The DRBD0.7 to DRBD8 upgrade script is just for converting the actual
metadata format and Ganeti configuration.

Before running the script, the following steps need to be done:
  - make sure all disks (remote_raid1) are fully synced and working
    correctly
  - shutdown all instances and make sure no operations on the cluster
    will be done until the end of the upgrade
  - remove the drbd 0.7 kernel module and userspace tools
  - install the drbd8 kernel module and userspace tools
  - run the upgrade script (no arguments)

After the upgrade, it is recommended to run ‘gnt-instance
activate-disks’ for all instances and make sure the disks have started
correctly.

Currently there are some race conditions in the activate-disks that are
triggered by the metadata conversion process. We're investigating these,
but there is no data loss or risk; activating drbd manually (as
secondaries on both sides) is enough to give the peers the chance to
reestablish connections and sync fully.

The script also saves the metadata from each device on the master under
the path it's run from as files named "metadump-$node-$lv_name". In case
of script failure, it should be possible to restore the original state
of the cluster by formatting and restoring the metadata on the drbd
devices and using the save copy of the config file. For this, it is
recommended to save the output of the script.