A simple website-based resource monitor for SLURM system.
It parses the results from sinfo
command every X seconds,
and host the statistics in an internal webpage.
If you are managing CPU/GPU resources with SLURM for your research group, you might find this repository helpful.
This repo is originally developed and maintained for Visual Geometry Group, Oxford.
- [Mar 2023] Monitoring GPU leaderboard by partitions
- [Feb 2023] Update leaderboard: show gpu memory
- [Aug 2022] Update documentation for command line usage
- [Mar 2022] Support monitoring Disk IO
- [Mar 2022] Support toggling light/dark mode
- [Feb 2022] Update leaderboard: show bash/sbatch jobs
- [Jan 2022] Initial release! Monitor three elements:
- GPU nodes (#gpu, #cpu-core, memory)
- Leaderboard,
- Waiting queue
Run pip install -r requirements.txt
to install the dependencies
For example, run the command: python app.py --host localhost --port 8080
.
Then the website will be hosted at localhost:8080/
.
You should change the host and port for your server. Also change the index.html for header/footer and formatting.
Running as a command-line tool (inherited from slurm_gpustat)
-
You can also use slurm_gpustat in the command line by
python slurm_web/slurm_gpustat.py
-
or by adding the following alias to your
.bash_profile
:alias slurm_gpustat='python ~/slurm_web/slurm_gpustat.py'
With some supports from slurm_gpustat.