How are the master components running? Some deployments use static pods on
the master which mean that they can never be relocated to a different
machine. If you are running them as pods in the cluster (i.e. via
self-hosting) then your deployment tool has likely put in the correct
tolerations so
With the master set to unschedulable, do you know what happens if one of
the pods already running on the master goes down?
Have you seen those system-type pods come back up?
It is OK to run these pods on nodes and not just the master; reference