Skip to content

Maintenance windows

A maintenance window covers a time range during which you don’t want alerts firing on the affected monitors. Use it for deploys, planned DB migrations, certificate rotations — anything where you’re expecting checks to fail and don’t want to wake anyone up.

Create windows in the WebUI at Maintenance → New maintenance window.

Configuration

FieldTypeDefaultNotes
NamestringFree-text label shown on the maintenance list and on linked status pages.
MonitorsUUID arraySpecific monitor IDs the window covers. Leave empty to cover every monitor in the org for the duration.
Starts atdatetime (UTC)ISO 8601 timestamp. The UI presents a local-time picker and converts to UTC on save.
Ends atdatetime (UTC)Must be after Starts at.
Suppress alertsbooltrueDon’t send notifications for affected monitors during the window. The check still runs and the result is still recorded.
Pause monitorsboolfalseDon’t even run the check during the window. Useful when the target endpoint will be torn down completely.
RecurrenceobjectOptional. See below.

Recurrence

Recurring maintenance covers repeated patterns — “every Sunday 03:00–05:00 UTC”, “first of the month 00:00–01:00”.

FieldNotes
Typedaily, weekly, or monthly.
IntervalEvery N days/weeks/months.
Days of weekFor weekly — array of 0 (Sun) through 6 (Sat).
Day of monthFor monthly — integer day.
End afterStop the recurrence after N occurrences.
End dateStop the recurrence at this date.

A recurring window evaluates the Starts/Ends times against the current recurrence instance — so a weekly window with Starts at 03:00 and Ends at 05:00 is a 2-hour suppression at the specified times every matching day.

Effect on status pages

Status pages that include a maintenance component (see Status pages) automatically show active and upcoming maintenance windows. Toggle the Show maintenance flag on the status page settings if you want that view; it’s on by default.

Tips

  • Prefer Suppress alerts over Pause monitors — recording the failure history during a planned event is useful retroactively.
  • Always set a maintenance window before a planned deploy, even if you think it’ll be fast. The window auto-expires; a forgotten page from a 90-second blip is the more annoying failure mode.
  • Recurring weekly windows for batch jobs (DB backups, log rolls) are cleaner than ack’ing the resulting alert every week.