RabbitMQ

RabbitMQ has been reliably passing messages between services since 2007. It sits between your services and makes sure messages get delivered, even if the receiving service is temporarily down.

It's a post office for your backend. Service A drops a message in a queue, Service B picks it up when it's ready. RabbitMQ handles routing, retries, and ordering so you don't have to. It supports multiple messaging patterns: simple queues, pub/sub (one message to many consumers), routing by topic, and RPC. The management UI is good: you can see queue depths, message rates, and consumer health at a glance.

Self-hosted is free under the Mozilla Public License. CloudAMQP offers managed hosting starting at $0 (free tier: 1M messages/month, limited connections) up to $500+/mo for production clusters. AWS, GCP, and Azure all offer managed RabbitMQ too.

The catch: RabbitMQ uses Erlang, which makes it incredibly reliable but painful to debug when something goes wrong at the system level. For simple job queues, it's overkill; BullMQ or Sidekiq are lighter. For high-throughput event streaming (millions of events/sec), Kafka or Redpanda are better suited.