Grafana integration - Dashboards with No Data

[Bartłomiej Poniecki-Klotz]

A number of dashboards show missing data:

*Charmed Kubernetes dashboard*
- kubernetes-master CPU idle (requires manual change from kubernetes-master to kubernetes-control-plane

- API server (apiserver_request_latencies_sum & apiserver_request_latencies_count metrics not found in Prometheus)

- Admission controller (missing metrics in Prometheus)

- Etcd (missing metrics in Prometheus)

- System services CPU usage (metrics in Prometheus but filtering removes all values)

- System services memory usage (metrics in Prometheus but filtering removes all values)

- Container Disk Utilisation (metrics in Prometheus but filtering removes all values)

*Kubernetes Metrics (via Prometheus)*
- Pods CPU usage (metrics in Prometheus but filtering removes all values)

- Pods CPU usage (metrics in Prometheus but filtering removes all values)

- Pods network I/O (metrics in Prometheus but filtering removes all values)

The environment is a bare metal deployment with CK8s 1.24 and LMA (hyper-converged).
The Targets in Prometheus were fixed by edit of prometheus yaml and are GREEN now. They provided data during scraping

[George Kraft (cynerva)]

Thanks for the report. The missing apiserver, admission, and etcd metrics have been previously reported in LP:1956611. You can find replacement apiserver and admission metric names in the previous issue's description. The etcd metrics no longer exist.

The other gaps you've reported look new. I can trivially confirm the kubernetes-master issue exists in the kubernetes-control-plane grafana dashboard templates[1] and will need to be fixed there. I'm not sure what's happening with the dashboard metrics for system services, containers, and pods. We should be able to reproduce those issues in a test deployment.

[1]: https://github.com/charmed-kubernetes/charm-kubernetes-control-plane/tree/40f5689f922c64929c2e3d5a093d1efb623fad1c/templates/grafana