cluster-api-provider-existinginfra
cluster-api-provider-existinginfra copied to clipboard
Can't repave original control plane node without a load balancer
It tries to kubeadm join itself, but at this point kubelet is not running and so it fails. With an LB kubeadm join is routed to another control plane node and succeeds.
#44 introduces an error message explaining this.
It might be nice to support this case w/out an LB. Need to add some more logic around the kubeadm join?