Conflicting error messages on a non-functioning cluster

[evilnick]

Summary

In a multi-node cluster where one of the control-plane nodes has disappeared:

ubuntu@able-antelope:~$ sudo k8s status
Error: The node is not part of a Kubernetes cluster. You can bootstrap a new cluster with:

  sudo k8s bootstrap
ubuntu@able-antelope:~$ sudo k8s bootstrap
Error: The node is already part of a cluster
ubuntu@able-antelope:~$ 

What Should Happen Instead?

The first error is wrong. Status should instead report that the cluster is not in a working state, rather than it has not been bootstrapped

Reproduction Steps

1. Set up a cluster with two or more control planes
2. Remove one of the control planes
3. run status on the one which still exists

System information

inspection-report-20240717_133154.tar.gz

Can you suggest a fix?

No response

Are you interested in contributing with a fix?

No response

[bschimke95]

@HomayoonAlimohammadi this should be addressed by #564, right?

[HomayoonAlimohammadi]

@bschimke95 I think we still had some problems which angelos fixed in #599
lemme try to reproduce this issue

[HomayoonAlimohammadi]

so here's what I did:

• created a 3 node cluster (all cp)
• killed one of the nodes (lxc delete --force)
• run k8s status on one of the remaining nodes:
  • the first time I got a deadline exceeded
  • second time I got the status but the IP of removed node is still there and we can see heartbeats fail for that node:

Aug 22 07:37:40 cp1 k8s.k8sd[1804]: time="2024-08-22T07:37:40Z" level=error msg="Received error sending heartbeat to cluster member" error="Post \"https://10.97.72.146:6400/core/internal/heartbeat\": Unable to connect to \"10.97.72.146:6400\": dial tcp 10.97.72.146:6400: connect: no route to host" target="10.97.72.146:6400"

I retried this but this time instead of killing one node, did a k8s remove-node and everything seems fine. k8s status shows correct message on all nodes (existing and removed nodes) and the IP is removed. even removed 2 nodes in a 3 cp setup and still everything works fine.

[maci3jka]

My tests confirm @HomayoonAlimohammadi observations looks like fixed now