[setup-storage] current state for v1.4 / fai v4.0 (was: Request for disk_configs (goodie: test-harness script))

[Thomas Neumann]

Hello

Caution:
I'm retracting any patches I posted to this list. I misunderstood some stuff
and introduced my own bugs. I posted an updated version for setup-storage to
the fai-devel list and asked on guidance on how to proceed / integrate these
changes into fai.
Full post: https://lists.uni-koeln.de/pipermail/linux-fai-devel/2012q3/001110.html


The following stuff is now working in setup-storage 1.4:
(It's definitely broken in fai experimental / 4.0.3+0~1346412696.57~1.gbp4b7b93
 but I don't know since when. Some of these problems only occured
 if there was a volume group present.)

preparation phase:
 - removal and wiping of existing LVM volumes
   (vg was deactivated before lv was wiped and removed)
 - removal and wiping of existing LVM volumes on MD
   (md was shut down and restarted, but not LVM -> setup-storage
    attempted to remove logical volumes which were not present)
 - wipefs for logical volumes failed due to missing '/dev/'

execution phase:
 - LVM               (complete reinstall & preserve logical volume)
 - LVM on MD         (complete reinstall & preserve logical volume)
 - LVM on LUKS on MD (complete reinstall)

I also added some debug info:
 - show a list of states which are required but not provided
 - show a list of states which are provided but unused
 - show list of all commands with the states they require and provide
   and provide a small stack trace to the code where they have been
   created (function names and line numbers)

 A 'state' is something like 'vgchange_a_n_VG_testvg'. Executing
 'vgchange -a n testvg' provides this state and 'vgremove [...]'
 requires the presence of this state before it can be run.


All test cases were run against
 - factory clean disks (no partition table)
 - empty partition table (msdos)
 - reinstall (simulate a previous installation with this disk layout)


There's one case from Brian's disk configuration which is still broken,
but judging from http://fai-project.org/doc/man/setup-storage.html it's
currently not possible to preserve an encrypted volume. Well it ~may~ be
possible if one is able to feed the original encryption key to
setup-storage and implement the required logic. [Personally I don't think
that's a good idea.]


Future plans:
The wrapper script I wrote and posted was a real big help, but I now find
it a bit lacking. I'm currently pondering on turning this bash beast into
a nice perl program with the goal of being easily extendable with anybody's
personal disk config. If nothing else it should provide a nice training to
hone my programming and design skills. =)

bye
thomas