I tried following test to see how network of brokers react to broker failure and new broker discovery (bringing dead broker back) :
Based on following test I think ActiveMQ-5.0-SNAPSHOT is handling this senario properly. Not sure abt 4.1.1. 2 store and forward brokers (B1, B2) 3 consumer (C1 connected to B1, C2 connected to B2, C3 not yet connected to any broker) 1 Producer (P1 connected to B1 and producing 2000 msgs) - While P1 was producing to B1, messages were being consumed by C1 and C2 equally (Store and forward working). - Bringing down B2 caused C2 to throw exception and stop consuming, meanwhile C1 continued consuming. - Bringing back B2 and connecting C3 to B2, C3 also started consuming messages. (New agent was located and msgs were forwarded to it as well). - On totalling the number of times onMessage was called on each consumer, total number was 2001, so there must have been a msg which C2 was consuming when B2 was brought down and this msg was then redelivered by B1. hbruch wrote: > > Hi, > > in our low bandwith, high latency network of AMQ-4.1.1 brokers sometimes > connection failovers after network problems or broker restart are not > successful. > It seems, that though the network connection itself exists between two > brokers, consumer registrations are not forwarded to the remote broker. > > Has anyone experienced the same issue or a hint what might cause this > behaviour? > > Thanks, > Holger > -- View this message in context: http://www.nabble.com/Consumer-forwarding-and-failover-tf4952446s2354.html#a14209771 Sent from the ActiveMQ - User mailing list archive at Nabble.com.