How to Extend an mdadm RAID1 to RAID5 on a UEFI Boot System

Does it work?

I carried out a number of tests while writing this article. Indeed, after migrating the root partition (which here includes the /boot/ partition) to RAID5, the system would no longer boot normally and switches to grub rescue mode. What's strange is that the system boots normally when you first enter the UEFI boot menu and manually select one of the EFI boot partitions. I've seen this behavior on VMware ESXi and Proxmox VE. Fortunately, it seems to work outside the virtual machine, in my case I tested it on a Dell PowerEdge T630. If you have an explanation, I'd be glad to hear it!

To sum up, always make backups. Ideally test before going into production.

Now that you've been properly warned, let's get on with the tutorial!

Adding the new disk

As explained in the introduction, we start with a working UEFI RAID1.

• Our disk configuration looks like this: two disks (sda and sdc) configured in RAID1 and one newly added disk (sdc):

Convert RAID1 to RAID5

• Identify the new disk to avoid deleting an existing one. Here we have our sdc disk with no partition configured, indicating that this is our new disk:

• An alternative way of identifying the new disk is to use the cat /proc/mdstat command. Here we see our RAID arrays for sda and sdb:

• Install the gdisk tool:

• Copy the partitions scheme from sdb to sdc using gdisk:

• Regenerate a GUID of the sdc disk to avoid having the same as that of sdb copied beforehand:

• After these commands, the partitions of our sdc disk should be defined:

• Add an EFI entry named debian3 for the EFI partition of our sdc disk:

• Add disk sdc1 to RAID1 md100 array. Note that we're not converting our UEFI partition to RAID5, because if we do, the boot system won't be able to read a RAID5 partition:

• Our md100 array is now a RAID1 composed of three partitions:

• This command converts the array to RAID5 and adds the sdc2 partition to the md0 array (SYSTEM). (Optional, but consider using the --backup-file option in case of a system crash during the operation):

• Check reshaping progress (⚠️ Do not reboot the system while reshaping is in progress⚠️):

• Our md0 array is now a RAID5 consisting of three partitions:

• Do the same with the sdc3 partition on the md1 array (SWAP):

• Our md1 array is now also a RAID5 with three disks:

Resize partitions

Resize the File System

• Once shaping is complete, resize the file sytem to the new size (example here with ext4 file system):

• Check that the new size has been taken into account:

Resize the Swap

For the swap partition, we need to disable and then recreate the swap partition. This change involves modifying the entries in /etc/fstab and /etc/mdadm/mdam.conf.

• Disable the swap:

• Create a new swap that takes the new size into account, and copy the new UUID:

• Note that we can retrieve the UUID with the blkid command:

• Edit /etc/fstab and modify the swap entry to include the new UUID:

• Next, edit the /etc/initramfs-tools/conf.d/resume file and replace the UUID as for the /etc/fstab file:

• Generate a new initramfs image for the changes to take effect:

• Activate swap:

• Check swap size:

With LVM

It's pretty much the same thing, the difference being that we need to use the pvresize command to make our LVM take the new space into account. As we can see, LVM brings flexibility.

• Let's suppose we're in this LVM configuration:

• Copy the partitions scheme from sdb to sdc using gdisk:

• Regenerate a GUID of the sdc disk to avoid having the same as that of sdb copied beforehand:

• Add an EFI entry named debian3 for the EFI partition of our sdc disk:

• Add partition sdc1 to RAID1 md100 array:

• Add partition sdc2 and convert the array to RAID5:

• Check reshaping progress (⚠️ Do not reboot the system while reshaping is in progress⚠️):

• Once fully reshaped, expand the LVM:

• Resize the file sytem of each logical volume you wish to resize: