Free guide
Planning an infrastructure migration: a step-by-step checklist
Most migration problems come from something nobody listed: a scheduled task, a hard-coded IP, a certificate that expires on the new server. The steps below are about finding those before the cutover.
Updated 2026-10-19
1. Inventory
- List every server, application and data set in scope, with its size, its owner and who uses it.
- Note licences that are tied to hardware, to a host or to a region.
2. Dependencies
- Find what talks to what: databases, file shares, printers, APIs, DNS names and anything with a hard-coded IP address.
- Look for the hidden parts: scheduled tasks, service accounts, certificates, firewall rules and integrations set up years ago.
3. Target and approach
- Choose, for each workload, whether it moves as it is, gets updated on the way, or is replaced (for example a file server by SharePoint).
- Size the target from measured use, not from the old hardware’s specs.
4. Downtime and success
- Agree on the downtime window and how much data change is acceptable to lose, if any.
- Write down what “done” means: the checks that must pass, and who signs them off.
5. Pilot
- Move something small and low-risk first, with real users, and time each step. The pilot is where the runbook gets corrected.
6. Rollback plan
- Decide how you go back and, just as important, the time by which you decide. A rollback that starts too late can’t finish inside the window.
- Keep the source untouched until the target is validated.
7. Before the cutover
- Lower DNS TTLs a few days before, so the change takes effect quickly.
- Take a fresh, tested backup and freeze changes on the source.
- Tell users what will happen, when, and who to contact.
8. Cutover
- Follow a written runbook: each step, its owner and its expected time. Sync the last changes, switch DNS or connections, and keep notes as you go.
9. Validate
- Have the application owners run their checks, not only IT. Watch logs, monitoring and helpdesk tickets for the first days.
10. After
- Keep the old environment read-only for an agreed period, then decommission it and wipe the data.
- Update the documentation and, in the cloud, review the first month’s costs against the estimate.
Want a second pair of eyes?
Tell me what you have and what worries you. I reply by e-mail with questions or a proposal.