I have been experimenting with the leader failure scenario and I am facing an issue.
I am using ZK ensemble of 3 servers and I have 2 clients connecting to the ZK cluster. In the connection lost callback, I have a timer which expires based on the session timeout after which I am doing a force close on ZK and calling ZK init. The 2 clients are connected to the same ZK server which is not the leader. When the leader is down, on one of the clients I see that it hits the connection lost callback and then it quickly reconnects, but the other client goes through the force close(basically the timer expired as it could not connect to one of the ZK servers) as I mentioned earlier. Why do I see a difference in the behavior when both the clients are connected to same ZK server?