All systems operational

How it's measured. What it can't show.

Every five minutes, a check inside the deployment queries the database every request depends on. If the service is up but cannot reach Postgres, that counts as a failure.

The check runs inside the deployment, so a total outage stops the recording. It shows as a gap, and this page says unconfirmed, not green. To check independently, poll /api/health. It returns 503 when the service cannot serve.