Closed smee closed 9 years ago
Please ask questions on rabbitmq-users.
Federation has to move messages between nodes and as long as it does so using "regular" means of client connections and consumers/exchanges/etc, it cannot adapt to route every possible message any other way than via default exchange. BCC
extension is the only option that may work even in theory.
I'd say relying on the routing key on a message on the consumer end is not a great idea in general. And if you don't do that, the fact that queue federation changes it when a message is re-published shouldn't matter (unless I'm missing something).
Queue federation seems to differ in behaviour from exchange federation quite significantly. According to my experiments and the source queue federation changes the routing key to the name of the source queue while exchange federation does not. The documentation does not mention this fact.
Background: I'm evaluating a strategy to update the version of a RMQ-cluster at runtime transparently for users. The logic would be:
The charm of this usage of queue federation would be that no matter to which cluster a client and a producer are connected to, they can still communicate without being aware of their location.
My question: I would prefer to keep messages transported over queue federation unchanged by the server. Is this possible in any way?