On 07/19/2010 07:14 AM, andiket a wrote:
Short answer: no
Long answer: support for this would defeat the purpose of volume groups. VGs have independent metadata so that you can remove them independently, move between system, and so on. If a misguided soul were to implement your feature, then the two volume groups would have to be "locked together" so that removing one or the other would invalidate the snapshot. For similar reasons, you cannot hardlink files between unix filesystems.
If you are using volume groups for the naming only, try using a separator in your LV names. (LVM lawyers - what are legal chars to use?) For instance, I know '_' is safe, so for example instead of vgabc/lv1 and vgdef/lv2, where you are willing to "entangle" vgabc and vgdef because they are always online together, just make a single vgroot and
use vgroot/abc_lv1 vgroot/def_lv2.
BTW, A feature that would make more sense would be hierarchical namespace for LVs within a VG, i.e. subdirs within the /dev/vg* directory. Very low priority, and tricky, but would be handy for naming purposes.