Cloud migration cutovers
The migration is not finished when the window closes
Support needs the final system state, temporary exceptions and retirement conditions. Hand over the operating service, including what still remains in the old environment.
In this article
Record where work now happens
State which environment owns each migrated capability and dataset. Include queues, scheduled jobs and integrations that remain elsewhere.
Link the actual deployed configuration and the cutover record. A planned architecture diagram can differ from the final arrangement after an issue was resolved during the window.
Explain any temporary routing or compatibility path. The next on-call engineer should not discover it by following an unexpected request into the old system.
Transfer the unresolved work
List discrepancies, delayed jobs and temporary controls with owners and next actions. Separate accepted limitations from defects still awaiting investigation.
For held work, provide the queue or operation identities and the policy for release. A generic note that some messages remain does not establish who can safely resume them.
Include business contacts for data questions. Technical teams may be able to identify a mismatch without having authority to choose the correct business value.
Define the observation period
Specify who watches the service, which workflows are checked and how incidents escalate. Include scheduled processes that run after the migration team has left the window.
Keep recovery capability available according to the agreed policy. Document whether rollback remains possible and how new target writes affect that decision.
Update support instructions and monitoring links so the receiving team uses the target environment's evidence. Old dashboards can remain reassuringly green because they no longer receive meaningful traffic.
Close the old environment with evidence
Name the owner of retirement and the conditions that permit it. Confirm data retention, backup access, dependency removal and completion of the observation period.
Remove temporary credentials, replication tasks and unused resources through the normal change process. Preserve the records needed to explain the migration without retaining unnecessary sensitive copies.
Ask the receiving team to trace one current transaction and locate the relevant recovery procedure. The handover is successful when they can operate the resulting service confidently, including its remaining transitional parts.
Primary sources
AWS: performing cutover tasksReferences checked 11 September 2026.