I'll try to explain my scenario (and problem) in more detail.
1) Nodes A, B, C have joined the group and work "normally".
2) Node A crashes;
3) Node A recovers and joins the group; meanwhile, nodes B and C have
worked normally without any failures, join, leave etc.
My point of view is that of node A.
I would expect to receive (at the application level) the
join-notification associated with nodes B and C. In contrast, the
following happens:
*) In some executions, I do receive both join-notifications
*) In some other executions, I receive only one join-notification
(sometimes from B and some other times from C)
In ALL executions, though, the log contains the correct node list
(i.e., A, B, C).
I am really lost
Thank you in advance
Enrico Sorio