Comment 7 for bug 1372375

The bug I submitted wants to be related more to the flow of the volume encryption procedure than to the LUKS header backup, that is also important.
I have no idea, at the moment, on where to store luks headers' backup, I'm looking around.
However a backup of the whole volume could be sufficient, but in my opinion it could become expensive in some cases.
I focused on luksHeaderBackup and luksHeaderRestore to highlight on how formatting a volume is not a valid choice when the device is not recognized as valid luks.
I agree with the decision to format the device in Nova, but I think that Cinder should ask luksFormat to nova at the moment of volume creation, and not when attaching for the problem above.