How to run a container

Running a container is similar to running a regular job in SLURM. Please visit here for further information on general job submission. 

Below is how to setup a job script to run Singularity.


#!/bin/bash

#SBATCH --partition=long     ### Partition
#SBATCH --job-name=SingularityRing  ### Job Name
#SBATCH --time=00:10:00      ### WallTime
#SBATCH --nodes=3            ### Number of Nodes
#SBATCH --ntasks-per-node=28 ### Number of tasks (MPI processes)

#
module load singularity

mpirun singularity exec --bind /cm/shared/:/cm/shared/ --bind /packages/:/packages/ ./container.img ./ring.out

Just as with a general job singularity is ran on allocated nodes based on the size and requirements specified in the job script. In this example, the --bind directive is used to allow for file systems mounted outside of the container be visible and usable as the running user and application inside the container.