Description of problem:
If you have 2 fedora installation, witch has both LVM and booth are bootable, then you could only start the first installed one only.
This bug prevents from saving data from USB rack, or from any external attached device - or to run anyway the external complete system. This seems to be NOT machine or device dependent. Also this bug has been seen at F11, 12, and above.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Have one normally IDE, or SATA HDD, installed inside the notebook. Have an Fedora installation on the disk.
2. Create an another bootable live, or have an another complete fedora installation with default partitioning on SATA, IDE HDD, or SSD.
3. Restart the machine, and ask from the machine to boot (mostly with f12, or Esc) the complete external system.
The boot process jumps back from the external device. You could only boot your external system when you removing the another one witch mounted inside, and this is mostly impossible.
You could boot any kind of fedora with default install where is lvm used.
Additional info: The gnomefs automatically recognises and mounts external devices, but doesn't handles LVM groups.
Do volume group names in these installations differ? If not, you need to rename one of them to make them unique (also changing the entry in boot loader's config accordingly).
> (also changing the entry in boot loader's config accordingly).
...and entries in /etc/fstab as well.
I have did the changes, and doesn't work. I have met this problem too with F13 beta live, where I couldn't started the live installation, until the internal disk hasn't been removed. I choose the booting device in computer, starts the proper hw, and jumps back to the internal disk. Btw, the device always gets an unique ID, what consists from lot of numbers, however still can't boot.
Else if I choose extFS, then I *could* boot differently. With LVM *not*.
PS: When I try to show the newest release at booth, this is really blame us.
If volume groups name differs, there is no problem from lvm point of view to handle it (I have system with 3 VG, every VG containt roo LV for different system - rawhide,RHEL5, etc... and even using one grub on one boot partition - and it works).
If you have two VGs with the same name, but with differnet UUID, lvm allows you to rename (but system will not boot properly - on many places it expect unique VG name.)
Hopefully installer now doed not use one common name for all VGs (which implicitly caused this problem.)
So if you cannot install _new_ system with different vg_name, it is not lvm problem, but installer one - please report bug to anaconda then.