How to migrate RHEL7 'Booting from Local Disk' to 'Boot from SAN' ?
Environment:
- RHEL 7.1 with 2 partitions: boot and LVM
- Destination disk smaller than source disk
- Destination disk use thin provisioning
Note:
This procedure assumes that the destination disk use thin provisioning
Therefore the data are copied at filesystem level using tar command
It also permits to change root filesystem type (ex from ext4 to xfs)
Procedure:
-
Install and configure device-mapper-multipath
-
Create and map only one LUN from the SAN. If other LUNs are attached, please unmap them during the time of the procedure and reattached them afterward.
-
Reboot the machine in rescue mode on RHEL7 media (DVD or PXE), add 'multipath' kernel option
-
Anaconda propose to mount Linux installation. Use "Skip" option.
-
Enable SSH access
mv /etc/ssh/sshd_config.anaconda /etc/ssh/sshd_config sshd-keygen /usr/sbin/sshd
-
Connect to the machine through SSH
-
Use lsblk / multipath -l to detect source and destination volumes
-
In this example, /dev/sda is the source disk and /dev/mapper/mpatha is the destination disk
-
Create partitions on destination disk
parted /dev/mapper/mpatha -s -a optimal mklabel msdos parted /dev/mapper/mpatha -s -a optimal mkpart primary 1 512 parted /dev/mapper/mpatha -s -a optimal mkpart primary 512 100% parted /dev/mapper/mpatha set 1 boot on parted /dev/mapper/mpatha set 2 lvm on
- Create /boot filesystem (XFS)
mkfs.xfs /dev/mapper/mpatha1 mkdir /mnt/source /mnt/clone mount /dev/sda1 /mnt/source mount /dev/mapper/mpatha1 /mnt/clone
- Clone /boot
tar -C /mnt/source -czf - . | tar -xzf - -C /mnt/clone umount /mnt/source /mnt/clone
- Ensure all LVs to replicate are 'ACTIVE'
lvscan
- Enable additional LVs if needed, ex:
lvchange -ay /dev/VolGroup00/datalv
- Create root volume group on target partition
vgcreate clone_VolGroup00 /dev/mapper/mpatha2
- Recreate all LVs with minimal size to store existing data, ex with 5GB LVs:
for lv in $( lvs VolGroup00 -o name --noheadings ); do lvcreate -y -L5G -n clone_$lv clone_VolGroup00; done
- Create cloned filesystems
for lv in $( lvs VolGroup00 -o name --noheadings | grep -v swap); do # use same FS than source FS=$(blkid /dev/VolGroup00/$lv | awk -F'"' '{print $(NF-1)}') # OR change to a new FS (uncomment to force) #FS=xfs # create and mount destination FS mkfs.$FS /dev/clone_VolGroup00/clone_$lv mkdir /mnt/clone/$lv mount /dev/clone_VolGroup00/clone_$lv /mnt/clone/$lv done
- Mount source filesystems
for lv in $( lvs VolGroup00 -o name --noheadings | grep -v swap); do mkdir /mnt/source/$lv mount /dev/VolGroup00/$lv /mnt/source/$lv done
- Check and adapt cloned filesystem size. Use 'df -h' to check.
For each LV, check the used size on the source and adapt the destination size. Ex:
df -h /mnt/source/varlv/ /mnt/clone/varlv/ Filesystem Size Used Avail Use% Mounted on /dev/mapper/VolGroup00-varlv 99G 9.8G 85G 11% /mnt/source/varlv /dev/mapper/clone_VolGroup00-clone_varlv 5.0G 33M 5.0G 1% /mnt/clone/varlv
Source data use 9.8GB, resize destination LV from 5GB to 20GB
lvresize -r -L 20G /dev/mapper/clone_VolGroup00-clone_varlv
- Clone all LVs
for lv in $( lvs VolGroup00 -o name --noheadings | grep -v swap); do echo "cloning $lv..." # copy data using tar - unmount if ok tar -C /mnt/source/$lv -czf - . | tar -xzf - -C /mnt/clone/$lv && \ umount /mnt/source/$lv /mnt/clone/$lv done
Note:
Following message may appear during copy, ignore them:
tar: ./spool/postfix/public/pickup: socket ignored
- Disable source VG
vgchange -an VolGroup00
- Set lvm filter to accept only cloned disk
echo 'devices { global_filter = [ "a|/dev/mapper/mpatha|", "r|.*|" ] }' >> /etc/lvm/lvm.conf
- Rename vg/lv to original names
vgrename clone_VolGroup00 VolGroup00 for lv in $( lvs VolGroup00 -o name --noheadings ); do lvrename VolGroup00 $lv ${lv/clone_/} done
- Run Anaconda menu and chose to mount Linux installation (do not use read-only)
anaconda
- When dropped into the shell:
chroot /mnt/sysimage
- Create swap
lvresize -L 8G /dev/VolGroup00/swaplv mkswap /dev/VolGroup00/swaplv swapon -a
- Update fstab on clone
UUID=$( blkid -o value -s UUID /dev/mapper/mpatha1 ) sed -i "s:UUID=.*/boot:UUID=$UUID /boot:" /etc/fstab
- If filesystem type has changed, update /etc/fstab accordingly.
Ex if migrating from ext4 to XFS, replace:
/dev/mapper/VolGroup00-LogVol00 / ext4 defaults 1 2
With
/dev/mapper/VolGroup00-LogVol00 / xfs defaults 1 2
- Install grub
grub2-install /dev/mapper/mpatha cp /boot/grub2/grub.cfg /boot/grub2/grub.cfg.ori grub2-mkconfig -o /boot/grub2/grub.cfg
- Update /boot/grub2/device.map, ex:
Replace
(hd0) /dev/sda
With:
(hd0) /dev/mapper/mpatha
- if friendly_names is disabled, create an alias to identify the correct disk, ex:
echo " multipaths { multipath { wwid 36005076300810a1f6000000000000185 alias mpatha } }" >> /etc/multipath.conf
- Check if lvm filter is used and update it accordingly, ex:
egrep "^[ \t]*global_filter[ \t]*=|^[ \t]*filter[ \t]*=" /etc/lvm/lvm.conf filter = ["a|^/dev/sda|", "r/.*/"] # replace with: filter = ["a|^/dev/mapper/mpatha|", "r/.*/"]
- Re-generate initramfs with multipath module
cp /boot/initramfs-$(uname -r).img /boot/initramfs-$(uname -r).img.ori dracut --add multipath -fv
-
Reboot and disable local disk in the bios (or remove them physically)
-
Reattached additional LUNs if any
-
Configure HBA bios to boot on the correct LUN and reboot.
Responses