A failed or stalled RAID rebuild is the single most common cause of permanent data loss we see. The instinct to swap a disk and rebuild again is exactly what destroys the array — here's what to do instead.
A rebuild onto a failing disk, or with the disks in the wrong order, can overwrite the data for good. Stop, power down, and don't change anything until it's assessed.
A rebuild that's stalled, aborted, or left the array refusing to mount is the commonest RAID emergency there is. To see why it's risky, it helps to know how RAID 5 works: your data is striped across the drives, with one rotating block of parity per stripe. That parity is what lets the array survive a single drive failing — the missing data is rebuilt from the surviving drives. But the moment one drive has failed, that safety net is gone. The array is “degraded,” running with no redundancy left — and the rebuild you start to fix it is the most demanding thing the array will ever do. That combination is exactly when a second problem tends to surface.
Nearly every failed rebuild comes down to one of these — and all of them stem from the array having no fault tolerance left while it works flat out.
The rebuild reads every sector of every surviving drive. Those drives are the same age and wear as the one that died, so the sustained load pushes a marginal one over its threshold — and now two drives are down.
To rebuild, the controller must read all the remaining data. Hit a single unreadable sector with no redundancy left, and the stripe can't be reconstructed — so the rebuild aborts.
A healthy consumer drive that pauses too long on a slow sector can be kicked out by the controller as if it had failed, collapsing an array whose drives are actually fine.
A rebuild started with the members in the wrong order, or the wrong settings, writes bad parity over good data — one of the few ways to make a recoverable array unrecoverable.
After a rebuild fails, the first correct action is inaction. A degraded array that dropped during a rebuild is usually very recoverable — but almost every “let me just try one more thing” makes it less so. Each forced retry, each repair, each attempt to bring the array back online risks overwriting the parity and file-system data that recovery depends on, turning a routine job into an impossible one. Power the array down and leave it exactly as it is.
A few careful steps now protect the data and make the recovery cleaner and quicker.
Shut the NAS or server down cleanly and stop using it. The array isn't going to fix itself, and every hour it runs degraded is another chance to lose a drive.
Before you remove anything, mark each drive with its bay or slot number. Drive order is critical to reconstructing the array — get it recorded while it's still known.
Every member, including the failed one, plus a note of which drive dropped first and anything you've already tried. We work from the complete picture.
The key is that we never rebuild on your original drives. Everything happens on read-only copies, in-house.
We check the health of each member and identify which have physical faults and which are fine. You get a clear picture and a fixed, written quote before any chargeable work.
Every drive is cloned individually through a write-blocked connection, so the originals are never touched. A mechanically failed member is repaired first, in clean-air conditions, just enough to image.
We reconstruct the stripe size, drive order and parity pattern in software from the images — assembling the array offline, without the controller and without risk to the disks.
We pull the file system and data from the virtual array, separating good pre-rebuild data from any bad post-rebuild writes, and send you a full list before you commit.
The questions we're asked most after a rebuild fails.
No. A degraded array has no redundancy left, so another rebuild pass risks a fatal read error or a second drive failing outright — and if it starts writing bad parity, it can destroy data that's currently recoverable. Stop, and let the drives be imaged first.
Usually not. A degraded array that dropped during a rebuild is the commonest RAID job we see, and imaging each drive then rebuilding virtually recovers it — as long as nobody forced it back online or ran repair tools over it.
Yes, a great deal. The reconstruction depends on knowing the original drive order, stripe size and parity pattern. Label the bays before you pull the drives — if the order is lost, we can work it out, but it's slower.
Yes — send the complete set. Even a failed member often holds data that makes the rebuild cleaner, and we image whatever can be read from it. Recovering an array from a partial set of drives is much harder.
RAID, NAS and server recovery starts from £500 + VAT, depending on the number of drives and what each one needs. There's a free diagnostic first, and we'll give you a fixed quote before any chargeable work begins.
Don't retry the rebuild or force the array online. Power it off, label the drives in order, and send the whole set in — we'll image each one read-only and rebuild it virtually, never on your originals.