Pooled NAS storage · job record · ADR-2025-2891
One Member Clicked, and Both Disks Went With It.
There were two hard drives in one big pool, and they were being managed by a FreeNAS server that had worked without much maintenance for a long time. The first indication that things weren’t right came as an unusual sound. He heard a clicking sound
before the volume went silent. When he said 1 drive is causing the problem
(he accurately identified the issue) and described what happened next, he also got that description correct, because when two disks form a single pool across each other, the moment either one fails it makes both drives inoperative
. After that came the sentence that decides how carefully everything after it is handled. He didn't back it up
.
Same story on yours? Tell us what it does — the look is free.
0203 0868633
What failed, and why.
You cannot treat two disks as separate units. FreeNAS creates one large storage system across multiple disks. This makes what's stored on each disk a portion rather than a full set, so a failure of one disk puts the entire pool outside of reach. This isn’t two simultaneous failures but normal behaviour of a pool. The reason the box stated nothing available was because even though only one disk had issues with it, it required access to both disks to function properly. The click narrowed down the diagnosis process. Clicks usually indicate mechanical problems with the heads of the disk instead of software problems. Also, each additional start-up erodes what’s left of the usable surface area of the disk. Therefore, the unit was not turned on again. At that point, the temptation is to replace the bad disk with a good one and allow the FreeNAS box to resilver. Resilvering writes brand new metadata directly over the old data that a rebuild depends on reading. Anything with a second disk in it costs from £500 + VAT on our published prices. Diagnosis costs nothing.
What the bench did with it.
The order of the job →| Equipment used | Its work on this job | The reason for it |
|---|---|---|
| DeepSpar Disk Imager 4 | Copied both disks while using a write-blocker, and only copied the disk that clicked after replacing its heads | Mapped the heads first, then imaged one face of a platter per pass, with every reset, timeout and the power itself kept under our own control |
| Atola TaskForce 2 | The two disks, the sound one and the repaired one, were read in parallel rather than sequentially | Copies every member of a set at once, so none sit waiting for another |
| UFS Explorer RAID Recovery | The pool's own labels were read off the images, and the shares were built from copies of those images | Recovers the stripe order, then reverse engineers the volume manager the NAS layered above it |
What the bench did with it.
Neither disk was started in the box again
Assessments were conducted on both disks prior to scanning them for their respective data. Each was placed on a bench supply to profile its power draw, and each had firmware queried. Health statistics for each disk were read off in succession. The clicking member behaved exactly as he had described; the other one functioned as it should. There was no intention of putting anything back into the FreeNAS to test behaviour when one member of a pool could not be located. When a pool cannot locate one of its members, the pool begins resolving the matter automatically.
Heads first
The faulty member was opened up and examined using a stereo microscope to determine the condition of the heads and platters. Although the surfaces of the disk were still intact, the head assembly did not survive, which is typically the reason why you hear continuous clicking. A matching replacement head assembly was placed into the faulty member. After that point, the disk was only ever used for reading purposes. It was never asked to be part of a volume again and nothing was ever written to it during the course of work.
Rebuild the pool away from the box
Each of the two images was sent to a software tool capable of understanding how the particular type of pool being addressed is configured. Using those images, the pool was reconstructed from duplicate images. The configuration was not inferred from assumptions. Roles within the pool and labels created by the pool itself were read directly from information provided by each disk regarding its configuration. If an assumption-based approach were taken, the resulting volume would appear to mount correctly, yet contain incorrect data. Each item within the pool was verified individually (file-by-file), before the pool’s contents were considered “recovered.”
What went home.
All data contained within the original pool was copied to new media and inspected before being sent home. At the beginning of our work together he clearly stated that none of these files existed anywhere else, and that statement has a huge impact on how jobs such as this are performed: one attempted recovery, done appropriately, without attempting anything on the originals. The box itself was never part of the answer.
Begin here if yours is doing the same thing.
Other RAID & NAS work on the log.
Yours doing something similar? Power it down first.
Published prices are fixed. Memory cards and USB sticks are £250 + VAT. Any stand-alone drive is £300 + VAT. Drive configurations containing multiple disks (RAID, NAS or server) start at £500 + VAT. First assessments are always complimentary. Thereafter, data recovery charges are also fixed in advance and in writing, so you'll always know how much you owe before paying any money. Shut it down, get a quote first (by phone or with the form), then send it in or drop it off at our Guildford intake laboratory, Monday to Friday, 9am to 5:30pm (no appointment needed). On most jobs you only pay if we recover your data.