Dead Letter Queues: Handling Poison Messages in Distributed Systems

Short summary

A dead-letter queue catches messages consumers cannot process, turning invisible failures into inspectable, alertable events. The article explains DLQ mechanics across SQS, RabbitMQ, Kafka, and Azure Service Bus, distinguishing poison messages from transient failures. It covers retry thresholds, replay strategies, and broker-specific implementation details for each platform.

• •DLQ is an ordinary queue that receives messages after repeated processing failures
• •Each broker implements redirect differently: SQS redrive policy, RabbitMQ x-death headers, Kafka consumer-side DLT pattern
• •Poison messages (malformed payloads, schema changes) must be quarantined and replayed only after root-cause fixes

Generated with AI, which can make mistakes.