What RabbitMQ Is For

A post office for your services

RabbitMQ is an open-source message broker: applications hand it messages, it stores them safely in queues, and delivers them to consuming applications. This decouples senders from receivers in time (the consumer can be offline), in load (a queue absorbs bursts that workers drain at their own pace) and in knowledge (the sender does not need to know which or how many services consume). Typical uses: background jobs (send email, resize images, generate PDFs), work distribution across many workers, integration between services through commands and events, and request buffering in front of slow systems. RabbitMQ's native protocol is AMQP 0-9-1; it also speaks AMQP 1.0 (a core protocol since RabbitMQ 4.0), MQTT and STOMP through plugins, and has its own stream protocol for streams. It is written in Erlang, known for reliability, and offers flexible routing, acknowledgements, dead-lettering and replicated queues.

Producers, broker and consumers

Producers hand messages to the broker; consumers take them when ready.

Without and with a queue

The web request returns immediately; a worker sends the email later.

# without a broker: the user waits for the slow SMTP call
def sign_up(user):
    save(user)
    send_welcome_email(user)          # 2 seconds, and fails if SMTP is down
    return "ok"

# with RabbitMQ: publish a message and return
def sign_up_queued(user, channel):
    save(user)
    channel.basic_publish(
        exchange="",
        routing_key="emails.welcome",
        body=json.dumps({"userId": user.id, "email": user.email}),
    )
    return "ok"                       # a worker consumes emails.welcome separately

A queue is not a database

Brokers are built to move messages, not to store them for months or query them. Keep queues short in normal operation and keep business records in a database.

Quick check: Which is a typical use of RabbitMQ?

• Distributing background jobs across many workers
• Long-term analytical storage
• Serving static web pages
• Replacing a relational database