I’ve set up a cluster using two nodes with OF 3.8.2 and Hazelcast plugin. The service works fine while there is no problems with the nodes.
When testing the cluster recovery behavior I started shutting down one server and monitoring how well the second handle the situation and recovers. Then starting the stopped server and verified consistency. I’ve found a problem that prevents me to use clustering.
Maybe you guys can help me understand why this happens and if there is any option to fix it.
The scenario I tested is as follows:
We start with two nodes working fine from scratch: “Node A” and “Node B”
We also have two users “User 1” and “User 2” connected to Node A.
User 1 create a chat room and invites User 2 to join and chat.
User 1 sends a message “Message 1”
Now we stop Node B.
User 2 sends a message “Message 2”
Now we starts Node B again.
User 1 sends a message “Message 3”
User 2 sends a message “Message 4”
Now we stop Node A.
Thanks to the Load Balancer, both users will have a disconnection and a reconnection to the Node B.
Once connected to Node B and joined to the chatroom both users receive the History (OF history settings was set to “Show entire chatroom history”)
They only receive as history the lines:
We should expect to receive:
Thank you very much.