03 / INFRASTRUCTURE
Queuehouse_
Queuehouse wraps RabbitMQ with the operational surface teams actually need: a job registry, per-queue concurrency limits, and a dashboard showing what is running, stuck, or failing.
It exists because background work is usually invisible until it breaks. Making queue depth and retry counts first-class metrics turned a recurring on-call mystery into an ordinary graph.
Highlights
- [ Declarative job registry with per-queue concurrency caps ]
- [ Automatic retries with jitter and poison-message quarantine ]
- [ Prometheus metrics for depth, latency, and failure rate ]
- [ Graceful worker shutdown that drains in-flight jobs ]