Why do Nodes Deleted from LKE Cluster via kubectl Still Appears in Cloud Manager?

kubectl is primarily designed for managing Kubernetes resources rather than the underlying infrastructure. When you use kubectl to delete a node from a node pool, the pool size remains unchanged, so a new node will automatically be created to replace the deleted one. As a result, running kubectl delete node $nodeLabel will initiate a node recycling process.

From the node labels in your LKE cluster, it looks like you aren't using custom labels. Therefore, when kubectl delete node $nodeLabel is executed, the node is recycled, but the label remains the same. This behavior can make it seem like nodes aren’t being recycled when, in fact, they are. You can confirm if a node has been recycled by running kubectl get nodes and checking the node's age.

Keep in mind that there may be a short delay between running the kubectl delete command and the node actually being recycled. This is similar to how nodes are queued for recycling when managed via the Cloud Manager.

If your goal is to permanently delete a specific node from the node pool and reduce the pool size, you’ll need to use the Linode CLI or API. Here’s how:

• Via the CLI:

  linode-cli lke node-delete $lkeID $nodeID 
  

• Via the API:

  curl --request DELETE \
       --url https://api.linode.com/v4/lke/clusters/clusterId/nodes/nodeId \
       --header 'accept: application/json'
  

For more information, you can refer to our Delete a node guide.

-- Koffi