In a message broker architecture, it is typically the clients connecting to the broker. Thus, as you correctly point out, they need to know where to connect. Bu

In a message broker architecture, it is typically the clients connecting to the broker. Thus, as you correctly point out, they need to know where to connect. But that is the same value for every single client. Using the failover transport of ActiveMQ, we have defined amq1, amq2 and amq3 - so it'll cycle through these until one of them answers. Once you're in, you only refer to the queue names. Which in Mats effectively are direct references to the endpoint name, typically in a Java-style class.method way, e.g. "OrderService.placeOrder".