I have a home media server that gave me a message today saying one of the drives is soon to fail (not happy, still under warranty, that's another issue)

I am running Mdadm raid 0 on a 3 drive system 4TB each.

I've got just enough space that I could remove the drive and still have some space left.

If I simply remove the drive from the array, will the data on it get moved to the other drives, or is just "lost"?

Since I have enough space on the remaining 2 drives, is it possible to "shrink" the array by 1 partition, and then remove it?

  • You can't shrink a RAID0 array this way. (Indeed hindsite is a wonderful thing - using LVM would provide most of the benefits of RAID0, but allow you to do just this)
    – davidgo
    Apr 20 '18 at 19:10

RAID 0 will implode as soon as a single disk fails, as the data is spread across the disks, regardless of the size the the partition/volume in comparison to the array (a volume that's 1/10th the size of the array is still spread across disks). No data will be automatically moved anywhere. Make a backup of the volume to a secondary data device, blow away the array, remove the drive, recreate the array as wanted, restore data.

  • so If I understand correctly, I can remove the drive, DD block copy it to a new drive, bring the new drive in and start the array once more?
    – JonYork
    Apr 20 '18 at 15:23
  • If DD copies all the RAID header information on the drive (like does a true block-by-block copy) then it will probably* work, but I wouldn't' bet on it (without a full backup first). ;) Apr 20 '18 at 15:26
  • Raid 0 is really a performance-only RAID and you should treat one failing drive as if you have one single hard drive and it's failing. Your smart move right now is to back up the contents of the array onto an external form of media.
    – CDove
    Apr 20 '18 at 15:42
  • @techie007 I don't, and will probably never use RAID0 - but assuming the old drive is still working (and not going offline) would you be able to add a new disk, then migrate the old disk to the new one using mdadm /dev/mdX --add /dev/sdX1 --replace /dev/sdY1 --with /dev/sdYX
    – davidgo
    Apr 20 '18 at 19:07
  • @davidgo would that be a one liner that would transfer the data to the new drive?
    – JonYork
    Apr 20 '18 at 20:10

since its raid 0 it doesn't have any back up I recommend writing a back up to another drive before doing anything and change the raid to a raid 10 it does have redundancy so at that point you can remove a drive and put a new one in and it will re write the info to the new drive. You do need 4 drives for raid 10 but I think it might be the answer for you because it provides the bonus performance while giving you a mirrored setup... with this set up you should even be able to hotswap 1 drive or any 1 drive could fail any you would have no issues.

check this out it has a bit more info on RAID: https://www.prepressure.com/library/technology/raid

  • I did not down voted you, but I imagine the person who did did do because you recommended RAID5. RAID5 is a pretty much dead technology because it can't reliably handle large disks when there is a failure (the chance of a second disk failure during rebuild is hugh and rebuilds are very low)
    – davidgo
    Apr 20 '18 at 18:56
  • Ah ... honestly didn't know that about Raid 5 honestly I recommend 10 because you get performance and mirror. Never understood why people would use 5 over 10 when you had to have the same amount of drives anyway XD ... thx for the info Apr 20 '18 at 19:00
  • In addition to the fact RAID 5 is bad advice, this reads more like a comment, and you not being sure of what you recommended doesn’t help. You also link to some random website but don’t even bother quoting anything from it.
    – Ramhound
    Apr 21 '18 at 0:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.