hpckit

A Python toolkit to interact with slurmrestd on an HPC cluster via SSH.

Features

• Send raw HTTP requests to slurmrestd over SSH
• High-level client (Restd) with get, post, put, delete, etc.
• Single-shot and shared connection modes
• Extensible backend system (start with NetSSHBackend)

Install

pip install .

from hpckit import Restd, NetSSHBackend
import json
import os

allocation = "plgallocation-name-cpu"
plgrid_user = "plgusername"

job_desc = {"job": {
    "environment": ["VAR1=VALUE1","VAR2=VALUE2"],
    "account": allocation,
    "partition":"plgrid",
    "nodes":"1",
    "tasks_per_node":1,
    "memory_per_node":"512M",
    "name":"py-hpckit",
    "time_limit":{"set":True,"infinite":False,"number":1},
    "current_working_directory":f'/net/people/plgrid/{user}'
    },
    "script":"#!/bin/bash\nhostname\n"
}

backend = NetSSHBackend("ares.cyfronet.pl", user=plgrid_user, key_filename=f'/home/{os.environ.get("USER")}/.ssh/id_ed25519')
client = Restd(backend)

response = client.post( "/slurm/v0.0.39/job/submit", json.dumps(job_desc), headers={"Content-Type": "application/json"})

if response.status == 200:
    job_details = json.loads(response.read())
    print(f"Job submitted with ID: {job_details['job_id']}")
    update=client.get(f'/slurm/v0.0.39/job/{job_details['job_id']}')
    print(update.read())
else:
    print(f"Job submission failed: {response.status} {response.reason}")