g. Summary

In this part of the lab you have seen how metrics can be collected by Prometheus and visualize them in Grafana. Prometheus provides a time series database and will connect to each node to collect metrics collected by the Node Export.

The Node Exporter you used here allows you to visualize system level metrics. Other metric exporters that you could consider using include the following:

  • DCGM Exporter enables your to usage and power metrics from your GPU based nodes.
  • PCP Exporter provides means to analyze system performance metrics at the kernel level.
  • eBPF Exporter gives you access to kernel level tracing with eBPF to visualize metrics such as the block IO latency, thread-blocks and more.

There are many Prometheus exporters readily available and you can even write your own with a few lines of Go, Python, C, C++ and other languages. In fact, nothing prevents you to build an exporter for your applications and visualize them in Grafana.