Uptime monitoring

A machine somewhere else on the internet requests your URL on a schedule (every 30 or 60 seconds is typical) and records whether it answered, how fast, and with what status. If a few checks in a row fail, it notifies you. The whole point is that it lives outside your infrastructure, so it still works when your servers, your dashboards, and your own alerting are the things that died. That outside-in view is called black-box monitoring. Uptime Robot, Better Stack, Pingdom, and Checkly all sell the same core loop.

Check from several regions and require two of them to fail before alerting, otherwise one flaky network path pages you for nothing. Assert on more than the status code: a login page can return 200 while rendering a stack trace, so match on a string that only appears when the page truly works. For scheduled jobs, flip the direction and use heartbeat monitoring, where your cron pings the monitor and silence is the alert.

Gotcha: a green uptime check is a low bar. It proves one URL answered, not that checkout works, that the queue is draining, or that the site is fast for real users. When you need that, graduate to synthetic monitoring (scripted journeys through signup and purchase) and real user monitoring. Also make sure the alert channel does not depend on the thing being monitored, and expect certificate expiry checks, which prevent one of the most common self-inflicted outages.