One of the most overlooked issues when working with BullMQ (or any job queue, really) is quietly building backlogs. It doesn’t crash the system – but it slows things down until something breaks.
After a few rough incidents (hello weekend alerts 😅), we started tracking how many jobs are waiting, prioritized, or delayed per queue. If the number crosses a certain threshold, we get alerted and scale up.
🔧 Pro tip: If you’re not monitoring queue backlog yet, set a soft threshold based on your infra. Even just 100 pending jobs on a medium-sized queue might be enough to indicate a bottleneck.
We ended up building a dashboard just for this – eventually turned it into Upqueue.io. Might help if you’re looking for something more lightweight than full observability platforms. No config hell, just watch your queues.
Would love to hear how others handle backlog monitoring. Manual scripts? Prometheus??
