Concurrency: Reliable In-Order Event Processing

Problem

Process events from multiple partitions concurrently while preserving per-key ordering, with at-least-once semantics.

Example

events keyed by id; same-key events processed in order across threads

Constraints

• No reordering within a key; no loss

What Confluent looks for

Kafka-grounded: partition-by-key, single consumer per partition, offset commit semantics.