Can't schedule workloads after upgrade of Kubernetes from 1.12.7 to 1.12.8
Fortunately I don't have mission critical services running on my cluster and now I likely never will be :)
I have started the automatic upgrade process of my DO-managed Kubernetes cluster from 1.12.7 to 1.12.8. After the upgrade of the Master control-plane went through I was expecting the worker nodes to get upgraded as well, however it somehow got stuck and the nodes are not being upgraded.
So currently I can no longer schedule new workloads as all new pods are going into the state ContainerCreating and are stuck there.
I tried to resize the node pool and this caused a new droplet to be spun up with the up-to-date Kubernetes version from DO (Debian do-kube-1.12.8-do.4). Using
kubectl get nodes I can see the old nodes still in status
Ready while the new node (even after around 30min) still shows up as
NotReady and the latest event is
In addition all the old but still running droplets in the node pool no longer report metrics into the DO web interface.
Any idea what I can do? If nothing works I will probably have to tear down the whole cluster and set it up from scratch again. To be honest, this is very worrying to me.