Bit late, but am I misunderstanding something here. Both RabbitMQ and Kafka aim to provide a pipeline for data between multiple distributed systems.

If you want asynchronous communication between nodes in a cluster (distributed applications) why not use technology specifically for that? Direct TCP, Akka, Vertx event bus, hazelcast.

If all you want is (async) communication between multiple deployed instances of your app, why are you adding extra complexity of rabbitMQ/kafka? It seems overkill for your use case. In theory you could even use async HTTP. Achieved same thing, way less complexity?