Skip to main content
Legia

IT Infrastructure & Cloud

Cloud Migration Checklist: What to Prepare Before Moving Servers

Cloud migrations usually go wrong in the preparation, not in the move itself. This checklist covers the six groups of checks that should be finished before the first server is moved.

SEO and Content Specialist

8 min read

A dark server rack with colored cables

Short answer: a cloud migration checklist covers six groups of checks: inventory and dependencies between systems, the real load on each server, compatibility and licensing of the software, security and access, backups with a tested rollback plan, and the cutover schedule including DNS. A server is ready to move when all six have an owner and proof of completion.

Migrations that stall often start the same way: the server is moved before the company knows what depends on it, how heavy its load is, and how to go back if it fails. This article is a cloud migration checklist that an IT team or business owner can use before approving a move date. The choice of strategy, whether to move as-is, re-platform, or rewrite, is covered separately in Rehost, Replatform, or Refactor.

The cloud migration checklist in one table

GroupWhat must be confirmedProof of completion
1. Inventory and dependenciesAll servers, applications, databases, and the links between systems are recordedA system list and dependency map confirmed by each system’s owner
2. Load sizingCPU, memory, disk, and network use is measured, including peaksSeveral weeks of usage data and the proposed resource sizes
3. Compatibility and licensingThe OS and software are supported at the destination, and licenses can moveA list of blockers with a plan to fix each
4. Security and accessService accounts, encryption, firewall rules, and admin accounts are mappedAn access design approved before the environment is built
5. Backups and rollback planThe latest backup can be restored, and there are failure criteria and steps to go backResults of a restore test and a rollback test in a test environment
6. Schedule and DNSThe cutover time, the order, and the lowering of DNS TTL are setA schedule approved by management, with an owner for each step
Illustration: server

1. Record every system and what depends on it

The inventory is the first check because a missed dependency is the cause of the hardest outages to trace. The Microsoft Cloud Adoption Framework advises discovering all dependencies first, because components that depend on each other and are not moved together can disrupt the service. The same guidance notes that automated discovery tools can miss undocumented dependencies, so the results need to be confirmed through interviews with system owners. Google Cloud’s migration guidance also places the inventory of applications and environments, dependencies, and performance baselines in the assessment stage, before planning begins.

What to record for each system:

  1. Name, function, and business owner.
  2. Other systems that call it or that it calls, including shared databases, login services, and third-party APIs.
  3. Supporting services that are often forgotten, such as DNS, automated email senders, scheduled jobs, and report printers.
  4. The environment: production, test, or development.

Systems that depend on each other are grouped and moved together. If a component cannot move yet, note the reason and how long the system will run split across two places. A tidy list also makes it easier to build the asset and license inventory covered in How to Build an IT Asset and Software License Inventory.

2. Measure the real load before choosing sizes in the cloud

Server sizes in the cloud should be set from actual usage, not from the old server’s specification. Microsoft advises collecting CPU, memory, disk I/O, network throughput, and peak load, and paying attention to daily or weekly peaks to understand capacity needs. The same data is later used to compare performance before and after the move.

Also record the server’s configuration: the number of cores and RAM, the operating system version, the type and capacity of storage, and the installed licenses. Resources that are too large inflate the monthly bill, while resources that are too small make the application slow at busy hours. The monthly cost components are worked through in Calculating Monthly Cloud Costs, and how to read the needs of an internal application server in Calculating Server Specifications for an Internal Application.

3. Check compatibility and licensing

Before choosing a destination, make sure the operating system and software that run today are supported there. Microsoft notes that assessment tools can flag components that are unsupported, deprecated, or nearing end of support, and advises separating the blockers that must be fixed before migration from those that can follow afterward.

  • Operating system version. A server whose operating system is past its support period needs a decision: upgrade first, or move as-is with an upgrade plan afterward.
  • Vendor software. Microsoft’s guidance states that third-party software compatibility must be confirmed before migration, through vendor documentation, a test environment, or directly with the vendor.
  • Licenses. Paid software licenses may need to be transferred or may fall under a special scheme in the cloud. Check the terms with the vendor and include the cost in the budget.
  • Application frameworks. The same guidance advises changing a framework only when there is a strong business reason, because the change demands extra development and testing.

4. Map security and access from the start

The old server’s security configuration has to be replicated or adjusted in the new environment, and that is only possible if it has been recorded. Microsoft advises listing service accounts, API keys, the accounts applications use to authenticate, the encryption methods for data, and firewall rules and access control lists. Check also whether any credentials are embedded directly in code or configuration files.

For access to the cloud environment itself:

  1. The cloud account is created in the company’s name, not in an individual’s or a vendor’s name.
  2. Administrator accounts use two-step authentication. The reasons and the types to choose are covered in What Is Two-Step Authentication.
  3. Access rights are divided per person and per role, and access logs are turned on.
  4. These rules are approved before the environment is built, not retrofitted after the system is already running.

5. Prepare tested backups and a rollback plan

The rollback plan must exist before the migration starts. Microsoft advises deciding in advance what counts as a failure, for example a failed health check, poor performance, or a security problem, then writing the steps to go back for each workload and testing them in a test environment. A plan that has never been tried is only proven faulty when it is needed.

Checks in this section:

  1. The last backup before cutover has been made and can actually be restored. How to test it is in Testing Server Backup Recovery.
  2. The failure criteria are written down, such as a response time limit or an error rate, along with who has the authority to decide to roll back.
  3. The steps to go back are written per stage, not only for the project as a whole.
  4. The old server is not shut down until the new system has proven stable.

6. Set the cutover schedule and prepare DNS

The order and timing of the cutover determine how big the risk is. Microsoft advises starting with simpler workloads, moving test environments before production, and scheduling critical systems after the team has shown it is capable. The schedule should avoid important business periods such as financial close, product launches, or peak season, with buffer time for testing and fixes.

The migration method is also chosen per system. A system that can stop at a scheduled hour can be moved with downtime, while a system that supports customer-facing service needs a near-zero-downtime method with continuous data replication, which must be tested first.

DNS is the step most often forgotten. Microsoft’s Azure networking guide calls DNS cutover the highest-risk step in a cross-cloud migration, and advises lowering the TTL of the records that will change to 60 to 300 seconds at least 48 hours before cutover, then raising it back to production values once resolution is confirmed stable. Record every entry that points to the servers being moved: A, CNAME, mail (MX), and others.

Short checklist before cutover day

  1. The system list and dependency map are confirmed by each owner.
  2. Several weeks of usage data have been used to set sizes.
  3. Compatibility and licensing blockers each have a decision.
  4. Cloud admin accounts use two-step authentication and are in the company’s name.
  5. The latest backup has been restore-tested, and the rollback steps have been tested in a test environment.
  6. DNS TTL has been lowered, and every record that changes is recorded.
  7. The schedule is approved and does not clash with an important business period.
  8. A budget owner has been named on the company side.

How Legia does it

In the Cloud Migration & Architecture service, the checks above form the early stage of the project. The readiness assessment maps the running systems and the dependencies between components, takes 1 to 2 weeks, and is followed by a placement design with a cost estimate over 1 to 2 weeks. The migration then runs in stages, alongside the old system, with a rollback plan for each stage. For mid-sized workloads, the whole process generally takes 2 to 4 months. The cloud provider’s bill is paid directly by your company, and the security measures cover per-person and per-role access rights, two-step authentication for administrator accounts, encryption, backups, and access logs. Other services in this pillar are on the IT Infrastructure & Cloud page.

Frequently asked questions

How long does preparation take before the migration starts?

It depends on the number of systems and how well documented they are. In Legia’s working stages, the readiness assessment and the design each take 1 to 2 weeks for mid-sized workloads. Automated discovery tools help, but their results still need to be confirmed through interviews with system owners.

Do all servers have to be moved to the cloud?

No. Some systems make more sense staying on your own infrastructure, especially when their usage pattern is stable and predictable. The results of the inventory and load measurement are the basis for sorting out which to move, which to restructure, and which to leave.

Will operations stop during the migration?

Not necessarily. With a staged migration, workloads are moved piece by piece and run alongside the old system until proven stable, then switched over in an agreed window. The length of downtime per system depends on the migration method chosen.

When can the old server be switched off?

Once the new system has run stably under real use and the last backup from the old server is stored. Decommissioning the source is indeed the final step in Microsoft’s migration sequence, after optimization in the cloud.

Sources

Category IT Infrastructure & Cloud

Keep reading

Related articles

All Articles