962 fixes - fix status of some targets are DOWN in prometheus at creating kind cluster step

[engchina]

Fix status of some targets are DOWN in prometheus at creating kind cluster step.

After investigation, the reason is that kube-controller-manager, kube-scheduler and etcd are listening on localhost.
And follow the document troubleshooting-prometheus, we also know metricsBindAddress of kube-proxy should be binding to 0.0.0.0, it's better to do it at creating kind cluster step.

So I've modified creating kind cluster script for solve this issue.

Signed-off-by: engchina [email protected]

[desagar]

Hello, thank you for researching this issue, and for your pull request. Your change does solve the problem of these monitors being down in Prometheus operator.

However, Kind is not meant to be used in production use cases, and we don’t want to recommend listening on 0.0.0.0 by default, in order to avoid any potential security risk, because those 4 pods use host networking in a Kind cluster. For example, Kind recommends against using 0.0.0.0 for the api server address in this document: https://kind.sigs.k8s.io/docs/user/configuration/#networking

Individual users can explicitly enable these settings based on the troubleshooting information, if they want metrics for these services, and they determine it is safe to do in their environment.

This github issue has some discussion about this topic: prometheus-community/helm-charts#204