Delayed retry with RabbitMQ

Dec 12, 2020

Let's imagine an application with a listener which receives messages from RabbitMQ's queue. At that point application sends a http request to some API but it responds with 500 status. What to do? Well, sometimes it's ok, logging the error is enough. But sometimes application should retry this request later, when API will be available.

So, here I want to show how delayed retry logic can be implemented using a RabbitMQ's features called dead letter exchange and message expiration.

Dead letter exchange is actually an ordinary exchange. It can be of any type and can be defined for any queue. If the queue has a messages expired due configured TTL then this messages will be routed to the dead-letter-exchange.

I believe message expiration does not need explanation. It can be configured on queue level using x-message-ttl parameter or specified on per-message basis using message parameters. RabbitMQ server will try to remove messages at or shortly after their TTL-based expiry.

I'm strongly recommend to read the official RabbitMQ docs about these topics after finishing this article.

Retry scheme

There is a diagram explaining the main idea

Example application can be found at github.com/misc/rabbitmqretry