I've setup a cluster in Google Container Engine and according to the Console, the API Endpoint for it seems to be completely different than the public APIs of either master or nodes in the cluster (which I thought was weird).
When trying the gcloud command line tool to manage the cluster, all commands issued with cloud preview container kubectl fail with a timeout reaching the API Endpoint.
I've tried changing Network Firewall rules and nothing seems to work.
I was trying to follow the guestbook tutorial to deploy my Docker instances in the Kubernetes managed GKE cluster and am currently stuck on not being able to issue any commands against the cluster.
Any ideas of what could be wrong?
I had the same problem. In my case, this helped: gcloud beta container get-credentials
As per our discussion on the google-containers mailing list, this was caused by the VMs' IP addresses changing when they were stopped and then re-brought up a few weeks later. Container Engine currently doesn't handle cases like this where all the IP addresses in a cluster change, and unfortunately the best option at the moment is to delete the cluster and create a new one.