🔌 Slurm Executor#

Executing tasks (electrons) in a remote cluster via SLURM. This executor plugin interfaces Covalent with HPC systems managed by Slurm. In order for workflows to be deployable, users must have SSH access to the Slurm login node, writable storage space on the remote filesystem, and permissions to submit jobs to Slurm.

To use this plugin with Covalent, simply install it using pip:

pip install covalent-slurm-plugin

The following shows an example of how a user might modify their Covalent configuration to support Slurm:

username = "user"
address = "login.cluster.org"
ssh_key_file = "/home/user/.ssh/id_rsa"
remote_workdir = "/scratch/user"
cache_dir = "/tmp/covalent"
conda_env = ""

partition = "general"
cpus-per-task = 4
gres = "gpu:v100:4"
exclusive = ""
parsable = ""

The first stanza describes default connection parameters for a user who is able to successfully connect to the Slurm login node using ssh -i /home/user/.ssh/id_rsa user@login.cluster.org. The second stanza describes default parameters which are used to construct a Slurm submit script. In this example, the submit script would contain the following preamble:

#SBATCH --partition=general
#SBATCH --cpus-per-task=4
#SBATCH --gres=gpu:v100:4
#SBATCH --exclusive
#SBATCH --parsable

Within a workflow, users can then decorate electrons using these default settings:

import covalent as ct

def my_task(x, y):
    return x + y

or use a class object to customize behavior scoped to specific tasks:

executor = ct.executor.SlurmExecutor(
        "partition": "compute",
    "cpus-per-task": 8

def my_custom_task(x, y):
    return x + y