$ open queue_ops_lab

Queue ops lab

A production-shape async-job queue you can poke at. Read the diagram, pick a real-world incident, and the lanes / KPIs / feed will play it out live. The Fastify worker on the other end of the SSE stream is real — same service that powers /sandbox, /telemetry, and /feedback.

idle

$ how it flows

The lifecycle, in one picture

Top row is the happy path. The two loops underneath are what makes a queue interesting — retries with backoff, and the dead-letter queue for jobs that gave up.

  1. Your apporder, charge, signup…
  2. ↓ enqueue
  3. Queuequeued · retrying
  4. ↓ pick
  5. Workerpicks · processes
  6. ↓ deliver
  7. External serviceStripe · webhook · email

  8. ↺ on failure (attempt < max)

    Backoffbase · 2^(n−1) + jitter

    ↑ re-queues into Queue

  9. ✕ attempts exhausted

    Dead-letter queueheld for human review

    ↑ manual replay re-queues into Queue

  • Happy path: producer → queue → worker → external.
  • Retry loop: on failure the worker schedules a retry with exponential backoff and re-queues the job.
  • Dead-letter: after max attempts the job is parked in the DLQ. An operator inspects it, fixes the upstream cause, and clicks replay.

Booting queue lab…