TT#15829 Wipe any possibly existing LVM/PV label from target disk

Starting with lvm2 v2.02.111 it includes a udev rules file named
/lib/udev/rules.d/69-dm-lvm-metad.rules (see
http://sources.debian.net/src/lvm2/2.02.111-2.2%2Bdeb8u1/udev/69-dm-lvm-metad.rules.in/
wheras its "(PVSCAN_RULE)" corresponds to
RUN+="/sbin/lvm pvscan --cache --activate ay --major $major --minor $minor", ENV{LVM_SCANNED}="1"
in the installed /lib/udev/rules.d/69-dm-lvm-metad.rules).

The udev rule gets triggered during disk changes and
automatically invokes `lvm pvscan [...]`, triggering the
following behavior if a LV is already present on a disk:

| # dmsetup ls # it's empty
| # parted -s /dev/sda mklabel msdos
| # parted -s /dev/sda mkpart primary 1048576B 8589934591B
| # dmsetup ls # this will list already existing LVs, like:
| vg0-swap ....
| vg0-root ...
| # blockdev --rereadpt /dev/sda # this will fail now with "Device or resource busy"

This causes deployments to fail if it's reinstalled on
a system that already had any LVs on the target disks.

Change-Id: I35cdd437e2c8de0ca956adce9891bb0b23f70d3b

deployment.sh

@@ -1165,7 +1165,12 @@ ${VG_NAME}-root     /       -95%      ext3 rw
 ${VG_NAME}-swap     swap    RAM:50%   swap sw $lv_create_opts
 EOF
 
-  # make sure setup-storage doesn't fail if LVM is already present
+  # make sure setup-storage/parted doesn't fail if LVM is already present
+  blockdev --rereadpt "/dev/${DISK}"
+  for disk in "/dev/${DISK}"* ; do
+    logit "Removing possibly existing LVM/PV label from $disk"
+    pvremove "$disk" --force --force --yes || true
+  done
   dd if=/dev/zero of="/dev/${DISK}" bs=1M count=1
   blockdev --rereadpt "/dev/${DISK}"