Cloud Migration Cutover Plan: What a Business Should Approve Before Moving
Gavior Editorial Team
Engineering and cloud editorial team
Published: 2026-10-15Updated: 2026-10-15
Plan a production cloud migration with a dependency inventory, tested recovery, data cutover rules, rollback decisions and evidence that customer workflows still work.
Approve a production migration only when the team has named the affected systems, rehearsed the move, verified recovery, defined how data stays consistent and agreed who can stop or reverse the change. A new server responding to a health check is not enough to prove that the business has migrated successfully.
Make your move
Ready to turn a better idea into a better business?
This is a planning guide for a business commissioning cloud or infrastructure work. It is not a command sequence for a particular database or hosting provider. Your engineering team must adapt the procedure to the application and the consequences of interruption.
Inventory the full service
Ask for a dependency map that includes more than the main website. Record databases, uploaded files, background workers, scheduled jobs, queues, email, payment callbacks, DNS records, certificates and third-party allowlists. Include internal access paths used by staff. Name the person who can provide access to each account without putting passwords in the migration document.
A hypothetical booking business might have a public site, an admin portal, a database, reminder jobs and a payment provider webhook. Moving only the public site could make the homepage look healthy while reminders stop or payment updates go to the old environment. This is an example to explain dependency risk, not a Gavior client outcome.
Agree the business limits
State the acceptable interruption window and what data loss, if any, the business can tolerate. These decisions affect the migration method. Do not let a technical team silently assume that restoring yesterday's backup is acceptable for a system taking today's orders.
AWS's pre-cutover guidance recommends planning dependencies, testing and contingency steps, including a rollback procedure. Use a rehearsal to measure whether the proposed window is realistic. Allow for validation and recovery time as well as copying data.
Ask for this cutover workbook
Section
Evidence to request
Scope and owners
Systems moving, systems staying and named responsibilities
Readiness
Target configuration, access, certificates and dependency checks
Recovery
Backup location and a recorded restore exercise
Data
Source of truth, final synchronisation and write-control rules
Execution
Ordered steps, checkpoints and expected duration
Decision points
Stop criteria, rollback owner and escalation route
Validation
Tests for the real customer and staff journeys
Keep the workbook usable during an incident. Each step should have an owner and a visible completion condition. Record configuration references securely, and keep secret values in an appropriate credential store. The business approver should be able to understand what the checkpoints mean without operating the infrastructure personally.
Resolve the data question before changing traffic
If both old and new systems can accept writes, the team needs a deliberate consistency strategy. Decide whether the source will be temporarily read-only, how the final data transfer happens and when the new environment becomes authoritative. Different workloads may need different approaches.
AWS's cutover-stage guidance distinguishes rollback before new data is written from rollback after the target has received transactions. Repointing traffic to an old database may lose new changes. Ask how those changes would be preserved, reconciled or restored before accepting “we can just switch back” as the recovery plan.
Test business behaviour, not only availability
Use approved test accounts and data to verify sign-in, record creation, file uploads, notifications and integrations relevant to the service. Confirm scheduled jobs and workers are actually processing work. Check that links and callbacks refer to the correct hostname and that important access restrictions still hold.
Compare counts or representative records using an agreed method. If the team declares a mismatch harmless, record the explanation. Monitor errors and queues during the initial observation period. Avoid deleting the old environment merely because the first smoke test passes; agree a retention period and the conditions for decommissioning it.
Close the migration with an operating handover
The handover should include monitoring, support contacts, renewal costs, deployment instructions and backup ownership. Identify which temporary access will be revoked and which legacy resources can be removed after approval. Update the dependency map so the next engineer does not rediscover the same system from scratch.
When commissioning cloud solutions or DevOps engineering, ask for this evidence as part of the scope. Start a migration brief with the current environment, applications affected, access owners and business limits. A concrete brief is more useful than requesting “zero downtime” without defining what must remain available.
G
Gavior Editorial Team
Engineering and cloud editorial team at Gavior
The Gavior Editorial Team shares practical guidance on product design, software engineering, cloud delivery and AI automation.