A factory’s twelve-drive Inspur RAID 10 was throwing constant I/O errors — files slow, saves failing, several disks degraded. RAID 10 mirrors everything, so it forgives one disk in a pair failing. What it cannot forgive is bad sectors creeping across both sides at once, which is exactly what was happening — and the array was still running.
A factory’s Inspur RAID 10 — twelve 2 TB drives — had begun throwing constant I/O errors: files slow to open, some refusing to save, several disks flagged degraded in the controller. RAID 10 is mirroring and striping together, and it is genuinely resilient — but its resilience assumes the two halves of each mirror do not fail at once. Here, bad sectors had crept across multiple drives, and on a mirrored pair that is the dangerous case: a bad sector on one side is covered by the other, until the other has one too, in the same place.
The array was still running, which was the immediate risk. Every read of a failing sector is stress it cannot spare, and every hour powered on was making a recoverable array less so.
Nothing was rebuilt on the Inspur hardware. All twelve disks were pulled, write-blocked, and imaged individually on the DeepSpar imager, whose whole purpose is exactly this: it reads a drive with bad sectors gently, working around the weak regions and taking the good data first, where a normal controller would hammer the failing areas until the drive gave out. The worst disks gave up their readable sectors; the mirrors filled in what any single disk could not.
The array was then reassembled virtually, from the images — the RAID 10 layout, stripe size and mirror pairing worked out from the data itself — and the file system rebuilt from the reconstructed set. Because it all happened on copies, a bad sector that landed in the same stripe across a mirror could be reconciled from whichever image had read it cleanly, rather than being a dead end on live hardware.
The production schedules, inventory databases and operational records were recovered, verified, and returned on fresh storage. The original twelve disks — several of them genuinely near the end — went back as they came, their job done.
Constant I/O errors and degraded-disk warnings mean drives that are failing, and the worst thing you can do is keep the array in service “until the weekend” while it works those failing sectors to death. Do not force a rebuild onto a degraded array either — on a set with bad sectors on multiple disks, a rebuild can be the push that finishes it.
Power it down, send every disk labelled with its position, and let them be imaged. The free diagnostic costs nothing; if the business is standing still, the emergency track exists for exactly this.