Remove a Node from a Cluster Locally

You can remove nodes from a cluster using the local CLI. The procedure to remove a node is different in a two-node cluster than in a cluster with three or more nodes.
  • Remove a worker node from a cluster with three or more nodes.
    1. Decommission the worker node from the worker node’s CLI:
      admin@WF-500> request cluster decommission start
      The decommission command only works with clusters that have three or more nodes. Do not use decommission to remove a node in a two-node cluster.
    2. Confirm that decommissioning the node was successful:
      admin@WF-500> show cluster membership
      This command reports decommission: success after the worker node is removed from the cluster. If the command does not display successful decommission, wait a few minutes to allow the decommission to finish and then run the command again.
    3. Delete the cluster configuration from the worker node’s CLI:
      admin@WF-500># delete deviceconfig cluster
    4. Commit the configuration:
      admin@WF-500># commit
    5. Check that all processes are running:
      admin@WF-500> show system software status
    6. Remove the worker node from the controller node’s worker list:
      admin@WF-500(active-controller)# delete deviceconfig cluster mode controller worker-list <worker-node-ip>
    7. Commit the configuration:
      admin@WF-500(active-controller)# commit
    8. On the controller node, check to ensure that the worker node was removed:
      admin@WF-500(active-controller)> show cluster all-peers
      The worker node you removed does not appear in the list of cluster nodes.
  • Remove a controller node from a two-node cluster.
    Each cluster must have two controller nodes in a high availability configuration under normal conditions. However, maintenance or swapping out controller nodes may require removing a controller node from a cluster using the CLI:
    1. Suspend the controller node you want to remove:
      admin@WF-500(passive-controller)> debug cluster suspend on
    2. On the controller node you want to remove, delete the high-availability configuration. This example shows removing the controller backup node:
      admin@WF-500(passive-controller)> configure 
      admin@WF-500(passive-controller)# delete deviceconfig high-availability
    3. Delete the cluster configuration:
      admin@WF-500(passive-controller)# delete deviceconfig cluster
    4. Commit the configuration:
      admin@WF-500(passive-controller)# commit
    5. Wait for services to come back up. Run show cluster membership and check the Application status, which shows all services and the siggen-db in a Ready state when all services are up. The Node mode should be stand_alone.
    6. On the remaining cluster node, check to ensure that the node was removed:
      admin@WF-500(active-controller)> show cluster all-peers
      The controller node you removed does not appear in the list of cluster nodes.
    7. If you have another WildFire appliance ready, add it to the cluster as soon as possible to restore high-availability (Configure a Cluster and Add Nodes Locally).
      If you do not have another WildFire appliance ready to replace the removed cluster node, you should remove the high availability and cluster configurations from the remaining cluster node because one-node clusters are not recommended and do not provide high availability. It is better to manage a single WildFire appliance as a standalone appliance, not as a one-node cluster.
      To remove the high availability and cluster configurations from the remaining node (in this example, the primary controller node):
      admin@WF-500(active-controller)> configure 
      admin@WF-500(active-controller)# delete deviceconfig high-availability 
      admin@WF-500(active-controller)# delete deviceconfig cluster 
      admin@WF-500(active-controller)# commit
      Wait for services to come back up. Run show cluster membership and check the Application status, which shows all services and the siggen-db in a Ready state when all services are up. The Node mode should be stand_alone.

Related Documentation