What is Scheduled Maintenance?

What is scheduled maintenance?

Scheduled maintenance is planned work on a service that is announced in advance, carried out in a defined window, and communicated as it happens. It covers database migrations, infrastructure upgrades, certificate rotations, dependency updates and anything else that may degrade or interrupt a service while it is under way.

The defining feature is not the work itself but the announcement. Unannounced work that breaks a service is an outage. The same work, announced ahead of time with a start, an end, and a description of the impact, is maintenance.

Why announcing it matters

An unannounced window costs you twice. Customers open support tickets asking whether something is broken, and your own alerting fires as if the degradation were unplanned. Announcing the window prevents both, and it protects your reliability record: a planned window is not a failure to meet an availability target, but only if it was declared as one beforehand.

Announcements normally go on a status page, where they appear before the work begins, during it, and in the history afterwards. If something goes wrong mid-window, the notice becomes an incident update, as covered in incident communication. Subscribers can be notified by email so they do not have to watch the page.

What a good maintenance notice says

• When - a start and end time, with a time zone. Ambiguous times are the most common failure in maintenance notices.
• What is affected - which components, not which servers. Customers recognise "the API", not "db-replica-3".
• What the impact is - fully unavailable, degraded, or no expected impact. "No expected impact" is a legitimate and useful category.
• What to do - whether anything is required of the customer, and usually nothing is.

One-off and recurring windows

Some maintenance happens once: a migration, a version upgrade. Other work repeats on a rhythm, such as a weekly batch job or a monthly patch cycle. Tools that support recurring windows let you declare the pattern once rather than re-announcing it every time, which makes the schedule predictable for customers who plan around it.

Example

A team plans a database upgrade for Sunday 02:00 to 03:00 UTC. On Thursday the notice goes up: component "API", impact "degraded, writes may fail", nothing required from customers. The window opens, the notice moves to "in progress", and at 02:40 it is marked complete. Support receives no tickets, and the hour shows as planned work, not an outage.

Maintenance and monitoring

During a declared window, monitoring should be handled deliberately. Suppressing alerts entirely risks missing a genuine failure that starts during the window; leaving them fully active guarantees noise. A common approach is to keep checks running and recording, so the data is there afterwards, while marking the period as planned so it reads correctly on the status page and in uptime reporting.

On LoadFocus status pages you can schedule maintenance as a one-off or as a repeating window, daily or weekly, and each occurrence can announce itself a day ahead. Visitors subscribe by email and are told about it. Alerting behaviour during a window is a separate decision, covered in the monitoring best practices guide.

FAQ: Scheduled maintenance

What is the difference between scheduled maintenance and an outage?

The announcement. Planned work that was declared in advance, with a start, an end and a stated impact, is maintenance. Unannounced work that breaks a service is an outage, even if the engineering was identical.

How far ahead should I announce maintenance?

Early enough that customers can plan around it, and again shortly before it starts. There is no universal number: a window with no expected impact needs less notice than one that makes the service unavailable.

Does scheduled maintenance count against uptime?

That depends on your SLA wording. Many agreements exclude declared windows, but only if the contract says so and the window was announced as required. Check your own terms rather than assuming.

Should I pause monitoring during maintenance?

Usually not completely. Keeping checks running preserves the data and catches a failure that starts inside the window. Mark the period as planned so reports and the status page read correctly.

Can maintenance repeat on a schedule?

Yes. Recurring windows suit regular work such as weekly batch jobs or monthly patching, and let you declare the pattern once instead of announcing it every time.