Control plane problems: static Pod manifests, crictl ps/logs when the API server is down, scheduler and controller-manager symptoms, etcd health with etcdctl

On a kubeadm cluster, the control plane components are static Pods defined in /etc/kubernetes/manifests. When one breaks, the cause is usually a bad edit to its manifest: a typo in a flag, a wrong certificate path, a wrong port, an invalid image or a YAML indentation error. The skill is working out which component is broken and reading its logs, even when kubectl does not work.