raid names incl. hostname are cut off after 31 characters.

The inventory script will use ipv4 for a host which has no ipv6 defined.

This allows specifying which branch (or commit) should be checked out by
the initial ansible-pull run after reboot.

GRUB can read /boot from a LVM logical volume on a md RAID 6 – except if
two disks are missing. RAID 6 should continue to work in this situation
(although in a degraded state), but GRUB probably lacks the code to
recover the missing data from the parity data (or something). In any
case tests with the previous RAID 6-only setup showed that GRUB couldn't
boot any more with missing disks (showing an "error: disk
`lvmid/.../...´ not found" error message). Thus /boot is moved to its
own RAID 1 which is easier to read even in a degraded state because all
disks contain all data.

Wiping and disassembling filesystems, LVM volumes, md raids, partitions
and partition tables has to be done from the inside to the outside,
telling the kernel to deactivate volumes etc. at the right stages. The
previous version of this code was erroneous and caused bootstrap.sh to
fail.