Convert Trixbox HVM installation to PVM
For some time I have been running an installation of Trixbox on a small mini-ITX computer. One goal when building the XEN cluster was to move Trixbox to a virtual machine. I didn't migrate the old system to a virtual machine, I just decided to go with a fresh install. Installing Trixbox from iso requires the setup of a HVM, but as I wanted to run it as a PVM I had to convert it. I will here describe how I did to convert the HVM installation of Trixbox to a PVM.
Prerequisites
I first installed Trixbox as a HVM using the iso file downloaded from www.trixbox.com.
I will not go through how to install Trixbox as a HVM, there are good resources already on the Net for how to install a HVM from ISO or CD/DVD. Just make sure you have a working installation of Trixbox running as a HVM. The creation of the LVMs below assumes that you have a Volume Group called: vg. If you followed my HowTo you should already have that.
Shutdown HVM
Make sure you shutdown your running Trixbox HVM. Login and run:
[trixbox ~]# shutdown -h now
Disk setup
We need to copy the files from our HVM disk to a file system that will be used by our PVM. We will use the tools kpartx, losetup and rsync to accomplish this. My HVM installation of Trixbox was done to a LVM called /dev/vg/trixbox which were mapped to a DRBD /dev/drbd6.
Install kpartx:
apt-get install kpartx
Create a loop device to your HVM disk, for me I use /dev/drbd6:
losetup /dev/loop0 /dev/drbd6
Run kpartx to be able to mount individual partitions:
kpartx -av /dev/loop0
My default installation of Trixbox comprised of 3 partitions, one for root "/", one for boot "/boot" and one for swap. Only the root partition will be of interest. In /dev/mapper/ you should now find 3 devices called loop0p1, loop0p2 and loop0p3. On my system loop0p2 was the root partition. You can run:
kpartx -l /dev/loop0
to check the size of the partitions. The biggest one should be the root.
Now we mount the root partition to /mnt:
mount /dev/mapper/loop0p2 /mnt/
Check with ls that /mnt is really the root partition:
ls /mnt
Now we configure two LVMs, one called trixbox-disk and one trixbox-swap. Trixbox-disk will be used to copy all data from the root partition of the HVM installation, the partition that is currently mounted to /mnt.
lvcreate -L 10G -n trixbox-disk vg
lvcreate -L 512M -n trixbox-swap vg
Next we will format both volumes:
mkfs.ext3 /dev/vg/trixbox-disk
mkswap /dev/vg/trixbox-swap
Create a temporary folder and mount it to the trixbox-disk LVM:
mkdir /tmp/trixboxmnt
mount /dev/vg/trixbox-disk /tmp/trixboxmnt
Now we use rsync to copy the whole root partition from the HVM installation to our new trixbox-disk LVM:
rsync -a /mnt /tmp/trixboxmnt
Edit /etc/fstab in the new LVM to reflect it will run as a DomU PVM. /dev/xvda will be our "/" and /dev/xvdb our "swap":
nano /tmp/trixboxmnt/etc/fstab
Below is a copy of my current fstab, I have kept the old values as comments for reference:
/dev/xvda / ext3 defaults 0 1
/dev/xvdb swap swap defaults 0 0
#LABEL=/ / ext3 defaults 1 1
#LABEL=/boot /boot ext3 defaults 1 2
#tmpfs /dev/shm tmpfs defaults 0 0
#devpts /dev/pts devpts gid=5,mode=620 0 0
#sysfs /sys sysfs defaults 0 0
proc /proc proc defaults 0 0
#LABEL=SWAP-hda3 swap swap defaults 0 0
Now we will copy the modules for our current xen kernel to the Trixbox PVM. Replace this with your current kernel:
rsync -a /lib/modules/2.6.24-19-xen/ /tmp/trixboxmnt/lib/modules/2.6.24-19-xen/
Unmount /mnt and /tmp/trixboxmnt
umount /mnt
umount /tmp/trixboxmnt
XEN configuration file
Create a xen config file for the Trixbox PVM DomU.
nano /etc/xen/trixboxpvm.cfg
Below is the output of my config file:
name = "trixbox"
kernel="/boot/vmlinuz-2.6.24-19-xen"
ramdisk="/boot/initrd.img-2.6.24-19-xen"
memory = 512
root="/dev/xvda ro"
disk = [ 'phy:/dev/vg/trixbox-disk,xvda,w', 'phy:/dev/vg/trixbox-swap,xvdb,w' ]
vif = [ 'mac=00:16:3e:aa:35:32' ]
extra="clock=jiffies"
Try to boot the new Trixbox PVM DomU: 
xm create trixboxpvm.cfg -c
Issue with network
When the Trixbox DomU booted I had an issue with the network, eth0 did not start. Inside the DomU I had to edit the /etc/sysconfig/network-scripts/ifcfg-eth0 file and configure it as below:
[trixbox ~]# nano /etc/sysconfig/network-scripts/ifcfg-eth0
DEVICE=eth0
ONBOOT=yes
BOOTPROTO=none
NETWORK=10.1.1.0
NETMASK=255.255.255.0
IPADDR=10.1.1.5
GATEWAY=10.1.1.1
MACADDR=00:16:3E:AA:35:32
Restart the network:
/etc/init.d/network restart
Hopefully the Trixbox DomU is working and is now running as a PVM.
Additional information
You can now go ahead and configure the PVM for DRBD if you have a xen cluster. Check the DRBD section in the xen cluster HowTo: http://www.asplund.nu/xencluster/xen-cluster-howto.html
I think there are basically three ways to run the kernel of the converted HVM.
- Use default Ubuntu Dom0 XEN kernel as I did above
- Use XEN kernel provided for CentOS and copy it to the Dom0
- Use XEN kernel provided for CentOS and use pygrub to boot the DomU
I have only tried #1 and #2 myself as #3 uses pygrub which doesn't run well with DRBD. My preference was to go with #2 as that will keep the original kernel and modules from the Trixbox installation. When doing it that way I could perfectly run the DomU, but I ran into trouble when trying to shutdown and migrate it. It was always complaining about issues to release the two disk devices. Before copying the kernel and initrd to the Dom0 I had to create a new initrd from within the HVM that loads the xennet and xenblk modules, I did that with this command:
[trixbox ~]# mkinitrd /boot/initrd-2.6.18-92.1.6.el5xenU2.img 2.6.18-92.1.6.el5xen --with=xenblk --with=xennet --preload=xenblk --preload=xennet
As I couldn't resolve the issues with #2 I went with #1. I am not 100% sure this is good solution, but at least it is working. 
The only issue that I notice is a failure during boot to load one module, but no side effects found so far.
Note: Some of the steps above are taken from my memory as I lacked proper notes. I plan to do another test install when time allows. Please let me know of any discrepancies.
