Creating and Submitting a Job

Creating SlurmJob object

SlurmJob is a class to manage SLURM job.

Required arguments for initiating SlurmJob object:

• workdir: Working directory for the job.

• template_file: The name of template file(not full path) to use. Available templates are listed in slurmer.list_templates()

• job_name, node_partition, num_nodes, num_tasks

Optional arguments:

• node_list: List of nodes to use in a batch job. It corresponds to --nodelist option in srun and sbatch

• node_exclude list: List of nodes to exclude in a batch job. It corresponds to --exclude option in srun and sbatch

• These two options cannot be used together.

Any additional keyword arguments will be used to substituting placeholder variables in user template, thus enabling the creation of arbitrary template. Here is an example:

from slurmer.job import SlurmJob


job = SlurmJob(
    workdir="my_job",
    template_file="default.sh",
    job_name="my_slurm_job",
    node_partition="g1",              # example.
    num_nodes=2,
    num_tasks=32,
    node_list=[1, 2],
    exec_command="echo Hello,world!", # optional, but this is required by default.sh
)

Submitting the job

To submit the batch job, call submit() method.

job.submit()

By default, the job script is temporarily created and won’t be written to the working directory. However, you can persist it with optional argument write_job_script. To write the job script in the file system:

job.submit(write_job_script=True, job_script_name="my_slurm_job.sh")