any login nodes (See Login nodes) using a text editor such as nano or vim (See Editors), and save it as submit.sh.

Tip

You can submit jobs from any login node to any partition. Login nodes are only segregated for build (CPU µarch) and scratch access.

Caution

#SBATCH directives must be at the top of the script, with no empty line(s) in between.

Slurm will ignore all #SBATCH directives after the first non-comment line (that is, the first line in the script that doesn’t start with a # character). Always put your #SBATCH parameters at the top of your batch script. For example :

#!/bin/bash
#
#SBATCH --job-name=big_job

#SBATCH --mem=16G
#SBATCH --time=0-00:15:00
#SBATCH --partition=E5

will cause –mem and all following #SBATCH directives to be ignored and the job to be submitted with the default parameters.

Spaces in parameters will cause #SBATCH directives to be ignored

Slurm will ignore all #SBATCH directives after the first white space. For instance directives like those:

#SBATCH --job-name=big job

#SBATCH --mem=16 G

#SBATCH --partition = Cascade

will cause all following #SBATCH directives to be ignored and the job to be submitted with the default parameters.

We provide a repository of examples scripts, with various conditions, that may help you write your own.

Preemption mode

Normal priority jobs running on partitions with preemption mode (see Clusters/Partitions overview) will be requeue when a high priority job take over:

if the job script (or runned software) handle restart, job will resume at last saved restart point,
if the job script (or software) does not handle restart, job will be restarted from the beginning…

Most of commonly used HPC softwares do handle restart point (vasp, gromacs, aspect, RAMSES, nextflow, etc).