Service status: All checks passing.

Two endpoints, every 10 minutes

A small program on Cloudflare checks the service every 10 minutes, from outside it. Cloudflare was chosen because it is neither the provider that runs the service nor the one that serves this website.

It requests the service's 2 public health endpoints, one for each of the 2 servers customers can reach. The second is the one every customer connection runs through, and it also checks its database.

Each answers in status words and nothing else: no filenames, counts or mailbox addresses. Both are the same addresses anybody can request.

Down, and degraded

Down means the monitor couldn't get a healthy answer: an endpoint didn't answer, or answered with an error. A day is recorded as down only when that lasted at least 10 minutes, which is 2 failed checks in a row. A lone failed check is a blip, not an outage.

Degraded means the service answered but reported a stale self-check, such as last night's backup marker being older than it should be. We're alerted, and nothing customer-facing is off.

The monitor alerts on both and records them apart, because a backup that quietly stopped is exactly the kind of thing a status page is for. The headline gives the same 2 words for right now, and says since when if the service is degraded as you read it.

No service level agreement, and what there is instead

The terms say so plainly. What there is instead:

• this page, drawn from the monitor's own 90-day record with a 60-second cache in front of it, and never from the service;
• an alert to the person who runs the service when the monitor sees consecutive failures;
• the monitor's own page, the fallback if this website is the thing that is down;
• a page for IT and security teams that says where the service runs, what is backed up and how an incident is notified.