On Fri, Mar 20, 2009 at 1:39 PM, jdow jdow@earthlink.net wrote:
From: "Dr. Michael J. Chudobiak" mjc@avtechpulse.com Sent: Thursday, 2009, March 19 05:10
Frank Cox wrote:
It looks like the machine can see the second drive and the lvm that's on it /dev/sdb2, but it has the same VolGroup name as /dev/sda2.
Yes, this is common and annoying. Here is the guide that I followed when it happened to me:
http://www.whoopis.com/howtos/linux_lvm_recovery.html
I filed bug 461682 (https://bugzilla.redhat.com/show_bug.cgi?id=461682), requesting that the default volume names not be so generic - they now incorporate the hostname, so this problem should be much less common in F11+.
That's a wrong solution. GUIDs were invented to handle this sort of problem. Suppose "Fred" has a machine he called "boundless". He has a disk problem. He asked "Judy" to fix it. And for unknown reasons "Judy" also has a machine called "boundless". GUIDs to the rescue.
Hmm... The wrong solution? *It is* the solution because I actually used the advice given in this thread to fix a system. I don't think I understand how hostnames or GUIDs figure in this kind of situation. If you have the time, would you share with us how you've done it?
The only time GUIDs will fail is when you use "dd" to create as good a back-up as you can of a dying disk. Sometimes this is a bad thing, as in the scenario under discussion. Sometimes it is a good thing, as when I performed that sort of a recovery on an NTFS laptop drive. I didn't even have to reinstall anything after NTFS chkdisk massaged the drive. (It had a directory block it could not update - on the "C:" drive.)
Sadly GUIDs are too complicated for the people naming disk partitions.
In this particular LVM case, we've looked at UUIDs as they refer to Logical Volumes, which are not disk partitions. They are not the same thing.
~af