How this page is measured

Back to status

What is checked

Every 60 seconds each service receives a real request from Cloudflare's network. A check passes only if the response status, its content and its latency all meet the expectations recorded for that service. A page that returns a success code while serving an error is treated as a failure.

When an incident opens

A single failed request is usually internet noise rather than an outage, so a service must fail several consecutive checks before an incident opens, and must pass several consecutive checks before it is marked resolved. Those thresholds are set per service and are visible in the source repository.

How uptime is calculated

Each bar is one day. Uptime for a day is the number of passing checks divided by the number of checks attempted that day. Days before this page began recording show as grey rather than as full uptime, because claiming uptime for a period that was never measured would be misleading.

Automated recovery

Some services can trigger an automated recovery action when they fail. Those actions are rate limited, capped per incident, and every one of them is recorded in the incident history whether it succeeded or not.

Honesty of the record

Check results are append only. Nothing rewrites a past measurement, and the same data drives the page, the JSON API and the feed.