Data center decommissioning: the sequence that keeps a project auditable

The sequence that keeps a project auditable

1. Discovery. Physical inventory, serials, and dependency mapping. Nothing is unracked before this is signed off.
2. Data classification. Which devices hold data, which do not, and what evidence each class requires.
3. De-installation. Power down, disconnect, label, stage. Labelling is what makes the rest reconcilable.
4. Data handling. Erasure, degaussing or destruction, on-site or off-site, with serial-level records. What that means across a hall is in data center destruction.
5. Removal and transport. Packing, dock scheduling, sealed loads, manifests.
6. Settlement and reporting. Certificates, recycling evidence, resale settlement, final reconciliation.

The order matters more than the speed. Skipping discovery to start pulling hardware is the most common way a project becomes impossible to close out cleanly.

Where projects actually go wrong

• An inventory taken from the CMDB rather than from the floor. The two rarely match.
• Dependencies discovered during removal, a forgotten jump host, a licence server, a monitoring collector.
• Data-bearing devices found in equipment nobody classified as data-bearing.
• Building access assumed rather than confirmed with facilities.
• Reporting agreed at the end, when the evidence has already been lost.

What to require in writing before work starts

• A serial-level inventory, produced on site, signed by both parties.
• The custody model: who holds the assets at each step, and how transfers are recorded.
• The destruction method per device class, and the evidence produced for each.
• Acceptance criteria for the room at the end of the project.
• The reconciliation report format, agreed before removal rather than after.

The data center decommissioning checklist covers what to settle internally before consulting anyone.

Frequently asked questions