DLQ and Retry - Professional¶

Kafka retry topics trade partition order for independent delay; RabbitMQ dead-letter exchanges and TTL queues provide routing-based delay; SQS exposes visibility timeouts and redrive policies. At scale, amplification and DLQ retention dominate. Dashboard attempts/original, retry age, DLQ bytes, replay rate, and downstream saturation.

Best practices¶

• Use stable event identity across every hop.
• Bound attempts, elapsed time, and retained bytes.
• Separate inspect, repair, replay, and audit permissions.
• Capacity-test dependency outage plus recovery surge.

  retry = transient + budget + backoff
  DLQ = durable quarantine + owner + replay control
  

Test yourself¶

1. How would you preserve order for one financial account?
2. What prevents replay from causing a second outage?
3. Which SLO makes DLQ neglect visible?

Further reading¶

• Kafka retry and delivery semantics documentation.
• AWS SQS dead-letter queue documentation.
• Enterprise Integration Patterns, Dead Letter Channel.