Details
-
Improvement
-
Status: Closed
-
Major
-
Resolution: Won't Fix
-
M4
-
None
-
None
Description
This code successfully failed over (upon one killed broker in a 3-cluster) 200 times, with both sender and listener failing over, and with the sending client running at full speed (no pauses between messages.)
This was tested on RHEL 5.2
Problems with it still outstanding:
1. lots of repetitive code in FailoverSession.cpp – aconway is planning to replace all this code with generated code.
2. It doesn't use the tagged-arguments interface – but I believe that will also be fixed by switching to generated code.
3. No message-replay as yet, to insure at-least-once delivery.