Replacing a XCP-NG Raid1 disk and growing the array

Check RAID status

cat /proc/mdstat

Identify RAID devices and members

mdadm --detail /dev/md0

If /dev/md0 doesn’t exist, check which md device is being used:

lsblk | grep md

Check physical disks and partitions

lsblk -o NAME,SIZE,TYPE,UUID,MOUNTPOINT

Check volume groups and logical volumes (if LVM is used)

vgs
lvs

Check mount points and filesystems

df -h

For Lenovo m720q host:

cat /proc/mdstat # check everything is synced

mdadm /dev/md127 --fail /dev/nvme0n1 # fail drive
mdadm /dev/md127 --remove /dev/nvme0n1 # remove drive

mdadm --detail /dev/md127 # double check

shutdown -h now # power off host

Swap disk out.

sgdisk -R=/dev/nvme0n1 /dev/sda #  clone partiion layout of /dev/sda
sgdisk -G /dev/nvme0n1   # Give it a new unique GPT GUID

This will mirror the partitions and metadata exactly. The commands assume that the new drive is still /dev/nvme0n1.

mdadm /dev/md127 --add /dev/nvme0n1 # add new disk to the RAID
watch cat /proc/mdstat # check rebuild progress

Wait until [UU] is shown again and rebuild completes (could take 15–60+ minutes depending on system load and speed).

mdadm --grow /dev/md127 --size=max # grow the raid array to use the full disk
mdadm --detail /dev/md127 # verify that its has been successful

Resize the Partition Table on /dev/md127

gdisk /dev/md127

In gdisk:
- p — print partition table (record start of partition 3)
- d — delete partition 3
- n — create new partition:
  - Partition number: 3
  - First sector: type the exact start sector (e.g., 75497472)
  - Last sector: just press Enter to accept default (uses remaining space)
  - Hex code: default (8e00 for LVM)
- w — write and exit

Double-check the start sector of md127p3 beforehand using lsblk -o NAME,START,SIZE or gdisk

pvresize /dev/md127p3 # resize the LVM PV

Expand Your Volume Group or Logical Volumes

vgs # verify free space
lvextend -l +100%FREE /dev/VG_XenStorage-xxx/VHD-xxxxx #example of estending the volume
resize2fs /dev/VG_XenStorage-xxx/VHD-xxxxx # grow filesystem if needed (EXT4)
xfs_growfs /mount/point # (if XFS)

You can leave the free space in the VG for XenServer to manage VM disks.

Final checks

mdadm --detail --scan >> /etc/mdadm.conf # update mdadm config

Double-check everything work;

cat /proc/mdstat
vgs
lvs
df -h

Other notes

VG_XenStorage-xxx: The Volume Group (VG) name

vgs

VHD-xxxxx: The Logical Volume (LV) name

lvs # can add VG name

Crafted with the assistance of https://chatgpt.com/share/68328b0c-5094-8001-b27a-d7a27a61cdfc