With ActiveMQ RAR, consumers do not reconnect after being closed by slow-consumer detection, with failover: protocol

1. Configure EAP with the ActiveMQ resource adapter, as described here: https://access.redhat.com/documentation/en-US/Red_Hat_JBoss_A-MQ/6.1/html-single/Integrating_with_JBoss_Enterprise_Application_Platform/#DeployRar-Failover The connection protocol should include failover: but only a single broker is needed to reproduce the problem. So in standalone.xml (or whatever), we have: <config-property name= "ServerUrl" > failover:(tcp: //localhost:61616)?maxReconnectAttempts=0 </config-property> 2. Enable slow-consumer detection in the broker: <slowConsumerStrategy> <abortSlowAckConsumerStrategy abortConnection= " true " maxTimeSinceLastAck= "3000" checkPeriod= "3000" ignoreIdleConsumers= " true " /> </slowConsumerStrategy I have set the slow consumer timeout to a short value (3 seconds) just to make the test quicker. 3. Build the attached project eap-ora-mdb-demo.zip and deploy the resulting JAR on EAP (e.g., copy to standalone/deployments/) This test project implements a single MDB that writes a log message, waits two seconds, and then writes another. Please ignore all the commented-out Oracle stuff – I thought that this issue was related to two-phase commit, but it is not. 4. Using an JMS client, post a message on queue/testqueue, and check that a message appears in the EAP log (so it's all working) 5. Post a large number of messages to queue/testqueue 6. While messages are being consumed, pause the EAP JVM by doing "kill -STOP" on its PID 7. Wait for the broker log to show that the consumer has been disconnected for being too slow. Check that the physical connection has been broker, e.g., using netstat 8. Resume EAP using "kill -CONT" 9. Note that, eventually, connections are reestablished, but no messages are consumed. The stack dump shows threads stuck in ResponseCorrelator.