Anatomy of a Modern Data Pipeline Quiz
Last Updated: September 12, 2026
- What must be checked before replaying Kafka offsets?
- A) Whether the needed records remain after retention/compaction
- B) Whether Kafka automatically deduplicates every external sink
- C) Whether all batch inputs were deleted
- D) Whether every event always remains forever
Show Answer
Answer: A) Whether the needed records remain after retention/compaction
Replay requires retained input and consistent state recovery. End-to-end exactly-once also depends on sink transactions, idempotency and external effects.
- What distinguishes ETL from ELT?
- A) Whether a lake is used
- B) ETL transforms before destination load; ELT transforms after destination load
- C) Whether Kafka is used
- D) ELT is always exact and ETL always approximate
Show Answer
Answer: B) ETL transforms before destination load; ELT transforms after destination load
Moving curated data into a warehouse does not by itself make the process ELT. Check the transformation order and destination.
- Which statement about batch and stream accuracy is correct?
- A) Streaming is always approximate
- B) Batch is always exact
- C) It depends on late data, duplicates, state and output contracts; fast provisional results can be followed by finalization
- D) Checkpoints automatically make external API effects exactly-once
Show Answer
Answer: C) It depends on late data, duplicates, state and output contracts; fast provisional results can be followed by finalization
Design event time, watermarks, allowed lateness and sink semantics. Spark and Flink are not limited to one processing mode each.
- Which statement about Schema Registry compatibility checks is correct?
- A) They automatically block every database DDL and semantic change
- B) They validate the registration/serialization/CI paths where applied, and consumer tests are still needed
- C) One registration automatically ensures transitive compatibility across every version
- D) Compatibility rules remove the need for data-quality checks
Show Answer
Answer: B) They validate the registration/serialization/CI paths where applied, and consumer tests are still needed
Bypass paths and business-semantic changes need separate validation. Check the scope of backward/forward and transitive settings.