berops / claudie

Cloud-agnostic managed Kubernetes
https://docs.claudie.io/
Apache License 2.0
643 stars 40 forks source link

Bug: Api endpoint node on k8s delete destroys the whole cluster #1500

Closed Despire closed 2 months ago

Despire commented 2 months ago

Current Behaviour

If the deletion of a node which is the api endpoint server is executed the whole clusters fails

Expected Behaviour

When the API endpoint is deleted, the diff functions should detect this edge case and handle it appropriately by selecting a new API endpoint node from the same node pool, or from a different node pool if there are no more nodes to choose from.

Steps To Reproduce

Autoscaler selects nodes to delete (not necessarily from the end of the nodepool), and if it lands on the API endpoint, it can destroy the whole cluster. this should be handled not only in the updateNodePool function that the autoscaler calls, but also in the Diff function that determines which steps to perform.

Despire commented 2 months ago

closed by https://github.com/berops/claudie/pull/1498