Stateless Brokers

In Tansu, all brokers are stateless and equal:

• Brokers do not replicate storage, so there is no concept of an in-sync replica
• Every broker is the leader of each topic partition
• Every broker is the coordinator of each consumer group
• Every broker is the coordinator of each transaction
• Every broker can be shutdown when not handling a Kafka API request

In Tansu brokers act without consensus (e.g., Raft, ZooKeeper). When an API request is made, the receiving broker optimistically executes the operation using its current state. At the end of the operational the new state is persisted to the storage engine. If the state was out of date, the storage engine rejects the request, responding with its current state. The broker then retries the operation with the newer state.