Case file · NAS & RAID · BDR-2025-0642
Two Failed Drives in a RAID 5 Array.
The enterprise storage shelf had a common problem: two failed drives in a RAID 5
. The user had spare disks
, and they asked an innocent-sounding question: could they rebuild the array from one disk to another
? The spares were the temptation, but the question was the bait.
Same symptoms on your desk?
01273 964902
The decode.
If two drives fail in a RAID 5 configuration, the controller does not know how to calculate what data is missing. If you try to use spare drives for rebuilding, then new parity will be written over the original stripes and any attempts to read those stripes by a recovery program will result in corrupted stripe reads because the original stripes were altered. There is a second trap when using enterprise level storage systems, where the standard sector size of a drive may vary depending on the manufacturer and the drive type used. In addition, most generic RAID utilities will fail to properly read the sectors of these drives due to the way they are formatted. Therefore, the proper use of spare drives would be to place recovered data onto them, not to help in recovering the lost data.
Equipment on this case.
How a job runs here →| The system | What it did on the day | Why this tool |
|---|---|---|
| PC-3000 SAS/SCSI | Addressed the enterprise SAS members that desktop hardware cannot talk to | Enterprise SAS and SCSI members that desktop hardware can't address |
| Atola TaskForce 2 | Parallel imaging of all members, compressing a week of sequential work | Images many drives at once — the difference between days and a week on an array |
| UFS Explorer RAID Recovery | Derived the array geometry and assembled the volume from the images | Reads NAS volume managers as they actually are, not as a flat array |
At the bench.
Image every member — including the survivors
Each unit in the group was placed onto an imager, the failed pair alongside the healthy ones. Enterprise level SAS hard drives require hardware interfaces designed to speak directly to those devices, and because the format of the sectors utilised on these particular drives required specific handling at the time of imaging (if done incorrectly, all of the steps performed after that point would be based upon bad assumptions), we handled that part of the process properly.
Recover the failed members enough to read them
Once the two failed units were assessed, they were restored to a readable condition. Not surprisingly, neither device was uniformly dead, and each had large amounts of usable data remaining on its surfaces. We chose to place both failed units into an imager since we wanted to allow re-assembly to select the best possible copy of any individual block of data rather than having to rely solely on parity calculations.
Derive the geometry, assemble in software
We determined stripe size, unit order, parity rotation and delay from on-disk metadata instead of relying upon default values. We then re-assembled the volume virtually using the images and pulled the file system from the re-assembly process.
Where it ended up.
The data within the array was successfully reconstructed and validated. Afterwards, we copied the data into the customer's spare drives as new storage: their first and only real use in this matter, and certainly a better application for them than if a rebuild had been performed as originally planned.
Related on the index.
More from NAS & RAID.
Recognise your own drive in this story?
As has always been the case with our previous situations, turn off power first, and allow diagnostic work to proceed free of charge prior to making a decision.