The AMQ 7 implementation on OpenShift includes a scale-down controller, which is necessary to redistribute messages when a broker is being withdrawn from a long time, or forever. Without this, messages in the withdrawn broker's message store would remain there indefinitely.

The current implementation does not allow any control over which broker is withdrawn – it just removes the highest-numbered replica in the stateful set.

It would be useful for some customers to be able scale down by removing a specific broker. Ideally, it shouldn't matter which broker is removed, as all nodes that host brokers should be equivalent. In practice, however, we sometimes find a situation where a particular node is suboptimal in resources, or is located in an unfavourable availability zone. In these circumstances, it would be helpful to be able to exercise more control over the scale-down process. Ideally, we should like to be able to say "Shut down broker X and move all the messages from broker X to brokers A, B, and C".