Chapter 01 · 12 min read
Planning a decommission
Most decommissions go wrong in the same place: someone starts pulling servers before anyone wrote down what was in them. By the time a valuation happens, the CPUs are in one box, the caddies are in another, and nobody can prove what came out of rack A14.
The order below exists to prevent that. It isn’t complicated, but the sequence matters more than the content — almost every expensive mistake in this process is a phase done out of order.
01.1Before you touch anything
Four things to establish while everything is still racked and powered:
- What’s actually in scope. Rack elevations, a CMDB export, or a walk-round with a phone. Photograph the front and rear of every rack before anything moves. This takes an afternoon and saves a fortnight.
- Who owns it. Leased, financed and hire-purchase kit cannot be sold. Find out now.
- What’s still live. Phased migrations mean some machines are still carrying workload. Mark them and physically separate them if you can.
- When the space has to be empty. If a colo contract or lease is ending, the dilapidation terms drive the entire timeline.
Read the dilapidation clause early. “Return the space as found” can mean removing containment, cabling, floor tiles and even the racks themselves. People discover this three weeks before hand-back, and then it’s an emergency instead of a schedule.
01.2Capturing the assets
This is the phase that determines what you get back financially. Per unit, aim to record:
- Manufacturer and model
- Serial number or service tag
- Processor model and count, memory total, drive count and size
- Any GPUs or accelerators — usually the highest-value thing present
- Rack and U position, so you can reconcile afterwards
- Known faults
Don’t gold-plate it. A rough list this week beats a perfect one next month, because a valuation can be revised and a hand-back date cannot. Our asset list template captures exactly these fields if it saves you building a spreadsheet.
01.3Deciding what happens to the data
Three options, and you can mix them across one estate:
- Erase and resell. Highest return. Media sanitised to NIST 800-88, certificated per device.
- Destroy the media, sell the hardware. The usual middle ground — a chassis keeps most of its value without drives.
- Keep the media yourself. You retain and destroy the drives; only hardware leaves. Perfectly reasonable and it barely dents chassis value.
Whichever you pick, remember the data that isn’t on the drives. Management controllers, boot modules, switch configuration and firewall policy all need handling separately, and none of them appear on a drive inventory.
01.4Getting it valued
- Send the list to two or three buyers and compare like for like.
- Ask for the offer itemised. A single lump figure can’t be checked against the audit later, which is precisely why some people quote that way.
- Ask what happens if the goods-in audit disagrees with the valuation — specifically whether you can decline the revised offer, and who pays to return the equipment.
- Confirm whether collection and de-racking are included.
- Ask for licences up front: waste carrier registration, treatment exemptions, insurance.
A reasonable question to ask any buyer: “What in this load is worth nothing, and what are you doing with it?” A straight answer tells you more about a supplier than the headline number does.
01.5Planning the physical removal
- Site access. Data centres need access requests days ahead, often with escorts. Book early.
- Loading bay and goods lift slots. Usually the real constraint on how fast a suite clears — not the number of people.
- Out-of-hours windows. Many DCs restrict removals to specific times.
- De-racking. Slower than everyone expects. Cable management and containment take longer than the servers do.
- Loose media. Drives pulled on site should travel in sealed, lockable containers, numbered and recorded.
- Sign a manifest on the day listing exactly what left the building.
01.6The paperwork you must keep
This is what an auditor asks for, and what discharges your legal obligations:
- Waste transfer note — required under your duty of care, issued at the point of transfer to an authorised carrier.
- Certificates of erasure or destruction — per device, by serial, stating method and verification result.
- A reconciled audit report — what was collected against what you listed, with discrepancies explained.
- Supplier licences — evidence you transferred waste to someone authorised to take it.
Chapter 05 covers what the law actually requires and why the serial numbers matter so much.
01.7Closing out
- Reconcile the audit report against your original list while people still remember the project.
- Retire the assets in your CMDB and cancel the support contracts — frequently forgotten, and frequently still billing a year later.
- Remove DNS entries, monitoring checks, backup jobs and firewall rules pointing at kit that no longer exists.
- File the certificates somewhere findable in three years’ time.
- Record the recovered value and the CO2e avoided. The second increasingly shows up in Scope 3 reporting and someone will ask for it.
01.8The five that catch people out
- Leased equipment found on collection day. Splits the load, delays everything, occasionally involves a lawyer.
- Asset capture after stripping. Value evaporates and reconciliation becomes guesswork.
- Forgetting the management controllers. Credentials leave inside the chassis.
- Dilapidation clauses nobody read. Containment and cabling removal is a project in itself.
- No waste transfer note. Duty of care isn’t discharged and the liability stays with you, whatever the other party did next.
About the author
Kamil Anwar — Founder, ServerGear. Kamil runs ServerGear, the data centre asset recovery arm of PYCO RENEW LTD. He spends most of his week looking at asset lists, arguing about what a four-year-old server is really worth, and making sure the drives that come with it are dealt with properly.