Hi, We are experiencing a strange issue where we from time to time end up with a stale network bridge from one broker to another, where the destination broker does not have a connected consumer. As a result, all messages sent to this broker ends up being expired, and causing timeouts.
We have the following configuration on the brokers: ActiveMQ Classic, latest version Transport connector settings: rebalanceClusterClients="true" updateClusterClients="true" Network Connector settings: conduitSubscriptions="false" decreaseNetworkConsumerPriority="false" messageTTL="15" suppressDuplicateQueueSubscriptions="true" We are running a cluster of three brokers (network of brokers configuration, multicast discovery), where consumers and producers will connect to any of the three brokers at random. In order to achieve an even load balancing between the consumers, we are using conduitSubscriptions="false" and decreaseNetworkConsumerPriority="false". >From time to time, we end up with stale network bridges between the brokers. >Given three brokers: A, B and C. We have two producers, P1 and P2. We have two >consumers C1 and C2. Sometimes we see a scenario where P1, C1 and C2 is connected to B, and only P1 is connected to A. When browsing active consumers on B, a network bridge towards A is listed, and we see messages being routed towards this bridge. Since A does not have any active consumers, the messages end up being expired on A. Even if we shut down all consumers on all nodes, the bridge still persist (?!). We use ActiveMQ as the message bus between our REST layer nodes and service nodes, so for every message routed down a stale bridge, we experience a timeout in our service, which is less than optimal. Has anyone seen anything similar before, or have any ideas to how we can avoid this issue? Best regards
