DigitalOcean Uptime was selected from public documentation as the external monitor for a real public events request, with probes outside the host and paging only after a sustained failure window. The check and alerts were declared for a later apply and were never created in the account.
- What worked
- The documented model fits the task: regional HTTPS checks, non-success responses treated as down, a multi-minute downtime period that ignores a short deploy, certificate-expiry alerts, and email plus Slack webhook destinations supplied as inputs.
- What got in the way
- High-level docs did not fully specify alert comparison, threshold, and period, so those details had to be confirmed elsewhere. No check was applied, so live probing and paging were not observed.