Panic Recovery and Dead Letter Queue in RabbitMQ Consumers

Problem

Without recover(), a panic in a message-processing goroutine crashes it silently. The message is never acknowledged, RabbitMQ redelivers it, and the same panic fires again — infinite loop for poison messages.

Fix: recover() + Dead Letter Queue

go func(msg amqp.Delivery) {
    defer func() {
        if r := recover(); r != nil {
            logger.Error("panic in message processing",
                zap.Any("panic", r),
                zap.String("stack", string(debug.Stack())))
 
            c.sendToDLQ(msg)
            msg.Ack(false)  // remove from main queue
        }
    }()
 
    // process message...
}(msg)

Why DLQ, not Nack?

Action	Result
Nack(requeue=false)	Message lost forever
Nack(requeue=true)	Infinite loop for poison messages
Send to DLQ + Ack	Quarantined for debugging, main queue unblocked

DLQ lets you inspect the failing message without blocking healthy ones.

See also

• parallel-polling-per-entity — publisher side: how messages are produced
• context-independent-cleanup — cleanup after panic must not use a cancelled context