badwulf 0.1.27

Minimal manager for Beowulf clusters and scientific data

Badwulf

Minimal manager for Beowulf clusters and scientific data

The goal of badwulf is a provide a minimal command line interface for accessing and managing experimental data on scientific computing servers and Beowulf clusters.

This tool is not intended to replace true cluster management and scheduling software for such as Slurm. Instead, badwulf is a lightweight package for simplying tasks such as:

• Connecting an port forwarded SSH session to a research server behind a login server

• Transfering files and directories between a local client and a research server

• Managing a simple repository of experimental data and metadata

• Searching experimental metadata for terms and keywords

• Syncing experimental data between a local client and a research server

Examples

A command line utility named wulf for a hypothetical "Badwulf" cluster with compute nodes named "Wulf-01", "Wulf-02", and "Wulf-03" could be set up as follows:

#!/usr/bin/env python3

import os
from badwulf.cli import clmanager

wulf = clmanager("Badwulf",
	nodes = {
		"01": "Wulf-01",
		"02": "Wulf-02",
		"03": "Wulf-03"},
	date = "2024-12-27",
	description = "Badwulf CLI utility",
	username = os.getenv("BADWULF_USER"),
	server = os.getenv("BADWULF_GATEWAY_SERVER"),
	server_username = os.getenv("BADWULF_GATEWAY_USER"),
	program = "wulf")

wulf.main()

This would read the environment variables $BADWULF_USER, $BADWULF_GATEWAY_SERVER, and $BADWULF_GATEWAY_USER to set up the SSH connection to the cluster.

A user could then connect to the node "Wulf-01" as:

wulf run -01