Dashboard improvements

[vlerenc]

Some weeks ago, we compiled a list of dashboards improvements that seem sensible:

• Let's add a KCM dashboard (we have CCM, MCM, etc. but no KCM dashboard so far)
• Users could get access to K8s DaemonSets/StatefulSets/Deployments/Pods (filtered by shoot cluster resources only)
• Users should get access to CCM, MCM, and the new KCM dashboards (ideally without resource usage, even if the dashboards then look empty, but CCM and possible KCM only look empty because some other insightful metrics that for instance MCM features are not yet shown there).
• User should get access to the CoreDNS dashboard (as-is)
• The nodes dashboard should also show the inodes as a metric that is eviction-relevant
• Prometheus lacks the CPU throttling information on Kubelet and the container runtime (see also issue Node details dashboard has missing metrics with containerd gardener#2800) and when we have it, we should add it to the nodes dashboard system components
• The pods dashboard should feature CPU throttling information ideally, as well

[vlerenc]

It may make sense to combine this ticket with gardener/gardener#2815.

[wyb1]

Copied task list since I cannot check the boxes

• Let's add a KCM dashboard (we have CCM, MCM, etc. but no KCM dashboard so far) -> Dashboard exists now
• Users should get access to CCM, MCM, and the new KCM dashboards (ideally without resource usage, even if the dashboards then look empty, but CCM and possible KCM only look empty because some other insightful metrics that for instance MCM features are not yet shown there).
  • CCM
  • KCM
  • MCM -> configure extensions see here
• User should get access to the CoreDNS dashboard (as-is) -> users have access
• The nodes dashboard should also show the inodes as a metric that is eviction-relevant -> check cadvisor metrics/research possible metrics
• Prometheus lacks the CPU throttling information on Kubelet and the container runtime (see also issue Node details dashboard has missing metrics with containerd gardener#2800) and when we have it, we should add it to the nodes dashboard system components -> check for existing metrics and research what is available
  • The pods dashboard should feature CPU throttling information ideally, as well

Currently out of scope for monitoring:

• Users could get access to K8s DaemonSets/StatefulSets/Deployments/Pods (filtered by shoot cluster resources only) -> Could be interesting in the future for user facing monitoring solution. Focus is only components managed by gardener + extensions.

[wyb1]

Many of these items have been completed. If new improvements are required we can create a new issue