Skip to content

dongyeoplee2/speek

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

13 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

πŸ” speek

speek is a suite of SLURM cluster monitoring tools β€” from quick one-shot snapshots to a full interactive TUI.

Installation

pip install speek

For the latest development version:

pip install --pre speek

Commands

Command Description
speek0 Classic one-shot cluster overview β€” GPU availability, per-user usage, job status
speek- Compact snapshot β€” per-model GPU bars, trends, pending pressure
speek+ Full interactive TUI β€” queue, nodes, users, stats, events, shell

speek0 β€” Classic Overview

speek0 [-u USER] [-f FILE] [-t T_AVAIL]
Option Description
-u USER Highlight a specific user (default: self)
-f FILE User info CSV file
-t T_AVAIL Time window for upcoming release, e.g. 5 m, 1 h

Shows a table of GPU usage per partition, ranked users with πŸ₯‡πŸ₯ˆπŸ₯‰, utilization-colored counts, and your current jobs.

speek- β€” Compact Snapshot

speek- [-u USER]

Per-GPU-model view with utilization bars, free/total counts, pending pressure (⏸N), availability trends (↑↓), and your running/pending jobs. Detects down nodes and shows them as DEAD.

speek+ β€” Interactive TUI

speek+

Full-featured Textual TUI with:

  • Cluster β€” speek0-style usage table (tab 1)
  • Queue β€” all cluster jobs grouped by partition, foldable
  • Nodes β€” per-partition node status with usage bars
  • Users β€” per-user GPU usage, fairshare, per-partition breakdown
  • Stats β€” GPU usage charts, per-user stacked view, issue dashboard
  • Logs β€” session CLI output (not persisted)
  • Settings β€” theme, refresh rates, cache management, log scanning
  • Info β€” cluster probe results, scheduling factors, error detection rules
  • Help β€” keybindings reference

Features

  • 70+ color themes (base16 standard)
  • OOM and error detection (11 error types) with log scanning
  • Job detail popup with stdout, stderr, GPU stats, analysis
  • Built-in shell with tab completion, history, sbatch suggestions
  • Per-job log hints in the table
  • Event notifications with read/unread tracking
  • Down node detection with DEAD indicators

Requirements

  • Python 3.8+
  • SLURM cluster with squeue, scontrol, sinfo
  • Optional: sacct, sprio, sshare, sreport, scancel for full features
  • rich (all commands), textual>=0.50.0 (speek+ only)

About

πŸ” speek lets you peek into slurm resource info such as GPU avaiability, usage per user, job status, and more.

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages