Server Migration & Management
Zero-downtime migrations with documented precision
Migrating a live server is not a file copy. The database and filestore have to arrive consistent with each other, DNS has to cut over without a window where half your users hit the old box, SSL has to be ready before the switch rather than after, and you need a way back if something only surfaces once real users arrive. Most bad migrations are not technically hard, they are unrehearsed.
- Pre-migration audit of the source: OS, application versions, custom code, cron jobs, certificates and their expiry, and current backup state
- Target server built fresh and validated on its own before any production data moves, never a live-to-live copy under time pressure
- Data transfer with checksums on both sides, so a partial copy is caught before cutover rather than after
- Parallel testing window on the new server using the real dataset, reached by hosts file or a temporary hostname
- Low-TTL DNS cutover during an agreed window, with the old server kept running and reachable
- Holdback period before decommissioning the old server, then a written post-migration report
From $250 (single server) or $150 per server for fleets
See the full rate card for fixed packages and managed retainer plans.
Deliverables
- Ubuntu 20.04 to 22.04 migration
- Database migration and validation
- Filestore and attachment migration
- DNS and SSL migration
- Post-migration performance audit
- Ongoing management retainers
How much downtime will we have?
We plan for none during business hours. The new server is built and validated in parallel, so the only switch is DNS, and the old server keeps serving until it propagates.
We have a whole fleet, not one server. Does that change things?
It changes the sequencing, not the method. We have migrated an 11-server Odoo fleet this way, one server at a time with the runbook improved after each, which is also why fleet pricing is lower per server.
What if something is wrong after cutover?
The old server is deliberately kept running for a holdback period, so a rollback is a DNS change and not a rebuild. That is written into the plan before we start.
Do you migrate Windows or MSSQL workloads too?
Yes. Most of our work is Linux and PostgreSQL, but we run production Windows and MSSQL workloads for clients as well.