Using VUL to backup bare metal RHEL 9.2 server, restore as VMware VM failing at dev/mapper-rhel-home.

[cerberus]

Case #07144666

I have a default installation of a RHEL 9.2 bare metal server, I am using VUL 6.0 to perform entire machine backup to VBR repository.

I then attempt to restore this backup to a VM successfully, whoever the VM is failing to boot, drops me right into dracut.



In reviewing https://www.veeam.com/kb2669 this seems common when going to completely new hardware, which would make sense going from bare metal to VM.

I follow the instructions in this KB and use Custom Veeam Recovery Media ISO created on the actual bare metal server.

Code: Select all

lvscan
vgchange -ay
mkdir -p /targetroot/boot/efi
mount /dev/rhel/root /targetroot/
mount /dev/sda2 targetroot/boot
mount /dev/sda1 /targetroot/boot/efi
mount --bind /dev /targetroot/dev
mount --bind /proc /targetroot/proc
mount --bind /sys /targetroot/sys
mount --bind /run /targetroot/run
chroot /targetroot /bin/bash
dracut -f -v
exit
umount /targetroot/dev
umount /targetroot/proc
umount /targetroot/sys
umount /targetroot/run
umount /targetroot/boot/efi
umount /targetroot/boot
umount /targetroot
reboot

I get further in the boot, but now its failing at...


I am not sure what to do here, /home was part of the restore selection..




On this page, it talks about some considerations/limitations when restoring LVM..

https://helpcenter.veeam.com/docs/agent ... tml?ver=60

• Since LVM volume group is a logical entity that spans across the physical disks, Veeam Agent treats the original disk and the LVM volume group as separate entities. Therefore, Veeam Backup & Replication will restore the original disk and the LVM volume group as 2 separate disks. This way, all data, including the data within the LVM volume group, is accurately restored.
  
  Restoring the original disk and the LVM volume groups as 2 separate disks requires an increased amount of storage space. For example, you restore a machine with 2 disks, and a separate LVM volume group is configured on each of these disks. In this case, Veeam Backup & Replication will restore 4 disks. The restored disks will consume the storage space equal to the size of the 2 original disks and the 2 LVM volume groups from these disks.

I am not exactly following, the restore seems to have restored both rhel/root and rhel/home, why is it not working... Has anyone ran into this?

[cerberus]

Ran into similar issue trying to use ReaR to backup/restore.

https://github.com/rear/rear/issues/3042
https://github.com/rear/rear/pull/3043

I am going to try removing system.devices via chroot when the system is restored, or disable use_devicesfile, and attempt a new backup/restore using Veeam Agent for Linux, will circle back with my findings.

TLTR: In RHEL 9.0, the use_devicesfile option in Logical Volume Manager (LVM) is enabled by default, checking whether the Physical Volume Identifier (PVID) of the LVM volume is identical to the last time the virtual machine was running. When it is not identical, the virtual machine may enter emergency mode, and errors related to /dev/mapper/rhel-home.

[rovshan.pashayev]

Hello Mirza,

Thank you for additional information and please continue with the support case for further assistance.

Rovshan.

[cerberus]

To be honest, support has been less than ideal; maybe it's because I filed this as a sev2 case, but one response email per day only to get a suggestion that is not relevant is not great.

The workaround (tested and confirmed) for this is to chroot, remove /etc/lvm/devices/system.devices and regenerate using vgimportdevices -a.

The other option is to disable use_devicefiles on the host being backed up as per: https://access.redhat.com/documentation ... al-volumes

I suggest publishing a KB on this in case others run into similar issue, this affect VAL 6.x when doing a bare metal restore of RHEL9 on dissimilar hardware.

[rovshan.pashayev]

Hello Mirza,

Thank you for providing this information. We will investigate and explore the mentioned options further.

Rovshan.

[rovshan.pashayev]

Hello Mirza,

KB is updated with possible causes and solutions.