Buying a Used Server: The Eight Questions and What's Missing
A used server can be an excellent purchase or an expensive mistake, and the difference is a handful of checks that take minutes. This covers what to ask before buying and what to verify when it arrives.
Before you buy: eight questions
A seller who answers all eight readily is telling you something useful.
What exactly is the configuration? Model, generation, processors, memory total and module count, drives, controller. Our naming guide covers identifying a generation.
How many drive bays, and what size? Bay count is fixed by the backplane and cannot be changed by adding drives. This is the specification most often misunderstood.
Are rails included? They stay in the rack when a server is removed, so used machines frequently arrive without them, and they are chassis-specific.
Are drive carriers included for every bay? Empty bays rarely have them.
Are risers included? Without them the machine takes no expansion cards.
Is the controller cache module and battery present? Without working cache backup the controller runs write-through and write performance collapses with no error reported.
What are the drive hours? The drives report this exactly. A seller who cannot say has not looked. Our hours guide covers judging the figure.
Is there a management licence? iLO Advanced or iDRAC Enterprise, for remote console. Licences do not always transfer, and this is the feature you need when a machine will not boot. Our out-of-band guide covers why.
The specifications that decide compatibility
Four things to confirm against what you already run, if the machine is joining an existing estate.
Processor generation and firmware level. A newer processor generation needs firmware support the board may not have. Our processor guide covers this.
Memory type and speed. RDIMM, LRDIMM or UDIMM, and the speed grade. Our memory guide covers reading labels.
Drive carrier generation, if you intend to move drives from another machine. Carriers are generation-specific and do not cross.
Whether it is a platform you already hold spares for. Adding a new platform multiplies your spares holding rather than adding to it. Our multi-vendor guide covers that cost.
When it arrives: what to check before it goes anywhere near production
Six steps. Skipping them is how a problem becomes your problem.
Inspect before unpacking fully if the box looks damaged, and photograph it. Our packing guide covers why the packaging matters for a claim.
Confirm the configuration matches what you ordered — processors, memory total, drives, controller.
Check it posts and reports everything. The management controller lists memory by slot, drives by bay, and flags anything it does not like.
Update firmware across BIOS, controller, drives and adapters in the vendor sequence. Our firmware guide covers sequencing.
Run it under load for a sustained period. Early-life failures and transit damage show up here, while you can still return it.
Record everything — serials, part numbers, firmware levels, drive hours at arrival. Our acceptance testing guide covers the checklist.
The parts that are usually missing
Budget for these rather than discovering them.
Rails. Drive carriers for empty bays. Risers. Cache battery. Bezel and blanking panels. Power cords — one per supply. Management licence.
None is expensive individually. Together they turn a two-hour job into a two-week one. Our build versus buy guide covers the full list.
When used is the wrong choice
Being honest about it.
Where manufacturer warranty is a requirement, not a preference.
Where the platform is close to losing operating system or hypervisor support. Check what your intended version supports before buying, not after. Our compatibility guide covers checking.
Where the machine will run continuously under load for years, and the power difference against a current generation exceeds the purchase saving. Our TCO guide covers the arithmetic.
Where per-core licensing makes older cores expensive relative to what they deliver.
Frequently asked questions
What should I ask before buying a used server?
The exact configuration, bay count and size, whether rails, carriers, risers and the cache battery are included, the drive hours, and whether a management licence is present. A seller who answers all eight readily is telling you something useful.
What is usually missing from a used server?
Rails, drive carriers for empty bays, risers, the controller cache battery, bezel and blanking panels, power cords, and the management licence. None is expensive alone; together they turn a two-hour job into a two-week one.
What should I check when a used server arrives?
Inspect the packaging, confirm the configuration matches, check it posts and the management controller reports everything, update firmware in the vendor sequence, run it under load, and record serials and firmware levels.
Can I add more drive bays to a used server?
Not by adding drives. Bay count and size are fixed by the backplane fitted, which is chassis-specific hardware with its own cabling. Confirm the bay configuration before buying.
When should I buy new instead of used?
Where manufacturer warranty is a requirement, where the platform is close to losing OS or hypervisor support, where it will run continuously for years and the power difference exceeds the saving, or where per-core licensing makes older cores expensive.
Do I need a management licence on a used server?
For remote console and virtual media, yes — and those are exactly what you need when a machine will not boot and nobody is on site. Licences do not always transfer with used equipment, so ask.
Tell us the model you are considering and we will list what it should include and what is usually missing before you commit.
