Maintenance window

An announced period when planned work may degrade or interrupt a service, with users warned and operators ready to roll back.

planned downtime slotwhen is it safe to restart productionscheduled time for server workthe site may be unavailable tonightwe need an hour of downtime, when can we take itdo the database upgrade after hoursmaintenence windowwarn users before taking the service down

What it is

A maintenance window is an announced start and end time when a service may be degraded or unavailable for planned work. It gives operators room for a database upgrade, network change, or restart, and gives customers a chance to move time-sensitive work away from the risk period.

Reach for one when a change cannot be made safely with a rolling or zero-downtime deployment. State the affected services, expected impact, time zone, update channel, owner, and rollback deadline, then send a completion notice even if the work caused no visible interruption. Recurring windows get names people memorize: Microsoft ships security updates on Patch Tuesday, the second Tuesday of each month, and plenty of corporate maintenance windows are scheduled around it.

Gotcha: the window is permission for bounded risk, not permission to run late. Start rollback early enough to finish before the end, and do not describe a global low-traffic time as convenient when customers in other regions are at their peak.

Ask AI for it

Build a maintenance-window runbook and automation for a PostgreSQL major version upgrade. Create an Atlassian Statuspage scheduled maintenance event with explicit `scheduled_for` and `scheduled_until` timestamps, affected components, expected impact, and UTC plus local-time wording. Add T-minus-24-hour and start notifications, a GitHub Actions workflow_dispatch job protected by the production environment, preflight checks for backups and replication lag, and a hard rollback decision point 30 minutes before the window ends. Update Statuspage when work starts, when it completes, or when it rolls back, and record the actual timings.

You might have meant

zero downtime deploydatabase migrationdeployment gaterunbookhealth check endpoint