The colors represent the node’s liveliness.
Green = healthy
Yellow = stale
Red = missing/dead
For the following info, I’m not a developer on the project, so I may not be 100% correct, but here’s how the states work as I understand them. If I’m wrong in any details I’m sure someone will correct me
If you have a cluster running and force one node to quit or if it becomes separated from the cluster, after 1 minute has elapsed, you should see that node transition to a stale (yellow) state.
The stale state indicates that the cluster is aware that the node is not responding. During this state, however, no re-replication of the ranges that exist on that node occurs. This is to prevent the heavy up-replication/rebalancing operation from happening during short-lived network partitions, where the cluster will reform shortly.
After a node has been unresponsive for some period of time (by default 10 minutes, or 9 minutes after being declared “stale”), it is declared actually missing/dead, and transitions to the red state in the UI. It is at this point that ranges which the failing node contained will be re-replicated to new, healthy nodes by the other nodes that contain replicas of those ranges.