A production database migration does not end when backup and restore commands succeed. It ends when the team has evidence that important data and user paths work after the service reopens.
Before the change: define the conditions
Record the data scope, source and destination, owner of each step, downtime window, and other activities that might conflict with the change, such as testing or a release. Agree on entry and cancellation criteria, together with a fallback plan that can be exercised.
A connection string or a proposed date is not proof of readiness. Confirm permissions, connectivity, configuration, and the decision maker for the change window before starting.
During the change: record checkpoints and decisions
Capture the start and finish of each step, migration results, and exceptions. Database, application, and operations teams should know who announces downtime, who authorizes reopening, and who decides to roll back when the window or checks fail.
After reopening: test the real user path
Check connectivity and agreed sample records, then exercise critical workflows through the same entry point users take. A responsive database or green deployment status does not establish that an API and interface can read and write the right data.
Record each result with its time, reviewer, and remaining limits. If results differ, distinguish observations from suspected causes and apply the agreed rollback criteria. The “complete” announcement should follow verification evidence, not just the restore command.