0

I have one server running several different virtual machines with Hyper-v for different clients. All of my clients have given the OK to a maintenance window over one day on a weekend in the future.

I want to preserve all of the virtual machines currently running while also rebuilding the server to use Arch Linux and KVM for virtualization. Some of the guest operating systems are Windows Server 2008, 2003, 2008, and 2012 as well as Debian Linux.

As of now, the virtual hard drives are all on the same RAID 1+0 and I have spare hard drives to make an identical RAID configuration.

I was thinking that in case nothing works in the end, I could at least just boot from the old hard drives. So in order to do that, I would set up the current drives in a disk array and attach it as a secondary to the system. The new blank hard drives would go into the server as the primary and would install and configure the OS and finally convert the VHD/VHDX files over and it should be fine.

Can anyone speak to this process, perhaps suggesting alternatives, or offer any advice on why this would or would not be an OK idea?

7
  • It may take much longer than you think to convert the disk images. Test first, then you can plan your maintenance windows. Jul 29, 2014 at 4:08
  • @MichaelHampton How would you go about testing it? Would you do a live backup of the image and time the conversion on a different server? Since the host OS doesn't have the conversion utilities needed, I can't do it on itself. Jul 29, 2014 at 4:09
  • You're not really going to put this all back on the same server it came from, are you, even without some intermediate server? You have no way to roll back! Jul 29, 2014 at 4:10
  • @MichaelHampton The rollback would be just powering off, removing the disks with the partially configured Arch Linux, and shoving the original disks back into the server after removing them from the drive array that was acting as the secondary so that I would be able to copy/convert the VHD files. The source would be on the original drive array, the destination on the live drive array for the new configuration. Hopefully that makes a bit more sense :) Jul 29, 2014 at 4:13
  • Arch Linux? I missed that the first time around. That's a truly bizarre choice for this application. Anyway, you still need to test, so you have a good idea how long your maintenance window must be. Jul 29, 2014 at 4:16

1 Answer 1

3

I would say you have a lot more to consider than just hard drives. You are intending to transfer to a completely different virtualization package, which will take much more than a days worth of work.

Theoretically you would need to: - Transcribe the VM configurations (more than likely by hand) from Microsoft's spec to QEmu's spec. - Convert the VHD images to one QEmu can understand (if VHD support isn't compiled into your version of QEmu). This will include Memory, Disk allocation, NICs, CPU flags, etc. File sharing probably won't be compatible. - Safely transfer all of the above to your new machine (HOPEFULLY YOU ARE USING A NEW SERVER FOR THIS MIGRATION) - Start each VM separately to ensure the configuration for each is sane. - For each VM, install/configure all of the QEmu-specific device drivers, and troubleshoot and resolve those that don't work out of the box. - Shut down all of the VMs, and cold boot the new server to ensure any auto-starting VMs can start successfully. - Test a reasonable amount of clients that would use the VMs (i.e. browse a website that may be hosted) to ensure nothing got mixed up. - Profit

Depending on the size and quantity and diversity of the VMs, this could take a minimum of a week to upwards of two months if BAD THINGS happen, and this is why it is recommended to attempt this on separate hardware, since it will be extremely difficult to recover a broken setup if you're attempting to migrate in place....

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .