I have just upgraded a Kubernetes cluster with (so far) two master nodes to v1.16.3. kubeadm get nodes marks only one of these nodes with ROLE master but the second "master" with "ROLE ". I'm not sure whether this was also the case before the upgrade. However the second "master" runs the prescribed static pods as it should: etcd, kube-apiserver, kube-controller-manager, kube-scheduler.

Could it be that this is only a matter of the node not being correctly marked but otherwise fully acting as a master? I strongly suspect this is the case (because of the presence of the static pods) and because kubectl continues to function even when I force it to contact not the load balancer for kube-apiservers, but the second master directly.)

Am I supposed to add a label node-role.kubernetes.io/master?