1

I'm using RKE on some local metal instances. If I kill a node (say, unplug it), kubectl doesn't respond to API requests anymore. It times out after 60sec.

What should I look for? FWIW, the IP listed in .kube/config isn't the node that is currently dead.

$ time kubectl get nodes
Error from server (Timeout): the server was unable to return a response in
the time allotted, but may still be processing the request (get nodes)

This means the kubernetes cluster is useless if one node goes down, which isn't a great thing.

I'm using Calico/Flannel/canal, nginx-ingress, and metallb.

tedder42
  • 833
  • 1
  • 9
  • 19
  • 1
    Did you shutdown master node or a worker node? – MWZ Jul 15 '19 at 09:14
  • it's one that is in the `controlplane`, `etc`, and worker. I have ~3 that are HA masters like this. – tedder42 Jul 15 '19 at 15:04
  • How do you advertise you control-plane's endpoint to clients like kubectl ? Seems like your kubectl configuration points to fixed apiserver endpoint, not LoadBalancerIP for K8S in HA mode. – Nepomucen Mar 02 '21 at 09:12

0 Answers0