Exchanging hard disks in a Software-RAID

Last change on 2021-11-23 • Created on 2020-03-19

Introduction

This article uses an example to describe the necessary steps involved in exchanging a defective drive in a software RAID (mdadm).

IMPORTANT NOTE: All commands are just examples. You should adjust them accordingly!

You can find a video how-to on our Hetzner YouTube channel

Example scenario

Here is the example configuration:

# cat /proc/mdstat
Personalities : [raid1]
md3 : active raid1 sda4[0] sdb4[1]
      1822442815 blocks super 1.2 [2/2] [UU]

md2 : active raid1 sda3[0] sdb3[1]
      1073740664 blocks super 1.2 [2/2] [UU]

md1 : active raid1 sda2[0] sdb2[1]
      524276 blocks super 1.2 [2/2] [UU]

md0 : active raid1 sda1[0] sdb1[1]
      33553336 blocks super 1.2 [2/2] [UU]

unused devices: <none>

There are four partitions in total:

  • /dev/md0 as swap
  • /dev/md1 as /boot
  • /dev/md2 as /
  • /dev/md3 as /home

/dev/sdb is the defective drive in this case. A missing or defective drive is shown by [U_] and/or [_U]. If the RAID array is intact, it shows [UU].

# cat /proc/mdstat
Personalities : [raid1]
md3 : active raid1 sda4[0] sdb4[1](F)
      1822442815 blocks super 1.2 [2/1] [U_]

md2 : active raid1 sda3[0] sdb3[1](F)
      1073740664 blocks super 1.2 [2/1] [U_]

md1 : active raid1 sda2[0] sdb2[1](F)
      524276 blocks super 1.2 [2/1] [U_]

md0 : active raid1 sda1[0] sdb1[1](F)
      33553336 blocks super 1.2 [2/1] [U_]

unused devices: <none>

You can perform the changes to the software RAID while the system is running. If proc/mdstat shows that the drive is failing, like in the example here, then you can make an appointment with our support technicians to replace the drive (see further below).

# cat /proc/mdstat
Personalities : [raid1]
md3 : active raid1 sda4[0]
      1822442815 blocks super 1.2 [2/1] [U_]

md2 : active raid1 sda3[0]
      1073740664 blocks super 1.2 [2/1] [U_]

md1 : active raid1 sda2[0]
      524276 blocks super 1.2 [2/1] [U_]

md0 : active raid1 sda1[0]
      33553336 blocks super 1.2 [2/1] [U_]

unused devices: <none>

Removal of the defective drive

Before you can add a new drive, you need to first remove the old defective drive from the RAID array. You need to do this for each individual partition.

# mdadm /dev/md0 -r /dev/sdb1
# mdadm /dev/md1 -r /dev/sdb2
# mdadm /dev/md2 -r /dev/sdb3
# mdadm /dev/md3 -r /dev/sdb4

The following command shows the drives that are part of an array:

# mdadm --detail /dev/md0

In some cases, a drive may only be partly defective, so for example, only /dev/md0 is in the [U_] state, whereas all other devices are in the [UU] state. In this case, the command

# mdadm /dev/md1 -r /dev/sdb2

fails because the /dev/md1 array is ok.

In this event, you need to execute the command

# mdadm --manage /dev/md1 --fail /dev/sdb2

first to move the RAID into [U_] status.

Arranging an appointment with the support team to exchange the defective drive

To exchange the defective drive, you need to make an appointment with the support team in advance. The support team will need to take the server off-line for a short time.

Please use the support request section on Robot to make an appointment with the technicians.

Preparing the new drive

Both drives in the array need to have the exact same partitioning. Depending on the partition table type you are using (MBR or GPT), you need to use appropriate utilities to copy the partition table. The GPT partition table is usually used in drives that are larger than 2TiB (for example, 3TB HDDs in EX4 and EX6).

Backing up the MBR/GPT

Before copying the MBR/GPT to a new drive, you need to back it up. That way, if something goes wrong during the copying, you will still be able to restore the original.

Backup with MBR

sfdisk --dump /dev/sda > sda_parttable_mbr.bak
sfdisk --dump /dev/sdb > sdb_parttable_mbr.bak
sfdisk --dump /dev/sdX > sdX_parttable_mbr.bak

Restore with MBR

sfdisk /dev/sda < sda_parttable_mbr.bak
sfdisk /dev/sdb < sdb_parttable_mbr.bak
sfdisk /dev/sdX < sdX_parttable_mbr.bak

Backup with GPT

sgdisk --backup=sda_parttable_gpt.bak /dev/sda
sgdisk --backup=sdb_parttable_gpt.bak /dev/sdb
sgdisk --backup=sdX_parttable_gpt.bak /dev/sdX

Restore with GPT

sgdisk --load-backup=sda_parttable_gpt.bak /dev/sda
sgdisk --load-backup=sdb_parttable_gpt.bak /dev/sdb
sgdisk --load-backup=sdX_parttable_gpt.bak /dev/sdX

Drives with GPT

There are several redundant copies of the GUID partition table (GPT) stored on the drive, so you need to use tools that support GPT (for example parted or GPT fdisk) to edit the table. You can use the sgdisk tool from GPT fdisk (pre-installed when using the Rescue System) to easily copy the partition table to a new drive. Here's an example of copying the partition table from sda to sdb:

sgdisk --backup=sda_parttable_gpt.bak /dev/sda
sgdisk --load-backup=sda_parttable_gpt.bak /dev/sdb

You then need to assign the drive to a new random UUID:

sgdisk -G /dev/sdb

After this, you can add the drive to the array. As a final step, you need to install the bootloader.

Drives with MBR

You can simply copy the partition table to a new drive using sfdisk

# sfdisk -d /dev/sda | sfdisk /dev/sdb

where /dev/sda is the source drive and /dev/sdb is the target drive.

(Optional): If the partitions are not detected by the system, then the partition table has to be reread from the kernel:

# sfdisk -R /dev/sdb

Naturally, you can create the partitions manually using fdisk, cfdisk or other tools. The partitions should be Linux raid autodetect (ID fd) types.

Integration of the new drive

Once you have removed the defective drive and installed the new one, you need to integrate it into the RAID array. You need to do this for each partition.

# mdadm /dev/md0 -a /dev/sdb1
# mdadm /dev/md1 -a /dev/sdb2
# mdadm /dev/md2 -a /dev/sdb3
# mdadm /dev/md3 -a /dev/sdb4

The new drive is now part of the array and will be synchronized. Depending on the size of the partitions, this procedure can take some time. You can check the status of the synchronization using cat /proc/mdstat.

# cat /proc/mdstat
Personalities : [raid1]
md3 : active raid1 sdb4[1] sda4[0]
      1028096 blocks [2/2] [UU]
      [==========>..........]  resync =  50.0% (514048/1028096) finish=97.3min speed=65787K/sec

md2 : active raid1 sdb3[1] sda3[0]
      208768 blocks [2/2] [UU]

md1 : active raid1 sdb2[1] sda2[0]
      2104448 blocks [2/2] [UU]

md0 : active raid1 sdb1[1] sda1[0]
      208768 blocks [2/2] [UU]

unused devices: <none>

Bootloader installation

Since the serial number of the disk changed, we need to generate a new device map with GRUB2:

grub-mkdevicemap -n

If you are doing this repair in a booted system, then for GRUB2, running grub-install on the new drive is enough. For example:

grub-install /dev/sdb

In Grub1 (grub-legacy), depending on which drive was defective, you may need to do more steps.

  • Start the GRUB console: grub
  • Specify the partition where /boot is located: root (hd0,1) (/dev/sda2 = (hd0,1))
  • Install the bootloader in MBR: setup (hd0)
  • To also install the bootloader on the second drive:

    • Map the second drive as hd0: device (hd0) /dev/sdb
    • Repeat steps 2 and 3 exactly (don't change the commands!)
  • Exit the GRUB console: quit
Probing devices to guess BIOS drives. This may take a long time.


    GNU GRUB  version 0.97  (640K lower / 3072K upper memory)

 [ Minimal BASH-like line editing is supported.  For the first word, TAB
   lists possible command completions.  Anywhere else TAB lists the possible
   completions of a device/filename.]
grub> device (hd0) /dev/sdb
device (hd0) /dev/sdb
grub> root (hd0,1)
root (hd0,1)
 Filesystem type is ext2fs, partition type 0xfd
grub> setup (hd0)
setup (hd0)
 Checking if "/boot/grub/stage1" exists... yes
 Checking if "/boot/grub/stage2" exists... yes
 Checking if "/boot/grub/e2fs_stage1_5" exists... yes
 Running "embed /boot/grub/e2fs_stage1_5 (hd0)"...  26 sectors are embedded.
succeeded
 Running "install /boot/grub/stage1 (hd0) (hd0)1+26 p (hd0,1)/boot/grub/stage2 /boot/grub/grub.conf"... succeeded
Done.
grub> quit
#

If you are doing the repair via the Rescue System, you need to first mount the installed system there.

After mounting the installed system also perform the following rebinds:

mount --rebind /dev /mnt/dev
mount --rebind /proc /mnt/proc
mount --rebind /sys /mnt/sys

You need to then do all the above GRUB installation steps in the chroot environment. You can safely ignore the warning grub-install couldn't find physical volumes.

Table of Contents