[openstack-dev] [ironic] ocata summit summary

Jim Rollenhagen jim at jimrollenhagen.com
Wed Nov 16 16:02:05 UTC 2016

Hi all,

Sorry for the late-ness of this email, but wanted to send a wrap-up of the
Ocata summit from ironic's perspective. It was another super productive summit
and I'm thankful for all of the people who showed up and made it so. :)

As always, we discussed priorities for the cycle and those have been documented
here: http://specs.openstack.org/openstack/ironic-specs/priorities/ocata-priorities.html
Notes from that session are here:

We discussed a few specs that are up for evolving our API to be more useful
for both humans and computers. Not much contention here, they're all pretty
obvious things to do. Those notes are here:

We talked about improvements for our QA code and CI jobs. Lots of different
topics there. Most notable was agreement to consolidate some of our CI jobs
so we aren't burning a nodepool node per feature (essentially). Another
good discussion was around lack of features in Cirros holding us back. A
couple of options forward here are convincing Cirros to enable these, or
building our own image with these features built in. Notes from this session:

Another session covered using callbacks and event handlers to handle async
actions in Neutron better. This improves those interactions by waiting until
they are done, rather than hoping things complete in time. It involves adding
code to Neutron or a Neutron plugin. Which we add code to was debated, but
Deva and I chatted with Armando later and agreed it should be okay to add it
to Neutron's tree, assuming it shares much of the code with the existing code
to do the same thing with Nova. Notes are here:

We joined the Nova team to discuss how a user might be able to define RAID
configuration during the server-create request. This ended with two options,
which Jay Pipes and Dmitry are going to explore. The first is adding to the
BDM v2 API, and the second is adding to a device metadata tag API. Ironic
will also need to expose a trait that says "can do RAID" to the resource
tracker. Last, we agreed that a flavor may have a default RAID config to be
passed to ironic. And of course, a spec will be needed. However, this work
probably won't happen until Pike or later. Notes:

The task framework session discussed how we might build an API to expose
status and progress for asynchronous operations triggered by the API, as well
as returning data for actions like "send this command to the BMC". We made
some progress, there are concerns about the database schema and how we purge
old data from the DB. Notes here:

Graphical consoles were another session. Much of the session was trying to
figure out how we secure things, and what should be in and out of tree.
Notes here: https://etherpad.openstack.org/p/ironic-ocata-summit-vnc-console

We spent a session figuring out what blockers exist for priorities, and how
to move around those, how to order things to avoid large conflicts, etc.
The notes are basically a plan to get each feature done. This session was
super valuable to me (and I hope, to everyone else), just to get in sync with
where everything is at.

Contributor meetup had three major discussions. First, our project logo. Most
folks were not huge fans of it and agreed to leave feedback. Second, what
projects belong in ironic governance or not. I sent an email about that
already: http://lists.openstack.org/pipermail/openstack-dev/2016-November/106569.html
Third, we decided that the entire core team should be core on specs, to increase
velocity there. I also emailed about that already:

Thanks again to everyone for making this summit awesome. :)

As always, questions/comments/concerns about this stuff are welcome.

// jim

More information about the OpenStack-dev mailing list