@krasserm: I'm working on a disaster recovery plan using eventuate and I'm looking for feedback. I want to use ephemeral storage for all application instances and use Cassandra to seed my recovery process. The idea is to set up a hub and spoke style replication graph of locations. The central hub location serves as the DR location and writes its log to Cassandra for durability. All other locations in the replication graph write their logs using LevelDB. If all locations are terminated, I would first bring up the DR location, and then recover each location using the hub and spoke pattern. I'm struggling with the recovery of hub locations because their replication connections include their spokes, which aren't recovered yet. Any thoughts on this idea?