Cloud Server Migration Runbook: Move a Website or App Between VPS Providers
Use this provider-neutral runbook to estimate migration effort, prepare a new VPS, synchronize files and databases, cut over DNS with controlled risk, and verify that your website or application works before decommissioning the old server.
Overview
A cloud server migration is more than copying a directory to a new machine. The destination server must match the application’s runtime requirements, preserve data, accept production traffic, and support a reliable rollback plan. A rushed move can create avoidable problems with database versions, scheduled jobs, email delivery, SSL certificates, file permissions, background workers, or DNS.
The safest general pattern is to keep the old server available while you build and test the new one. First inventory the current environment. Then provision the destination, transfer a copy of the data, validate the application on a temporary hostname or restricted route, perform a final synchronization, and change DNS only after the new server is ready. Keep the source server intact until the verification period is complete.
This process applies to a brochure website, content management system, ecommerce application, API, or developer project hosted on VPS infrastructure. The exact commands depend on the operating system, control panel, web server, database engine, and deployment method, so treat the steps below as a planning framework rather than a copy-and-paste script.
Before starting, create a current backup and test that it can be restored. A snapshot alone may not be sufficient for every application, particularly when data is changing during the snapshot. Use the cloud server backup and restore plan to document recovery points, offsite copies, and restore tests.
How to estimate
Estimate the migration in four parts: preparation, data transfer, cutover, and verification. A simple planning formula is:
Total migration window = preparation time + transfer time + final sync time + cutover time + verification buffer
Preparation and verification are usually better predictors of effort than raw bandwidth. A small site can still require careful work if it has several integrations, custom services, or a busy database. Conversely, a large static file collection may transfer predictably if it is well organized and the network path is stable.
To estimate transfer time, divide the amount of data by the effective transfer rate, not the advertised connection speed. Include a buffer for encryption overhead, small files, disk performance, compression, retries, and files that change during the copy:
Estimated transfer time = data volume ÷ effective transfer rate × safety factor
For example, if a site has 80 GB of files and database exports combined, and your observed effective rate is 40 MB per second, the theoretical transfer time is about 34 minutes. A planning estimate should be longer because the transfer may include many small files and a second pass for changes. Treat the result as an estimate to test, not a promise.
Estimate downtime separately. If the application supports read-only mode or a brief maintenance window, the final cutover may be limited to the time needed to stop writes, complete the last database synchronization, switch traffic, and run smoke tests. If it cannot tolerate that pause, use a replication or blue-green deployment approach appropriate to the database and application architecture.
Record these estimates in a change plan:
- Planned start and end times, including the time zone.
- People with access to the source server, destination server, DNS provider, and application credentials.
- Expected maintenance window and the maximum acceptable interruption.
- Backup locations, restore instructions, and the person authorized to invoke rollback.
- Success checks and a specific deadline for deciding whether to continue or revert.
Inputs and assumptions
Build an inventory before provisioning the new VPS. Do not rely on memory or a control-panel summary. Capture the following inputs.
Application and traffic
- Domains, subdomains, redirects, and alternate hostnames.
- Application language versions, package managers, framework versions, and environment variables.
- Web server configuration, reverse proxies, load balancers, container definitions, and exposed ports.
- Scheduled tasks, queue workers, webhooks, third-party APIs, cron jobs, and monitoring checks.
- Expected CPU, memory, storage, and network requirements during normal and peak periods.
Data and state
- Database engine, version, size, character set, extensions, users, and connection settings.
- Uploaded files, generated assets, logs, caches, search indexes, and persistent container volumes.
- Data that can be regenerated and data that must be copied exactly.
- Write activity during the migration and the point at which writes can be paused.
DNS, email, and security
- A complete DNS zone, including A, AAAA, CNAME, MX, TXT, and verification records.
- Current TTL values and the provider responsible for authoritative DNS.
- SSL certificate method, private-key handling, renewal automation, and HTTP-to-HTTPS redirects.
- Firewall rules, SSH keys, user accounts, sudo access, backups, and monitoring.
- Mail services and authentication records such as SPF, DKIM, and DMARC, if email is part of the domain.
Use the same operating system family and compatible runtime versions where practical. If you are also changing the database version, web server, operating system, or application version, treat the project as both a migration and an upgrade. Test those changes separately when possible. A provider change is easier to troubleshoot when fewer variables change at once.
Prepare the destination with least-privilege accounts, key-based administration, time synchronization, firewall rules, required packages, log rotation, backups, and monitoring. Do not copy unnecessary secrets or temporary files. Store credentials in the destination’s intended secret-management method rather than placing them in a public web directory.
Worked examples
Example 1: A content site with a short maintenance window
Suppose a content site has 12 GB of uploads, a 2 GB database, a small number of scheduled jobs, and no user checkout process. The migration plan could be:
- Lower the relevant DNS TTL in advance, subject to the DNS provider’s behavior and your change policy.
- Provision the new VPS and install the required web server, runtime, database, firewall, monitoring, and SSL tooling.
- Copy the application files and perform a database export and import.
- Point a temporary hostname or local test entry at the new server and check pages, administration, forms, redirects, media, logs, and scheduled tasks.
- At cutover, enable maintenance mode or stop writes, run a final database export, copy changed files, and import the final data.
- Update the A or AAAA record, then verify the site through the domain, HTTPS, and key user journeys.
Because the site has limited transactional activity, the final synchronization may be simple. The rollback condition should still be explicit: revert DNS and restore normal service on the old VPS if critical pages fail, data is missing, or the application cannot serve HTTPS within the agreed verification window.
Example 2: An application with active user transactions
Now consider an application that accepts orders or writes frequently to a database. A one-time export followed by a long DNS change can lose recent transactions or create conflicting records. Use a migration design that keeps the destination current, such as database replication, a tested managed migration feature, or a short controlled write pause with a final export. The correct method depends on the database engine, versions, application behavior, and operational skills available.
For this example, estimate the final cutover as a separate event. Define when writes stop, how the final position is confirmed, how workers are drained, and how the destination is promoted. Test the exact sequence in a staging environment. Keep the old application available in a read-only or stopped state until transactions, queues, payments, and external callbacks have been checked.
Example 3: A domain with email dependencies
A website migration does not automatically require an email migration. If mail remains with the existing provider, preserve its MX and mail-related TXT records. Copy the web records deliberately rather than replacing the entire DNS zone with a simplified template. If nameservers are also changing, compare the old and new zones record by record and plan for resolver caching. The DNS records cheat sheet and zone-file template can help structure that comparison.
When to recalculate
Revisit the estimate whenever an input changes. Recalculate before the migration if data volume grows, traffic patterns change, a database is upgraded, the destination’s storage or network plan changes, or the application gains new integrations. Recheck the transfer estimate with a small test copy instead of relying on an old benchmark.
Update the runbook after every migration. Record actual transfer durations, final synchronization time, DNS change time, verification findings, and rollback decisions. These measurements make the next move more predictable and expose recurring work, such as manual SSL setup or undocumented scheduled jobs.
Use this final action checklist:
- Confirm tested backups and a documented restore path.
- Freeze or document configuration changes during the migration.
- Compare source and destination DNS, application settings, users, permissions, and scheduled tasks.
- Complete a staging validation before the final synchronization.
- Set a named rollback owner and a deadline for the go or no-go decision.
- Monitor HTTP responses, HTTPS certificates, logs, CPU, memory, disk, database health, queues, and critical user flows after cutover.
- Check email delivery and authentication records if the domain has mail dependencies.
- Keep the old VPS available until the verification period and backup requirements are satisfied.
For broader launch verification, use the website launch checklist. If the project also changes registrar ownership or nameservers, separate that work from the server move where possible and follow the domain transfer checklist. A migration is complete when the new server is serving real traffic, data is accounted for, operational controls are working, and the rollback decision is no longer needed—not merely when the DNS record has changed.