Handle absence of scheduler and controller-manager in a more convenient way

[beorn7]

On managed K8s platforms, the metrics for scheduler and controller-manager are not necessarily accessible.

In that case, there is no good choice to set kubeSchedulerSelector and kubeControllerManagerSelector. For both, there are absence alerts (cf. #200 where the use case is to filter out the absence alert for the scheduler) and dashboards. The dashboards are useless, but they could just be ignored, while the absence alerts will always fire.

Currently, I guess, those have to be hidden in the local config, which is quite tedious. Would be nice if there was a more convenient way, e.g. setting the selectors to empty strings.

Or I am missing something here…

[brancz]

Filtering out the alert group should be easy enough no?

[beorn7]

In principle, yes. However, you have to filter out the alert group and you have to disable the dashboard, and the whole thing feels like low-level tweaking. However, I'd expect this issue to be very common (probably relevant for everybody using a managed K8s offering), and it would be really nice if there were an easy "flip this switch if you don't have a scheduler to scrape".

[brancz]

Sorry I should have been more precise. We could write a pretty easy jsonnet patch that removes the alert group and dashboard. Then anyone who uses a managed kube can just apply that patch. What do you think?

[beorn7]

Depends on the usability. As long as it boils down to "flip this switch", I think it's just an implementation detail if we use a jsonnet patch or a conditional.

[brancz]

Agreed, I tend to prefer patches when possible but you’re right.

[github-actions]

This issue has not had any activity in the past 30 days, so the
stale label has been added to it.

• The stale label will be removed if there is new activity
• The issue will be closed in 7 days if there is no new activity
• Add the keepalive label to exempt this issue from the stale check action

Thank you for your contributions!