We have 2 nodes running on seperate windows machines. The first node starts up and builds the cache. The second node starts afterwards and get's the initial state. This works great for a little while. If the node 1 stays alive while node 2 is shutdown and restarted after a few hours, node 2 is no longer able to retrieve the state. We have disabled the Media Sense on both machines and set the loopback flag to true in the cluster config.
After starting node 2, node 2 sends the initial membership broadcast, receives no reply and continues, after awhile it does correctly join the cluster. In the meantime, node 1 sends heartbeats, and eventually identifies node 2, and the two nodes join the cluster and further replication happens correctly, but at that point it's too late, and no initial state replication took place.
Any ideas would be greatly appreciated...
