Alibaba: How Does RocketMQ Guarantee Message Reliability?

RocketMQ guarantees reliability at three levels. At the producer level, synchronous send mode waits for broker acknowledgment before returning success, with automatic retry on failure (default 2 retries). For scenarios requiring local-transaction-plus-message atomicity, RocketMQ's transaction message feature uses a half-message mechanism: the message is invisible to consumers until the local transaction commits, and a check-back mechanism handles ambiguous states. At the broker level, synchronous flush mode ensures messages are written to disk before acknowledging the producer, and Dledger-based replication synchronizes data across replicas before returning. At the consumer level, messages are only marked as consumed after the business logic succeeds — if processing fails, the message re-enters the retry queue with exponential backoff, and after max retries, it moves to the dead-letter queue for manual inspection. I'd also implement consumer-side idempotency using a message-ID deduplication table to handle redeliveries safely.