A RAID 5 tolerates one failed member. This one had two drives making contact with their own platters and was still being used, because it still worked — which is exactly what makes that state the most expensive one an array can sit in.
A pharmaceutical company called about noise. Their Synology DiskStation had started making a grinding, scraping sound, files had begun opening with corruption, and the interface was reporting the array as degraded. It was also still serving data, which is why it had been left running while somebody looked into it.
On it were research reports, experiment data, regulatory compliance records and trial results — material that is difficult to reproduce and, in places, impossible.
Two of the five drives had failed mechanically, and both were producing the worst noise a drive makes. Clicking means the heads cannot find their servo data; grinding or scraping means they are in contact with the platter surface, and that causes permanent damage with every rotation.
The array's arithmetic had therefore already run out. RAID 5 survives one lost member by computing it back from parity; with two members failing there is nothing left to compute from where both are unreadable in the same stripe. That the volume was still mounting was not evidence of health — it was the array serving what it could while two disks destroyed themselves.
The RAID metadata was still partly intact, which mattered. The remaining three drives were sound.
The two failing drives were powered down and not run again until they were on the clean-air bench. Each was opened, its damaged head assembly replaced with a matched donor set, and the internal filter changed — a head crash generates particles that circulate and cause further harm if left in place. Where the platters had been scored, that damage was permanent and set the ceiling before any imaging began.
All five drives were then imaged read-only, worst-condition first so the weakest were read while they still responded. The repaired pair gave up what lay outside their scored bands. From those images the RAID 5 was reassembled virtually — geometry derived from the data rather than trusted from the Synology — with parity filling any stripe where only one member was unreadable.
Research documents and spreadsheets were opened in their applications, and databases checked for structural consistency rather than merely copied. Files falling in stripes where both damaged drives were unreadable in the same place were reported individually by name.
Because the material included regulated and confidential records, the work ran under a signed data processing agreement, through one named contact, with all data remaining in the UK and working copies destroyed on an agreed schedule.
About 97% recovered, including the regulatory and trial documentation. The shortfall sat where both damaged drives had scored platters in the same stripes — the honest limit of RAID 5 once two members have physically failed.
The company replaced the array and moved to a configuration with two-disk tolerance. The more useful change was procedural: the first drive had been reported as failed weeks earlier and the notification had gone to an address nobody read.
An array that has lost a member keeps serving data with no margin left, which is precisely what encourages people to carry on using it. If a drive is grinding, every rotation is scoring the surface holding your files — and on a degraded array there is no second copy to fall back on. Power it down, keep bay order, and send every disk. RAID and NAS recovery from £500 +VAT, with a 50% deposit where clean-air work is needed.
Get the device to us for a free diagnostic and a quick note on what went wrong — an engineer looks it over and puts your exact quote in writing before anything is started.
The route to your data starts with the device reaching us. Box it up securely, tuck your contact details inside, and send it across — after the free diagnostic, we put your exact price in writing before a single step is taken.
Sending it by post? Go with a tracked, insured service. Prefer to bring it round? Our door is open Monday to Friday, 9am to 5:30pm — just pack the device as described above beforehand.
Prefer to get a sense of things first? Complete the form with a bit more about the fault and an engineer will look it over and send back a tailored quote.
We’ll get back to you soon. Anything pressing, call 01865 593000.
It is the most expensive decision available. A degraded array has no redundancy left, so a fault on any remaining member means total loss — and if a drive is grinding, continued use is actively scoring the platter. Copy what you can, then stop.
Often, yes, though it depends on how long they ran. Each is repaired with donor heads under clean air and imaged individually, then the array is reassembled from the images. Parity fills stripes where only one member is unreadable; where both are, that data is gone.
Clicking means the heads cannot locate their servo data. Grinding or scraping means they are touching the platter surface, which removes the magnetic coating permanently. It is the noise that costs you the most per minute of runtime.
Yes. Data processing agreements and NDAs are signed as standard, work runs through a single named contact, data remains in the UK, and working copies are destroyed on a retention schedule you set.
From £500 plus VAT after a free 48-hour diagnostic and a written quote. Where members need clean-air work and donor parts there is a 50% deposit, with the balance due only on a successful recovery.
Kick off with an instant online quote, or ring us and talk it through first. Either way you’ll know a clear, fixed price before any work starts.