According to Microsoft’s lifecycle page, extended support for Windows Server 2016 ends on 12 January 2027. After that date, no standard security updates will be released for this version. The timeline may look long; it isn’t. Getting a support statement from the application vendor, building a test environment, finding maintenance windows and refreshing DCs one by one quickly use up those few months. The plan below was prepared to keep the migration under control from inventory to the rollback step.
1. Inventory: which server, which role?
List all physical and virtual servers with their versions. A quick start for domain-joined servers: Get-ADComputer -Filter 'OperatingSystem -like "*2016*"' -Properties OperatingSystem,LastLogonDate. This list will be incomplete; workgroup servers, DMZ machines and old virtual machines nobody remembers must also be pulled from the hypervisor inventory. Surprises usually turn up here.
Collect the following for each server:
- Its roles (AD DS, DNS, DHCP, file, print, RDS, IIS, Hyper-V)
- Business applications and databases running on it
- The application owner and vendor
- Dependencies: which systems connect to this server?
- Hardware age and warranty status (if not virtual)
Don’t leave the dependencies column empty. A file server’s name may be hard-coded in an ERP report written ten years ago or in a printer’s “scan to folder” setting. A practical way to find these is to record the active SMB sessions (Get-SmbSession) and network connections on the server for a few weeks before the migration.
2. Application compatibility
The most time-consuming part of the migration is usually not the operating system but the applications on it. Get a support statement from vendors for the target version (Windows Server 2022 or 2025), and plan an upgrade to the current application version if needed. For unsupported legacy applications, consider alternatives or isolation measures.
Ask for a written support statement. A verbal “it works” can turn into the vendor saying “we don’t support that version” at the first error. At this stage, also ask about the database version the application uses, the .NET Framework version and whether the license key is tied to hardware or a MAC address. A hardware-locked license won’t open the application on the new server.
3. Choosing the target and method
A single method is chosen for each server, based on the server’s role, how portable the application is and the need for rollback:
| Method | When it fits | Watch out for |
|---|---|---|
| Clean install + moving roles/applications | Domain controllers, file and print servers, systems where a clean start is wanted | Parallel running period and resource needs |
| In-place upgrade | Application servers that are hard to move or have complex configurations | A supported upgrade path, old settings carried over, always a backup and a snapshot |
| Moving to the cloud | Suitable workloads, systems that need capacity flexibility | Cost model, connectivity and data residency requirements |
Before deciding on an in-place upgrade, check Microsoft’s current upgrade matrix; it shows whether there is a direct upgrade path from 2016 to the target version and which edition changes (such as Standard to Datacenter) are supported. Take a snapshot of the virtual machine before upgrading, but a snapshot does not replace a backup. Even if the upgrade succeeds, don’t leave the snapshot open for days; the disk chain grows and performance drops.
For file servers, Storage Migration Service can move shares, permissions and even the server identity (name and IP) to the new server. This keeps mapped drives on the user side unchanged.
4. Domain controllers
Domain controllers should be refreshed by adding new servers rather than upgrading in place: new controllers are installed, replication and DNS are verified, FSMO roles are transferred and the old controllers are removed cleanly. Before starting, run an Active Directory health check and verify that SYSVOL replication uses DFSR.
This is where common mistakes appear:
- The old DC is shut down but metadata cleanup isn’t done; the dead server stays in the replication topology and DNS.
- The old DC’s IP address is set as a DNS server in the DHCP scope, on servers with static IPs and on printers; when the DC is shut down, name resolution breaks.
- The PDC emulator is moved but the external time source isn’t configured on the new server.
- The functional level and SYSVOL conditions the target version expects from the forest aren’t checked beforehand; the new DC gets stuck in the promotion wizard.
Giving the old DC’s IP to the new DC reduces the work of finding hard-coded DNS settings one by one. With this method, the IP should be taken over only after the old server has been completely removed. The same IP must never be on two machines at once.
5. Licensing
Server core licenses and client access licenses (CALs) should be reviewed for the new version. Windows Server is licensed per core; under Microsoft’s rules, at least 16 cores are licensed per server and at least 8 per processor. In virtualized environments, the difference between Standard and Datacenter editions directly affects cost depending on the number of virtual machines: when all cores are licensed, Standard grants rights to two Windows Server virtual machines, while Datacenter grants unlimited virtual machines.
Check whether existing agreements include upgrade rights (such as Software Assurance). When the version goes up, CALs must also be for the same or a newer version; connecting to a new server with old CALs is a licensing non-compliance. In environments using RDS, RDS CALs are calculated separately.
6. Testing, migration and rollback
The migration plan should be one page per server: prerequisites, steps, verification and rollback. A rollback step that isn’t written down turns into improvisation in a midnight maintenance window.
- Migration and rollback steps should be written down for each server
- A verified backup should be taken before the migration
- Critical applications should be tried in a test environment or with pilot users
- The migration should be done in maintenance windows that fit the business calendar
- After the migration, check that monitoring, backup and security agents are running on the new servers
The last item is the most often skipped. The new server is built, the application works, everyone relaxes; three weeks later someone notices the backup software is still trying to back up the old server name. Before declaring the migration “done,” you should see the first successful backup job and a green status in the monitoring system. The old server isn’t deleted immediately either: it is shut down, kept for a while and then removed.
7. Servers that won’t be finished in time
Some applications may not make the deadline. For these servers, record the risks, restrict network access to the necessary systems only and assess Microsoft’s Extended Security Update options with their current terms. This should be used for a controlled interim period, not to postpone the migration.
Practical isolation steps: moving the server to a separate VLAN, opening only the ports the application needs on the firewall, cutting internet access, allowing RDP only from the management network and reducing the number of accounts that sign in to the server as administrator. The risk register should record the server owner, the target closure date and the manager who accepted the risk. An exception without a date becomes permanent.
Frequently asked questions
Does the server keep working after end of support?
Yes, the server keeps working, but no new security updates are released. That means vulnerabilities that are found can’t be patched, and it shows up as non-compliance in audits.
In-place upgrade or clean install?
An in-place upgrade can shorten the timeline in some scenarios, but it also carries over old configurations. For domain controllers and role-based servers, building a new server and moving the roles is usually cleaner and easier to roll back.
What can be done if migrating right away isn’t possible?
Check the current Extended Security Update (ESU) options and terms Microsoft offers. During that period, restrict the server’s network access and firm up the migration schedule.