After I fixed the nodeID issue between replicated Netty servers, some tests progressed further and now they (almost) all fail trying to crash the session.
I've been trying to make sense of what happens when using Netty. One oddity is that, at the backup, we use NodeManager.awaitForLive(), to monitor when the live goes out. But AFAICT both NodeManagers in the source tree are only suitable for 'shared storage' replication as both are based on file locks, which are simply not going to work.
Is there any negative consequence of replacing that nodeManager.awaitForLive() (in the remote replication case) for a cluster connection failure listener? (FWIW any better alternatives?)
Edit: Assuming I keep using ClusterTopologyListener to detect a live server going down. Is there any way to tell appart an orderly shutdown of the live server from a crash? I can't find any, and it would be better to just avoid the quorum voting in the event of a orderly shutdown.