sbatch

Submit batch jobs to the Slurm scheduler

TLDR

Submit batch job

$ sbatch [job.sh]

Submit with custom name

$ sbatch --job-name=[myjob] [job.sh]

Set time limit

$ sbatch --time=[00:30:00] [job.sh]

Request multiple nodes

$ sbatch --nodes=[3] [job.sh]

SYNOPSIS

sbatch [--job-name name] [--time limit] [--nodes n] [--cpus-per-task n] [script]

sbatch submits batch job scripts to the SLURM workload manager. Jobs are queued and executed when requested resources become available on the cluster.
Job scripts contain SLURM directives (lines starting with #SBATCH) specifying resource requirements, followed by commands to execute. SLURM handles job scheduling and resource allocation.

PARAMETERS

--job-name _name_

Job name

--time _time_

Time limit (HH:MM:SS)

--nodes _n_

Number of nodes

--ntasks _n_

Number of tasks

--cpus-per-task _n_

CPUs per task

--mem _size_

Memory per node

--partition _name_

Partition/queue to use

--output _file_

Output file (%j for job ID)

--error _file_

Error file

--mail-type _type_

Email notifications (BEGIN, END, FAIL)

CAVEATS

Resource requests affect queue wait time. Exceeding time limit kills job. Environment may differ from submission shell. Check partition limits with sinfo.

HISTORY

sbatch is part of SLURM (Simple Linux Utility for Resource Management), developed at Lawrence Livermore National Laboratory. SLURM is now the most widely used HPC job scheduler.