Monitoring¶

Research Computing¶

What tools can you use to monitor the health of a cluster? Here are some recommendations:

• Grafana: provides operational dashboards to monitor cluster health.

• Slurm-web: web frontend and REST API to [Slurm]({{ site.baseurl }}/docs/schedulers/slurm) workload manager. You can see job states, reservations, and node metrics.

• Prometheus: open source monitoring solution used outside of the HPC space

• Telegraf: open source server to collect system metrics.

• InfluxDB: time series data platform

• Netdata: real-time node-level performance, however does not provide much history

• XDMoD 17: provides historical performance of jobs and queues, and resources

• XDMoD SUPReMM: an extension that also provides job-level performance

• Ganglia: Not real-time, but provides node-level performance and history. Note that in 2019 they were looking for maintainers.