Data is redundantly replicated between TiKV nodes using the Raft consensus algorithm to ensure recoverability when a node failure occurs.
At the bottom layer, TiKV uses a model of replication log + State Machine to replicate data. For the write requests, the data is written to a Leader and the Leader then replicates the command to its Followers in the form of log. When the majority of nodes in the cluster receive this log, this log is committed and can be applied into the State Machine.