When a new Cassandra node joins the cluster for the first time it is assigned a random token. But it can match with an already existing token in the ring. What is the token conflict detection and resolution mechanism of Cassandra?
1 Answers
The token ranges really aren’t assigned randomly. When a new node joins the cluster, the token ranges of existing nodes are bisected and reassigned to the new node. Therefore, a token “collision” really isn’t a possibility.
I should note that if you are running with multiple nodes and a replication factor > 1 (which you should be), that token ranges are replicated to neighboring nodes as a secondary range. This helps to enforce high availability of your data in the event of a hardware failure. But that still wouldn’t result in a collision.
Related