Chapter 18. Removing a cluster from management
When you remove an OpenShift Container Platform cluster from management that was created with Red Hat Advanced Cluster Management for Kubernetes, you can either detach it or destroy it.
Detaching a cluster removes it from management, but does not completely delete it. You can import it again, if you want to manage it. This is only an option when the cluster is in a Ready state.
Important:
- Destroying a cluster removes it from management and deletes the components of the cluster.
When you detach a managed cluster, the related namespace is automatically deleted. Do not place custom resources in this namespace.
18.1. Remove a cluster by using the console
- From the navigation menu, navigate to Infrastructure > Clusters.
- Select the option menu beside the cluster that you want to remove from management.
Select Destroy cluster or Detach cluster.
Tip: You can detach or destroy multiple clusters by selecting the check boxes of the clusters that you want to detach or destroy. Then select Detach or Destroy.
Note: If you attempt to detach the hub cluster, which is named local-cluster, be aware that the default setting of disableHubSelfManagement is false. This setting causes the hub cluster to reimport itself and manage itself when it is detached and it reconciles the MultiClusterHub controller. It might take hours for the hub cluster to complete the detachment process and reimport. If you want to reimport the hub cluster without waiting for the processes to finish, you can enter the following command to restart the multiclusterhub-operator pod and reimport faster:
oc delete po -n open-cluster-management `oc get pod -n open-cluster-management | grep multiclusterhub-operator| cut -d' ' -f1`
You can change the value of the hub cluster to not import automatically by changing the disableHubSelfManagement value to true, as described in Installing while connected online.
18.2. Remove a cluster by using the command line
To detach a managed cluster by using the command line of the hub cluster, run the following command:
oc delete managedcluster $CLUSTER_NAME
Note: If you attempt to detach the hub cluster, which is named local-cluster, be aware that the default setting of disableHubSelfManagement is false. This setting causes the hub cluster to reimport itself and manage itself when it is detached and it reconciles the MultiClusterHub controller. It might take hours for the hub cluster to complete the detachment process and reimport. If you want to reimport the hub cluster without waiting for the processes to finish, you can enter the following command to restart the multiclusterhub-operator pod and reimport faster:
oc delete po -n open-cluster-management `oc get pod -n open-cluster-management | grep multiclusterhub-operator| cut -d' ' -f1`
You can change the value of the hub cluster to not import automatically by changing the disableHubSelfManagement value to true, as described in Installing while connected online.
18.3. Remove remaining resources after removing a cluster
If there are remaining resources on the managed cluster that you removed, there are additional steps that are required to ensure that you remove all of the remaining components. Situations when these extra steps are required include the following examples:
-
The managed cluster was detached before it was completely created, and components like the
klusterletremain on the managed cluster. - The hub that was managing the cluster was lost or destroyed before detaching the managed cluster, and there is no way to detach the managed cluster from the hub.
- The managed cluster was not in an online state when it was detached.
If one of these situations apply to your attempted detachment of a managed cluster, there are some resources that cannot be removed from managed cluster. Complete the following steps to detach the managed cluster:
-
Make sure you have the
occommand line interface configured. Make sure you have
KUBECONFIGconfigured on your managed cluster.If you run
oc get ns | grep open-cluster-management-agent, you should see two namespaces:open-cluster-management-agent Active 10m open-cluster-management-agent-addon Active 10m
-
Download the
cleanup-managed-clusterscript from thedeployGit repository. Run the
cleanup-managed-cluster.shscript by entering the following command:./cleanup-managed-cluster.sh
Run the following command to ensure that both namespaces are removed:
oc get ns | grep open-cluster-management-agent
18.4. Defragmenting the etcd database after removing a cluster
Having many managed clusters can affect the size of the etcd database in the hub cluster. In OpenShift Container Platform 4.8, when you delete a managed cluster, the etcd database in the hub cluster is not automatically reduced in size. In some scenarios, the etcd database can run out of space. An error etcdserver: mvcc: database space exceeded is displayed. To correct this error, reduce the size of the etcd database by compacting the database history and defragmenting the etcd database.
Note: For OpenShift Container Platform version 4.9 and later, the etcd Operator automatically defragments disks and compacts the etcd history. No manual intervention is needed. The following procedure is for OpenShift Container Platform version 4.8 and earlier.
Compact the etcd history and defragment the etcd database in the hub cluster by completing the following procedure.
18.4.1. Prerequisites
-
Install the OpenShift CLI (
oc). -
Log in as a user with
cluster-adminprivileges.
18.4.2. Procedure
Compact the
etcdhistory.Open a remote shell session to the
etcdmember, for example:$ oc rsh -n openshift-etcd etcd-control-plane-0.example.com etcdctl endpoint status --cluster -w table
Run the following command to compact the
etcdhistory:sh-4.4#etcdctl compact $(etcdctl endpoint status --write-out="json" | egrep -o '"revision":[0-9]*' | egrep -o '[0-9]*' -m1)
Example output
$ compacted revision 158774421
-
Defragment the
etcddatabase and clear anyNOSPACEalarms as outlined in Defragmentingetcddata.
