Deleting ZooKeeper from Streams Messaging clusters

The ZooKeeper instances deployed in Streams Messaging clusters can be manually deleted if the cluster uses KRaft for metadata management. Learn how to delete ZooKeeper from a Streams Messaging cluster.

If you use the default cluster definitions to deploy a Streams Messaging cluster, you have the option to provision the cluster with KRaft nodes. If KRaft nodes are deployed in a cluster, Kafka runs in KRaft mode and uses KRaft for metadata management instead of ZooKeeper. By default, clusters like this still include ZooKeeper instances. However, the ZooKeeper instances are not required and can be deleted. The ZooKeeper instances can be removed using Cloudera Manager.

  • Ensure that the cluster, its hosts, and all its services are healthy.
  • Deleting the service does not remove the service data from your hosts. It only removes the service from management by Cloudera Manager. All role groups under the service are removed from host templates.
  1. Access the Cloudera Manager instance managing the cluster.
  2. Remove ZooKeeper service dependencies:
    1. On the Cloudera Manager Home page, go to Configuration > Service Dependencies.
    2. Locate and clear the ZooKeeper Service property.
    3. Click Save Changes.
  3. Go to Clusters and select the ZooKeeper service.
  4. Stop the ZooKeeper service:
    1. Select Actions > Stop.
    2. Click Stop to confirm the action.
    3. Click Close after the process is finished.
  5. Delete the ZooKeeper service:
    1. Select Actions > Delete.
    2. Click Delete to confirm the action.
The ZooKeeper service is deleted from the cluster.
  • If you configured Attached Volume per Instances and the volume count is not identical for the Broker and Core_brokers host groups, continue with Configuring data directories for clusters with custom disk configurations.
  • If you did not customize Attached Volume per Instances, or customized it in a way that the volume count remained identical for the Brokers and Core_brokers host groups, continue with Give users access to your cluster.