3

The server has a 7-disk RAID 0 array, and sdf is starting to die.

Is there a way to remove sdf while keeping the array intact?

# df -h
Filesystem  Size  Used Avail Use% Mounted on
/dev/md1     14T  6.6T  7.0T  49% /var
[...]

# cat /proc/mdstat
Personalities : [raid0]
md1 : active raid0 sda4[0] sdf1[5] sdd1[3] sdb1[1] sde1[4] sdg1[6] sdc1[2]
      14482788352 blocks 512k chunks

Looking to keep downtime to a minimum.

Clarification:

I understand there is no redundancy, and I've made a backup of important data.

But is it possible to move the "stripes" off sdf and onto the other drives? Just to get it into a stable condition without having to wipe everything. So far, it's just a few bad sectors I could forcefully repair.

If it is possible, afterwards I would plan to do the same for 3 healthy drives, and mirror them. Ultimately I need to convert this into RAID1.

I do not believe it is possible to add another hard disk to this Hetzner server. Also, they cannot mirror the defective drive and replace it:

"Please note that we can only exchange your defective hard disk for an empty hard disk. We do not carry out any data exchange or backups."
-- Hetzner

Improve this question
2
  • 5
    No. RAID 0 has no redundancy at all. If you don't have backups, start right now because once sdf is gone, it's all gone. Commented Nov 8, 2013 at 16:16
  • RAID is not an arbitrary storage pool; you have to design drive replacability in when you build it, and you designed it with none. So, no, absolutely not. Commented Nov 13, 2013 at 2:38

1 Answer 1

Reset to default
5

RAID 0 has no redundancy so the array actually becomes more fragile with more disks since a failure in any of them will render the entire array unrecoverable.

If you want to continue with your RAID 0 (for performance reasons presumably), and minimize downtime, boot your system with a rescue OS, e.g., SystemRescueCD, and use 'dd' or 'ddrescue' to make the best copy of /dev/sdf1 that you can. Replace the old /dev/sdf1 with the new /dev/sdf1 and continue to worry about the next drive failure.

Improve this answer
3
  • 2
    [Updated the OP with clarification] But is it possible to do this without adding a drive? There's enough free space in the array and I imagine there would be some way to just distribute the load, no? Commented Nov 8, 2013 at 17:53
  • 1
    No, there is no way to do that. If you want to remove a drive, then you are basically down to: backup, rebuild your volume, restore. Commented Nov 8, 2013 at 19:21
  • 1
    @Tim: Depends on what you mean by "is it possible to do this without adding a drive?". To make a copy of sdf you could poweroff and replace one of the other drives with a new drive and then boot into a rescue OS. Then you could use the rescue OS to copy (dd or ddrescue) the failing sdf to the new drive. After the copy, poweroff, replace the old sdf with the new copy and return the drive you removed to make the copy to its proper place. That should restore the array and booting normally should work. Permanently removing a drive will require a complete backup and restore to a new array. Commented Nov 8, 2013 at 21:46

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.