Disable Tungsten Fabric analytics services

Available since MOSK 23.3 TechPreview

By default, analytics services are part of basic setups for Tungsten Fabric deployments. To obtain a more lightweight setup, you can disable these services through the custom resource of the Tungsten Fabric Operator.

Warning

Disabling of the Tungsten Fabric analytics services requires restart of the data plane services for existing environments and must be planned in advance. While calculating the maintenance window for this operation, take into account the deletion of the analytics DaemonSets and automatic restart of the tf-config, tf-control, and tf-webui pods.

To disable Tungsten Fabric analytics services:

  1. Open the TFOperator custom resource for editing:

    kubectl -n tf edit tfoperators.operator.tf.mirantis.com openstack-tf
    
    kubectl -n tf edit tfoperators.tf.mirantis.com openstack-tf
    
  2. Disable Tungsten Fabric analytics services in the TFOperator custom resource:

    spec:
      settings:
        disableTFAnalytics: true
    
    spec:
      services:
        analytics:
          enabled: false
    
  3. Clean up the Kubernetes resources. To free up the space that has been used by Cassandra, ZooKeeper, and Kafka analytics storage, manually delete the related PVC:

    kubectl -n tf delete pvc -l app=cassandracluster,cassandracluster=tf-cassandra-analytics
    kubectl -n tf delete pvc -l app=tf-zookeeper-nal
    kubectl -n tf delete pvc -l app=tf-kafka
    
  4. Remove the tfanalytics=enabled and tfanalyticsdb=enabled labels from nodes, as they are not required by the Tungsten Fabric Operator anymore.

  5. Manually restart the vRouter pods:

    Note

    To avoid network disruption, restart the vRouter pods in chunks.

    kubectl -n tf get pod -l app=tf-vrouter-agent
    kubectl -n tf delete pod <POD_NAME>
    
  6. Delete terminated nodes from the Tungsten Fabric configuration through the Tungsten Fabric web UI:

    Caution

    With disabled Tungsten Fabric analytics, the Tungsten Fabric web UI may not work properly.

    1. Log in to the Tungsten Fabric web UI.

    2. On Configure > Infrastructure > Nodes > Analytics Nodes, delete all terminated analytics nodes.

    3. On Configure > Infrastructure > Nodes > Database Analytics Nodes, delete all terminated database analytics nodes.

  7. Since MOSK 24.1, disable monitoring of the Tungsten Fabric analytics services in StackLight by setting the following parameter in StackLight values of the Cluster object to false:

    tungstenFabricMonitoring:
      analyticsEnabled: false
    

    When done, the monitoring of the Tungsten Fabric analytics components will become disabled and Kafka alerts along with the Kafka dashboard will disappear from StackLight.

Now, with the Tungsten Fabric analytics services successfully disabled, you have optimized resource utilization and system performance. While these services are deactivated, related alerts may still be present in StackLight. However, do not consider such alerts as indicative of the actual status of the analytics services.