RAID 5 With Two Disks Failed? Stop. Don't Rebuild.
UK RAID data recovery specialists since 2002. Hardware controllers (Dell PERC, HP Smart Array, LSI/Broadcom), software RAID, ZFS, Btrfs and ext4. Trusted by UK enterprises including the BBC, Tesco, HSBC and Williams F1.
What this means and what to do next
RAID 5 spreads data across all disks in the array along with parity information that lets the array tolerate a single disk failure. When one disk fails, the controller drops it from the array and continues serving data from the survivors using parity reconstruction; a replacement disk slotted in then triggers a rebuild and the array returns to full redundancy. That's the design. The catastrophic case — and it's a much more common case than RAID 5's marketing suggests — is when a second disk fails before the first rebuild completes, or when the second disk's failure is what reveals that it had been silently developing bad sectors for months.
What's actually happened in most two-disk-failed scenarios we see is this: a disk failed weeks or months ago, the controller dropped it from the array, and the user either didn't notice the warning email or postponed the disk replacement. During that time, the array operated in degraded mode with no redundancy. A second disk then developed a bad sector during normal reads, the controller dropped it too, and the array went offline. The disks themselves are usually not catastrophically dead — they just have specific damaged regions. The data is still substantially recoverable. But only if nobody panics and triggers a rebuild on the wrong assumptions.
The single worst thing you can do at this point is what the RAID controller's UI will likely suggest: "force online" the disks and try a rebuild. The controller will calculate parity from whichever disks it considers valid and write that parity over whichever disks it considers stale. If its assessment is wrong — and after a two-disk failure it usually is wrong, because the controller's metadata is itself partially out of sync — the rebuild writes corrupted parity onto good data. We've recovered hundreds of two-disk failures successfully and we've also seen what arrives at the lab after a forced rebuild went wrong. Both are recoverable, but the latter takes longer and recovery rates are lower.
The three two-disks-failed scenarios we see most often
1. Single original failure, second disk failed during the rebuild attempt. The classic case. A disk failed cleanly. The user replaced it, started a rebuild, and partway through the rebuild a surviving disk hit unreadable sectors and was dropped. The rebuild aborted, the array dropped offline. This is the highest-recovery-rate scenario for us — typically 95 to 100% — because both "failed" disks usually still have most of their data intact; one had a clean failure point and the other has localised bad sectors that we can read around with imaging tools that the controller's standard read path cannot.
2. Long-degraded array — second disk silently failed and was masked by RAID. The most common case in real-world RAID 5 deployments. A disk failed weeks or months ago and was never replaced (sometimes the email alert was missed; sometimes the IT contractor moved on). The array operated in degraded mode for a long time, and during that time another disk developed bad sectors that RAID 5 couldn't detect because there was no redundancy left to compare against. When the array finally fails, two disks are dropped at almost the same time. Recovery requires careful imaging of all disks (including the original failure) plus forensic reconstruction of the RAID parity to determine which disk had which sectors readable at the moment of failure.
3. Catastrophic event — power surge, controller failure, multiple SMART failures. A power surge takes out two disks simultaneously. A failing power supply slowly damages disks over weeks. A RAID controller dies and writes garbage on its way out. These cases are rarer but they do happen, and the recovery work is similar to the long-degraded case: image every disk, including the dead-looking ones, and reconstruct from the most complete set of images. We've recovered RAID 5 arrays with three or more disks reporting failed; the disks usually have more readable data than they appear to.
How Data Clinic recovers a RAID 5 with two disks failed
First and most important: we work from images, not the original disks. Every disk in the array is imaged individually using a PC-3000 or Atola DiskSense, with bad-sector handling tuned for the specific drive model. Crucially, this includes both "failed" disks — they're often the most informative source because they were dropped from the array at known points and their on-disk metadata helps us reconstruct the array's history. The original disks are never written to during recovery and are returned to you untouched.
Once we have full images of every disk, we reconstruct the RAID layout in software. From the on-disk metadata (or, if the metadata is corrupted, from statistical analysis of the data patterns) we determine the stripe size, the disk order, the parity rotation pattern, and the chunk count. RAID 5 has a very specific parity layout — left-symmetric, left-asymmetric, right-symmetric or right-asymmetric depending on the controller — and getting this wrong leaves you with garbage. Our forensic toolkit checks all four against the actual data and validates the correct layout against known file signatures.
With the array virtually reassembled, we expose the underlying logical volume and run filesystem recovery. NTFS, ext4, XFS, VMFS (for VMware datastores) and ZFS are all supported by our tooling. For each, we walk the on-disk filesystem metadata, fall back to redundant copies where the primary is damaged, and extract files. We also handle file system layers stacked on top of the RAID — LVM, virtual machine disks (VMDK, VHDX), and Microsoft Storage Spaces virtual disks. More about our RAID data recovery service →.
Recovered files are returned on a new drive or array of your choice. For business customers we can also deliver onto a temporary loan server while you sort out replacement hardware, and we can structure the directories to match your existing share layout so users see no difference.
Get a free initial diagnosis in 60 seconds
In the tool below, choose RAID / NAS → Multiple disk failure for the right escalation path.
What our customers say
"Three years of family photos on a drive that suddenly failed. Data Clinic collected next day, kept me updated through the cleanroom work, and got everything back. Worth every penny."
"Honest, fixed-price, no-fix-no-fee. Quoted by another lab at three times the price. Recovered 100% of my files."
"Reasonable cost, clear communication, and they were straight with me about what was recoverable and what wasn't. Recommended."
Frequently asked questions
My controller is offering to "force online" the failed disks. Should I?
Almost never on a two-disk failure. Force-online tells the controller to ignore the metadata that says these disks are stale, and use them anyway. If the disks are stale for a good reason (because they were dropped before further writes happened), force-online will mix old and new data in unpredictable ways. The safe path is to image everything first and assess the disks individually. Get a quote before you click that button.
How much does RAID 5 two-disk-failure recovery cost in the UK?
Most cases fall in our standard RAID tier: typically £995 to £2,495 including VAT depending on the number of disks (typically 3 to 12 in a RAID 5), total capacity, the filesystem on top (NTFS and ext4 are routine; VMFS and ZFS are more complex), and damage extent. Free diagnosis. No fee if we cannot recover your data.
Should I send the whole server or just the disks?
Just the disks is usually fine, but label them clearly with their bay numbers (1, 2, 3, 4, etc.) before removal. If the RAID metadata is on the controller card rather than the disks (some older Dell, HP and IBM controllers do this), send the controller card too. If you're unsure, send everything — we'd rather have too much information than not enough.
My RAID controller email said one disk failed. Now two are showing as failed. What happened?
Almost certainly: the second disk had been silently developing bad sectors for some time, but RAID was masking the problem because the parity from the surviving disks could correct any reads from the bad sectors. When the first disk failed and you started a rebuild, every sector on the surviving disks had to be read end-to-end for the first time in years — and one or more of those reads hit the bad sectors that had been hiding. The controller dropped the second disk and the array went offline. This is one of the most common reasons RAID 5 fails in practice, and it's why scheduled scrubbing of the array (which proactively reads every sector and fixes errors during normal operation) is so important.
How long does RAID 5 recovery take?
Standard turnaround is 5–20 working days from receipt of the disks. Emergency turnaround is available — 48 to 72 hours for an additional fee — and we routinely meet enterprise SLAs for business customers. Call 0800 151 2207 to discuss.
Can you do this remotely? My disks are in a data centre 200 miles away.
Sometimes. For business customers with the disks already imaged on local infrastructure, we can work from images shipped to us electronically. For most cases we still need the physical disks. We can also arrange same-day collection by secure courier from any UK location.