Check cluster health?
Make sure cluster is not in a degrade state?
Check if range replication status allows for it (i.e can survive a full failure)
On the problem node:
-
Delete the node's store directory
-
Observe the node being "rehydrated" (up-replicated)