[0] raw/~ # mdadm --create --verbose /dev/md0 --level=mirror --raid-devices=2 /dev/sda /dev/sdb
mdadm: /dev/sda appears to be part of a raid array:
    level=raid1 devices=2 ctime=Thu Aug 22 12:15:05 2013
mdadm: Note: this array has metadata at the start and
    may not be suitable as a boot device.  If you plan to
    store '/boot' on this device please ensure that
    your boot-loader understands md/v1.x metadata, or use
    --metadata=0.90
mdadm: /dev/sdb appears to be part of a raid array:
    level=raid1 devices=2 ctime=Thu Aug 22 12:15:05 2013
mdadm: size set to 130944K
Continue creating array? y
mdadm: Defaulting to version 1.2 metadata
mdadm: array /dev/md0 started.

[0] raw/~ # pvcreate /dev/md0
  Physical volume "/dev/md0" successfully created

[0] raw/~ # vgcreate vg /dev/md0
  Volume group "vg" successfully created

[0] raw/~ # lvcreate -l1 vg
  Logical volume "lvol0" created

[0] raw/~ # dmsetup info -c -o name,attr
Name             Stat
fedora_raw-swap  L--w
fedora_raw-root  L--w

[0] raw/~ # vgchange -ay vg
  1 logical volume(s) in volume group "vg" now active

[0] raw/~ # dmsetup info -c -o name,attr
Name             Stat
fedora_raw-swap  L--w
vg-lvol0         L--w
fedora_raw-root  L--w

[0] raw/~ # vgchange -an vg
  0 logical volume(s) in volume group "vg" now active

[0] raw/~ # dmsetup info -c -o name,attr
Name             Stat
fedora_raw-swap  L--w
fedora_raw-root  L--w

[0] raw/~ # mdadm -S /dev/md0
mdadm: stopped /dev/md0

[0] raw/~ # dmsetup info -c -o name,attr
Name             Stat
fedora_raw-swap  L--w
vg-lvol0         ---w
fedora_raw-root  L--w


And assosciated log from dmesg after mdadm -S:

[ 5046.297812] md: md0 still in use.
[ 5046.333193] device-mapper: table: 253:2: linear: dm-linear: Device lookup failed
[ 5046.336647] device-mapper: ioctl: error adding target to table
[ 5046.500320] md0: detected capacity change from 134086656 to 0
[ 5046.503839] md: md0 stopped.
[ 5046.505986] md: unbind<sdb>
[ 5046.506862] md: export_rdev(sdb)
[ 5046.507715] md: unbind<sda>
[ 5046.510534] md: export_rdev(sda)

And udev log after mdadm -S:

[1] raw/~ # udevadm monitor --kernel    
monitor will print the received events for:
KERNEL - the kernel uevent

KERNEL[5046.296878] change   /devices/virtual/block/md0 (block)
KERNEL[5046.330424] add      /devices/virtual/bdi/253:2 (bdi)
KERNEL[5046.331208] add      /devices/virtual/block/dm-2 (block)
KERNEL[5046.511436] change   /devices/virtual/block/md0 (block)
KERNEL[5046.511777] change   /devices/virtual/block/md0 (block)
KERNEL[5046.512331] remove   /devices/virtual/bdi/9:0 (bdi)
KERNEL[5046.512496] remove   /devices/virtual/block/md0 (block)


So there's a CHANGE event followed by REMOVE event when stopping the MD array. The CHANGE event for md0 causes the autoactivation to trigger. At the time the LVM on MD is autoactivated, the MD itself is removed. That explains why we end up with the vg-lvol0 created, but not loaded - the MD underneath is already gone when trying to load the table:

[ 5046.333193] device-mapper: table: 253:2: linear: dm-linear: Device lookup failed

So what we need is to filter out that CHANGE even that's just before REMOVE... That's going to be a little bit harde since the CHANGE even before REMOVE is exactly the same as the CHANGE event that notifies about MD array activation (there's no variable we could make a comparison with).is lvmdump I hope)

--- Additional comment from Peter Rajnoha on 2013-09-11 13:42:32 CEST ---

This patch should fix the issues reported:

  https://git.fedorahosted.org/cgit/lvm2.git/commit/?id=8d1d83504dcf9c86ad42d34d3bd0b201d7bab8f6