Linux migration / Engineering resource
Linux migration readiness checklist: find blockers before cutover
A Linux VM is ready for migration when its intended target is confirmed, its workload dependencies are understood, and owners can demonstrate recoverability and acceptance—not merely when an image can be copied. Build an evidence-backed checklist before booking the outage. Separate a hypervisor move from an OS upgrade or rebuild: those changes have different risks and different tests.
01Define the move and confirm the exact path
Write down the source distribution, release, architecture, kernel, boot mode and image format; then record the target OS release and platform. Moving an unchanged guest to another hypervisor primarily changes the virtual hardware and platform integration. Rebuilding on a newer Linux distribution also changes package availability, service behavior, libraries and security policy. Do not treat success at one as evidence for the other.
Confirm the specific source-to-target conversion with the current supported list and your scoping contact. Separate source and target release lists are not a full support matrix: a listed source does not necessarily reach every listed target. Record exclusions, manual steps and the version of the tooling used. Include the intended downtime, recovery objectives and the person authorized to stop the change.
02Inventory the workload, not just the VM
Identify the application owner and the people responsible for infrastructure, networking, security and databases. Ask what the VM does, who consumes it, and what breaks if it disappears. Capture inbound and outbound dependencies: identity providers, queues, storage, mail relays, APIs, licensing endpoints and monitoring. Record their owners and a representative transaction for each critical path.
List installed packages, third-party repositories, language runtimes, extensions, native libraries and locally compiled software. A replacement package with a similar name is not automatically compatible. Check supported application/runtime combinations, library interfaces, configuration locations and repository trust. Mark unsupported components for replacement, rebuild or retirement before the rehearsal.
03Record services, schedules, identity and access
- Capture enabled services separately from running processes. Include custom unit files, startup scripts, dependencies, service accounts and environment files. Record expected behavior rather than assuming all enabled units should run continuously.
- Inventory user crontabs, system cron directories and timers, including timezone, overlap protection and external destinations. Plan how to prevent both old and new systems from running the same production job.
- Record UID/GID requirements, group membership, directory integration, sudo policy and access methods. File ownership may refer to numeric IDs even when names look correct. Keep credentials and private keys in your approved secret-management process, not in a planning document.
04Check network identity and data consistency
Document interfaces, addressing, routes, DNS records, resolver settings, firewall rules and the network paths clients actually use. Include proxies, load balancers, TLS certificate names, renewal mechanisms and trusted certificate authorities. A successful ping from the VM does not establish that an external customer can complete an HTTPS transaction. Assign someone to test from the relevant client locations.
Map mounts, filesystem types, device naming, mount options, capacity and ownership. Identify application data separately from rebuildable binaries. For databases, agree a database-consistent backup or replication procedure and the point at which writes stop or switch over. A copied live database directory or a VM snapshot alone is not evidence of a consistent recovery point. Record how lag, final synchronization and duplicate writers will be handled.
05Prove recovery and establish a performance baseline
Confirm backups exist, can be read and can actually be restored into an isolated environment. Record the restore procedure, required keys and dependencies, elapsed recovery time and the checks that showed usable data. Verify that rollback remains possible after any database-format or application-version change; a reverse VM move cannot undo incompatible writes.
Measure representative CPU, memory, storage latency, disk growth and application response time on the source. Include peak periods and batch work rather than one idle sample. Compare the target capacity and virtual hardware assumptions with those needs. Decide acceptable tolerances before seeing the new results, and agree who investigates a regression.
06Turn findings into owned decisions
Use a small register that links to evidence kept in your own controlled change record. These are example blocking findings, not results from a customer environment. Close an item only when its owner records a tested resolution or an explicitly approved exception.
| Evidence to retain | Blocking finding | Accountable owner |
|---|---|---|
| Exact OS pair and confirmed scope | Conversion not confirmed; manual work unowned | Migration lead |
| Runtime and dependency inventory | Required extension unavailable on target | Application owner |
| Schedules and service identity | Duplicate job execution or UID mismatch | System/application owner |
| Client-path and TLS test plan | Required route, trust or name missing | Network/security owner |
| Consistent recovery point and restore result | Restore untested or data recovery exceeds objective | Database/backup owner |
| Rehearsal and acceptance results | Critical transaction fails; rollback cannot be rehearsed | Change approver |
07Optional Virtual Move planning examples
The official quickstart documents the following CentOS 7 image and Rocky 9 examples. Use them only for a confirmed path, on an image you are authorized to inspect, with the documented prerequisites. They are planning/inspection examples, not a command to execute cutover. Assess reads image magic bytes; the migrate preflight performs analysis, reports counts and blocking steps, and writes no output files.
Review findings alongside your manual inventory. Automation can expose mapping gaps, but workload ownership, data handling and acceptance criteria still require decisions. The current supported-path page and command documentation are the authority for the specific release and mode; do not extrapolate these examples to every VM.
vrtmv assess centos7.vmdk
vrtmv migrate --image centos7.vmdk --target rocky9 --preflight08Rehearse, then sign off the cutover
Rehearse the rebuild, configuration, data recovery and application tests with representative inputs. Time the procedure and note where it differs from production. Define the observation window and the point after which rollback becomes a different recovery operation. Before cutover, name the operator, acceptance owners and final approver; leave a clear stop condition for every critical test.
Workload / change ID: __________
Source / target / confirmed scope: __________
Evidence register and rehearsal date: __________
Data consistency and restore evidence: __________
Acceptance tests / owners / tolerances: __________
Cutover window / observation window: __________
Rollback trigger / last safe point / operator: __________
Open exceptions / risk owner: __________
Decision: GO / HOLD Approver / date: __________Questions engineers ask
- Does a supported source and target mean this pair is supported?
- No. The official list explicitly names releases, not every conversion. Confirm the exact pair and workload scope before committing a schedule.
- Is preflight a migration execution?
- No. The documented --preflight option analyzes feasibility and reports counts and blocking manual steps without writing files. Your team still reviews and executes the eventual migration.
Sources
Product details checked against the current official documentation. Confirm your installed version, exact conversion and operational prerequisites before use.
Continue reading
Apply this to one workload
Virtual Move helps assess, plan and verify supported Linux workloads. Your team owns the migration and acceptance decision.
