Background processing
Overview
Everything that takes longer than a request runs in the background: mail delivery, document generation, imports, syncs with external systems, scheduled reports. The application puts the job on a queue, a worker processes it.
From that follows the most important operating rule: without running workers none of this happens — and without an error message in the interface. The job was accepted, after all.
Operation
Workers run as a service and start by themselves after a reboot. They deliberately terminate after a while and restart — that way they pick up the new code after an update without anyone having to remember.
An update explicitly stops the workers at the end. That is not a fault but the handover point.
Spotting trouble
The health endpoint says whether the system as a whole responds. For the queues there is a separate operations view: it shows how much is waiting and how much has failed.
Two numbers suffice day to day:
- Waiting grows continuously → no worker is running, or too few.
- Failed grows → a job fails repeatedly; the cause is on the entry.
Common problems
Mails do not arrive, without an error message. Check the queue first, then the mail credentials. A silent failure is almost always a worker that is not running.
A document PDF is not produced. Check the queue and whether Chromium is present.
After an update jobs behave as before. Then a worker with old code is still running. Restart the service.