The Refurbished Server Inspection Checklist

A good refurbisher has already burned the machine in, updated firmware, and cleared the previous life's traces. The first hour after delivery is where you verify that story — and enterprise hardware makes verification easy, because unlike consumer gear, these machines keep receipts about their own health.

Before power: the physical pass

Rails and ears intact, no shipping damage to the chassis lips where rails seat. Every bay either carries a drive or a blank — missing blanks wreck airflow. Both PSUs seated, matching wattages. Pop the lid: heatsinks torqued, no missing DIMM blanks, shrouds present, no dust drifts (a caked interior contradicts the word refurbished), cables routed like a professional was paid to route them.

First boot: read what the machine says about itself

Go straight into the BMC — iDRAC or iLO — before installing anything. System Event Log: should be short and recent; entries from the previous owner mean cleanup was skipped, and historical ECC or thermal events deserve explanation. Memory health: per-DIMM correctable-error counters should read zero; a nonzero slot on day one is tomorrow's failure being polite. Thermals at idle: inlet in the low 20s room-dependent, CPUs settling in the 40s, fans dropping from boot roar to steady state — a fan pinned high at idle signals a sensor, shroud or firmware issue. PSU status: both present, both green; then the honest test — pull one power cord under a light load and confirm the machine sails on.

Drives: the odometer check

SMART tells the truth listings do not: power-on hours, reallocated sectors, media errors — and for SSDs, percentage-used/wear-leveling. Enterprise drives run fine to 40–50k hours, but zero-hour drives at wear 3% and 55k-hour drives at wear 80% should be priced like what they are. Anything with pending or reallocated sectors on arrival: replacement conversation, immediately.

Firmware: current or scheduled

BMC, BIOS, controller and NIC firmware at or near current is the mark of a real refurb process. Old firmware is not a defect — it is an afternoon — but it tells you how much of the seller's checklist was aspirational.

Burn-in: one honest day

A memory test overnight (the platform's built-in diagnostics or memtest), then a few hours of combined CPU+disk load while watching BMC temperatures. What you are hunting is the intermittent: the DIMM that errors warm, the fan that surges, the drive that resets under sustained writes. Cheap to find this week; expensive to find in month eleven.

Red flags worth a return

Mismatched or missing caddy screws and rails, SEL history someone tried to hide by clearing after shipping damage events, DIMM slots that lose capacity between boots, single-PSU units sold as dual, and any seller who cannot say what their refurbishment process actually includes. The good ones can — and their machines pass this whole checklist before lunch.