Topics should have replication factor greater than 1 (usually between 2 and 3), commonly 3.
If broker goes down or goes for maintenance, another broker can take over and serve the data.
Replication is synchronous if fast enough i.e In sync Replica (ISR).
At any time, one broker can be a leader for a given partition.
Producers can only send data to broker that is leader of the partition.
Consumers by default, will read from leader broker for a partition.
But as of Kafka 2.4+ it is possible to configure consumers to read from closest replica. This may help with latency and network cost.
For replication factor of 3, topic data durability can withstand 2 brokers loss. Recommended for production.
General formula is for replication factor of N, can still recover data even after permanent loss of N-1 brokers.
Last updated 23 days ago
