Skip to content
Essential Data Recovery

RAID 5 failure and rebuild mistakes

RAID 5 survives one failed drive, but degraded still means one fault away from losing everything. Why rebuilds are the dangerous step, the four documented mistakes, and what to do instead.

Written by Essential Data RecoveryPublished 2026-09-14Safety-critical: owner-reviewedRevision 4
Four identical external drives standing in a neat row on a pale oak shelf with long morning shadows

Do not rebuild, reinitialize, change the RAID mode, or press repair until you know which drives failed and what state the pool is in. A rebuild normally reads from the surviving drives and writes the reconstructed data to the replacement drive; controllers and software RAID differ in the details. Done in the wrong state or the wrong order, it can overwrite data that a professional could otherwise have read.[1][2]

RAID 5 survives exactly one failed drive. In that degraded state the data is still there and readable, and the urgent job is to back it up, not to rebuild. Degraded is safer than crashed, but it is not safe: redundancy is gone, so a second failure or an unreadable sector at the wrong moment ends the pool. A rebuild is a bet: it reads every sector of every surviving disk for hours, and on modern large drives an unlucky unreadable sector, or a second drive that was quietly failing, ends it. The documented ways people lose RAID 5 data are not the first failure; they are the rebuild with two drives gone, the array recreated in the wrong order, and the RAID mode changed on a whim.[3][4][1]

This guide covers RAID 5 and Synology's SHR-1 on consumer and small-business NAS, and the same ideas on small servers. RAID 0 and JBOD have no redundancy at all; RAID 1, RAID 6 and RAID 10 tolerate faults differently and need different advice; ZFS and btrfs pools follow their own rules. Nothing here is a rebuild tutorial, and that is deliberate.

Not sure which situation you are in? Guided help walks you through it.

Degraded, rebuilding, crashed: three different emergencies

RAID 5 pool states on a typical NAS
StateWhat it meansWhat the vendor says to do
DegradedOne drive defective; the pool still works, but a single further fault ends itConfirm which drive is defective, replace it with a healthy one, then repair
RebuildingParity is being recalculated onto the replacement; every surviving disk is read for hoursDo not use the device during the rebuild
CrashedMore faults than RAID 5 toleratesYou can no longer repair it yourself; vendors point at professional recovery

The vendor documentation is unusually clear here. Synology describes Degraded as the state where one drive is defective but no data has been lost, and performance as severely impacted while the array is degraded. Western Digital's manual adds the rule people break: do not attempt to use the device during the rebuild. And the crashed state has a documented ceiling: once a pool is Crashed, Synology says you can no longer repair it yourself, and its tutorial points owners at local data recovery companies.[5][3][1][6]

Why the rebuild itself is a gamble

A rebuild is not a gentle copy. It reads every sector of every surviving disk to recompute what the dead drive held, which is why Synology deliberately lowers the default repair speed to keep the array usable while it runs. The risk comes from drive physics: NAS drive specifications quote an unrecoverable read error rate around one error per 10^14 bits read, and 10^14 bits is about 12.5 terabytes, so a rebuild that reads tens of terabytes is rolling those dice across every surviving surface. A datasheet error rate is a specification, not a measured probability for your rebuild: real outcomes depend on drive age, firmware, workload and luck, which is why this page refuses to quote rebuild failure percentages. The conservative reading is the same either way. A degraded array holding irreplaceable data is the moment to copy or image what you can reach, and to treat the rebuild as the risky step, not the routine one.[7][8]

The four rebuild mistakes

Rebuilding with two drives gone

RAID 5 tolerates one fault, full stop; QNAP's own guide lists the maximum removed drives for RAID 5 as one. Replacing a second failed drive and forcing a rebuild asks the array to reconstruct from a drive that holds unreadable data, and the write-heavy rebuild can destroy the surviving consistency that is the one thing a professional would want preserved. Synology documents the threshold: once removed drives reach the maximum fault tolerance, the volume crashes.[4][9]

Recreating the array instead of reassembling it

After a controller swap or a reshuffle of bays, the array should be reassembled or imported, never created anew. The mdadm manual is candid that its Build mode cannot distinguish creation from assembly and cannot check that the right components were requested, and that adding a spare to a degraded array starts rebuilding onto it immediately; a recreated array with two disks swapped stripes the wrong way and the resync overwrites the old layout in place. Hardware controllers encode the same lesson: Broadcom's documentation describes importing a foreign configuration as the way to activate a virtual drive moved from another system. And the consumer version, changing the RAID mode in a NAS interface, carries Western Digital's own warning that it will delete all of your data.[2][10][1]

Repairing or writing to a half-healthy array

Repair belongs after a failed drive has been replaced, and starting one while a second drive is marginal is exactly the wrong medicine. The legitimate maintenance is different: Synology recommends data scrubbing periodically on healthy pools to catch inconsistency before a failure, not repair tools on a degraded one. Continuing to write to a degraded array has its own cost: after a second failure, any stripe that depended on the missing drive cannot be reconstructed.[11][5]

Pulling the wrong drive

Synology's repair procedure begins with confirming which drive the vendor reports as defective before anything is pulled, and QNAP ships a recovery feature for accidentally removed drives precisely because this happens. Pulling a healthy member leaves the defective one in place plus a vacancy, which is how one real failure becomes a crashed pool by human error.[5][9]

What to actually do, in order

If the array is degraded

  1. Stop writing to it. The degraded array still reads, and it is the only copy of anything you have not backed up.
  2. Back up anything important from the pool to somewhere entirely outside it, now, while it still reads.
  3. Photograph or note which bay holds which drive before touching anything.
  4. Confirm which drive the vendor reports as defective, then remove only that one.
  5. Fit the replacement, then let the vendor's own repair or rebuild run once, and leave the NAS alone while it does.
  6. If a second drive shows abnormal, if the pool is Crashed, if files read with errors while you copy, or if the data is irreplaceable: stop, and take the professional route the vendors themselves name.
[5][1][6]

RAID 5 vs RAID 6, and the backup question

The honest summary of RAID 5 versus RAID 6 is the rebuild window. RAID 6 carries a second parity block on every stripe, so the array keeps fault tolerance while it is being rebuilt from one failure, which is the window when a RAID 5 pool has none. Microsoft's Storage Spaces documentation draws the same boundary for software pools: parity spaces protect against a single disk failure and need at least three disks, while simple spaces protect against nothing at all. None of it is a backup: RAID tolerates drive faults, not deletion, ransomware or a unit that dies. The backup guide covers the version of this story that starts before the failure.[12]

Backups after data loss: setting up the copy RAID cannot give you.

Choosing a data recovery service: quotes, privacy and honest guarantees.

Sources

Facts on this page are cited to the publishers’ own documentation.

  1. [1] Western Digital — My Cloud EX2 Ultra user manual. captured 2026-09-12.
  2. [2] mdadm project — mdadm(8) manual page (upstream source). captured 2026-09-12.
  3. [3] Synology — Choose a RAID type (Synology DSM help). captured 2026-09-12.
  4. [4] QNAP — QNAP TS-364 user guide. captured 2026-09-12.
  5. [5] Synology — Repair a storage pool (Synology DSM help). captured 2026-09-12.
  6. [6] Synology — What can I do if one of my volumes crashed?. captured 2026-09-12.
  7. [7] Synology — Why does it take a long time to repair a storage pool?. captured 2026-09-12.
  8. [8] Toshiba — N300 NAS hard drive specification. captured 2026-09-12.
  9. [9] Synology — What causes a volume to crash?. captured 2026-09-12.
  10. [10] Broadcom — View and import a foreign configuration (MegaRAID documentation). captured 2026-09-12.
  11. [11] Synology — Data scrubbing (Synology DSM help). captured 2026-09-12.
  12. [12] Microsoft — Deploy Storage Spaces on a stand-alone server. captured 2026-09-12.

Related pages