Open Stack

On 01/06/2014 02:30 PM, Vishvananda Ishaya wrote:
> 
> On Jan 6, 2014, at 11:02 AM, Jay Pipes <jaypipes at gmail.com> wrote:
> 
>> Hello Stackers,
>> 
>> I was hoping to get some answers on a few questions I had
>> regarding the Gantt project [1]. Specifically, here are my
>> queries:
>> 
>> 1) Why was Nova forked to the http://github.com/openstack/gantt 
>> repository? Forking Nova just to then remove a bunch of code
>> that doesn't relate to the scheduler code means that we bring
>> 10K+ commits and a git history along with the new project... this
>> seems to be the wrong origin for a project the aims to be a
>> separate service. There's a reason that Cinder and Neutron didn't
>> start out as a fork of Nova, after all…
> 
> Authorship history is nice, but this does seem a bit excessive. The
> cinder strategy of a single squashed fork would have been/still be
> fine I’m sure.

That's not exactly what was done here.

It's a new repo created with the history filtered out.  The history
was only maintained for code kept.  That seems pretty ideal to me.

>> 
>> 2) Why is Gantt in the /openstack GitHub organization? Wouldn't 
>> the /stackforge organization be more appropriate for a project
>> that isn't integrated? If I understand some of the backstory
>> behind Gantt, the idea was to create a scheduler service from the
>> existing Nova scheduler code in order to "complete the work
>> sometime in our lifetime". While I understand the drive to start
>> with something that already exists and iterate over it, I don't
>> understand why the project went right into the /openstack
>> organization instead of following the /stackforge processes for
>> housing code that bakes and gets iterated on before proposing for
>> incubation. Some explanation would be great here.
> 
> This is split-out of existing code so it is following the same path
> as cinder. The goal is to deprecate the existing nova scheduler in
> I. It currently a new project under the nova program I believe.

Correct (compute program, technically).  It's just a mechanical thing,
not new code.  Also, it's not an incubated or integrated project yet.
 It's just an official repo under the compute program.

> 
>> 
>> 3) Where is feature planning happening for Gantt? The Launchpad
>> site for Gantt [2] is empty. Furthermore, there are a number of
>> blueprints for improving the Nova scheduler, notably the
>> no-db-scheduler blueprint [3], which even has code submitted for
>> it and is targeted to Icehouse-2. How are improvements like this
>> planned to be ported (if at all) to Gantt?
> 
> Not sure about the launchpad site. There is a regular scheduler
> group meeting and as I understand it the hope will be to do the
> no-db-scheduler blueprint. There was quite a bit of debate on
> whether to do the no-db-scheduler stuff before or after the
> forklift and I think the consensus was to do the forklift first.

The planning is just being done in nova blueprints right now.  Once
gantt has enough momentum, we can start using a separate launchpad
project.  But we haven't even finished step 1 of making the thing run yet.

>> 
>> 4) Is the aim of Gantt to provide a RESTful HTTP API in addition
>> to the RPC-based API that the existing Nova scheduler exposes?
> 
> In the short term the plan is to just replicate the rpc api, but I
> think a REST api will be considered long term.

Yep.

-- 
Russell Bryant