On this page

01 · Watch the wait
02 · Split by urgency
03 · Safe to run twice
04 · Alert on silence
The checklist

Every Laravel app we ship runs queued work: emails, webhooks, exports, image processing. Queues are also where failures go to hide — a job that never runs doesn't throw an exception in front of anyone. This is the checklist we put in place before launch.

The failed_jobs table tells you what broke. It doesn't tell you a queue is quietly backing up. We alert on wait time per queue, so someone hears about it before a customer does.

One queue means a ten-minute export can block a password-reset email. We split work into at least critical, default and long — each with its own workers, timeouts and retry policy.

Workers die, deploys restart them, retries happen. Jobs have to be idempotent: check state before acting, give them a unique ID, and back off between attempts instead of hammering a failing API.

A scheduled command that stops running produces no errors at all. Every scheduled task pings a heartbeat monitor when it finishes, and we alert when the ping doesn't arrive.

+ Wait-time alerts per queue, not just failure alerts
+ Separate critical, default and long-running queues
+ Idempotent jobs with unique IDs and backoff
+ Heartbeat pings on every scheduled command

#laravel #horizon #monitoring

[Author name] — [Role], xDevLabs. One-line author bio.