Durable subscribers behave a little differently across a cluster of ActiveMQ brokers. The two main issues surround messages getting stuck on other brokers in the network after a durable subscriber has disconnected and reconnected to a different broker in the network without fully unsubscribing. The reasons for disconnecting/reconnecting can be voluntary (where the subscriber is using the failover transport), or involuntary (one of the brokers in the cluster fails).

For example, if you have two brokers A and B networked together in both directions to form a cluster, and a durable subscriber connects to broker B, a producer P on broker A will have its messages properly forwarded to broker B and the durable subscriber. However, if the subscriber disconnects and reconnects to broker A, any messages sent by P while the subscriber was away will be stuck on B until the subscriber reconnects to B.

The solution is to use Virtual Destinations

Virtual topics use queues under the covers and this allows ActiveMQ to treat each subscriber as a plain old Queue subscriber. This allows ActiveMQ to replay messages that have been orphaned on another broker in the cluster using this configuration as explained in the Networks of Brokers documentation:

    <destinationPolicy>
      <policyMap>
        <policyEntries>
          <policyEntry queue="TEST.>" enableAudit="false">
            <conditionalNetworkBridgeFilterFactory replayWhenNoConsumers="true"/>
          </policyEntry>
        </policyEntries>
      </policyMap>
    </destinationPolicy>

Also, take a look at this unit test to see the full configuration and usage