In Redhat Enterprise Linux environment, when you restore a virtual machine which has multiple volumes by using CentOS-based recovery media (*), there is a case the OS does not start with the intended root volume after the restore even if the restore was successful.
In this case, please confirm the following points.
(※) Customized CentOS-based recovery media, which is not standard Linux-based recovery media created during installing Arcserve (UDP_Agent_Linux-LiveCD.iso).
(1) How to connect volumes to the restore instance.
As described in the configuration guide, the procedure for connecting multiple volumes to the restore instance is implemented by CLI command. If you connect the volumes from the portal GUI screen, there is a possibility that the volumes will be connected in the unintended order. So you should not connect multiple volumes to the restore instance by GUI operation.
(2) When the size of the volume differs between the backup volume and the restore volume.
Even if you connect multiple volumes to the restore instance by CLI operation, when the size of the volume differs between the backup volume and the restore volume, there is a possibility that the boot area will be restored to the unintended volume. Please do the following tasks according to the version you use.
 When using Arcserve UDP 6.5
Please update to "Arcserve UDP 6.5 update 2" and carry out the procedure of  below.
<Upgrade Arcserve UDP Agent (Linux)> http://documentation.arcserve.com/Arcserve-UDP/Available/V6.5/ENU/Bookshelf_Files/HTML/Agent%20Online%20Help%20Linux/default.htm#AgentforLinuxUserGuide/udpl_upgrade_d2d_linux.htm
 When using Arcserve UDP 6.5 update 1 or update 2
Check the following URL, apply the patch, recreate CentOS-based recovery media and restore again.
<arcserve-KB : [Solution] Can not restore to the intended disk as design in Linux BMR with Centos Based LiveCD> https://arcserve.zendesk.com/hc/en-us/articles/360000339426