Useful logs
A good log says what happened, when and for which request, without exposing personal data. It lets you reconstruct an incident without guessing.
A few key indicators
Response time, error rate, processing queue: a handful of indicators is enough to spot a drift before users do.
Alerts worth reading
Too many alerts and nobody looks at them any more. Every alert should call for an action. That’s what we set up when keeping the product alive and in our back-end & APIs projects.