RAID 5 Data Recovery: What Happens When Your Array Fails
Our founder and lead technician, Joseph Montanti, handles cases personally.
RAID 5 can keep your files online after one disk fails. If a rebuild fails or a second member drops out, set aside every original disk and any partly rebuilt replacement. RAID 5 data recovery then depends on reading enough matching blocks from those members to reconstruct the stripes your files occupy.
On this page
A missing disk with the array still online
A RAID 5 stripe holds data blocks and one calculated parity block, spread over at least three disks. The parity position rotates between drives. When one disk goes missing, the controller uses the blocks it still has to work out what the missing one held. That's how all the folders can still open when the front of the server shows a failed drive.
Shut down if a disk clicks, beeps or grinds, a copy stalls, a second disk drops out or the array crashes. Do not run CHKDSK or fsck on a degraded array, or delete and recreate it. Leave the disks in their present order and retain the replacement as well as the removed disk.
If one member has failed and the remaining disks sound normal and read steadily, save anything newer than your last backup to separate storage. Check that the copied files open. On QNAP, a group that specifically says degraded read-only is different from an ordinarily degraded array. Keep a crashed or offline set powered down until the members are preserved. QNAP advises copying out the readable data on a degraded read-only array. Replacing disks in that state doesn't restore the missing redundancy. QNAP's degraded read-only guidance.
To replace one failed disk, you need a verified backup, an otherwise healthy degraded array and a positive identification of the failed bay. Confirm that the disk reported by the controller is the disk you're about to remove. Taking out a healthy neighbor creates a second missing member at exactly the wrong time. A failed disk that comes out should keep its original bay label.
- Use the controller or NAS management screen to identify the failed member and check the others for read errors or disconnects. If QNAP shows degraded read-only, another member has read errors, or your backup hasn’t been checked, skip Repair. Save what still reads steadily or preserve the set offline. Otherwise, confirm the replacement model and size. Use the unit's own shutdown procedure if its bays don't support hot swapping.
- Replace only the identified disk with a healthy compatible one. On a Synology RAID 5 the replacement must be at least as large as the smallest disk in the pool. Select the degraded pool's Repair action and confirm the replacement is the disk the system will erase.
- Follow the rebuild status in the controller or Storage Manager. Once it completes, check the status of the volume and open files from it. Keep the verified backup through that check. Synology's instructions for repairing a degraded pool.
Why a RAID 5 rebuild can fail halfway through
A normal file copy reads the blocks belonging to the files you select. A full rebuild has a different job. It has to produce the missing disk's contents across the array, including stripes that haven't been read by a user in a long time. For four equal 8 TB members, a full-capacity rebuild may read roughly 24 TB from the three survivors to reconstruct the fourth. The manufacturer's read-error specification is a limit under specified conditions, and each disk has its own history. The condition of the surviving disks determines how much of it they can supply.
Suppose disk A is out and disk B returns an unrecoverable read error on one stripe. The controller now lacks two blocks from that stripe and has a single parity relationship. It can't determine both missing values. Other stripes that read cleanly still have enough information. A second disk going entirely offline affects the rest of the volume much more broadly than one unreadable sector.
On Dell PERC systems this local double fault can produce what Dell calls a puncture. The rebuild can continue with an affected stripe marked bad. That's why a rebuild marked complete and a file that gives a read error can be part of the same incident. Which file uses that stripe matters more to its owner than the progress bar. Dell's explanation of double faults and punctures.
A force-online command can put stale data back into service. Initialization or a new volume writes fresh configuration. CHKDSK and fsck work on the file system they are presented with and can write corrections to the wrong layout.
When two RAID 5 disks are marked failed
Consider a project folder saved before either disk dropped out. Where the stripe has stayed the same, a readable block from the older disk can still fit. Now consider a database updated after the first disk dropped out. Substituting that older member can join last week's blocks to today's directory. The filenames may look right even though the database doesn't open. This is why both originals, their bay positions and the partly rebuilt replacement can matter to a single file recovery.
Reconstruction starts with separate images of the readable areas of the members. The order of the disks, the size of the chunk written per disk, the starting offset and the rotation of parity determine how those images fit together. A technician can vary those parameters on a virtual array and examine the resulting file system. If two unknown values remain in the same stripe, software needs another source for at least one of them. How UFS Explorer represents a RAID layout.
A hot spare adds another possibility when you find an array offline in the morning. It is a standby disk until a failure makes the controller start a rebuild onto it. Nobody has to remove a disk for that process to begin. How far the controller got before the interruption determines which stripes the spare can supply. IBM's RAID 5 spare-takeover procedure.
Other reasons a controller can lose a RAID 5
After a controller change, a foreign configuration means the metadata on the disks differs from the controller's current configuration. Use Preview only. Leave Import and Clear alone on the original disks, especially on a running degraded array. Reject initialize, format, create volume and force-online options, and skip file-system repairs. The original disks may carry the layout the new controller needs. A deleted volume or a newly initialized array is a different problem. Keep any previous images and snapshots because the new configuration can already occupy areas that described the old volume. Dell's PERC foreign configuration guidance.
A power loss can interrupt a write after one member has changed but before its parity update reaches disk. Which writes reached storage depends in part on the controller's cache and power protection. If the array failed during the outage, leave it powered down when power returns so the controller cannot start an automatic rebuild. A replacement controller with the same model number doesn't settle that question.
A server with a protected write cache can retain data for a virtual disk that went offline. Dell calls this preserved cache on its PERC controllers. The disks may contain the previous version of a block while the newer write is still in the controller. Keep the cache hardware with the original controller if it is available, and leave any option to discard preserved cache unselected. That is separate from preserving the physical member disks. Dell's guidance on preserved cache.
RAID 5 recovery costs by fault
| Fault | RAID 5 recovery cost | Work that affects the estimate |
|---|---|---|
| Failed controller with readable disks | Quoted per case | Image the members and reconstruct the original layout. |
| Single disk lost, then a rebuild failure | Quoted per case | Determine how far the replacement was written and which other sectors still read. |
| Two failed drives | Quoted per case | Read the available parts of both disks and reconcile stale stripes. |
| Multiple disks with mechanical damage | Quoted per case | Address the affected mechanisms before assembling the readable sectors. |
| Firmware or array metadata damage | Quoted per case | Recover the member data or establish the array parameters from copies. |
| Accidental reconfiguration or deletion | Quoted per case | Separate the old layout and directory information from newer writes. |
The $1,000 to $3,000+ range is for comparison with single hard drive recovery. The pricing breakdown lists other recovery services. A RAID estimate covers multiple members and their layout. Allow for a standard recovery estimate of 4–6 weeks, with timing dependent on the condition of the array. For a RAID assessment, keep the controller and enclosure available along with the full disk set. If the disks travel separately for mail-in recovery, mark the bays before removal and pack each disk in its own antistatic bag and padding so the drives can't strike one another.
How RAID 5 compares with RAID 6 and RAID 10
| Level | Minimum disks | Usable capacity with four equal drives | Which losses it tolerates |
|---|---|---|---|
| RAID 5 | 3 | Three drives' worth | Any one member |
| RAID 6 | 4 | Two drives' worth | Any two members |
| RAID 10, striped mirror pairs | 4 | Two drives' worth | One member of each pair, provided each pair has a survivor |
For a four-disk business array, this choice trades usable capacity against what happens during the next drive replacement. RAID 6 still has a second parity block after one disk fails. RAID 10 copies the surviving member of the affected mirror pair. Neither arrangement keeps yesterday's version of a spreadsheet somebody deleted today. That requires a separate version or backup. Dell's comparison of RAID levels.
Get help with data recovery
Bring your device to our Lincroft lab during opening hours, or mail it in.
How to mail your deviceThe $100 diagnostic is non-refundable once diagnostics start, and donor drives or replacement parts you approve are non-refundable once ordered, even if recovery fails.
Published by MDrepairs
MDrepairs is a data recovery and electronics repair lab in Lincroft, New Jersey. Our founder and lead technician Joseph Montanti films real cases for our YouTube channel, showing tools, repair steps and recovery outcomes.
Keep reading
- Windows Storage Spaces Failure: Preserve the Pool Before Resetting Disks
- Synology Volume Crashed: What to Do While Files Are Still Readable
- RAID 1 Recovery: Why a Mirror Does Not Save Every Deleted File
