A common scenario is a newly created kubernetes control node, in which there is a small race condition in the initial startup of the static pods (kube-apiserver, kube-controller-manager, kube-scheduler), in which we need to manually restart the last 2.

Other scenario is a kubeadm cert renewal, in which to make sure some of the static pods are updated for real with the newer certs, we need to manually restart them.

Restarting a static pod is a matter of:

• moving away the corresponding /etc/kubernetes/manifest/<file>A rename with a heading . (dot) is enough.
• waiting for kubelet to detect the missing file, for it to kill the pod
• put again the file in place
• wait for kubelet to detect the file and create the static pod again