I HAD a fedora 11 server with md RAID 1 across two 1TB SATA drives. The md0 space was set up to be an LVM PV and the single LVM VG was carved up into 5 or 6 LVs. The MB on this system died and I wound up buying a new one.

Now I want to recover the data from the RAID1 setup on the new server. However, when I attach the two 1TB drives to a new fedora 13 setup, mdadm is only able to find one of the two drives. The partition on the second drive shows "busy" during an mdadm -A -s -v to scan for md volumes.

Well, one drive should be enough since this is RAID1, right? Well, when I do a pvscan -v, the other drive shows up as a "NEW" pv not allocated to a VG. In addition, vgscan does print "Invalid metadata header checksum" when it runs but it doesn't point at any particular PV. I'm afraid to go any further with LVM since I can't afford to lose the data on this system. It is backed up offsite, but the restore will take several days and I can't afford to be down that long.

Are there any tools or techniques where I can dig deeper into what each drive, in the RAID1 pair, has right and wrong with it and pick one that I can force into a usable VG so that I can recover the data?

Thanks in advance,
Marc

mdadm raid1 is transparent in this case. In mdadm raid1, what is written to one drive is just written to the other as well. You can take either drive and put it into another computer and it will boot up just fine without mdadm even installed. This is probably more of an issue with lvm. I don't know how you set it up, but when I run "pvscan -v" on my system, I only see my raid1 md0, not any single disk. Perhaps fedora 13 while installing wrote to the raw disk or something? Or maybe fedora 13 has incompatible mdadm and lvm versions with fedora 11?

That's good information. I wanted to think that mdadm RAID1 was that simple. I just need to find out whether the ext3 filesystem is intact inside the LVM LVs. Then all I need is a way to find the superblocks, right?