slurmy

A terminal UI for SLURM. Browse your completed and running jobs, read queued jobs, and tail stdout — all without leaving the terminal.

slurmy-demo-v04.mov

Features

• Three tabs — Jobs, Cluster, and Users, switched with tab or 1/2/3
• Jobs — your jobs from the last 30 days (sacct) with pending jobs and their wait reason pinned to the top (squeue); cancel the selected job with c
• Live stdout/stderr tail — follows the selected job's output like tail -f; press o/e to switch stream and enter to scroll back through the log. Handles multi-GB files and strips ANSI / tqdm progress bars
• Cluster — at-a-glance GPU summary (used / free / down) plus a colour-coded grid of every node with its state, per-GPU allocation, and CPU usage (sinfo)
• Users — ranked bar chart of GPUs each user is running and has queued, across the whole cluster (squeue)

Keybindings

Key	Action
tab / shift+tab	Cycle between tabs (Jobs / Cluster / Users)
1 / 2 / 3	Jump straight to a tab
↑/↓ or j/k	Navigate job list / scroll the active tab
c	Cancel selected job (Jobs tab, running/pending only)
y	Confirm cancellation
n or Esc	Dismiss confirmation
q or Ctrl+C	Quit

Requirements

• A SLURM environment with sacct, squeue, and sinfo in $PATH
• Terminal with 256-colour support

No Go toolchain needed on the cluster — just copy the pre-built binary.

Installation

Download a release binary (recommended)

Grab the latest binary from the Releases page and copy it to your cluster:

# Linux amd64 (most clusters)
wget -O slurmy https://github.com/bertaveira/slurmy/releases/latest/download/slurmy-linux-amd64
chmod +x slurmy
./slurmy

Add to your PATH (optional)

Move the binary to a directory in your $PATH, or add an alias to your shell config:

# Option 1: Move to ~/bin (create it if needed)
mkdir -p ~/bin
mv slurmy ~/bin/

# Option 2: Add an alias to your .bashrc
echo 'alias slurmy="~/path/to/slurmy"' >> ~/.bashrc
source ~/.bashrc

Build from source

git clone https://github.com/bertaveira/slurmy
cd slurmy
go build -o slurmy .
./slurmy

Cross-compile from macOS to Linux:

make build-linux-amd64   # → slurmy-linux-amd64
make build-linux-arm64   # → slurmy-linux-arm64
make build-all           # both architectures

Then scp the binary to your cluster and run it.

Demo mode

Run with --demo to launch the UI on synthetic data — no SLURM, no cluster, no network. Useful for screenshots, recordings, or just trying it out:

slurmy --demo

It shows a made-up set of jobs (with a live-streaming log), a fake node grid, and fake per-user GPU usage. Synthetic log files are written under /tmp/slurmy-demo/.

License

MIT