Skip to content

A migration runbook: from an office server room or the cloud to a data center cabinet

How Telnetrix moves equipment or workloads into a colocation cabinet: inventory, network first, proven backups, a rehearsed cutover and checks afterwards.

Weeks out: inventory and dependencies

Every move starts with a list of what exists and what depends on what. For an office room that is the devices, their roles, their power draw and the services on them. For a cloud environment it is the usage export and the resource list. In both cases the output is a dependency map: which systems talk to which, over which ports, and which must move together.

  • Device or resource list with model, age, support status and role.
  • Services and the dependencies between them, including external ones: licensing servers, payment processors, identity providers.
  • Measured power draw and measured utilization, not nameplate figures.
  • Who uses each system, from where, and what they would notice if it stopped.

Design the destination before touching the source

The footprint, the power, the cabinet layout and the network are designed and agreed before anything is ordered. The network design is the critical one: addressing, routing, the private WAN from the offices, the multi-homed BGP at the cabinet, what stays in the cloud and how the cabinet reaches it. If the business brings its own address space, the move is planned so that public addresses do not change.

New hardware is specified from the measured utilization with headroom, not from the sizes of the cloud instances or the age of the office servers. Equipment that should not travel, because of its age or its warranty, is replaced rather than moved.

Build the network first

The cabinet, circuits, cross-connects, routers and firewalls are installed and tested while nothing depends on them. Failover is tested by pulling a carrier, a router and a power feed, in that order, and watching what happens. The private link from the office to the cabinet is live and monitored before any server moves, so the destination is reachable from the day the first system arrives.

Prove the backups

Nothing moves until a current backup of it has been restored somewhere else and checked. For an office room this is often the first time a restore has been attempted. The restore test produces the two numbers the cutover plan needs: how long a restore takes, and how much data changes per day, which sets how long the final synchronization will run.

Rehearse the cutover

Where possible, the new environment is built in the data center first, so that only data moves on the day. Each system has a runbook entry: the order of shutdown, the synchronization step, the start order, the checks that prove it is working, and the person who signs it off. The sequence is rehearsed against a copy, and the rehearsal sets how long the cutover window needs to be.

  1. 01Freeze changes on the source environment.
  2. 02Final synchronization of data, with the delta measured against the rehearsal.
  3. 03Shut down the source in dependency order.
  4. 04Start the destination in reverse dependency order, checking each system before the next.
  5. 05Switch DNS, routing or the private link so users reach the destination.
  6. 06Run the acceptance checks with the people who use each system.

Rollback criteria, decided in advance

The decision to roll back is made against written criteria, not in the moment: a system that fails its acceptance check, a synchronization that overruns by more than the agreed margin, a dependency that was missed. The source environment stays intact and powered until the acceptance checks pass and a period of normal operation has run, so a rollback is a routing change rather than a rebuild.

After the move

  • Monitoring on every moved system and on the circuits, with alerts routed to people who can act.
  • Backups running from the new location to a second site, with a restore tested again.
  • Documentation updated: the inventory, the network design, the runbooks, the access list for the cabinet.
  • The source room or cloud account decommissioned on a date, not left running out of caution.
  • A first monthly report comparing what the environment costs now against what it replaced.

Telnetrix follows this sequence for every move into a data center cabinet, whether the equipment comes from an office server room or the workloads come from the cloud.

Contact Telnetrix

If this reads like your situation, tell us what you run and where. We reply by email or phone, whichever you prefer.

Opens your email app, addressed to sales@telnetrix.com.