RabbitMQ Lab¶

Hands-on companion to the RabbitMQ overview and lab guide. Targets RabbitMQ 4.3.

Quick start¶

make env          # .env with random credentials and Erlang cookie
make up           # single node, waits until healthy
make venv         # Python virtualenv with pika
make creds        # show the generated login for http://localhost:15672

source .venv/bin/activate
cd py
python basics.py consume      # terminal 1
python basics.py produce      # terminal 2

Requirements: Docker with Compose v2, Python 3.10+, openssl, jq (health check only), make.

Layout¶

.
├── compose.yaml              single node (+ --profile monitoring)
├── compose.cluster.yaml      3 nodes behind HAProxy (+ --profile monitoring)
├── config/
│   ├── rabbitmq.conf         shared broker settings  -> conf.d/20-lab.conf
│   ├── cluster/rabbitmq.conf peer discovery          -> conf.d/30-cluster.conf
│   ├── enabled_plugins       management + prometheus
│   ├── haproxy.cfg
│   ├── prometheus/           scrape configs + alert rules
│   └── grafana/provisioning/ Prometheus datasource
├── py/                       one script per exercise, each with --help
└── scripts/
    ├── healthcheck.sh        Nagios-style API health check
    ├── triage.sh             read-only troubleshooting sweep
    └── perftest.sh           PerfTest benchmark scenarios

Script	Lab	Topic
`basics.py`	2.1	topic exchange, quorum queue, confirms, manual acks
`work_queue.py`	2.2	competing consumers, prefetch and fair dispatch
`pubsub.py`	2.3	fanout, exclusive server-named queues
`slow_worker.py`	2.4	long jobs, heartbeats, thread-safe acks, consumer timeouts
`retry.py`	3.1	poison messages, delivery limit, DLX, 4.3 delayed retry
`delayed.py`	3.2	scheduled delivery with TTL tiers
`priority.py`	3.3	quorum vs classic priorities
`rpc.py`	3.4	request/reply with Direct Reply-to
`streams.py`	3.5	streams and offset replay
`confirms.py`	5.2	the cost of publisher confirms
`failover.py`	6.3	node failure, at-least-once delivery, duplicates
`orders.py`	7	event-driven pipeline with compensation

Stacks¶

Only run one stack at a time; both bind the same host ports (all on 127.0.0.1).

	Single (`make up`)	Cluster (`make cluster-up`)
AMQP	localhost:5672	localhost:5672 (HAProxy)
Management	http://localhost:15672	http://localhost:15672 (HAProxy)
Prometheus	http://localhost:9090	http://localhost:9090
Grafana	http://localhost:3000	http://localhost:3000
HAProxy stats		http://localhost:8404

make clean removes both stacks and their volumes.

Not production¶

Credentials and the Erlang cookie live in .env and are passed as environment variables, TLS is off, and the monitoring images float on latest. The overview's production checklist covers what changes for real deployments.