ServiceExternalIPs has nothing to do with metrics. It is really a regression test for a feature we restricted; you're unlikely to do anything to trigger it.

MetricsApiProxy is the diagnostic that checks the metrics route is working.

@lmeyer  Since MetricsApiProxy can not tell the metrics route status outside openshift-infra namespace, I thought the option ServiceExternalIPs is used to do it...One thing I want to confirm is: Are we designed to not able to diagnostic the metrics route when metrics is NOT deployed in openshift-infra?

Yeah, sorry for the confusion, but ServiceExternalIPs is completely unrelated, I just happened to add the two diagnostics in the same PR.

The main purpose of the MetricsApiProxy diagnostic is not testing metrics, but testing that the HPA can work. If metrics is not deployed in openshift-infra, we assume the user didn't want it there - either didn't want metrics at all, or didn't care about the HPA (which is the part that cares about it being in openshift-infra), so there is no point in nagging them via diagnostics.

However, if we happen to have both the master config and cluster-admin context, it would probably be worthwhile to find the route configured for metrics in the master config (if it's not in openshift-infra) and test the deployment there as well. I created a Trello card for that enhancement https://trello.com/c/MkpqJU6O

@lmeyer Thanks for the info. The new card looks pretty well! I'm accepting this completed trello cards.