Replacing a node but keeping data disk as-is

ideally you stop indexing into the shards on that node and call synced flush before you shut down the node just to be on the safe side