Data Recovery Case File · NAS & Network Storage · It Fails Gracefully

Spanned Rather Than Striped, and Why That Helps

This enquiry came from an IT team and reported the configuration apologetically, as though it made things worse. A network unit with "two physical drive failures, but unfortunately the drive has been configured as JBOD rather than a usual RAID configuration." That word unfortunately is doing the wrong work. A spanned volume has no redundancy at all, which sounds like the worse position — but it also has no striping, and that single difference changes everything about what a double failure costs. Where a striped array with two members down yields nothing, a spanned one still yields whole files from every surviving disk.

The decode: what spanning does differently, and why it fails better

What a spanned volume is: several disks joined end to end and presented as one large volume. The system fills the first disk, then continues onto the second, and so on. There is no parity, no redundancy and — this is the part that matters — no striping. Files are not broken into fragments distributed across every member. Each file is written wherever the volume happens to be filling at that moment, which means it sits wholly on one disk, except for the small number that straddle a boundary between two.

Why that changes a double failure completely: in a striped array with parity, every file is interleaved across all members, so losing more members than the parity can cover means no file can be assembled — including files whose data is sitting perfectly intact on healthy disks. The loss is total rather than proportional. In a spanned volume the opposite holds: a failed disk takes down the files that lived on it, and the files on every other member are complete, whole and readable. Two failures out of several members costs the contents of those two, not the contents of the set.

What that means for the work here: the surviving members can be read for complete files immediately, which establishes what is safe before anything difficult is attempted. Then each failed disk is recovered on its own merits — as an individual drive with an individual fault — and whatever comes back is added to the picture. Every disk recovered restores its own files rather than being a precondition for the others.

Why the volume structures still need care: the filesystem describing the whole spanned volume — the directory tree, the file names, where things sat — is written across the set and typically begins on the first member. If that member is one of the failed ones, the structures need reconstruction before the surviving disks can be presented with their original folders and names. The content is still there either way; the difference is whether it comes back organised or as recovered files.

What must not happen: no attempt to rebuild or re-initialise the volume on the appliance, which writes fresh structures across the members, and nothing plugged into a machine that offers to initialise a disk it does not understand.

The note for afterwards: a spanned volume gives capacity and no protection, and it is often chosen without that being understood. It is worth saying plainly to whoever configures the replacement.

On the bench

All members were removed, labelled by bay and imaged individually write-blocked on the Atola TaskForce 2. The surviving members were read for complete files before the failed members were addressed — since a concatenated volume stores each file wholly on one disk, the healthy members hold intact files immediately, and establishing what was already safe came before any difficult work. Failed members were then recovered individually and their contents reintegrated, and the volume assembled offline from the images with its structures reconstructed.

The outcome

The healthy members read for whole files first, the failed disks recovered individually and the volume assembled offline from the images. Free assessment, one fixed written figure including VAT; where a drive has to be opened, 50% of parts and labour is payable upfront with the balance only on success — otherwise no recovery, no fee. The decode, for anyone with a spanned volume rather than a striped one: it is better news than it sounds. Spanning has no redundancy, but it also has no striping — each file sits wholly on a single disk rather than being interleaved across all of them. So where a striped array beyond its parity yields nothing at all, a spanned volume loses only what lived on the failed disks and every other member gives back complete files.

Our case files are drawn from genuine enquiries received by our laboratory over the past ten years, anonymised to protect client confidentiality. Each one describes the diagnostic and recovery procedure our engineers apply to that fault, using the equipment listed.