LVM/RAID recovery question

[Rich Braun]

Matthew Gillen offered:
> Add a device to your
> LVM, then do a 'pvmove' (the man page has a good example).  Then remove
> your broken "RAID" device from the LVM and re-create it properly, add it
> back to LVM and 'pvmove' back.  I think you can even do all that without
> taking the system offline.

This thing is just totally *gonzo*.  I'm pretty amazed at the behavior I'm
seeing.

So I plug in a new drive (hot-swap, no reboot so far).  Its name becomes
/dev/sdh.  I set up the partition table to match /dev/sda (two RAID1
partitions).  I do this:

 # mdadm --create --level=raid1 --raid-devices=2 --chunk=256 /dev/md5
/dev/sdh2 missing

That creates in theory a new, untainted device into which I should be able to
move the system volume group.  Wrong!  Here's what I get:

 # pvmove /dev/sda2 /dev/md5
 Physical Volume "/dev/sda2" not found in Volume Group "system"

So where is the system vg?

 # pvdisplay -C
  PV         VG     Fmt  Attr PSize   PFree
   ...
  /dev/md5   system lvm2 a-   232.00G   4.00G

The LVM thinks it's already on the newly-generated lvm (with the old size)! 
(The first time I tried this, I used the original md device name /dev/md1, but
the same thing happened when I used a new device /dev/md5!)  If I then do
this, I see the original:

  # mdadm --stop /dev/md5
  # pvdisplay -C
  PV         VG     Fmt  Attr PSize   PFree
   ...
  /dev/sda2  system lvm2 a-   325.97G  97.97G

I'm not sure it's safe to do a pvremove to wipe /dev/md5 (pvdisplay seems to
show the UUID of the /dev/sda2 device for some reason), and if I use pvmove to
there is some other underlying corruption of this volume.

So I have no way of telling pvmove where the source data lives.  I suppose I
could move it from /dev/sda2 to /dev/sdh2 (leaving /dev/md5 stopped) but that
defeats the purpose:  I'm trying to convert this from a raw partition into a
RAID1 volume.

I'll try a reboot to see if that clears this weirdness.

-rich