slappt
generates and submits Slurm job scripts for Apptainer workflows. Jobs can be configured in YAML or via CLI.
slappt
requires Python3.8+ and a few core dependencies, including click
, pyaml
, paramiko
, and requests
, among others.
To submit a job script, the host machine must either run slurmctld
with standard commands available, or must be able to connect via key- or password-authenticated SSH to the target cluster.
slappt
is available on the Python Package Index and can be installed with pip
:
pip install slappt
Say you have access to a Slurm cluster with apptainer
installed, and you have permission to submit to the batch
partition.
Copy the hello.yaml
file from the examples
directory to your current working directory, then run:
slappt hello.yaml > job.sh
Alternatively, without the configuration file:
slappt --image docker://alpine \
--shell sh \
--partition batch \
--entrypoint "echo 'hello world'" > hello.sh
Your hello.sh
script should now contain:
#!/bin/bash
#SBATCH --job-name=0477f4b9-e119-4354-8384-f50d7a96adad
#SBATCH --output=slappt.0477f4b9-e119-4354-8384-f50d7a96adad.%j.out
#SBATCH --error=slappt.0477f4b9-e119-4354-8384-f50d7a96adad.%j.err
#SBATCH --partition=batch
#SBATCH -c 1
#SBATCH -N 1
#SBATCH --ntasks=1
#SBATCH --time=01:00:00
#SBATCH --mem=1GB
apptainer exec docker://alpine sh -c "echo 'hello world'"
If already on the cluster, use the --submit
flag to submit the job directly. (Standard Slurm commands must be available for this to work.) In this case the job ID is shown if submission was successful.
You can provide authentication information to submit the script to remote clusters over SSH. For instance, assuming you have key authentication set up and your key is ~/.ssh/id_rsa
:
slappt ... --host <cluster IP or FQDN> --username <username>
There are a few things to note about the example above.
For most image definitions, specifying the shell
is likely not necessary, as the default is bash
. However, for images that don't have bash
installed (e.g., alphine
only has sh
) a different shell must be selected.
If your cluster still uses singularity
, pass the --singularity
flag (or set the singularity
key in the configuration file to true
) to substitute singularity
for apptainer
in the command wrapping your workflow entrypoint.
Note: if apptainer
or singularity
are not available by default on your cluster's compute nodes, you may need to add --pre
commands (or a pre
section to the configuration file), for instance --pre "module load apptainer"
, or:
...
pre:
- module load apptainer
...
Documentation is available at slappt.readthedocs.io.
This project is not affiliated with Slurm, Apptainer or Singularity and cannot guarantee compatibility with all cluster configurations.