RabbitMQ — Reliable Open-Source Message Broker

What RabbitMQ Does

• AMQP 0.9.1 — primary protocol, plus AMQP 1.0
• Exchanges — direct, fanout, topic, headers
• Queues — classic, quorum (Raft-based), streams (append-only log)
• Bindings — flexible routing rules
• Persistence — durable exchanges/queues + persistent messages
• Clustering — high availability across nodes
• Mirrored/Quorum queues — HA with replication
• Plugins — MQTT, STOMP, Federation, Shovel, Web STOMP
• Management UI — http://host:15672
• RabbitMQ Streams — append-only log (Kafka-like)

Architecture

Erlang processes are cheap and fault-tolerant. A RabbitMQ node runs thousands of Erlang processes per queue/connection. Cluster nodes share metadata via Mnesia DB (or Khepri in newer versions). Messages route through exchanges by binding rules.

Self-Hosting

# docker-compose.yml with clustering
version: "3"
services:
  rabbit1:
    image: rabbitmq:3-management
    hostname: rabbit1
    environment:
      RABBITMQ_ERLANG_COOKIE: "secretcookie"
  rabbit2:
    image: rabbitmq:3-management
    hostname: rabbit2
    depends_on: [rabbit1]
    environment:
      RABBITMQ_ERLANG_COOKIE: "secretcookie"

Key Features

• AMQP, MQTT, STOMP protocols
• Multiple exchange types
• Quorum queues (Raft)
• Streams (Kafka-like append log)
• Federation across clusters
• Shovel for cross-cluster forwarding
• Management UI and HTTP API
• Plugins ecosystem
• TLS and SASL auth
• Mature client libraries for every language

Comparison

FAQ

Q: RabbitMQ vs Kafka? A: RabbitMQ is a traditional MQ (flexible routing, task queues, RPC, low-latency async); Kafka is a distributed log (event streaming, long retention, high throughput). The choice depends on whether you need queue semantics or stream semantics.

Q: Classic vs Quorum queue? A: Quorum is the new Raft-based HA queue type in v3.8+ and is recommended for production. Classic mirrored queues are deprecated after v3.12.

Q: How large does it scale? A: A single node supports tens of thousands of connections and tens of thousands of messages per second. Clusters scale higher. For very high throughput (millions/sec), Kafka is recommended.

Sources

• Docs: https://www.rabbitmq.com/documentation.html
• GitHub: https://github.com/rabbitmq/rabbitmq-server
• License: MPL 2.0