Wes Parish wrote:
I have seen similar problems (although no out of order messages so
far). I am testing on RHEL5.3 using the C++ API with a
MessageReplayTracker to handle failover, and after a failover I am
calling replay().
This scenario is a consumer consuming as fast as possible, a producer
producing 5k messages per second and a "service qpidd stop" on a the
broker that both the producer and consumer are connected to. Both
clients seem to fail over very quickly to the other broker in the
cluster, although the consumer is receiving duplicate messages.
There is a snippet from the sequence number log below. Is this
expected?
Yes, there are two sources of possible duplication at present. One is
from the publisher to the broker as it replays any in-doubt messages;
the other is from the broker to the consumer as the broker will requeue
and redeliver any unacknowledged messages when the clients session
ends/dies.
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:users-subscr...@qpid.apache.org
