Open, and logging new jobs · weekdays, 9am–5:30pm Phone is quicker: 0800 6890668
HDR Hull Data Recovery 0800 6890668 Price it up
HDR / The work we take on / Stalled RAID 5 rebuilds

RAID 5 · a single parity disk

RAID 5 recovery, Hull. One dead disk survives; the rebuild is where you lose the next.

It happens quite frequently. A member leaves. No one sees the alert. A spare is installed many months later. And the rebuild process is halted after the next disk fails. Sets arrive from freight offices on Hedon Road, accountants and lawyers in the Old Town, and workshops on Sutton Fields, all of whom have a Hull address. Typically, the news is more positive than the front panel makes it appear. "Failed" is determined by the decision of the card itself, and not based upon a diagnostic test. What is returned to you is read from an image. Your own disks remain untouched.

Only pay if we recover your data, on most jobs Free to look at, then one fixed price Parcels come in from Beverley, Grimsby and Skegness

An engineer will take the details
0800 6890668

Putting a name to the symptom.

Nothing similar? Go to the triage →
What you seeWhat the fault isWhat to try
Card states one member missingParity has been exhausted. The loss of one additional disk will result in the loss of the entire volumeTurn it off immediately, before losing another
The rebuild process had run for hours and stopped.The full read operation identified a weakened area on another disk of similar age. This is typically how it ends.Leave it turned off. Do not attempt to restart.
Card shows two members as failedTypically, this represents a single dead disk along with a disk containing some unreadable sectors. It does not represent a matched pair.Obtain a quote prior to sending each disk (labelled per bay). The images will allow us to identify which disk is which.
A previously ejected member has been manually brought back online.Old data blocks are being intermixed with current stripes.Do not perform any other actions.
A replacement card cannot detect the existing arrayThe disagreement concerns metadata. The files are located on the system exactly as before.This is a simple reassembly job.
Somebody has already run some type of file recovery program on the drives.The original disks have been written to.Get a quote before you send them anyway. Identify the tool that was used.
Send it to us: get a quote first, either over the telephone or via our online quote form. Then carefully package it to prevent damage in transit, insure it for its full value and post the package using a trackable service to the Leeds intake laboratory; alternatively you may drop it off there (Monday to Friday, 9am to 5:30pm, no appointment needed). For laptops, PCs, Macs and recorders, we require only the drive unit itself (storage soldered to the board is not taken on; if you're not sure which kind you have, ring us first); for RAID arrays and servers, we just need the individual disks, labelled by their respective bays; for a NAS, send the drives if you can remove them, but if not, please include the entire NAS device. A single external drive goes whole, with its cable and power adapter. We pay for return postage. If you would like an engineer to inspect the packaging prior to taping the box closed, please give us a call first. Further information is available on our sending page.

What a RAID 5 rebuild demands of every disk left.

The reading a rebuild has to doThere is but one method to rebuild, and it is flat out. Every surviving member of the array gets read completely (call this 48TB where there are three 16TB disks in a four-bay enclosure), and they've all experienced the same amount of time running hot as the one that failed.
Reading the URE ratingConsumer disks can be defined at about one unreadable sector for every 1014 bits. That's roughly one for every 12.5 terabytes. Enterprise disks can be defined at 1015, or ten times the margin. This isn't a countdown, but rather the floor the manufacturer will stand behind. And many disks still function well beyond this number. Nothing is wrong with the number. What is wrong is the scare graphs and charts that were created around this number.
How the card respondsThat depends on how old the card is. Older cards can throw the entire rebuild away if they hit an unfinishable read anywhere along the line. Modern MegaRAID and PERC firmware will puncture the stripe and continue. mdadm will mark the offending block as "bad", but it will also allow the rebuild to proceed. To lose a stripe due to one bad sector is a good trade-off. Losing the volume never is.
Why the next disk goesAll of the disks came in one box and went into one server. They have been sitting in the same environment at the same temperature ever since. If you ask all of the disks to do one large read, then whatever disk is quietly failing will stop. Instead we're going to do a series of smaller reads, and move to areas that have weak spots later in the process.

From the box arriving to the files going back.

01

A case number first, then a free diagnosis Free

Every piece of gear that shows up receives a case number on the very same day it arrives. An engineer examines it and determines what is really wrong with it and there is no charge for this. We give you straight language about which files we will be able to copy from the disk and which ones we won't be able to. Then we provide you with a single price quote, in writing, based upon these findings. We don't begin working on anything until you agree to the quoted price.

The diagnosis is freeOne price, put in writingNothing to pay yet
02

Image every disk first

We create a copy of each disk in its entirety onto our storage environment. That includes the failed disks too. Any errors encountered while reading from these disks are logged as soon as they occur. After that point, we put off completing the worst parts of the array until last. Nothing at all gets written back to your original media.

Each disk clonedIncluding the write-offs
03

Work the layout out from the disks

Two things reveal the disk order, the stripe size and which direction parity travels: the metadata each device contains, and the pattern created by the data. None of these are guesses. None of them require your controller.

Disk order and stripe size fixedVerified before assembly starts
04

Build the volume over the images

The members were put into a set using the images. Wherever each block was most readable to an image, it came from that image. If they all failed to find a readable block, parity supplied the missing piece; thus files stranded on a failing disk become available again. Then the file system is repaired, and the volume is mounted to verify everything is okay.

Every block from its best imageMounted and checked
05

You sign it off, then it comes back

You will not be charged while you think about this. First, you receive a list of everything that was recovered from the drive. Once you have seen the list and given us the go-ahead, the bill comes next. Files are returned to you on media purchased specifically for this job and we cover the return postage as well. Your job is kept open by us until you can view them on your personal computer.

You approve the listing firstCopied onto new mediaReturn postage at our cost

What turns up most often

  • Degraded simply means it has already happened: RAID 5 maintains one disk's worth of redundancy. Once a member fails, it uses up the redundant space. At this point, the RAID volume is no stronger than a single drive alone.
  • Your fear sells: arrays are the most expensive items on the bench, and much of what you read about them is meant to scare you. We would rather read out exactly what our imager finds as long as your disks remain stationary.
  • A puncture loses only a stripe, and not the entire array: newer versions of the firmware now mark an unreadable sector within a stripe when they encounter it, write down the information, and continue on. Older cards would stop rebuilding when they encountered an unreadable sector. Once the array is rebuilt, striped blocks with punctures can be identified by their address.
  • mdadm logs history: between the superblock(s) and the bad-block log, you should be able to determine which member was removed from a Linux-based software RAID, and when. A hardware controller typically tells you very little about what occurred.

Calculate it just once and then stop: the manufacturer's published rating for a consumer disk is one unrecoverable read error (URE) for every 1014 bits it reads. So, approximately one bad sector every 12.5TB. Enterprise disks have a rating of 1015. Rebuilding a four-bay chassis with 16TB drives requires a total of 48TB from the three survivor disks without errors during the process, so the risk is very real and not theoretical. There are many disks that remain operational past their ratings. Both points clearly illustrate the same thing: image first, and the rebuild after that costs nothing.

One case from the casebook.

HU · HUL-2026-0822LOGGED ✓

A RAID 5 down at an accountancy firm, year-end a week away

Disk 2 went offline in June and nobody took a look. Disk 4 went offline during a rebuild in August and appeared from the outside as two dead disks. They weren't. There were only 19 bad sectors on disk 4, so the remaining 3 good disks, along with the nearly good disk, were sufficient to rebuild the stripe. The end-of-year filing was completed on time.

19 bad sectors, and no second failure6 days after it landed

Before you box it up.

Do these first

  • As soon as a disk goes offline shut the box down!
  • When pulling a disk, label the disk with the correct bay number.
  • Obtain a quote first, then send all members including dead ones.
  • Identify which controller is involved and what has already been attempted.

What not to do

  • Begin or restart a rebuild on a RAID set that is missing a disk
  • Bring a previously rejected disk back online via the card utility.
  • Run off-the-shelf recovery software against the disks.
  • Install a new disk into the system and allow the card to initialise the set.

Asked most weeks, answered here.

The card has flagged two of the disks. Is the volume gone?

Not usually. The card can remove a disk based on how slow individual sectors respond. This does not necessarily mean there is a dead drive. When all disks are copied to our internal storage, if there is something that a single copy cannot determine, it is built from parity held on other disks. Losing 2 disks because a controller thinks they are bad is normal. Losing 2 disks that are actually dead is not.

We are running degraded with a member missing. Rebuild or not?

Only when the disks have been imaged. All disks will be forced to "read" themselves completely during rebuild, which can finish off any disk already close to being finished. Also, rebuilding the disks will overwrite the parity that was intended to restore the volume to its original state. When the images have been taken, rebuild all you want; the images do not change.

How much do the URE figures matter?

One URE represents one sector that could not be corrected using the built-in error correction mechanism within the drive and therefore was abandoned. The published specification indicates approximately 1 URE for every 12.5TB on consumer disks and 1 URE for every 125 TB on enterprise class disks. Read this as the worst figure that the manufacturer will support, not a countdown. Currently, if a URE occurs, the firmware will simply puncture that specific stripe and continue reading.

Does the controller need to come as well?

No. Your disks contain everything needed to configure your RAID. Wherever damage has occurred to these configurations, the data will reveal the layout to us. We perform rebuilds in software using the images we take. The disks you send are only read and never written.

What people read after this.

Multi-disk systems → A stopped server → Not seen by the PC → Everything we take in →

A drive left switched off loses nothing more.

Leave it off. There are only a few start attempts left in your failing drive and each attempt uses up a little more of what remains. Our complimentary diagnosis will determine what files are readable.

0800 6890668

Have it looked at How to post it to us