We often get "etcdserver: leader changed" responses from kubectl, which
can be ignored.
Rather than special write special code for that one error, just log and
ignore all errors on 'kubectl get nodes'.