Run a single node job in AWS PCS - AWS PCS

Run a single node job in AWS PCS

To run a job using Slurm, you prepare a submission script specifying job requirements and submit it to a queue with the sbatch command. Typically, this is done from a shared directory so the login and compute nodes have a common space for accessing files.

Connect to the login node of your cluster and run the following commands at its shell prompt.

  • Become the default user. Change to the shared directory.

    sudo su - ec2-user
cd /shared

  • Use the following commands to create an example job script:

    cat << EOF > job.sh
#!/bin/bash
#SBATCH -J single
#SBATCH -o single.%j.out
#SBATCH -e single.%j.err

echo "This is job \${SLURM_JOB_NAME} [\${SLURM_JOB_ID}] running on \${SLURMD_NODENAME}, submitted from \${SLURM_SUBMIT_HOST}" && sleep 60 && echo "Job complete"
EOF

  • Submit the job script to the Slurm scheduler:

    sbatch -p demo job.sh

  • When the job is submitted, it will return a job ID as a number. Use that ID to check the job status. Replace job-id in the following command with the number returned from sbatch. 

    squeue --job job-id
    squeue --job 1

    The squeue command returns output similar to the following: 

    JOBID PARTITION NAME USER     ST TIME NODES NODELIST(REASON)
1     demo      test ec2-user CF 0:47 1     compute-1

  • Continue to check the status of the job until it reaches the R (running) status. The job is done when squeue doesn't return anything.

  • Inspect the contents of the /shared directory. 

    ls -alth /shared

    The command output is similar to the following:

    -rw-rw-r- 1 ec2-user ec2-user 107 Mar 19 18:33 single.1.out
-rw-rw-r- 1 ec2-user ec2-user 0 Mar 19 18:32 single.1.err
-rw-rw-r- 1 ec2-user ec2-user 381 Mar 19 18:29 job.sh

    The files named single.1.out and single.1.err were written by one of your cluster's compute nodes. Because the job was run in a shared directory (/shared), they are also available on your login node. This is why you configured an FSx for Lustre file system for this cluster.

  • Inspect the contents of the single.1.out file. 

    cat /shared/single.1.out

    The output is similar to the following:

    This is job test [1] running on compute-1, submitted from ip-10-3-13-181
Job complete