I am testing the following scenario : Machine Foo has a Master/Slave pair
Machine Bar has another Master/Slave pair Brokers on Foo are linked to brokers on Bar via a network connector containing the static addresses of the two brokers on Bar and with failover=true. There is no reverse network connector from Bar to Foo. Each pair of brokers happily form a Master/Slave pair when started. Start just the master on each and the network connector correctly transmits messages. Now start the slaves and both Master/Slave pairs appear to form correctly. Messages continue to flow from Foo to Bar, however, both the Slaves throw the following exception for each message: "Cannot lookup a consumer from a connection that had not been registered." This seems related to a bug that had been reported as fixed for 4.0.2. Now comes the interesting bit - kill the Slave on Foo and the Master on Foo stops sending messages across the network connector to Bar. That shouldn't really happen, should it? If I kill the Master on Bar and leave the Slave running, then it fails over resiliently as expected, but if a Slave dies in the background it will apparently drag down the Master. Terry -- View this message in context: http://www.nabble.com/Networks-of-Master-Slave-brokers-with-failover-acting-strangely-tf2345265.html#a6528386 Sent from the ActiveMQ - User mailing list archive at Nabble.com.
