Moving KVM machine with LVM disks fails

Submitted by Hal9000 on Mon, 05/21/2012 - 15:17

Very scary message actually... looked like it did not copy the disks then deleted them, but actually it also failed at deleting the disks so i could start the vm on the old machine again. I will have to find out how to migrate the lvm volumes manually now...

Moving KVM system from discovery.xxx.yyy to atlantis.xxx.yyy ..
Shutting down on original host system ..
.. done
Copying disk images to new host system ..
Copying disk /dev/raid/mercury_xxx_xxx_img ..
.. not a disk image, so cannot be copied
Copying disk /dev/ssd/mercury_ssd ..
.. not a disk image, so cannot be copied

.. no disk images were copied
Deleting disk and config files from original host ..
.. done

Starting up on new host system ..
.. done

Re-fetching system status ..
.. done. New status is : Down
Status: 
Closed (fixed)

Comments

Submitted by JamieCameron on Mon, 05/21/2012 - 16:56

I think I know what the cause is here - was this mercury system re-imported in Cloudmin?

Cloudmin incorrectly doesn't know that those LVs are logical volumes which should be copied, as opposed to being regular devices that were attached to the VM. A fix for this issue (which only happens to re-imported KVM systems) will be included in the next release.

I have a work-around to fix other systems before you move them though, if you are interested..

Submitted by Hal9000 on Mon, 05/21/2012 - 17:07

I tried moving from the original cloudmin install, but that fails too. I would be very interested in your fix, as my old server is going to be shut down in a few hours and i am so far failing to manually copy over the lvm volumes...

Submitted by JamieCameron on Mon, 05/21/2012 - 17:30

The fix is as follows :

  1. Get the unique ID for the VM with the command cloudmin list-systems --host mercury.whatever --id-only
  2. Edit the file with that ID under the directory /etc/webmin/servers
  3. At the end, add the lines :
kvm_filesystem_lv=mercury_xxx_xxx_img
kvm_filesystem_vg=raid

You should then be able to re-do the move.

However, in the case of the system that failed, you should first do a "manual" move to tell Cloudmin that it is back on the original system. The command for this is :

cloudmin move-system --host mercury.whatever --host discovery.xxx.yyy --manual

This will not actually move anything, but instead make Cloudmin think that the host system is still discovery, which is really the case as it still has the disk images.

Submitted by Hal9000 on Mon, 05/21/2012 - 18:05

Thank you for your quick support. I already had those two lines in the config file. However, my second try of manually copying over the LVM volumes appears to have worked and the machine is up and running. So no urgent needs for now :)

Submitted by JamieCameron on Mon, 05/21/2012 - 18:06

Great! I just realized that the first failed move would have set those lines for you ..

Submitted by Issues on Mon, 06/04/2012 - 18:18

Automatically closed -- issue fixed for 2 weeks with no activity.