Checking

This page keeps a record. Once a minute the control plane writes down how each part of the service is doing, in the same words this page shows, and keeps ninety days of those minutes. Everything above is computed from them: the uptime percentages, the day by day cells, the last day at five-minute resolution, and the incidents. Nothing is typed in by hand except a note on an incident, and a note is marked as one.

What the words mean. Normal is every part answering as it should. Degraded is one part unwell while the rest still works. Unreachable is a minute the control plane could not write its record at all, which is what an outage of the control plane itself looks like from the inside, so it is counted against uptime rather than left out. An incident opens when a part has been degraded for two minutes in a row and closes after two normal minutes; a gap of three or more minutes in the record is an incident too.

If this page says Down, it could not reach the control plane just now, and it says what happened: no response, a timeout, or the HTTP status it got back. It retries once before saying so, checks again every minute, and shows the last reading it managed to fetch, marked with its age, because an old answer labelled old is better than none. Down can also mean the problem is between you and us: a network that blocks the control plane's host, or a browser extension that blocks the request, gives the same verdict. The raw data is at /status as JSON. For the story of an incident that affected you, write to support@kept.solutions and a person will answer.