Hi,
we have found suspect behavior of HornetQ cluster. We have the following scenario with default production profile (BLOCK policy). We have 3 servers (A, B, C), there is queue and simple MDB which reads messages from queue and writes information into standard output on each server. After startup HQs form cluster. Client starts to send messages into queue and we can see information from MDB on each server, that is correct behavior. If we shutdown server (clean shutdown, Ctrl+c) server B is stopped and servers A and B continues with its work, that is correct behavior.
Problem is when we kill (-9) server B. MDBs on A and C are stopped and client is also stopped (because of full queue - BLOCK policy). MDBs on A and C should not be stopped. We think that this is not correct behavior. Because you will loose whole cluster in case of crashing one node. If I recall correctly, this configuration is used by our customers. It is not scenario with HA but we think that current behavior is not correct.

Server A (MDB) <- Client, sends messages into queue

Server B (MDB)

Server C (MDB)

1. Start servers
2. Start client
3. Kill server B with kill -9
4. All MDBs are stopped (but should not be)
5. If server B is restarted, A and B will continue in work