[SOLVED] migration from raw to qcow2, Windows XP guest won't boot

[tafkaz]

Hello!
on our Porxmox VE 4.2 we have a working windows XP guest with a "raw" disk image.
In order to being able to use live snapshots, i want to migrate the raw image to qcow2. I did that by using the PVE GUI (after preparing the storage to be mounted).

Now, when i want to start the XP guest, i get
"A disk read error has occurred. Press CTRL+ALT+DEL to restart."
right after the start.

Switching back to the raw disk, solves the problem, the VM boots perfectly.
I have another windows 7 guest on the same storage, where this migration worked like a charm.
Any ideas?
Thank you very much for your help!
Sascha

 

[linkstat]

When you sayd migrate, you refer to use the "Move disk" from the Web UI?
And, the Bus/Devices combination is EXACTLY the same? (for example: VirtIO/0 instead of IDE/1). This is the most commonly problems with XP Guests.

Greetings!

 

[tafkaz]

Hi,
yes exactly, the move disk wizard.
in both cases i used virtio_0 as device. I later tried with other combinations as well, changing the boot order accordingly, but unfortunately to no avail. Semms like the vdisk is actually loosing its boot record, after the migration.
thanks
Sascha

 

[tafkaz]

can't find any clue here. Any help appreciated, really strange.
Thanks
Sascha

 

[cosmos]

Can you post your VM config (qm config vmid from a shell) before and after the changes? A newbie here, but just wondering if I can help.

 

[tafkaz]

hi, sure:
working with raw:

qm config 103
boot: dc
bootdisk: virtio0
cores: 1
cpuunits: 1000
freeze: 0
hotplug: disk,network,usb
kvm: 1
memory: 4000
name: winxp-server
net0: virtio=8A:FF:B2:44:29:58,bridge=vmbr0
onboot: 1
ostype: wxp
sockets: 1
virtio0: VM-DATA:103/vm-103-disk-1.raw,size=150G

not working with qcow2:

qm config 103
boot: dc
bootdisk: virtio0
cores: 1
cpuunits: 1000
freeze: 0
hotplug: disk,network,usb
kvm: 1
memory: 4000
name: winxp-server
net0: virtio=8A:FF:B2:44:29:58,bridge=vmbr0
onboot: 1
ostype: wxp
sockets: 1
unused0: VM-DATA:103/vm-103-disk-1.raw
virtio0: VM-DATA:103/vm-103-disk-2.qcow2,size=150G

at the end of the move disk wizard i find, that the last bits seem to do something weird:
there's a ton of

transferred: 161061273600 bytes remaining: 0 bytes total: 161061273600 bytes progression: 100.00 %
transferred: 161061273600 bytes remaining: 0 bytes total: 161061273600 bytes progression: 100.00 %
transferred: 161061273600 bytes remaining: 0 bytes total: 161061273600 bytes progression: 100.00 %
transferred: 161061273600 bytes remaining: 0 bytes total: 161061273600 bytes progression: 100.00 %
transferred: 161061273600 bytes remaining: 0 bytes total: 161061273600 bytes progression: 100.00 %

at the end, before i get the "Task OK" message

 

[tafkaz]

whoops... i think you pointed me to something there...
i deleted the
cpuunits: 1000
line, and t seems to be working now.
strange...

 

[cosmos]

Under /var/log/messages you should find which command is used to move the raw to the qcow file and paste it here.

Another approach:

1) keep only the original VM-DATA:103/vm-103-disk-1.raw as boot disk
2) Create a new qcow disk of the same size on another virtio controller

If the system boots alright with this setup, use some cloning software to clone one disk to the other. If the cloning software can not "see" these 2 disks, it's because it does not understand the virtio controller. Move them to ide0/ide1 and then use some free cloning software to move one to the other.

 

[cosmos]

Glad it was solved, please set the title to SOLVED. And do paste here what prompted you to remove the cpuunits from the configuration.

 

[tafkaz]

actually the hint to post my vm config made me stumble upon this cpuunits thingie...
So i just tried without et voilá!
thanks for your help.
Sascha