Jerry, after disassembling our client, pipeline and broker configuration I finally found out what's causing troubles: When working with old AS 2.3.0 (and AMQ 4.x) we were used to work with a failover transport (http://activemq.apache.org/failover-transport-reference.html) by adding "failover:" as prefix to the broker URI. That was recommended, since it was said it would help on recovering from lost connections when having high message load (even when working with a single broker and no backup broker at all). However, when deploying a pipeline using failover transport, this pipeline shows the behaviour I described above. I could reproduce it using the MeetingDetectorTaeQueue, which comes with the UIMA AS bundle.
I have not yet tried whether that affects "real" failover scenarios (using multiple broker uris) as well. Does anyone of you have experiences using UIMA AS in a scenarion with multiple brokers and failover transport? All the best, Frank
