Vantage points and confirmation

A site is observed from more than one place. When one place says the site is down and the others say it's up, the disagreement is information, and OctoPing uses it to decide whether to alert. This page describes the vantage points, what each one is for and how a failed check is confirmed.

The vantage points

All of them send the same request, so a difference between them describes the site and not the monitor. The primary checker decides when a site goes down and when it recovers. The probe regions provide the second opinion.

• Each probe region reports about every 5 minutes. A result counts as current for 20 minutes, and a region whose newest result is older is shown as stale (no recent data), which is different from up or down.
• A region that has stopped reporting is left out.
• The primary checker has its own freshness allowance, which scales with the site's check interval, so a site on a slow interval isn't marked stale between checks.

What happens when the primary check fails

When the primary check fails, OctoPing looks at what the probe regions reported since the failure began. Reports from before the failure, and reports made during a maintenance window, can't confirm or contradict it. There are three verdicts.

"The probes have nothing to say" covers two cases: no probe has reported for the site within 24 hours, and probe silence that has outlasted the 10-minute wait. In both of them the failure is confirmed.

• A declared outage is never reversed by later healthy probes. Recovery is decided by the primary checker.
• A contradicted run becomes a real outage the moment any probe region reports a failure.
• When only some regions are failing, the down alert names them, for example that the site is down from Europe, so a regional routing problem can be told apart from a global outage.

Partial: only the primary checker is refused

Partial means the probe regions reach the site and OctoPing's primary checker doesn't. A firewall or bot rule that blocks the checker's address is the usual cause. Because visitors can reach the site, Partial isn't counted as downtime, and a public status page shows the site as Operational.

OctoPing sends one informational notice per run of failures, instead of a down alert, once the notification's failure threshold is met. It names the regions that see the site up and the status the checker received, and it lists the addresses to allow when they are known. There's no recovery alert, because no down alert was sent.

There are two ways to clear it. Allow the addresses named in the notice on the firewall, or add a Custom HTTP Header under the site's Advanced request settings and write a firewall rule that allows requests carrying that header. The header is sent by the primary checker and the probe regions.

Reading disagreement

Each vantage point's latest result and response time are on the site's detail page. For how failures are worded, see the failure reasons page. For how these decisions affect the numbers, see uptime and outages.