[Soc-coordination] Compute Clusters Integration for Debian Development and Building - Report 3

[Rudy Godoy]

Hello this is the third report regarding my project. As usual, quick
summary first. From my last report I have progressed much little than I
would like but, as stated previously also, I'm finishing semester
and between exams and projects it takes almost all my time. However, I'm
free of academic duties next week and I'll double the speed on the project. 


What was done so far
--------------------
- Started working on the modifications to the Eucalyptus code, as
  expected. 
  - I've almost complete the code for the Libvirt's XML Domain
  definition. I've abstracted the wrapper in order to support a broader 
  number of arches. Previously it was tailored for supporting x86/amd64
  based machines. The idea is to have a list of supported arches and
  generate the XML file according that. For ARM some parameters and
  elements are different than for the others. This would allow for
  people to support another architectures, given they are well supported
  by KVM-Qemu.
  - Started working on changes to util/data.c to make sure the correct
  parameters are being passed to the XML generator and to libvirt kvm
  calls.
- Define and start to setup the testing / demo environment. Basically
  involves having one machine to run KVM as hypervisor so the Eucalyptus
  handler uses it to run ARM kvm-qemu images.
- Over this weekend I'll be publising the code changes on my
  website. Steffen suggested to create some docs regarding the progress
  too. I'll do that once I feel I'm on track again for my current
  pending work, that's coding.

What I'll be working over the next two weeks
--------------------------------------------
- My primary goal is to catch up on the work pace and finish the neccesary
  code for Eucalyptus to support ARM images. This means to finish
  modifications to:
  - node/handlers_kvm.c - handling nc creation. Little changes are need
  here from what I've indentified already. 
  - util/data.{c,h} - handling instance creation. This is the part where
  most of the work shall be done. I've started to work here already.
- Test the code I've complete already and send the patches to the teams
  involved. To date I'm still working on a local environment, because no
  usable patches have been generated already, just little pieces here
  and there.
- Since this is a sort of gluing project, that is code here and there
  and make it work together. I'm in the process of finding a machine to
  be able to test parts of it once they are complete. I'm still
  contacting people to get such machine, in anycase I gues I can manage
  to get one myself. I can't use my current machine since it involves
  reinstalling and working at the OS level.
- For the project's "demo" I'll use both machines, one as the front-end
  and the kvm one for NCs (nodes). That's the goal.

What is required after that
---------------------------
- I'd like to release a package that includes my changes, so it can be
  installed either from official packages or my own packages, this
  involves: 
  - Updating the packaging bits to the pkg-Eucalyptus repo.
  - Talking to upstream about it.
- Write the reference Steffen suggested.


I believe that I can finish the first part in one week, because of the
work I've done in the past weeks to understand the Eucalyptus code. I'm
also will be talking more to the Eucalyptus team and also to the
maintainers about such changes.

I guess that's all for now.

Best regards,
Rudy