cassandra node replacement via backup or node removal?

Question

Hi im new to cassandra and let me put my understanding here. So if a node goes down for some reason in a cassandra cluster. If that node is gracefully removed and a new node is added, is the lost data recreated on the new node by law of replication factor ? i mean say i have a 3 cluster node with a replication factor of 2 and i loose a node. if i spin up a new node, is the the data which is lost on my dead node recreated on this new node ? If this is the case can also explain any reason why i should go with restoring from backups as opposed to let the c* architecture handle the graceful recovery for me ?

Answer 1

If you loose a node and replace it (decommissioning the dead one and adding a new) with a new one the data for the lost node will be streamed to this new node and after the streaming terminates you will have the same data as before. However you can still loose data if for example 2 out of the 3 nodes dies or due to network partitioning previous to loosing a node. So you should definitely do backups even if you have RF>1.