Delete Management Cluster via Backend (CLI) ​

Uninstall a management cluster. During installation, the node agent bkeagent first listens to the bootstrap cluster, then switches to listening on the management cluster itself. Therefore related BKECluster (bc) objects exist in both places:

1. Bootstrap cluster (objects created at install time)
2. Management cluster itself (reconcile entry after the listener switch)

For a full uninstall, both bc objects must be handled.

Prerequisites ​

• Operate with an account that has sudo privileges and kubectl access to the target APIs.
• Access to both the bootstrap node and a management control-plane node.
• Service clusters under the management cluster should already be uninstalled.
• Read Cluster Uninstallation Overview.

Procedure ​

Step 1: Delete the local bc on the management cluster (clean target nodes) ​

After the listener has switched, node reset/cleanup is driven by the bc on the management cluster. On a management control-plane node:

# kubeconfig points to the management cluster
kubectl get bc -A
kubectl edit bc -n <namespace> <name>

For full node cleanup:

metadata:
  annotations:
    bke.bocloud.com/ignore-target-cluster-delete: "false"
spec:
  reset: true

Watch progress briefly if needed. As uninstall proceeds the management-cluster API often becomes unavailable—that is expected. Do not use kubectl get bc on the management cluster as the success check; continue with step 2.

Step 2: Delete the corresponding bc on the bootstrap cluster ​

On the bootstrap node (kubeconfig for the bootstrap cluster), remove the matching management-cluster bc to clear leftover objects on the bootstrap side:

# kubeconfig points to the bootstrap cluster
kubectl get bc -A
kubectl edit bc -n <namespace> <name>

If nodes were already cleaned in step 1, you may keep ignore-target-cluster-delete at the default "true" and only set spec.reset: true (avoid repeating reset on an already cleaned environment):

metadata:
  annotations:
    bke.bocloud.com/ignore-target-cluster-delete: "true"   # Can keep the default if nodes have been cleaned in step 1
spec:
  reset: true

Or delete directly:

kubectl delete bc -n <namespace> <name>

Note:

• Namespace/name on each cluster may differ; always check kubectl get bc -A on that cluster.
• If the listener has not switched yet and reconcile is still on bootstrap only, you can first delete from bootstrap with ignore-target-cluster-delete: "false" and spec.reset: true; after switchover, still confirm and clean any local bc on the management cluster.

What this deletes ​

Success criteria ​

• Former management-cluster API / management plane is unavailable.
• On the bootstrap node: kubectl get bc -A no longer lists that management cluster.

Follow-up ​

• Annotations / bkeagent-listener: cluster-api-provider-bke Configuration.
• Other uninstall paths and delete order: Cluster Uninstallation Overview.