Open Stack

Hello Tony,

it's already implemented twice :-)

[1] is mostly exactly what your propose (except of the removal of
sfdisk).  The problems with this kind of patches:
* If you want to do it somewhat general (i.e. allow more than
  one disk partition) the changes are getting (too) big.
* It is not possible to add things like LVM, MD or encryption -
  because of some basic problems handling parameters with
  dib-run-parts.

[2] is the more general approach: it is completely independent
of the used tool, but describes the partition table as a JSON
structure. An example can be found in [3]: to use GPT you just
need to change 'msdos' to 'gpt' in line 26 (already tested this).

A proposal how to handle block device setup can be found in [4].

Because also [2] is a somewhat big change, we currently try to
split off some general usable things first, like python logging [5],
move hook generation to python [6], or introduce exit phase [7].

As soon as these patches get merged, the original patch needs to
be rebased (and reworked). Maybe there is the need to extract
additional things, so that in the end, it's small and does
exactly one thing.

Would be great if you'd support this refactoring / migration
phase: it's some work.

Kind regards

Andre


[1] https://review.openstack.org/#/c/313938/
[2] https://review.openstack.org/#/c/322671/
[3] https://review.openstack.org/#/c/322671/6/elements/block-device/environment.d/07-block-device-config
[4] https://etherpad.openstack.org/p/C80jjsAs4x
[5] https://review.openstack.org/325409
[6] https://review.openstack.org/271139
[7] https://review.openstack.org/324811