Buying used enterprise storage is one of those things that makes perfect sense when you look at the price and becomes more complicated once you actually put the hardware into a machine.Enterprise HDDs that originally cost a small fortune can appear on the used market for surprisingly little money.For a homelab, where storage capacity is always useful and budgets are rarely unlimited, it is difficult to ignore those prices.
I recently bought a batch of 8 used 2 TB Seagate 3.5-inch SAS hard drives and one HGST Ultrastar DC SN200 after finding a seller offering them cheaply.The seller claimed that the SAS drives were spare parts and had barely been used apart from testing their functionality.For a homelab, that sounded like exactly the kind of deal worth taking a chance on.
Two drives had a non-standard sector format The drives had previously been used with hardware RAID Close Two of the drives appeared to have previously been part of a hardware RAID system.They had been formatted with a non-standard sector size, using 520-byte or 528-byte sectors rather than the 512-byte sectors that a conventional HBA expects.This is common with some enterprise storage systems.
The additional bytes can be used by the RAID controller for metadata or checksum information while the storage system presents a conventional logical sector size to the operating system.Related I ran the same SSD health check on Linux and Windows—only one gave me the full picture A four-year-old SSD, 6,605 unexplained errors, and one OS that actually had answers.Posts 45 By Dibakar Ghosh That configuration became a problem when I connected the drives directly to my HBA.
They had worked in the storage environment they were originally configured for, but they would not work correctly in the direct-attached, software-managed setup I was building.The solution was a low-level format that returned the drives to a standard sector size.I used sg_format, which is part of the sg3_utils package, to perform the reformatting.
The process is much slower than a normal filesystem format.Depending on the drive, the low-level format takes roughly 2.5 to 3 hours per TB.With large enterprise disks, that can turn what initially looks like a simple compatibility problem into an entire afternoon of waiting.
Once the formatting completed, the drives could be used normally through the HBA.The particular SN200 drives were a different experience Low power-on time didn’t make them easy to use I have always been more cautious about buying secondhand flash storage, and my experience with Western Digital’s Ultrastar SN200 series did little to change that.The particular SN200 drive I bought was a U.2 NVMe device.
The seller claimed that the flash storage had been powered on for fewer than 12 days, which initially sounded excellent, but I tried using it with Debian.It couldn’t properly recognize the device, although smartctl could communicate with it.The drive identified itself as a Cisco-rebranded device using customized firmware.
Getting the drive working became considerably more complicated than the power-on figure suggested.I eventually had to deal with firmware, vendor-specific tools, different operating systems, and driver issues before getting the device into a usable state.Even then, the problem wasn’t completely resolved.
During later testing, the drive could return to Diagnostic Mode after multiple restarts, with the data on the device becoming unavailable.That made it unsuitable for the caching workload I had planned.Related Please stop putting desktop hard drives in your NAS Don't start your NAS journey off on the wrong foot.
Posts 51 By Patrick Campanale The seller had listed the drive as non-refundable, but after I explained the problem, they agreed to accept the return.I wouldn’t take that experience as evidence that SN200 drives are inherently bad.There was a problem with the particular drives I bought, but it did show how little a power-on-hour figure can tell you about a secondhand enterprise NVMe device.
The real cost is the testing Cheap drives require time before they earn your trust The biggest lesson from buying used enterprise storage is that the purchase doesn’t end when the package arrives.Testing is part of the cost, for HDDs, I want to inspect SMART information, verify firmware, run full-disk writes and reads, watch temperatures, and look at latency under sustained workloads.If I’m buying several drives, I want to test them individually before putting them into an array.
That matters even more with large ZFS pools.A failed disk means a replacement and a resilver, and a long resilver puts sustained load on the remaining drives.If those drives are also old, one failure can expose another weak disk at exactly the wrong time.
Related Half of failed hard drives die in year one—here is the one thing you must do before trusting yours Your new hard drive is hiding a fatal flaw, and SMART data won't save you Posts 25 By Monica J.White ZFS provides checksums, redundancy, scrubs, and excellent visibility into storage problems, but none of those features makes questionable hardware reliable.Redundancy gives you room to recover from failures.
It doesn’t remove the possibility of those failures.A backup remains necessary regardless of how expensive or reliable the drives are.I still think used enterprise HDDs are worth buying My experience didn’t convince me to stop buying used enterprise HDDs.
In fact, the batch I bought turned out to be a pretty good deal after extensive testing.The drives except the 2 used in hardware RAID had low power-on hours, little recorded write activity, and no significant abnormalities during repeated tests.The SN200 experience was much less pleasant, but it also demonstrated why I treat used enterprise NVMe differently from used HDDs.
A drive can have very little recorded use and still present firmware, compatibility, or operational problems that aren’t obvious from its health statistics.That is the real risk with secondhand enterprise storage.You aren’t simply buying a drive at a discount.
You’re buying hardware with an unknown history and taking responsibility for establishing whether it is suitable for your system.
Read More
