Ordering and keys
A fictional parcel service has six partitions and keys records by parcel ID. Two events for the same parcel arrive in order. Where is ordering guaranteed? Within a partition. Kafka does not provide one global order across the six logs. A stable key normally keeps that parcel on one partition while the partition count and partitioner remain unchanged.
Commits and recovery
A worker processes offset 7 and crashes before committing offset 8. What can its replacement read? If the stored commit still points to 7, it can read record 7 again. The committed offset identifies the next record to read. Explain how an idempotent destination handles the duplicate before claiming exactly-once processing.
Rebalances and lag
A group has two consumers and three partitions. Adding a fourth consumer cannot give all four consumers a partition from this topic. Membership changes cause reassignment. Describe what happens to in-flight work, and distinguish the consumer position from the durable committed offset.
Replication and acknowledgment
A producer uses acks=all, replication factor 3 and a minimum in-sync replica count of 2. If the in-sync set drops below 2, a write cannot meet that requirement. Explain the availability trade-off and why lowering the setting changes the durability promise.
Practice these decisions in the Kafka course and compare your reasoning with the Interview Hub. For a broader pipeline discussion, combine Kafka with Airflow and a warehouse.