Kafka Retry Patterns vs AWS SQS for Shipment Notification Recovery

The article compares Apache Kafka and AWS SQS for recovering failed customer support shipment notifications. It argues that a durable work queue with explicit retry state, bounded attempts, and a dead-letter path is the simplest design that meets operational needs. Kafka suits shipment events that several consumers must replay at their own pace, while SQS suits single-worker tasks acknowledged once finished. The author stresses that neither model makes external notifications exactly-once, so each destination needs a delivery ledger and idempotency keys. The approach trades extra state management for recovery that operators can inspect and act on.