Plan a change window: task time, rollback reserve, verification and whether it fits before the impact deadline.
Execution time scales down with parallelism but preparation and verification do not, so they dominate small changes. The rollback reserve is subtracted from the end of the window to give the point of no return — the moment after which backing out no longer fits. Changes that overrun their window turn into unplanned outages with no rollback path, which is why the point of no return matters more than the total duration.
Maintenance Window
total = preparation + (devices × time per device ÷ parallel streams) + verification; rollback reserve = half the execution time + verification.
total = preparation + (devices × time per device ÷ parallel streams) + verification; rollback reserve = half the execution time + verification. Execution time scales down with parallelism but preparation and verification do not, so they dominate small changes. The rollback reserve is subtracted from the end of the window to give the point of no return — the moment after which backing out no longer fits.
Changes that overrun their window turn into unplanned outages with no rollback path, which is why the point of no return matters more than the total duration.
This calculator takes 6 inputs: Devices to change, Time per device, Devices worked in parallel, Pre-checks and preparation, Post-change verification, Window length available. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.