Maintenance windows

A maintenance window tells OctoPing that a site or heartbeat is expected to be unavailable for a while. Checks keep running and are stored, but the downtime isn't an outage, the alerts that would only repeat what you already know are held back and your clients see the planned work on your status page. This page covers how to schedule a window, what changes while it runs and what stays the same.

What a window does

Schedule a window

Open Maintenance in the sidebar and choose to create a window. A Viewer can read the list. A Member, Admin or Owner can create, edit, delete and end windows.

The form lists the next occurrences as you fill it in, so you can check them before you save. A one-time window must end in the future, and a weekly window can't have a Repeat until date that has already passed.

How long a window can be

A Member can schedule up to 24 hours. An Admin or Owner can schedule up to 7 days. The form disables the durations your role can't use. Anyone with write access can rename a longer window an Admin created, but only an Admin or Owner can add more monitors to it.

Daylight saving time

A window follows the clock in its own timezone. If the start time doesn't exist on a day the clocks move forward, the window starts after the gap. If it happens twice because the clocks move back, the window uses the first one. The duration is elapsed time, so a four hour window across a change lasts four hours.

Repeating windows

A weekly window repeats on its days and creates its occurrences about 35 days ahead, adding more as time passes.

Alerts

While a window runs, OctoPing holds back site down alerts and their reminders, heartbeat missed alerts, the checker blocked notice, slow response alerts and HTTP status change alerts for the monitors it covers. Each is logged as skipped with the reason maintenance, so the notification log shows what was held back. Contacts don't receive an alert that your own notifications hold back.

Alerts about things a window doesn't explain still send. An SSL certificate or domain that expires during the window is not planned, and a change to content, DNS records, robots.txt or hosting is worth knowing about.

Around the edges of a window

• An outage you were already alerted about. It stays declared, its reminders stop and its downtime counting stops when the window begins. The recovery alert is sent as soon as the site recovers, even inside the window.
• A failure that starts inside the window. Nothing is declared and nothing is sent. The failed checks are stored but don't count as an outage.
• A site still failing when the window ends. OctoPing queues a check right away. The first failing check after the window starts a normal failure, and the alert follows the usual rules, including the retry threshold and the region confirmation described on Alert rules.
• A heartbeat. No missed alert is sent inside the window. After it, the next ping is due one interval plus the grace period after the later of the last ping and the end of the window, so a job that was paused doesn't alert the moment the window ends.

A window begins for the checks within about a minute of its start time, so a check in the first seconds can still be stored as outside it, and two windows back to back can leave a gap of up to a minute. The end is exact.

Uptime, outages and the monthly summary

• Checks made during a window leave both sides of the uptime calculation. A period that is only maintenance has no uptime figure and doesn't read as 0%.
• A failure during a window isn't an outage, so it adds nothing to downtime or to the outage count.
• The daily bar shows a day with maintenance in blue.
• The monthly summary mentions windows and their total time, and says that time isn't counted as downtime.

The calculations are on Uptime, downtime and outages, and the status on Statuses. A paused or disabled monitor keeps that status in place of Maintenance, because it isn't being checked.

On your status page

A status page shows a Scheduled maintenance section for windows that are running and for those that start within 7 days, up to 50 entries. A window shows only when it covers at least one visible item on that page, and its list of affected components names only items on that page. The name you gave the window is visible to the public. The window's creator and the monitors on other pages never appear.

An affected item reads Under Maintenance and the page banner reads as maintenance while nothing worse is happening. A major outage or a degraded item still ranks above maintenance, and an incident that marks a component down or degraded takes over the status of that component.

Maintenance windows aren't emailed to subscribers or contacts. They appear in the JSON as scheduled_maintenance, described on Status page JSON.

Change, end or delete a window

• Edit a window to change its name, monitors or schedule. The occurrences that haven't started are replaced and finished ones stay as history.
• End now cuts the running occurrence short. It's available only while the window is running, and a weekly window carries on with its next occurrence.
• Delete removes the window and all of its occurrences, including finished ones. The monitors it covered leave maintenance at once, unless another window still covers them.

Finished occurrences are history and are trimmed with the rest of the workspace's activity history, in line with the retention of your plan. The windows themselves are configuration and are never trimmed. A banner on the monitor's page shows an active window and, during the week before, the next one.

Limits

• Emailing subscribers about a window.
• Monthly recurrence, such as the first Monday of each month.
• A calendar view.

Related pages

• Alert rules lists what each alert does and how it is decided.
• Status pages covers the page the window appears on.