- For each partition there is an lead node (Chosen at random).
- The other replicas of the partition are called as follower nodes.
- The write by the producer only happens on a single broker which becomes the leader node and it replicates the data into other brokers
- ISR: In Sync Replica (The follower does that the leader does)
- In time of failure the next ISR becomes the new leader node
Kafka Topic Architecture - Replication, Failover, and Parallel Processing - DZone Big Data