Skip to content

Dear customers! Please send orders and inquiries to [email protected] or to Telegram @servix_uz instead of calling by phone: due to the high workload, the operator cannot answer all incoming calls.

Servix

Used server: what to check when buying — acceptance checklist

Servix technical team · Updated

In brief. Before paying, check the serial number on the manufacturer's website and read the server logs: the IML in iLO for HPE, the SEL and Lifecycle Log in iDRAC for Dell. At acceptance, take the SMART readings and the remaining SSD endurance, the RAID cache battery status, firmware versions and the management license type, and check power supply redundancy and the fans. Then put the server under full load for at least 24 hours, and for production for 48–⁠72 hours, and open the logs again: there should be no new errors.

This article is the acceptance procedure for a used or refurbished HPE ProLiant Gen10 or Dell PowerEdge 14G–15G server, bought from Servix or any other seller. It is for the administrator who will receive the box and for the manager who signs the acceptance report. Whether to buy a refurbished server at all is covered in the articles refurbished server or new and where savings on refurbished are justified. Here the decision has already been made, and the server is on the desk.

“Used” and “refurbished”: what they mean at Servix

“Used” refers to origin: the server has already worked somewhere. “Refurbished” refers to condition: a previously used server that the seller has checked, tested and provides a warranty for. A used server from a private seller without any checks is “as is.”

In the Servix catalog, every product has a condition field: new, refurbished or used. All 90 previously used servers from the selection guide used and refurbished servers are published as refurbished: the listing says “Condition: refurbished,” the answers to questions say “checked and tested,” warranty 12 months, serviced by Servix. There are no “as is” items in the catalog. All 90 are made to order, 2–⁠3 weeks.

The listings show what to take into account before ordering:

  • in 89 configurations out of 90, drives are not installed and are selected for the task;
  • 26 have a single power supply, and the listing advises buying a second one for production;
  • 89 specify the basic management level — iLO 5 Standard or iDRAC 9 Express;
  • 13 rack configurations have “Rack rails: no” in their specifications.

Acceptance is needed for a refurbished server too: the seller checked the machine on their side, while you make sure it arrived intact and works in your rack, and record its initial condition.

Acceptance checklist: summary table

What to check

Where to look

Warning sign

Serial number

Label on the chassis, iLO or iDRAC, HPE or Dell website

The number on the chassis and in iLO/iDRAC differ; the seller does not give the number before payment

Original configuration

HPE PartSurfer; Dell — Original Configuration by Service Tag

Components differ from those declared without explanation

Completeness

Inspection, specifications in the listing

No rails, drive trays or second power supply, and this was not stated

Logs

iLO: IML, Event Log, Active Health System; iDRAC: SEL, Lifecycle Log

Repeated errors of one component, overheating, recent clearing

Memory

iLO: System Information → Memory; IML, SEL

Correctable error threshold exceeded on a module; for Dell — MEM0701, MEM0702

HDD

SMART via RAID controller tools

Growing counters of reallocated sectors and defects

SSD

Remaining write endurance in iDRAC, HPE SSA, iLO; Percentage Used for NVMe

The endurance will not cover your write volume

RAID cache battery

iLO: Smart Storage Energy Pack; iDRAC: Storage → controller

Failed, Predicted failure, volumes in write-through

Firmware

iLO: Installed Firmware; iDRAC: firmware inventory

BIOS and BMC are years behind

Management license

iLO: Administration → Licensing; iDRAC: Configuration → Licenses

Advanced or Enterprise was promised, but Standard or Express is installed

Fans and power

iLO: Power & Thermal; iDRAC: Cooling, Power

Degraded; Not Redundant with two power supplies connected

Stress test

Diagnostics, MemTest86, CPU and array load

New errors in the logs, reboots, overheating

1. Before payment: serial number and original configuration

Ask for the serial number (HPE) or Service Tag (Dell) before the invoice. It is used to check the warranty status on the manufacturer's website; details are in the article on warranty and support for HPE and Dell. The factory warranty of a refurbished server has usually already expired, and that is normal. What matters is that the number exists and the model it shows matches the one being sold.

Then — the original configuration. On Dell's support page, the Service Tag gives “View product specs” with an Original Configuration tab: what the server left the factory with. At HPE, PartSurfer plays a similar role — it shows the parts list by serial number if data for that number exists. Memory and processors are often changed to order, but you should be told about this in advance.

At acceptance, compare the label on the chassis with the number in iLO or iDRAC. If they differ, ask whether the system board was replaced.

2. External inspection and completeness

  • Chassis. Dents, signs of a drop, bent mounting “ears,” corrosion on connectors.
  • Rails. Servix listings have a “Rack rails” line. Where it says “no,” the rails are ordered separately.
  • Drive trays. An empty bay may be closed with a blank rather than a drive tray, and then a drive cannot be installed. Their number is clarified together with the drives.
  • Power supplies and small items. How many power supplies there are, whether the model and wattage are identical; whether power cables and PCIe slot blanks are included.

3. iLO and iDRAC logs: the server's history

The management controller logs are the main source of truth about a used server. They are stored in iLO or iDRAC itself, not on the drives, and reinstalling the OS does not erase them.

HPE iLO 5. Integrated Management Log (IML) — system events: fans, power supplies, temperature and emergency shutdowns, drive failures, firmware updates, Smart Storage energy pack status. The iLO Event Log stores the events of iLO itself. The Active Health System Log is detailed telemetry exported as a file.

Dell iDRAC9. The System Event Log (SEL) and Lifecycle Log are in the Maintenance section; the SupportAssist report collects logs and inventory into one archive. All of this is available at any license level, including Basic and Express.

An empty log does not mean a clean history. Clearing the IML and iLO Event Log leaves an entry in the log itself, and clearing the SEL in iDRAC is recorded in the Lifecycle Log with the user name and IP address. Logs are often cleared after repairs, but the seller should tell you what was replaced.

The warning sign is not a single entry from three years ago, but repeated errors of one component, emergency thermal shutdowns, loss of power redundancy, drive failures. Export the logs on acceptance day: the IML to CSV, Active Health System and SupportAssist as files. This is the baseline for the entire warranty period.

Memory: ECC errors

ECC corrects single errors, the server keeps working, and they are easy to miss. In iLO, open System Information → Memory: the status of each module and the Advanced Memory Protection mode. In the IML, look for entries about exceeding the correctable error threshold with the processor and slot number. In iDRAC these are MEM0701 (warning) and MEM0702 (critical) with the text “Correctable memory error rate exceeded.”

If errors repeat in one slot, move the module. If the error moved with the module, replace the module; if it stayed in place, the problem is in the slot or processor, which is more serious.

4. Drives: SMART, power-on hours, SSD wear

At Servix, in 89 refurbished configurations out of 90, drives are not installed and are selected for the task, so you will need to check the drives you choose. Ask right away whether they are new or refurbished.

  • HDD. Power-on time: Power_On_Hours for SATA, Accumulated power on time for SAS; 8,760 hours is a year of non-stop operation. Reallocated sectors (Reallocated_Sector_Ct), pending sectors (Current_Pending_Sector), uncorrectable sectors (Offline_Uncorrectable), and for SAS — Elements in grown defect list. Zeros are good. Non-zero values that grow between two readings mean the drive needs replacing.
  • SSD. The main thing is the remaining write endurance: Remaining Rated Write Endurance in iDRAC, Usage Remaining in the SmartSSD Wear Gauge report in HPE Smart Storage Administrator, Media Life in the Storage section in iLO 5. For NVMe, look at Percentage Used: 100% means the rated endurance has been used up.
  • How to read it. Behind a RAID controller, the OS does not see SMART directly. You need controller tools — SSA or ssacli for HPE, iDRAC or perccli for Dell — or smartctl with support for your controller.

Compare the remaining SSD endurance with your workload: the model's endurance is given in TBW or DWPD, and the daily write volume is visible in the monitoring of your current system.

5. RAID cache battery or capacitor

68 refurbished configurations out of 90 have a controller with a write cache protected against power loss: Dell PERC H730p, H740p, H745 and H755 with a battery, HPE Smart Array P408i-a, P408i and P816i with FBWC flash cache. Software RAID such as Dell S150 and HPE S100i has no cache and no battery.

HPE Gen10. The Smart Array cache is protected by an energy pack — an HPE Smart Storage Battery or a Smart Storage Hybrid Capacitor. On power loss it gives the controller time to save the cache to flash memory; usually one module is enough per server. The status is in iLO on the Power & Thermal → Power page, in the Smart Storage Energy Pack section; events are in the IML.

Dell PERC. If the battery has degraded or failed, the controller itself switches volumes from write-back to write-through: the data is safe, but writing is noticeably slower. About once every 90 days the controller calibrates the battery (Transparent Learn Cycle), and “Battery failed” for a few minutes during calibration is normal. The same error outside calibration means the battery must be replaced; “Predicted failure” means the battery does not hold a charge for 24 hours.

At acceptance you need two facts: the module is in OK status, and write-back is enabled on the volumes. Controller names, cache sizes and cables are covered in the reference guide to Dell PERC and HPE Smart Array RAID controllers.

6. Firmware: SPP and Dell Update Package

Current versions are visible in iLO on the Firmware & OS Software → Installed Firmware page and in the iDRAC firmware inventory. Compare the BIOS, iLO or iDRAC, RAID controller, drives and power supplies with the latest versions from the manufacturer.

  • HPE. For Gen10, the Service Pack for ProLiant (SPP, “Gen10 only” releases starting with 2019.12.0) and System ROM firmware are downloaded with an HPE Passport account without a warranty check. For Gen9 and older, a valid warranty or contract is required — as stated in HPE notice a00094704.
  • Dell. Dell Update Packages (DUP) are taken from the Drivers & Downloads section by Service Tag and installed via iDRAC or Lifecycle Controller, which can load them directly from downloads.dell.com.

Update before the stress test: then the test will check the server with the firmware it will go into production with.

7. iLO Advanced and iDRAC Enterprise licenses

The license type is visible in iLO on the Administration → Licensing page, and in iDRAC under Configuration → Licenses.

  • HPE. Without a license, on rack and tower servers the iLO graphical console works only until the OS boots. An iLO license cannot be transferred to another server.
  • Dell. iDRAC9 Express for rack and tower servers has no virtual console and no virtual media; that is the Enterprise level. A Dell license is perpetual and tied to the Service Tag of one server.

89 of the 90 Servix listings specify iLO 5 Standard or iDRAC 9 Express. If the OS will have to be installed and repaired remotely, include the license in the order. If the seller promises Advanced or Enterprise, check the licenses page before signing the acceptance report.

8. Fans and power supplies

In iLO, the Power & Thermal → Fans page shows the status and speed of each fan in percent; in iDRAC the same is in the Cooling section. At idle with the Optimal Cooling profile, the fans should not be howling at maximum. A fan noticeably faster than its neighbors or a Degraded status means it needs replacing.

For power, look at Power Status in iLO: with two power supplies connected it should say Redundant, while Not Redundant means one power supply is not delivering power. Test under load: pull the cable of one power supply — the server keeps working, and an event about loss of redundancy appears in the log. Plug the cable back in and repeat with the second power supply.

9. Stress test: what to run and for how many hours

Our guideline, not a manufacturer requirement: at least 24 hours of full load, and 48–⁠72 hours for a production node. During this time, overheating, memory faults under load and intermittent power errors show up.

  1. Built-in diagnostics. On Dell: F10 at boot → Lifecycle Controller → Hardware Diagnostics, or F11 → Launch Diagnostics (ePSA); the test runs on all detected devices.
  2. Memory. MemTest86 from a bootable USB drive. By default it does 4 passes, and the developer, PassMark, considers this sufficient for most systems. The pass time depends on the amount of memory.
  3. Processors. stress-ng or mprime on all threads, for several hours in a row.
  4. Disk array. Full volume initialization and a consistency check with the controller tools, then fio with mixed writes and reads.

Watch the temperatures in iLO or iDRAC throughout. The test is passed if there are no new errors in the IML or SEL, the server has not rebooted on its own even once, and the temperatures are normal.

Example: what to check in specific configurations

Configuration (refurbished)

Price, November 2026

What to look at

HPE DL20 Gen10 2LFF, 1×Pentium G5420, 16 GB

9,425,000 UZS

iLO 5 Standard; one 290 W power supply; S100i software RAID; rails — “no”

Dell T150 4LFF, 1×Xeon E-2324G, 16 GB

14,329,000 UZS

Tower; iDRAC 9 Express without virtual console; one 300 W power supply; S150 software RAID

HPE DL380 Gen10 8SFF, 2×Xeon Silver 4215R, 64 GB

price on request

P408i-a, 2 GB FBWC — energy pack status; one 800 W power supply, a second one is needed for production

Dell R740xd 24SFF, 2×Xeon Silver 4215R, 128 GB

47,165,000 UZS

H730p, 2 GB with battery; two 750 W power supplies — redundancy test; drive trays for 24 bays; rails — “no”

Dell R650 10SFF, 2×Xeon Silver 4310, 64 GB

71,883,000 UZS

H755, 8 GB with battery; two hot-swap 800 W power supplies; iDRAC 9 Express

Prices are from the product pages as of November 2026, including 12% VAT. Drives are not included in these configurations.

What the warranty of 12 months gives you

Servix provides a warranty of 12 months on new and refurbished equipment and services it itself. This is a seller's warranty: in case of failure you contact Servix, not HPE or Dell. The factory warranty and vendor contracts are a separate matter; on a refurbished server they have usually already expired.

The logs and SMART readings from day one show the condition in which the server started working for you — acceptance complements the warranty. Delivery in Tashkent is free. Unopened goods can be returned within 10 days at the office — decide on this before opening the box for tests.

Frequently asked questions

How many hours should the stress test run?

At least 24 hours of full load, 48–⁠72 hours for production, with no new errors in the logs.

The seller cleared the logs. Is that bad?

Not necessarily; they are often cleared after repairs. But clearing leaves a trace: an entry in the IML for HPE, an entry in the Lifecycle Log with the user name and IP address for Dell. Ask what was replaced.

What SSD wear is acceptable?

It depends on the workload. Multiply the remaining endurance in percent by the model's TBW and compare it with how much you will write over the server's service life. For infrequent writes, even a drive with half its endurance left will do; for a database with active writes, it is better to take a new one.

Do you need an iLO Advanced or iDRAC Enterprise license?

If you can walk up to the server with a monitor, the basic one is enough. If the OS will have to be installed and restored remotely, you need a console after the OS boots, and for Dell also virtual media — that is a paid license.

Can HPE firmware be updated without a contract?

For Gen10 — yes: the SPP for Gen10 and System ROM are downloaded with an HPE Passport account. For Gen9 and older, a valid warranty or contract is required.

Send to [email protected] your task (1C, virtualization, file server, video surveillance), the number of users or virtual machines, the required drive capacity and budget. We will select a refurbished or new configuration, note what needs to be added — a second power supply, rails, a management license — and send a quote. Prices include 12% VAT, made to order — 2–⁠3 weeks, the price is fixed in the invoice, warranty 12 months.

See also: server drives: HDD or SSD, SAS, SATA or NVMe, drives and RAID controllers for servers: prices and ready-made sets, iLO and iDRAC: remote server management and which license you need.

Related reading