marcemv90
commented
2 years ago Hello @rvowles 👋,
I don't think is worth putting the Prometheus installation as a Chart dependency of the Chart. Normally a Production installation would have an existing Prometheus installation (or other monitoring system). I know you can turn on and off the dependencies, but think if every Chart developer would add such "infrastructure" components into the Chart.
I would say is best to just mention it in the README.md something like follows, instead of putting it as a dependency:
In order to try Prometheus metrics you would need a Prometheus installation, we recommend using to the `kube-prometheus-stack, as it already comes with all the monitoring, alerting and dashboarding tools, and it is probably the one that you will find in a Production installation.
https://github.com/prometheus-community/helm-charts/tree/main/charts/kube-prometheus-stack
From my experience, I would say best here would be:
- Make the services to expose Prometheus metrics
- I know some services are already doing, maybe scope it just to add more metrics.
- Add documentation on the metrics, the meaning of each metric
- Wheter it is a counter, a gauge, etc, and what does it measures.
- Recommended Prometheus Alerts from the application standpoint
- As Grafana is the usual companion of Prometheus, create some dashboards and either:
- Upload the JSON of the Dashboard into this repo
- Upload it to https://grafana.com/grafana/dashboards to allow users easily import it directly by its ID.
- Documentation of the Grafana dashboards.
- What each panel means, how it should be interpreted, etc.
Hope you can find this comment helpful,
Regards!
rvowles
This is designed to allow these two services to be in the cluster to allow iteration on supporting useful Grafana style dashboards for FeatureHub.