Managing incidents

An incident is Stector's record of a monitor's downtime, opened automatically when a monitor goes down (see Monitoring HTTP endpoints for when that happens) and closed automatically when it recovers, except in the cases described below. This guide covers the incident lifecycle and what you can do manually.

How it works

An incident moves through up to four states: Investigating, Identified, Monitoring, and Resolved. Stector creates one automatically when a monitor goes down, starting in Investigating with a system-posted update: "Monitor went down, incident auto-created by system." It resolves automatically when the monitor comes back up, which Stector is set up to decide on one successful run from enough of its regions: one region for a one-region monitor, both for a two-region monitor, and any two for a three-region monitor, even if the third is still failing. Going down takes two failing runs in a row in each region that counts, so recovery is detected faster than failure. A new or just-resumed monitor is the exception: its first check decides, so one failed run marks it down and one passing run marks it up.

While a maintenance window that lists a monitor under Affected monitors is in progress, Stector holds back that monitor's down and recovery alerts. As a result, if an incident opened before the window began and the monitor comes back up during the window, the incident isn't resolved automatically on any page, then or after the window ends. The same applies if you resume a monitor during such a window and its first check passes. Resolve these incidents yourself with the Close incident button (see Closing an incident below) or the resolve endpoint.

An incident on a paused monitor doesn't resolve automatically, because the monitor isn't being checked. Until you resume it, the incidents list shows it as Monitoring paused. After you resume, the first check decides: if it passes, the incident resolves, unless a maintenance window covers the monitor as described above; if it fails, the monitor stays down and the incident resolves when it recovers under the rule above.

Posting updates

From an incident's page at /dashboard/incidents/[id], anyone with the Member, Admin, or Owner role can post an update: a status change plus a message. The status dropdown offers Investigating, Identified, or Monitoring — Resolved isn't one of the options here. To close an incident, use the Close incident button instead.

Closing an incident

Closing an incident is restricted to Admin and Owner roles. This is deliberate: closing is a highly visible action that affects your status page, and keeping it separate from the update form reduces the chance of an incident being marked resolved prematurely during an active outage.

Incidents on your status page

Which incidents a public status page shows depends on how they were created:

  • Automated incidents, the ones Stector opens when a monitor goes down, appear on every status page that has a component linked to that monitor, and on no other page. While a maintenance window on any of your pages lists that monitor under Affected monitors, no automated incident opens for it, so none of those pages shows one (see Maintenance windows). An incident on a monitor that isn't linked to any component stays in your dashboard and isn't shown publicly.
  • Incidents you create by hand appear only on the status page you post them to, whether or not they're tied to a monitor. In the dashboard you choose the page once your team has more than one public page. Through the API and the MCP server you can read only your team's oldest status page, and incidents you create through them are posted to that page; to post an incident to another page, create it in the dashboard. The oldest page is the first one in your Status Pages list. An incident can't be moved to another page after it's created. If your team has no public status page when you create one, it isn't shown anywhere, and it doesn't start showing later when you create a page. If you delete a page, the incidents posted to it stay in your dashboard but aren't shown on any page.

Open incidents show on the status page itself, and resolved ones stay in the incident history for 90 days after they close. Each page's subscribers follow the same rule: they're emailed only about incidents that page shows.

There's no per-incident visibility toggle. To keep an automated incident off a page, leave its monitor unlinked from that page's components. Take care when changing links during an outage: if you unlink a monitor while its incident is open, the incident disappears from that page and its subscribers won't get any further emails about it, including the resolution.