RAID 5 · single-parity arrays
We see the same story most weeks. A disk drops out, the alert goes unread, a replacement goes in, and the rebuild stalls when the next member will not read. These sets come from Digbeth edit suites and Colmore offices across Birmingham, and Black Country machine shops. The position is better than it looks. Failed is the card's opinion, not a post-mortem, and what we hand back comes off a full image, never off the disk you sent.
Tell an engineer what it is doing
0800 6890668
| What you see | What is behind it | What to do |
|---|---|---|
| The card reports one member missing from the set | Parity is spent — the next fault takes the volume with it | Shut it down before another goes |
| The rebuild ran for hours and then stopped | A full-surface read has found the weak patch on the next disk of the same age — the usual ending | Leave it alone. Do not try again |
| Two of the disks now carry a failed flag | In nearly every case that is one dead drive plus one with a few bad sectors, not two dead drives | Send every disk — the images tell us which |
| An ejected member was pushed back online | Old data is now mixed into current stripes | Stop there now |
| A replacement controller cannot see the array | The disagreement is in the metadata; your data has not moved | Ordinary reassembly job |
| Someone has already pointed a recovery tool at the disks | Writes have gone onto the original members | Post them, and name the software used |
Every item gets a case number of its own the day it lands. An engineer then works out what has actually failed and tells you plainly which files have a real chance of coming back and which do not. Only then does a price follow: one figure, fixed, in writing, and it costs you nothing to see it. Nothing is charged until you say yes.
Every member is copied end to end onto our hardware, including the ones the controller had written off. Bad areas get mapped while the copy runs and the worst of them are left for a last pass. Your disks stay read-only throughout.
Two sources settle the disk order, the stripe size and the direction parity rotates: what the card wrote onto each member, and the shape of the data itself. We guess at none of it, and we do not need the controller.
The array is assembled across the images. Each block comes from the copy that read it best, and where every copy came back short, parity supplies the missing piece — that is how data stranded on a condemned member survives. The file system is then mended and mounted for checking.
No invoice is raised until the full list of recovered files has sat in front of you and you have said go ahead. Your data comes back on media we buy new, posted at our expense, and the job is not closed here until every file has opened on your own machine.
Do the sum once and then stop worrying at it: a consumer disk carries a published limit of one unrecoverable read error, a URE, per 1014 bits — near enough one bad sector in every 12.5TB. The equivalent enterprise figure is 1015. Rebuild a four-bay set of 16TB disks and the three survivors have to hand over about 48TB without stopping, so the figure is not academic. Plenty of drives also beat their rating by a wide margin. Both things are true, and both point the same way: clone first, and a rebuild can no longer cost you anything.
June was when disk two quietly dropped out, and nobody saw it go. August was when disk four gave up mid-rebuild, which made it look like two had died at once. It had not. Imaging turned up nineteen bad sectors on that fourth disk and nothing else, so three clean members and one nearly clean fourth rebuilt the stripe. The filings went in on time.
Usually not. A controller marks a disk failed on its own terms, and that verdict often rests on a few sectors that time out rather than on a dead drive. Once every member sits on our storage as an image, anything one clone cannot supply is worked out from the parity held by the others. Genuinely dead pairs are rare. Pairs the card has thrown out are ordinary.
Not before the disks are imaged. Rebuilding drags every remaining member from first sector to last, and that is the exact demand that kills a disk already struggling. It also overwrites parity on the stripes we would otherwise read to piece the volume together. With images on the bench, rebuild as many times as you want.
A URE is a sector the drive gives up on after its own correction code has run out of options. The published rate on consumer disks works out at about one per 12.5TB read, and on enterprise disks one per 125TB. Read them as the worst the manufacturer will stand behind, not as a countdown. Current firmware treats a URE by puncturing the one stripe and keeping the rebuild alive.
No. Every member carries the card's own record of how the set was laid out, and where that record is damaged the layout can be read back out of the data itself. The set goes together in software over the images. Your disks are only ever read from.
Every extra power-up takes something off a drive that is already going. Leave it switched off and let the free diagnosis say what still reads.