RabbitMQ work queues

Work queues distribute tasks among multiple consumers attached to the same queue. RabbitMQ delivers round-robin by default; prefetch and consumer priority adjust fairness. Use durable queues, persistent messages, and manual ack for tasks that must survive restarts.

Design visibility: expose queue depth, consumer count, processing latency, and failure rates. Bound retries with DLX and cap requeue attempts. For ordered tasks per entity, use consistent-hash exchange or shard into multiple queues instead of one global work queue.

On interviews: describe scaling workers for an image-processing backlog, explain competing consumers versus pub/sub, and list metrics you would alert on.

Common pitfalls: single queue for heterogeneous job durations; no idempotency when workers retry; scaling consumers without broker capacity planning; treating work queues as a log with no retention policy.

The trade-off is flexibility versus complexity—know when the simpler path is enough.

Checklist:

• Use durable queues and persistent messages for tasks.
• Tune prefetch for job duration mix.
• Add DLX and bounded retries.
• Monitor depth, lag, and consumer health.