DE Mastery GitHub

Apache Kafka · KFK-04

Consumers & groups — manual commits, at-least-once, no loss across crashes

From your world. SOX completeness: a crash must never drop an unprocessed event

Practice quiz with instant feedback. For a scored run that counts toward your local progress, use uv run de quiz KFK-04.

1You want at-least-once processing (no loss, possible duplicates). What do you do?
2A consumer uses enable.auto.commit=true. It polls a batch. Auto-commit runs. Then the process crashes before it finishes the batch. After the restart, what happens to those records?
3How does Kafka divide work within a consumer group?
4What does KIP-848 (the new consumer group protocol in Kafka 4.x) mainly change?
5How is consumer lag for a partition defined?
6At-least-once can produce duplicates. What is the standard way to make the end result correct anyway?
7How do share groups (KIP-932) differ from classic consumer groups?