Skip to main content
Skip table of contents

Cloud Director Delete a Cluster

Prepare to Delete a Self-managed Workload Cluster

A self-managed cluster cannot delete itself. If your workload cluster is self-managed, you must create a bootstrap cluster and move the cluster lifecycle services to the bootstrap cluster before deleting the workload cluster.

If you did not make your workload cluster self-managed, as described in Make the New Cloud Director Cluster Self-Managed, proceed to the Delete the workload cluster section below.

  1. Create a bootstrap cluster:

    The bootstrap cluster will host the Cluster API controllers that reconcile the cluster objects marked for deletion:

    (info) NOTE: To avoid using the wrong kubeconfig, the following steps use explicit kubeconfig paths and contexts.

    CODE
    dkp create bootstrap --vcd-bootstrap-credentials=true --kubeconfig $HOME/.kube/config

    The output resembles:

    CODE
    ✓ Creating a bootstrap cluster
    ✓ Initializing new CAPI components
  2. Move the Cluster API objects from the workload to the bootstrap cluster: The cluster lifecycle services on the bootstrap cluster are ready, but the workload cluster configuration is on the workload cluster. The move command moves the configuration, which takes the form of Cluster API Custom Resource objects, from the workload to the bootstrap cluster. This process is also called a Pivot.

    CODE
    dkp move capi-resources \
        --from-kubeconfig ${CLUSTER_NAME}.conf \
        --to-kubeconfig $HOME/.kube/config
  3. Use the cluster lifecycle services on the workload cluster to check the workload cluster status:

    CODE
    dkp describe cluster --kubeconfig $HOME/.kube/config -c ${CLUSTER_NAME}

    CODE
    NAME                                                                      READY  SEVERITY  REASON  SINCE  MESSAGE
    Cluster/vcd-example                                                       True                     34s
    ├─ClusterInfrastructure - vcdCluster/vcd-example
    ├─ControlPlane - KubeadmControlPlane/vcd-example-control-plane            True                     34s
    │ ├─Machine/vcd-example-control-plane-6fbzn                               True                     37s
    │ │ └─MachineInfrastructure - vcdMachine/vcd-example-control-plane-62g6s
    │ ├─Machine/vcd-example-control-plane-jf6s2                               True                     37s
    │ │ └─MachineInfrastructure - vcdMachine/vcd-example-control-plane-bsr2z
    │ └─Machine/vcd-example-control-plane-mnbfs                               True                     37s
    │   └─MachineInfrastructure - vcdMachine/vcd-example-control-plane-s8xsx
    └─Workers
      └─MachineDeployment/vcd-example-md-0                                    True                     37s
        ├─Machine/vcd-example-md-0-68b86fddb8-8glsw                           True                     37s
        │ └─MachineInfrastructure - vcdMachine/vcd-example-md-0-zls8d
        ├─Machine/vcd-example-md-0-68b86fddb8-bvbm7                           True                     37s
        │ └─MachineInfrastructure - vcdMachine/vcd-example-md-0-5zcvc
        ├─Machine/vcd-example-md-0-68b86fddb8-k9499                           True                     37s
        │ └─MachineInfrastructure - vcdMachine/vcd-example-md-0-k8h5p
        └─Machine/vcd-example-md-0-68b86fddb8-l6vfb                           True                     37s
          └─MachineInfrastructure - vcdMachine/vcd-example-md-0-9h5vn

    (info) NOTE: After moving the cluster lifecycle services to the workload cluster, remember to use dkp with the workload cluster kubeconfig.

Delete the Workload Cluster

  1. Evict all workloads that use Persistent Volumes.
    We recommend using dkp with the bootstrap cluster to delete all worker node pools. This evicts all workloads that use Persistent Volumes. See information on Cloud Director Delete Node Pools.

Persistent Volumes (PVs) are not deleted automatically by design in order to preserve your data. However, they take up storage space if not deleted. You must delete PVs manually. Information for backup of a cluster and PVs is on the page in documentation called Back up your Cluster's Applications and Persistent Volumes .

  1. Use dkp with the bootstrap cluster to delete the workload cluster. Delete the Kubernetes cluster and wait a few minutes:

    Before deleting the cluster, DKP deletes all Services of type LoadBalancer on the cluster. To skip this step, use the flag --delete-kubernetes-resources=false.

    CODE
    dkp delete cluster --kubeconfig $HOME/.kube/config --cluster-name ${CLUSTER_NAME}

    CODE
    ✓ Deleting Services with type LoadBalancer for Cluster default/vcd-example
    ✓ Deleting ClusterResourceSets for Cluster default/vcd-example
    ✓ Deleting cluster resources
    ✓ Waiting for cluster to be fully deleted
    Deleted default/vcd-example cluster

    After the workload cluster is deleted, delete the bootstrap cluster.

Delete the Bootstrap Cluster

  1. Delete the bootstrap cluster using dkp delete:

CODE
dkp delete bootstrap --kubeconfig $HOME/.kube/config
CODE
✓ Deleting bootstrap cluster

Delete a Managed Cluster

In the CLI, using the kubeconfig for the Managed cluster that you want to delete. In order to properly delete a Managed cluster, you must first delete the PVCs using these steps:

  1. Export your workspace namespace:

    CODE
    export WORKSPACE_NAMESPACE=<your-workspace-namespace>
  2. Get all of the PVCs that you need to delete from the Applications deployed by Kommander:

    CODE
    kubectl get pvc --namespace ${WORKSPACE_NAMESPACE}
  3. Delete those PVCs and the Managed cluster deletion process will continue:

    CODE
    kubectl delete pvc --namespace ${WORKSPACE_NAMESPACE} <pvc-name-1> <pvc-name-2>

Known Limitations

The DKP version used to create the workload cluster must match the DKP version used to delete the workload cluster.

Next Step:

Day 2 - Cluster Operations Management

JavaScript errors detected

Please note, these errors can depend on your browser setup.

If this problem persists, please contact our support.