Is the site up?¶
Besides counting visits, the app can check that a site answers: every 5 minutes, the app requests an address of the site, and records whether it answered, how fast, and until when its TLS certificate is valid. It's optional, site by site.
Monitoring a site¶
The Monitoring page of the project's settings lists its sites. Start monitoring checks a site at the address next to it, its home page unless you change it: give another address of the same site if the home page is slow or redirects, such as a health page. The address must be on the site's own hostname. The page then shows whether the site is up, the share of checks it answered over 30 days, its response time and its certificate's end. Stop monitoring ends it; the incidents stay. A site stays at "Waiting for the first check" until the next round of checks, within 5 minutes; if no check comes, the page says so.
A check fails when the site doesn't answer within 10 seconds, or answers with an error (a status of 400 or more). Redirects are followed.
Incidents¶
Two failed checks in a row open an incident, dated from the first; the first success closes it. A short hiccup thus raises nothing. While a site is down:
- the bell counts it, and the Needs attention page lists it;
- the project's report says so above its figures;
- the project's alerts, by webhook or email if it has them, say when the site goes down. Another follows when it comes back, with how long it lasted (alerts).
A certificate that expires within 14 days shows on the bell too.
On the report¶
The report's Uptime card lists the monitored sites: whether each is up now, the share of checks it answered over the last 30 days, its average response time, and its certificate's end. The chart marks each incident of the period, with how long it lasted. Checks are kept 30 days, and incidents for good.
What a monitor can't see¶
The checks run on the server that runs the app. If that server goes down, or loses its network, nothing checks the sites, and no alert leaves: a monitor can't watch itself. When the sites and the app share a server, the sites' outages that matter most go unseen.
- Run the app on another server than the sites it watches, if you can.
- Or watch the app itself from outside, with a "dead man's switch" service that alerts you when it stops hearing from it, or an outside check of its
/healthpage.
The checks come from one place: they can't tell a site that's down from a network fault between it and the server. Two failures in a row filter out the shortest faults.