slurm

To execute resource intensive commands, submit the commands as jobs to the slurm workload manager

Note

The login nodes are limited to a single CPU core and 2 GB of of RAM and are not suited to running many bioinformatics tools. Also, attempting to run anything beyond a text editor will likely result in receiving a curt email telling you to cut it out.

On Walnut, quick, simple jobs can be submitted using the srun command while longer and more complex jobs should be written as batch scripts and submitted using sbatch. Additionally, use the command squeue to list the current job queue and use scancel to terminate one or more jobs.

srun

Most terminal commands can be prefixed with srun to offload them to the. cluster. Specify the necessary resources by passing arguments to srun:

--mem=GBMEMORY

memory to request for the job

--cpus-per-task=CPUS

number of CPUs to request for the job

--partition=NAME

hardware partition type to run job on

Warning

srun functions like a typical command line program in that disconnecting from the terminal session will result in the command exiting early. If you plan to use srun for a command that will take some time or there is the possibility of the terminal disconnecting, start a tmux session first. Or, just use sbatch.

sbatch

Batch scripts have the following format:

#! /bin/bash -l

#SBATCH --account {ACCOUNT}
#SBATCH --job-name {JOBNAME}
#SBATCH --output bcl2fastq_demux.log
#SBATCH --mail-user={YOUR EMAIL ADDRESS}
#SBATCH --mail-type=END,FAIL
#SBATCH --mem={GB MEMORY}
#SBATCH --partition={PARTITON}
#SBATCH --nodes=1
#SBATCH --cpus-per-task={# CPU CORES}
#SBATCH -t 1:6:15

module load {NECESSARY MODULES}
{LIST OF COMMANDS...}

--account

Account names for tracking usage

--output

Output file name to redirect STDOUT into

--mail-user

Address to email updates

--mail-type

Types to updates to email about

--job-name

Job a name

--mem

Amount of memory to request

--partition

Hardware node type to request. Options include serial, debug, highmem, and gpu. Unless otherwise necessary, use serial.

--nodes

Number of compute nodes to request for job

--cpus-per-task

Number of cpu cores on that node to request

--time

Maximum amount of time a job can run before being killed. In {days}:{hours}:{minutes} format.

Submit the job by:

user@walnut:/Volumes/guth_aci_informatics/$ sbatch job_script.sh