slurm-term 0.1.2

A Terminal User Interface for the Slurm workload manager

SlurmTerm

A keyboard-driven Terminal User Interface for the Slurm workload manager, built with Textual.

Monitor jobs, compose batch scripts or interactive sessions, explore cluster hardware, and inspect running jobs with real-time metrics, all without leaving your terminal.

Installation

pip install slurm-term

Requires Python 3.10+ and a working Slurm installation on the target machine.

Quick Start

slurm-term                     # on a Slurm login node
slurm-term --demo              # simulated cluster (no Slurm required)
slurm-term --since now-3days   # set initial history window

Launch slurm-term on any Slurm login or compute node. The interface opens with five tabs accessible via number keys 1–5.

Features

Monitor

Auto-refreshing job queue with color-coded states. Cancel, hold, and release jobs directly from the table. Supports multi-select (Space / Ctrl+A) for bulk actions, searchable filter bar (/), and quick-peek output (o). Press Enter on any job to open the Inspector.

Composer

Unified job submission form supporting both sbatch (batch) and srun (interactive) modes:

• Live script preview that updates as you type
• Core resource fields with inline partition summary: partition, time limit, nodes, tasks, CPUs, memory, GPUs
• Searchable catalog of 30+ additional Slurm parameters with built-in documentation
• Inline validation for time formats, memory specs, and GPU configurations
• Save and load job templates for repeated workflows
• Import existing .sbatch files (Ctrl+I) — parses directives, modules, env vars, and commands
• Five built-in default templates seeded on first run
• Auto-switch to Monitor tab after successful submission

Hardware

Cluster overview with two sub-tabs:

• Partitions: availability, time limits, node counts, CPUs, memory, GRES
• Nodes: per-node state, CPU count, memory, free memory, GRES, load

History

Completed job accounting from sacct with configurable time window (1–30 days). Displays elapsed time, CPU usage, peak memory, and exit codes. Press s to resubmit a completed job.

Inspector

Visual job inspector with three sub-tabs:

• Overview: status badge, metadata grid (partition, user, submit time, working directory, nodes), resource summary, time remaining progress bar
• Output & Error Logs: live stdout/stderr tailing with toggle (e), log file paths display
• Live Metrics: real-time CPU, memory, and GPU utilization charts

Press s to resubmit the current job — parameters are extracted and loaded into the Composer.

Key Bindings

Key	Action
q	Quit
1–5	Switch tab
Ctrl+R	Reload configuration
/	Search / filter jobs (Monitor)
Space	Toggle select (Monitor)
Ctrl+A	Select all (Monitor)
r	Refresh current view
k	Cancel selected job(s)
h / u	Hold / release selected job(s)
o	Peek output (Monitor)
i or Enter	Inspect selected job
s	Resubmit job (Inspector / History)
e	Toggle stdout/stderr (Inspector)
Escape	Back / clear selection
Ctrl+S	Submit job or launch interactive session
Ctrl+T	Save template
Ctrl+L	Load template
Ctrl+I	Import .sbatch file
Ctrl+Y	Copy script preview

Safety

All Slurm commands use subprocess.run with list arguments — no shell interpolation. Job IDs, parameter keys, parameter values, and job names are validated before use. User-controlled strings are escaped before rendering. Template names are sanitized to prevent path traversal.

Requirements

• Python >= 3.10
• Textual >= 0.90.0
• textual-plotext >= 1.0.0
• A Slurm cluster (commands: squeue, scontrol, sbatch, srun, sacct, sinfo, sstat)

Links

• Source Code
• Issue Tracker

License

MIT