lvm over raid confusion
Jack Tanner
ihok at hotmail.com
Tue Mar 28 06:31:15 UTC 2006
Clyde E. Kunkel <clydekunkel7734 <at> cox.net> writes:
> > Maybe I can separate the VGs so that the RAID array isn't part of VolGroup00.
> >
>
> Try this: after your ctrl-alt-F2 and start md0, do a vgscan, then
> vgchange -ay --ignorelockingfailure. Make sure the LVs look right.
> Then, before you chroot, bind mount sys, dev and proc from the FC5
> installed system. When doing the mkinitrd, include the raid1 module
> just to be sure...dunno...I have done similar and WFM. Good luck!
>
Well, for all those playing along at home: problem solved. Thank you all for
your advice, esp. Paul. For posterity's sake, here's roughly what I did (the
command syntax below is recreated from memory; don't take my word for it):
As above, boot off the rescue CD, ctrl-alt-f2 before bringing up the linux
partitions, mdadm -AR /dev/md0 /dev/hdb /dev/hdc, then flip back to rescue UI,
scan for linux partitions (which now succeeds).
chroot /mnt/sysimage
mkinitrd -v /boot/initrd-2.6.... (make sure raid modules are loaded)
pvmove /dev/md0 VolGroup00
pvchange -x n /dev/md0
pvremove /dev/md0
Edit /etc/lvm/lvm.conf to comment out removed PV (saving a backup, of course).
vgcfgrestore /etc/lvm/lvm.conf # Make sure the metadata reflects PV removal.
reboot, and voila! The box boots, doesn't complain about lack of PV.
To test, I created a separate VG, and added /dev/md0 to it... The system boots
fine and mounts the new VG's LVs.
Once again, thanks guys, you were a lot of help.
More information about the fedora-list
mailing list