Observability
Uptime and the status page
Check your sites over HTTP from the panel, get alerts when they're down, and show them on a public status page (the Uptime page).
Monitors
A monitor has a name, an http(s) URL, how often it's checked (30 seconds to an hour; 60 by default) and the timeout (less than that; 10 by default), optionally the app it's about, whether it's on the status page, and whether it's checked.
API: GET/POST /api/monitors, PUT/DELETE /api/monitors/{id}.
A check is a GET from the panel, as a visitor would make it: redirects
followed (5 at most), up when it answers below 400 within the timeout.
Checks are kept 90 days. The list gives each monitor's last check, its
uptime over 24 hours and 7 days, and its average latency;
GET /api/monitors/{id}/checks gives the checks of the last hours (24 by
default, up to a week).
Private addresses
Checks come from the panel's servers: a monitor on a URL only reachable inside the private network shows what the panel sees, not visitors.
Only a monitor an admin saved (last) may check private addresses: loopback, private and shared ranges (the mesh, a LAN, CGNAT), link-local (cloud metadata), IPv6 unique and link local. Anyone else's is refused them, whether written in the URL, resolved from its name, or redirected to, so members can't probe what only the panel reaches.
Alerts
A monitor whose last two checks failed fires monitor_down (critical,
see Alerts), resolved once it answers again.
The status page
Admins turn it on and title it (GET/PUT /api/status/settings). It's
then public at /status, with no sign-in (GET /api/status, 404 while
off): each public monitor's name, whether it's up now, its uptime over 90
days and a bar per day (green at 99.9%, amber at 95%, red below). Its URL
and errors aren't shown.