I am testing pure master/slave with 4.0.1 and 4.0.2 with this scenario: 1. start the master and slave brokers 2. start a producer, sent 50 messages to a Queue, stop the producer 3. start a consumer, let it consume 20 messages from the Queue 4. kill the master broker, see that the consumer reconnects to the slave 5. let the consumer consumer 10 more messages from the queue 6. kill the consumer 7. start a new consumer, expect that it will consume the remaining 20 messages from the Queue 8. restart the producer, send more messages, expect that those messages will not be consumed until all 50 of the first messages have been consumed
Note that the consumer prefetches messages from the original master. If I set the prefetch limit to 1000, meaning that the first consumer will prefetch all 50 messages from the original master before the failover, I find again the behavior described in the linked thread but also that the second consumer will NOT receive the remaining 20 original messages. If I stop the first consumer after the failover but before it has consumed all of its prefetched messages, then those messages are lost. If I set the prefetch limit to 1, then I find the behavior described in this thread: http://www.nabble.com/Bug-in-failover-transport-tf2587218.html#a7258236 on the first consumer -- including the potential for new messages to be consumed before the first 50 are consumed, but the second consumer will consume the remaining 20 messages. -- View this message in context: http://www.nabble.com/Master-Slave-and-Failover---pefetched-messages-are-lost-tf2721384.html#a7589021 Sent from the ActiveMQ - User mailing list archive at Nabble.com.
