Cluster operation
The managed Kafka instances can be configured with different cluster operations like pausing reconciliation or stopping the cluster. See cluster operations for more details.
Restarts caused by scaling
Changing the number of replicas of the KRaft controllers restarts all controller Pods of the cluster, because every controller lists all controller Pods in its configuration (controller.quorum.bootstrap.servers).
The restart controller detects the changed ConfigMap and restarts the controllers one after another.
The brokers are not restarted.
Changing the number of replicas of the brokers does not restart any other Pods.