Maintenance Preview simulates a planned maintenance window against the live environment. Shut down service X, drain pool Y, restart database Z, what would break? How many customers? Which downstream services? The preview answers before you start, so the maintenance is planned with eyes open.
Maintenance Preview uses the same digital-twin engine as Simulation Engine. The twin captures the live environment's state. The preview applies the maintenance plan step-by-step against the twin and predicts the SLI movement, customer impact, and downtime distribution. Production is never touched during preview.
For every preview, the page estimates customer impact: how many users will see degraded service, in which regions, and for how long. Estimate uses traffic patterns over the planned window. Maintenance during low-traffic windows shows lower estimates; mid-day estimates show the cost of bad timing.
The preview includes a window recommendation: the best 2-hour slot in the next 7 days based on traffic patterns and SLO budget remaining. Recommendations weight low-traffic windows, avoid recent incidents, and respect known holidays. Override always available; the recommendation is a starting point, not a directive.
For maintenance windows that cross customer-facing thresholds, the preview produces a draft status-page entry. Pre-filled with the planned window, the affected services and regions, and a calibrated tone. Edit, schedule, done. The same channel publishes the live status during the actual window if anything goes long.
Subscribe to Nova AI Ops on YouTube for demos, tutorials, and feature deep-dives.
A 5-minute preview saves a 5-hour postmortem. Maintenance Preview is the cheapest insurance for planned changes.