Apache Kafka Practice

Messages in Apache Kafka are considered committed when all in-sync replicas have received them. This means that not only has the leader broker received the message, but also that the designated follower brokers, which are part of the replication group for a specific partition, have acknowledged receipt of the message. This approach ensures data durability and consistency across the Kafka cluster. By requiring acknowledgment from all in-sync replicas, Kafka can provide stronger guarantees about data safety in case of broker failures. If the leader broker fails after sending messages to the replicas, as long as the messages are committed, they will still be accessible due to the replicas' data. This commitment process is integral to Kafka's reliability model. It ensures that messages that reach this state are safe from being lost, assuming the remaining replicas are operational and in sync. Thus, the system can withstand certain types of failures without data loss, which is critical for many applications that rely on Kafka. Other options, such as simply when the producer sends the messages or when at least one replica has received them, do not provide the same level of data guarantees, as they do not ensure that the data has been reliably stored across the necessary nodes in the cluster. Acknowledging messages by Zookeeper is not part of the message commitment process