By the time most arrays are received here after failing they are less likely to be recoverable than at the point of their failure. Typically damage does not occur due to the original fault; rather, it is what occurs during the first few hours. RAID 10 explained provides additional details regarding specific failure patterns unique to striped mirrors in RAID 10 configurations.
Although each level has its own mathematical formulas, the steps taken to achieve recovery follow a consistent pattern: shut down system power; maintain bay order; create a non-destructive image of all drives and use these images to reconstruct. Reconstruction is the only step that varies based upon the type of RAID configuration being recovered.
Power down and leave it down. Your array is degrading when you are running it as is, and every hour you run an array with bad sector spread is costing you future reads. Preserve bay order. When taking disks out of the RAID system label each disk with its bay prior to removal, because disk order is critical to every single possible reconstruction. Never write to the members. No rebuild. No resync. No initialise. No repair utility. No recreating the RAID “just to see if it mounts.” Image before anything else. Each member will be cloned read-only, and all reconstructions happen against these copies so nothing you attempt will eliminate any future option.
RAID 0 has no redundancy at all therefore every member is required and the outcome tracks the condition of the failed disk (covered fully on our RAID 0 recovery guide). RAID 1 provides two copies that can be merged into one good image making it the most winnable level and easiest to ruin with careless rebuilding (see the RAID 1 guide).
RAID 5 recovers if one member is lost using the parity information to rebuild it. It will fail once a second member loses its integrity (and therefore can’t be used), and also in cases where surviving members have enough bad sectors on them that a stripe cannot be built/recovered, although all disks nominally appear. RAID 6 contains two parity blocks and thus typically tolerates two member losses. This means most arrays we receive have lost at least three members, or their parity does not verify. RAID 10 is simply striping across pairs of mirrors. There is no parity to calculate; the challenge here is determining which physical disk pairs up with which. As long as none of the losses are part of the same mirror pair RAID 10 can survive multiple failures.
Since controller metadata represents an educated guess (not factual knowledge), and since it is precisely the controllers that have failed or been reset upon which we should not rely, we derive our parameters including stripe size, member ordering, parity rotation and data offsets from the disks themselves. Candidate configurations are then run through until reassembly produces recognisable filesystem structure vs. random noise. Then we assemble the array virtually from the images, and build the filesystem from that virtual array.
Where there is initial physical maintenance required of a member (donor heads, board repairs, firmware work), this occurs prior to imaging because an unresponsive member simply does not provide any data in contributing to the overall image.
Include all disks which were previously identified as “failed” (a common occurrence when a disk is removed from a system). Data may still exist on these drives, and could result in either a successful recovery of some data, or a complete recovery. Each drive should be labelled with its corresponding “bay”. Provide information about your controller/NAS model, the RAID configuration used (if known), and specifically, what has already been tried to recover the data. It is preferable to advise us of any rebuild attempts made and whether they were completed or cancelled. The method of recovery differs significantly based upon whether or not a rebuild was initiated and then ceased.
The cost of RAID/NAS/server recovery starts at £500 +VAT, which will remain as stated in a written quotation that has been produced based upon our 48-hour no-charge diagnostic service. For those members requiring physical drive level work we require a 50% deposit. The balance is not payable until we have recovered your lost data successfully.
Switch off the system and keep it switched off, then label each hard disk with its bay number. Do not allow the system to attempt to rebuild, resync or repair the array; nor should you create an alternative configuration to try to recreate the array. We almost always see arrays that could have been recovered at the time they failed but were instead made unrecoverable by what happened next.
Often, yes; it really does depend on how much damage was done to the second disk during failure. If one disk fails entirely, the other disk is likely still largely intact because most controllers “eject” a disk once it stops responding to commands. In many cases, an image of the ejected disk may be enough to reconstruct the entire array using the images of both disks.
Only when each of the other members is actually good. Rebuilds write to disk; and on a RAID with defects, this can mean writing what may appear to be correct data (and hence uncorrected) where previously there were errors that would have prevented the reading of parity, which is the most common way a recoverable array becomes unrecoverable. So always image first if you want your data back rather than up-time.
All, including those which have been identified by another mechanism as being faulty. Each type of RAID (except RAID 1) requires all members in order to recreate the volume. Even when a member has been removed, we normally get usable information from the member. Identify them by their bays prior to ejection.
Recovery costs for RAID/NAS/server systems start at £500+VAT and are provided as part of a written quote generated after our initial 48-hour free diagnostic. When a component of the system requires physical drive-level work, we require a 50% deposit prior to commencing the repair. The balance will only be invoiced upon a successful recovery.