KIP-353 — Improve Kafka Streams Timestamp Synchronization

Accepted Kafka 2.1 Streams

Formalizes Kafka Streams stream-time tracking by defining partition time as the max timestamp seen on each partition independently, and task time as the min across all partition times, with configurable behavior for empty partitions. The ad-hoc timestamp synchronization logic was non-deterministic when a task consumed from multiple input partitions — processing order depended on fetch scheduling, producing non-repeatable results.

Details