Maintenance windows

A maintenance window tells Stector you're expecting planned downtime, so it holds back down and recovery alerts for the monitors you list and shows the maintenance on the status page you schedule it from. This guide covers how to schedule one and what it does and doesn't hold back.

How it works

Maintenance windows are scheduled by Admin or Owner roles, from a status page's maintenance section. A window belongs to that one page: only that page shows its banner, and only that page's subscribers are emailed about it. A window has two lists, and they do different things:

  • Affected monitors controls alerts. While the window is in progress, Stector holds back the down and recovery alerts for each monitor listed here. A down alert is what opens an automated incident, so a listed monitor that goes down during the window doesn't open one. Stector still runs every check and records the monitor's real status, so its status history is unaffected.
  • Affected components controls this status page. While the window is in progress, each listed component shows Under Maintenance in place of the status its linked monitors would give it, and the maintenance banner names it. Listing a component holds back no alerts, including for the monitors linked to it.

A window that lists only components holds back nothing. To hold back alerts, list the monitors under Affected monitors.

Holding back alerts isn't limited to the window's own page. While the window is in progress, a monitor you list under Affected monitors sends no down or recovery alerts to any of your alert channels and opens no automated incident, whichever page the window belongs to. So none of your status pages shows a new incident for it, including pages that link that monitor and have no window of their own. On those pages, components linked to the monitor keep showing its real status, with no maintenance banner, and their subscribers get no email about it. If an incident was already open for the monitor and the monitor comes back up during the window, that incident isn't resolved automatically on any page (see Incidents during a window below). If two of your pages link the same monitor, a window on one of them holds back that monitor's alerts and incident for both.

Scheduling a window

From a status page's maintenance list at /dashboard/status-pages/[id]/maintenance, click Schedule maintenance and fill in:

  1. Title and an optional Description.
  2. Start and End, in your browser's local time; there's no separate timezone selector. The start can't be in the past, and the end must be after the start.
  3. Affected components and Affected monitors. The first lists this page's components and the second lists every monitor on your team. Each list appears only if it has something to show.
  4. Whether to Notify subscribers by email.

Click Schedule maintenance to save it. A window can't be edited from the dashboard once it's scheduled. To change one, cancel it and schedule a new one.

During and after

A window moves through its own states automatically: Scheduled, then In progress once its start time passes, then Completed once its end time passes. You can cancel a window with the Cancel maintenance button while it's Scheduled or In progress. Once a window is Completed or Cancelled, it can no longer be changed.

The status page the window belongs to shows a banner for it, with its title, description and affected components, and its end time in UTC (its start time too, until it begins). The banner appears up to 14 days before the window starts and stays until it ends or is cancelled.

When an in-progress window ends or is cancelled, its affected components go back to the status their linked monitors give them. A status you had set on one of those components by hand is cleared at the same time. A component can also go back early: if a linked monitor that isn't listed in the window recovers and its incident resolves during the window.

If you ticked Notify subscribers by email, that page's verified subscribers get an email when you schedule the window (with its title, its start and end in UTC, and the names of the affected components), when it starts, when it ends, and if you cancel it. The box is ticked by default.

Incidents during a window

Because alerts are held back in both directions, a listed monitor's incidents behave differently during a window:

  • If the monitor goes down during the window and is still down after it ends, the down alert is sent and an incident opened on the next check after the window ends. The incident's start time is when the monitor went down, not when the alert was sent.
  • 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 the window and its first check passes. Resolve these incidents yourself with the Close incident button or the resolve endpoint.

What is not held back

  • SSL and domain expiry warnings still fire during an active window, including for listed monitors. They're informational and don't open incidents, so the window doesn't apply to them.
  • Alerts for any monitor not listed under Affected monitors, including monitors linked to an affected component.
  • Status for components not listed under Affected components. They keep reflecting their monitors' real state.