Instrument a website with actionable alerts

Start with service objectives

Measure uptime, error rate and latency for the pages or endpoints that users actually need. Add queue failures, backup age and storage capacity if relevant.

Three essential signals

• Availability: Is the service reachable and functioning?
• Failures: How often do critical requests error?
• Latency: Are real users experiencing unacceptable delays?

Privacy-safe instrumentation

Keep personal information, credentials and form bodies out of logs and traces. Set retention and access restrictions according to the project.

Alerting policy

Choose a small number of actionable alerts with known owners. Avoid sending alerts that nobody intends to act on. Verify how the system behaves when its monitoring provider is unavailable.

Drill

Simulate a stopped worker or database outage, confirm detection and document the recovery path.