We are booking in new cases · Weekdays, 9am–5:30pm Urgent case? Call 0800 6890668
MHDR Maidenhead Data Recovery 0800 6890668 Send it in
MHDR / Specialist cases / RAID 5 array rebuilds

Array recovery · RAID 5 sets

RAID 5 recovery in Maidenhead. Parity covers one lost disk, never the second.

Most often we see very few variations in how the RAID 5 sets that come to us from Maidenhead fail. Usually one unit goes down and there is no buffer left, so the rebuild begins and at some point, while the rebuild process is still reading, a second disk stops working. It may seem like there was little room to lose, but "failed" is an opinion of the controller, not a diagnosis. And all items sent back are taken directly from sector-by-sector images, not the original disks.

Only pay if we recover your data, on most jobs Diagnosis free, then a fixed quote Get a quote, then post it in from Windsor, Slough or anywhere else

Describe the fault to an engineer
0800 6890668

RAID 5 symptoms, and what each one means.

None of these? Run the triage →
The symptomThe likely causeStart here
There has been one disk failure and the RAID system is currently running in degraded modeCurrently there is nothing protecting the data; each additional hour spent waiting for something to happen could result in lost dataTurn off the power supply before the second disk also fails
Started rebuilding, and never finishedEnd-to-end read detected weak surfaces on a second old disk which is how most of these eventually failDon't start it again, let it alone
The card reports that two units are flagged as failedAlmost always one disk completely failed and another had a small number of unreadable sectors; not two failuresFirst receive a quote, then send all disks; imaging will settle it once and for all
The dropped disks were forcibly returned to the setNow there are old (stale) blocks in the middle of active (live) stripesDo not continue any further
After installing a new RAID card, the array disappearedThe metadata for the RAID disagrees, but all of the information stored on the RAID is intactNormal reconstruction activities
Recovery software has already been applied to each memberEach original hard disk has been written toFirst, get a quote, and then send them in, and let us know what you used to repair your RAID
Where to send it: first get a quote, by phone or with the form, then send it using a tracked, insured service to the Guildford intake laboratory, or drop it off there yourself (Monday to Friday, 9am to 5:30pm, no appointment needed). We will cover the cost of the postage back to you. From a laptop, PC or Mac, we just require the drive out of your computer; storage soldered to the board (T2 and Apple silicon Macs, soldered laptop flash) is not taken on, and if you're unsure as to type, call us first. From a server or RAID set, just send the individual drives, all clearly labelled by bay. From a NAS, send the disks, each labelled by bay, if they come out; otherwise, send the entire NAS unit. A single external drive comes whole, with its cable and power supply. If you wish to review your packaging before posting, contact an engineer who will walk you through the process over the phone. All steps involved with the postage process are listed on the postage page.

Why a RAID 5 rebuild is one long read.

The price of a rebuildA rebuild does not have a gentle option. Every single remaining hard disk will be scanned from the very beginning through to the end, some 48TB in a four-bay set using 16TB members. And every single remaining hard disk has operated for the exact amount of time as the one that has recently died.
The URE figure, in contextThe consumer specification is one unreadable sector every 1014 bits. This is roughly one in every 12.5 terabytes. Enterprise disks are rated at 1015, ten times better than that. The specification is how bad the manufacturer will allow a disk to be and does not describe a clock running down. Many disks outlive the specified figure comfortably. The specifications are accurate. However, the graphs drawn from them can often be salesmanship.
What the card does nextIt is age which determines whether or not a card will carry on with a rebuild after a failed read. Older cards will give up on a rebuild at the first failed read. Newer PERC and MegaRAID firmware punctures the affected stripe and continues on. mdadm also notes where there is a problem with a block (bad-block list) and carries out the same actions. It would make sense to sacrifice one stripe to save all data if only one sector was bad. It would not make sense to sacrifice an entire volume just because one bad sector exists.
Why the next one failsAll members were built in the same batch and have been powered up exactly the same number of times. The members are located inside the same warm chassis. When a single large read is applied across all members, the one that had quietly deteriorated stops working. We read off the same surfaces but do so slowly, leaving the fragile parts until last.

What happens between arrival and return.

01

Booked in, then assessed at no cost Free

Nothing is chargeable until you say yes. Every item gets a case number of its own immediately upon receipt, an engineer determines what has failed, and you get a clear picture of which files have a real possibility of being recovered. The single quote that follows is fixed, written down and does not cost anything.

Diagnosis at no costOne fixed quote, in writingNothing signed
02

First job: image every disk

Each disk is copied in its entirety into our own storage space, including those the controller had identified as "bad". Troublesome areas are documented as the copy process runs along, poor performing surfaces are held back for a final pass over those areas and at no point is any data written to the media you sent.

Every member clonedEven the failed members
03

Find the geometry

The three pieces of information regarding member order, stripe size and parity rotation all have two sources: the metadata that the controller placed on each disk as well as the layout of the data itself. No assumptions are made, and no single surviving working card is required.

Member order and stripe establishedVerified before reassembly
04

Reassemble the set virtually

When the array is rebuilt across the clones, every block will come from the image that can recover it best. Where there are blocks that cannot be recovered cleanly from any of the members, they will be filled in with parity information. This allows the data held only by a failed or "condemned" drive to come back. After the array has been rebuilt, the file system residing on top of it is repaired and then mounted for testing.

Each block from the clearest imageMounted, then read back
05

Opened, verified, and sent home

You don't get charged anything until you've confirmed and authorised that everything is fine after seeing the entire list of what was recovered. Any recovered items are returned via brand-new storage media and the postage costs for these are covered by us. We do not consider a project complete here until we receive confirmation that you have been able to successfully access and review every item you received back.

You approve the list firstReturned on fresh mediaThe return post is ours

What tends to show up first

  • Degraded means that the "margin" is already gone: RAID 5 has used up the one member failure that parity allows it to recover from (RAID 5 is able to continue operating after the loss of one member). With one member missing, RAID 5 now operates exactly like an unprotected disk.
  • Alarm is a sales technique: RAID arrays cost the most to recover in this trade, so writers frequently try to create fear. Instead we report: you receive the results of imaging as soon as they happen, and we never handle anything except clones.
  • A punctured stripe is a small wound: when new RAID firmware encounters an unreadable sector, it flags that one stripe, logs it and continues on (as opposed to older cards, which would abandon the entire rebuild process). The reconstruction process identifies each stripe that was affected.
  • mdadm creates its own history: the superblock and the bad-block log provide information on which member left the set first, and in what order. Hardware cards do not even begin to approach this level of detail.

Work the numbers through once, and then let them go: the consumer specifications permit only a single URE (unrecoverable read error) per 1014 bits, which amounts to approximately one bad sector every 12.5TB read. Enterprise disks are rated at 1015. In theory, three good disks out of a four-disk set of 16TB disks must give up approximately 48TB in a single sitting, so the arithmetic is not imaginary. In practice, a great many drives run well past their specification. While neither of these observations negates the other, the result is still the same: clone everything at the start of your operation, and a rebuild can't harm you.

From the casebook, lately.

MH · MHD-2026-6822ON RECORD ✓

An accountancy practice lost its RAID 5 a week from year-end

The fact that the second disk dropped out in June did not get noticed. So, by the time the fourth disk dropped out during a rebuild in August, it looked like there were two failures. However, imaging identified nineteen bad sectors on the fourth drive, no worse than that, and therefore the remaining three good disks (plus the almost complete fourth) formed the stripe again. The customers made their deadline.

19 bad sectors, not two dead drives6 days from arrival

Before it comes to the lab.

Do these now

  • Once a disk is reported as dead turn off the entire system
  • Write the "bay" number for each disk you pull out before you take them out
  • First get a quote and then send all the disks, the failed disks above all
  • Record the controller type/model and everything else you have tried up until now

Things best not done

  • Starting or restarting a rebuild on an existing array that is already degraded
  • Forcing the card to bring back a removed disk
  • Running consumer recovery software across your members
  • Putting in a new disk and allowing the card to initialise the array

The questions that arise most often.

Two drives have failed in one RAID 5. Is anything recoverable?

Generally yes. There is a big difference between a disk that the card has removed and a disk that has simply stopped spinning. When the card ejects a disk, most times it's because there are only a handful of sectors that won't read back. Since everything is being constructed using clones, if one of the members refuses to give up something, we take it from the remainder using their parity. It's quite rare for there to be two completely destroyed disks; however, two disks that are being evicted arrive at least once per week.

The array is already degraded. Can we safely rebuild it?

Only once all members have been cloned. When a rebuild occurs, each remaining disk is asked to perform an entire end-to-end read (which finishes off a disk nearing the end), and then writes new parity across those exact stripes needed for a reconstruction. Once you have your clones, go ahead and rebuild as much as you want.

URE — how much weight should the figure carry?

Unrecoverable read errors (UREs) occur when there is a sector whose contents can't be returned by the drive after its built-in error correction has failed. The published rate of URE occurrence for consumer drives is approximately one in every 12.5TB read. Enterprise drives are quoted at one in 125TB. These numbers represent a floor rather than a forecast of what can happen, and modern firmware will answer a URE by writing off one stripe, rather than failing to create a usable copy of the array.

Must the original RAID card come with the disks?

No. The metadata from the controller describing the configuration of each member of the array and how they fit together is held on each member. If this record has become damaged, the layout of the data itself gives the same answer. Assembly back into the original configuration is accomplished on the workbench using images of the original data; the disks you provided are only ever read.

More of the jobs we've closed.

Servers & RAID sets → Data off servers → No drive detected → Media we work with →

Switched off, it keeps every chance it has.

Every additional power-up asks for even more of a drive that is already starting to fail. First open a case with us, and let the free diagnosis tell you how much remains.

0800 6890668

Ask for a free diagnosis Posting it to us