I'm running into a slight problem trying to upgrade a CentOS 4.4 install to 5.0.
The configuration: CentOS 4.4 is an old install that used to boot natively. That has since been backed up and restored to two seperate LVs (as the original install had two partitions). The host system now runs Debian, and I've been able to successfully boot the old CentOS 4.4 install under Xen (with a Debian domU kernel, but that's besides the point).
I've decided to upgrade the CentOS domU to CentOS 5. After determining that the recommended upgrade path is to do so via the installer, I grabbed the CentOS installer kernel/ramdisk and configured my domU to boot from those. The installer comes up without a problem.
Things get hairy when the installer takes a look at the 'disks'. I receive the following message: /dev/xvda currently has a loop partition layout. To use this disk for the installation of CentOS, it must be re-initialized, causing the loss of ALL DATA on this drive.
It looks like the installer doesn't like the fact that there are no partitions (which is correct, as the 'disk' is actually a logical volume under lvm). As I'm trying to do an upgrade, and not blow away my existing setup, this isn't going to work.
My question, therefore, is - how do I tell the CentOS 5 installer to ignore the fact that there are no partitions, and to simply do the install? As mentioned previously, the CentOS domU does boot perfectly fine, and given the prevalence of volume groups, logical volumes and virtual machines these days, it seems rather silly to have an installer absolutely rely on the existence of partitions before it'll let one do an upgrade or install.