This small project provides a Terminal User Interface (TUI) of your SLURM jobs, showing CPU and GPU allocations per node using characters. It is intended for use on clusters like ROMEO.

Simple version

Enriched version

• Lists all SLURM jobs for the current user
• Interactive scrolling with mouse wheel
• Auto-refresh every 30 seconds
• UI with a sidebar menu for job selection
• Shows detailed job information:
  • Job ID, Name, Submission time
  • Number of nodes, Elapsed/Max time
  • Partition, Status (color-coded), Constraints
  • PENDING reason decoding with actionable suggestions
• Visualizes node allocations for running jobs:
  • CPU usage (■ = allocated, . = free)
  • GPU usage (● = allocated, ○ = free)
  • Dynamic expansion of compressed node lists (e.g., romeo-a[045-046])
  • Nodes grouped by APU type (CPU/GPU architecture)
• Partition view (p): cluster-wide partition status (like sinfo)
• Debug view (d): raw scontrol show job output with syntax highlighting
• Log viewer (l): view stdout/stderr files with scrolling (↑↓ or wheel)
• History view (a): job history via sacct with:
  • Filter by status (ALL/RUNNING/PENDING/COMPLETED/FAILED/CANCELLED/TIMEOUT)
  • MaxRSS memory usage
  • CPU count and elapsed time
  • Exit codes
• User quota (u): view your resource limits via sacctmgr:
  • CPU/Node/Job limits
  • Current usage with progress bars
  • Remaining resources
• Cancel jobs (c): cancel selected job via scancel (with confirmation)
• Copy job ID (y): copy selected job ID to clipboard
• Sort jobs (s): cycle through sort modes (ID/Name/Entry)
• Color-coded status:
  • RUNNING → Green
  • PENDING → Yellow
  • COMPLETED → Blue
  • FAILED → Red
  • CANCELLED → Purple
  • TIMEOUT → Red

• SLURM commands available (squeue, scontrol, sinfo, scancel, sacct, sacctmgr)
• C++17 compiler (GCC or Clang recommended)
• CMake ≥ 3.14
• Terminal supporting ANSI colors

Note: FTXUI library is automatically fetched during build via CMake FetchContent.

./dev.sh all    # Clean build from scratch
./dev.sh run    # Run the application

mkdir build
cd build
cmake ..
make

This produces the rsv executable. Libraries are statically linked, so the executable is portable and can be transferred to your cluster.

The dev.sh script simplifies the development workflow:

To build locally and deploy to a remote cluster:

./deploy.sh <username> <host>
# Example: ./deploy.sh jdoe romeo.univ-reims.fr

This script:

1. Builds the project locally
2. Copies the executable to your home directory on the cluster
3. You can then SSH and run ./rsv directly

A test job script is provided to create a sample SLURM job for testing RSV:

./testjob.sh

This submits a simple 5-minute job (4 nodes, 8 tasks, 2 GPUs/node) that sleeps, allowing you to test RSV's visualization features. Modify the script parameters as needed for your cluster.

1. Run the program:

./rsv

2. The program displays:

• A sidebar menu listing all your jobs
• Details of the selected job in the main panel
• Node allocations with CPU/GPU usage visualized in a grid