Skip to content

Check: AMQP delivery controls

Knowledge check · self-test

Check: AMQP delivery controls

Check bounded retries, rejection and dead-letter handling.

Back to the chapter →

Check bounded retries, rejection and dead-letter handling.

Checkpoint: Acknowledgment Failure Windows

You now know:

  • Auto-ack removes work before processing finishes, so a consumer crash can turn a delivered message into permanent loss.
  • Manual acknowledgment keeps work visible to the broker until the consumer sends basic_ack, and basic_nack decides whether the message requeues or moves to dead-letter handling.
  • The throughput tradeoff is explicit: auto-ack favors 50K-100K msg/sec paths, while manual acknowledgment fits 20K-50K msg/sec workloads where lost orders, payments, or alerts are unacceptable.

Checkpoint: Durability Is Two-Part

You now know:

  • A durable queue preserves the queue definition, but persistent messages are what preserve the message contents.
  • Transient delivery is appropriate for high-volume telemetry where the chapter accepts brief loss, such as the vibration stream using 100K msg/sec throughput and sub-1 ms latency.
  • Persistent delivery is the safer fit for production orders, where the chapter's broker-restart example makes the loss visible as 50 orders/sec times the order value during the outage.

Pump retry and restart exercise

Send one command identity, lose the acknowledgement, repeat the command, stop the receiver, and restart it. Record queue state, receiver decision, physical result, and final acknowledgement. Treat delivery and external action as separate claims.

Back to AMQP Reliability: Delivery Controls: Choose When to Acknowledge