Transitioning from slurm to LSF

Overview

• This page discusses the similarities between slurm and the LSF system when it comes to managing and submitting jobs.

• This page won’t cover everything about LSF but is designed to be a stepping stone to go from when transitioning to the Compute Platform.

Comparible Commands

• The following list is slurm commands and their equivalent within LSF.

  • sbatch - bsub

  • squeue - bjobs

  • scancel - bkill

  • sinfo - bhosts

• More information about job commands can be found in our job execution examples documentation.

Job Scripts

• Most of the documentation and use of the Compute Platform assumes submitting jobs directly with bsub.

• Jobs can also be submitted via job scripts just like in slurm.

• An example slurm job script might look like the following.

#!/bin/bash
#SBATCH --job-name=jobname
#SBATCH --output=results.txt
#SBATCH --ntasks=1
#SBATCH --time=10:00
#SBATCH --mem-per-cpu=100

commands

• While an example job script in LSF would look like the following.

#!/bin/bash
#BSUB  -n 1
#BSUB -q submission_queue
#BSUB -G my_group
#BSUB -M 8000000
#BSUB -oo ${HOME}/path/to/log_file
#BSUB -a 'docker(container_name:container_tag)'

commands

• Users can create job scripts just like they would for slurm.

• The -a option must be included to specify which Docker image the job is using.

• Unlike a more traditional HPC system the Compute Platform makes use of Docker images instead of modules.

• You can read more about Docker and it’s association with the Compute Platform at the following links.

  • Docker and the RIS Compute Service

  • Docker Basics: Building, Tagging, & Pushing A Custom Docker Image

  • RIS Docker Workshop

• You can submit a job script in LSF using the following command.

bsub < job_example.bsub