Metrics on Kubernetes with Prometheus

[talon-vonneudeck]

This a nice overview. Thanks for making it.

Prometheus is not built-in as a running server, but service discovery and metrics are all exported by default. So when you start a Prometheus inside a Kubernetes cluster with a "this is a kubernetes cluster"-option, you will see that it will gather plenty of metrics by default.
So it feels very smooth and almost as if it was completely built-in. That is definitely worth mentioning.
It is also very easy to activate service discovery for the prometheus metrics of the software you deploy to the cluster. Just add a specific annotation to the pods, done.

[GuillaumeRochat]

Hi @talon-vonneudeck, do you have a source backing that? It does like a very interesting option considering what you mentioned. I'll try to add it over the weekend.

[talon-vonneudeck]

Here. The Kubernetes part starts at slide no. 26:
https://www.slideshare.net/brianbrazil/monitoring-kubernetes-with-prometheus-kubernetes-ireland-2016

If you just need an official statement for your report, this is more helpful:
https://prometheus.io/docs/instrumenting/exporters/#directly-instrumented-software