Your question seems to be confusing the producer and consumer responsibilities in message delivery durability. Neha from Confluent wrote a blog post about this which should help your understanding of the how the Kafka clients handle this.
https://www.confluent.io/blog/exactly-once-semantics-are-possible-heres-how-apache-kafka-does-it/ There is a clear description in the section titled "Overview of messaging system semantics." On Fri, Mar 30, 2018 at 10:03 AM Victor L <vlyamt...@gmail.com> wrote: > Can someone clarify it: > - "At most once": messages may be lost but are never redelivered > Does it mean: "commit, when message is received by leader partition, > retries disabled for producer no matter of ISR status"? > - "At least once": commit only when message is ackd by all partitions in > replication set, if not - keep retrying by producer"? > > - "Exactly once - each message is delivered only once"... What does that > mean? Delivered to leader partition and never replicated? > > Thank you, > V >
