[openstack-dev] [tc][kolla] Adding new deliverables

Michał Jastrzębski inc007 at gmail.com
Wed Jan 11 20:26:35 UTC 2017


Hey Brandon,

So couple comments to your mail.
Kolla at it's core is community which took on preparing deployment of
openstack using docker containers. This same community now is working with
both ansible and k8s as means to deploy these containers. So far we're
preserving that single community to allow full cooperation and let's be
honest, we still learn and experiment, especially in k8s space.

As for k8s being abstraction layer to container runtime, it indeed is, but
that's only part of the story. Container runtime is less important than
it's ABI for k8s deployment mechanism. With OCI container, what we today
have as docker containers can (don't know really, never tested) be
compatible with RKT, maybe with a bit of work. What's more important is how
to interact with these containers. Kolla honed our containers ABI over
multiple releases and we are still working on it. While k8s can run
multiple container formats, how do you interact with them depends on how
containers are built. While I can clearly see benefit of having
multi-runtime mechanism like that, all containers should follow same ABI
for deployment code to consume, and as far as I know (please, correct me if
I'm wrong), there is no alternative to Kolla's images that would be
compatible with Kolla ABI. So question about multiple runtimes becomes
hypothetical until one these appear. If there is community that is working
on alternative image format, I'd love to talk to them so we can try to keep
our ABIs compatible so deployment projects like one you describe can have
this choice too. I'd go further still, if such project would appear
(alternative container format), I'd be happy to discuss kolla-ansible and
kolla-k8s being able to consume it too! Just...nobody did that, doing that
or plant that as far as I know.

Cheers,
Michal

On 11 January 2017 at 12:09, Steven Dake (stdake) <stdake at cisco.com> wrote:

> Sure – you asked me and I thought you wanted an answer from me (which fits
> under the do not use OpenStack properties (i.e. this mailing list) for
> promotion of candidates email that Mark sent out).
>
>
>
> Others are able to answer in the broader Kolla community.
>
>
>
> Regards
>
> -steve
>
>
>
> *From: *"Brandon B. Jozsa" <bjozsa at jinkit.com>
> *Date: *Wednesday, January 11, 2017 at 1:01 PM
> *To: *"Britt Houser (bhouser)" <bhouser at cisco.com>, "Steven Dake
> (stdake)" <stdake at cisco.com>, "OpenStack Development Mailing List (not
> for usage questions)" <openstack-dev at lists.openstack.org>
>
> *Subject: *Re: [openstack-dev] [tc][kolla] Adding new deliverables
>
>
>
>
>
> I’m not entirely sure how the two relate, but anyone from Kolla can
> respond.
>
>
>
> Brandon B. Jozsa
>
>
>
> On January 11, 2017 at 2:49:07 PM, Steven Dake (stdake) (stdake at cisco.com)
> wrote:
>
> Brandon,
>
>
>
> Your question is a mix of political and technical aspects that I am not
> permitted to answer until Monday because of my parsing of this email from
> Mark Collier:
>
>
>
> http://lists.openstack.org/pipermail/foundation/2017-January/002446.html
>
>
>
> I will answer you Monday after the individual board of directors elections
> conclude.
>
>
>
> Regards
>
> -steve
>
>
>
>
>
> *From: *"Brandon B. Jozsa" <bjozsa at jinkit.com>
> *Reply-To: *"OpenStack Development Mailing List (not for usage
> questions)" <openstack-dev at lists.openstack.org>
> *Date: *Wednesday, January 11, 2017 at 12:36 PM
> *To: *"Britt Houser (bhouser)" <bhouser at cisco.com>, "OpenStack
> Development Mailing List (not for usage questions)" <openstack-dev at lists.
> openstack.org>
> *Subject: *Re: [openstack-dev] [tc][kolla] Adding new deliverables
>
>
>
> To your point Steve, then I’d image that Kolla would have no objection to
> the introduction of other Openstack-namespace projects that provide
> alternative image formats, integration choices, or orchestration variances
> for those in the larger community who do not want to use Kolla images. All
> of the Kolla-x projects point to this one source of truth in the end. This
> results in large to the many projects falling under the Kolla umbrella:
> Kolla, Kolla-Mesos, Kolla-Ansible, Kolla-Kubernetes, Kolla-Salt, and I’d
> assume whatever else wants to consume Kolla, if things continue as they are.
>
>
>
> My immediate ask is "what are the potential negative impacts to Kolla
> having so many projects under one mission”: fragmentation of goals,
> misunderstanding of mission, increased developer debt across each
> inter-twined project (cross-repo commits and reviews), complex gating
> requirements? #kolla has been a place of spirited debate with the recent
> addition of Kolla-Kubernetes, and I think some of this is the result of the
> problems I’m alluding to. It’s very difficult to preserve what Kolla is at
> it’s core, and in turn preserve the benefits of something like Kubernetes
> which has a Runtime Interface abstraction model. It’s a tough sell for the
> larger Openstack community, and this is a critical time for Openstack and
> CNCF interoperability; would you not agree?
>
>
>
> I’m failing to see the benefits you mention outweighing what others might
> see as potential pitfalls. My viewpoint is not news to those in Kolla. I’ve
> expressed this in Kolla already, and this is why I’m disappointed when
> Kolla-Kuberntes drops Secs in favor of quicker ad-hoc IRC
> architecturally-focused discussions.
>
>
>
> So my question now becomes; "How is Kolla addressing these issues, and
> what has Kolla been doing with the assistance of the Openstack Foundation
> to gain the confidence of those who are watching Kolla and looking for that
> next cool container project”?
>
>
>
> Brandon B. Jozsa
>
>
>
> On January 11, 2017 at 1:46:13 PM, Britt Houser (bhouser) (
> bhouser at cisco.com) wrote:
>
> My sentiments exactly Michal. We’ll get there, but let’s not jump the gun
> quite yet.
>
> On 1/11/17, 1:38 PM, "Michał Jastrzębski" <inc007 at gmail.com> wrote:
>
> So from my point of view, while I understand why project separation
> makes sense in the long run, I will argue that at this moment it will
> be hurtful for the project. Our community is still fairly integrated,
> and I'd love to keep it this way a while longer. We haven't yet fully
> cleaned up mess that split of kolla-ansible caused (documentation and
> whatnot). Having small revolution like that again is something that
> would greatly hinder our ability to deliver valuable project, and I
> think for now that should be our priority.
>
> To me, at least before we will have more than one prod-ready
> deployment tool, separation of projects would be bad. I think project
> separation should be a process instead of revolution, and we already
> started this process by separating kolla-ansible repo and core team.
> I'd be happy to discuss how to pave road for full project separation
> without causing pain for operators, users and developers, as to me
> their best interest should take priority.
>
> Cheers,
> Michal
>
> On 11 January 2017 at 09:59, Doug Hellmann <doug at doughellmann.com> wrote:
> > Excerpts from Steven Dake (stdake)'s message of 2017-01-11 14:50:31
> +0000:
> >> Thierry,
> >>
> >> I am not a big fan of the separate gerrit teams we have instituted
> inside the Kolla project. I always believed we should have one core
> reviewer team responsible for all deliverables to avoid not just the
> appearance but the reality that each team would fragment the overall
> community of people working on Kolla containers and deployment tools. This
> is yet another reason I didn’t want to split the repositories into separate
> deliverables in the first place – since it further fragments the community
> working on Kolla deliverables.
> >>
> >> When we made our original mission statement, I originally wanted it
> scoped around just Ansible and Docker. Fortunately, the core review team at
> the time made it much more general and broad before we joined the big tent.
> Our mission statement permits multiple different orchestration tools.
> >>
> >> Kolla is not “themed”, at least to me. Instead it is one community with
> slightly different interests (some people work on Ansible, some on
> Kubernetes, some on containers, some on all 3, etc). If we break that into
> separate projects with separate PTLs, those projects may end up competing
> with each other (which isn’t happening now inside Kolla). I think
> competition is a good thing. In this case, I am of the opinion it is high
> time we end the competition on deployment tools related to containers and
> get everyone working together rather than apart. That is, unless those
> folks want to “work apart” which of course is their prerogative. I wouldn’t
> suggest merging teams today that are separate that don’t have a desire to
> merge. That said, Kolla is very warm and open to new contributors so
> hopefully no more new duplicate effort solutions are started.
> >
> > It sure sounds to me like you want Kolla to "own" container deployment
> > tools. As Thierry said, we aren't intended to be organized that way any
> > more.
> >
> >> Siloing the deliverables into separate teams I believe would result in
> the competition I just mentioned, and further discord between the
> deployment tool projects in the big tent. We need consolidation around
> people working together, not division. Division around Kolla weakens Kolla
> specifically and doesn’t help out OpenStack all that much either.
> >
> > I would hope that the spirit of collaboration could extend across team
> > boundaries. #WeAreOpenStack
> >
> > Doug
> >
> >>
> >> The idea of branding or themes is not really relevant to me. Instead
> this is all about the people producing and consuming Kolla. I’d like these
> folks to work together as much as feasible. Breaking a sub-community apart
> (in this case Kolla) into up to 4 different communities with 4 different
> PTLs sounds wrong to me.
> >>
> >> I hope my position is clear ☺ If not, feel free to ask any follow-up
> questions.
> >>
> >> Regards
> >> -steve
> >>
> >> -----Original Message-----
> >> From: Thierry Carrez <thierry at openstack.org>
> >> Organization: OpenStack
> >> Reply-To: "OpenStack Development Mailing List (not for usage
> questions)" <openstack-dev at lists.openstack.org>
> >> Date: Wednesday, January 11, 2017 at 4:21 AM
> >> To: "openstack-dev at lists.openstack.org" <openstack-dev at lists.
> openstack.org>
> >> Subject: Re: [openstack-dev] [tc][kolla] Adding new deliverables
> >>
> >> Michał Jastrzębski wrote:
> >> > I created CIVS poll with options we discussed. Every core member
> should
> >> > get link to poll voting, if that's not the case, please let me know.
> >>
> >> Just a quick sidenote to explain how the "big-tent" model of governance
> >> plays in here...
> >>
> >> In the previous project structure model, we had "programs". If you
> >> wanted to do networking stuff, you had to join the Networking program
> >> (neutron). If you worked on object storage, you had to join the Object
> >> Storage program (swift). The main issue with this model is that it
> >> prevented alternate approaches from emerging (as a program PTL could
> >> just refuse its emergence to continue to "own" that space). It also
> >> created weird situations where there would be multiple distinct groups
> >> of people in a program, but a single PTL to elect to represent them
> all.
> >> That created unnecessary political issues within programs and tension
> >> around PTL election.
> >>
> >> Part of the big-tent project structure reform was to abolish programs
> >> and organize our work around "teams", rather than "themes". Project
> >> teams should be strongly aligned with a single team of people that work
> >> together. That allowed some amount of competition to emerge (we still
> >> try to avoid "gratuitous duplication of effort"), but most importantly
> >> made sure groups of people could "own" their work without having to
> >> defer to an outside core team or PTL. So if you have a distinct team,
> it
> >> should be its own separate project team with its own PTL. There is no
> >> program or namespace anymore. As a bonus side-effect, it made sure
> teams
> >> would not indefinitely grow, and we all know that it's difficult to
> grow
> >> core teams (and trust) beyond a certain point.
> >>
> >> This is why we have multiple packaging project teams, each specialized
> >> in a given package orchestration mechanism, rather than have a single
> >> "Packaging" program with a single PTL and Ansible / Puppet / Chef
> >> fighting in elections to get their man at the helm. This is why the
> >> Storlets team, while deeply related to Swift and in very good
> >> collaboration terms with them, was set up as a separate project team.
> >> Different people, different team.
> >>
> >> The fact that you're having hard discussions in Kolla about "adding new
> >> deliverables" produced by distinct groups of people indicates that you
> >> may be using Kolla as an old-style "program" rather than as a single
> >> team. Why not set them up as separate project teams ? What am I missing
> >> here ?
> >>
> >> --
> >> Thierry Carrez (ttx)
> >>
> >> __________________________________________________________________________
>
> >> OpenStack Development Mailing List (not for usage questions)
> >> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:
> unsubscribe
> >> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
> >>
> >>
> >> __________________________________________________________________________
>
> >> OpenStack Development Mailing List (not for usage questions)
> >> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:
> unsubscribe
> >> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
> >
> > __________________________________________________________________________
>
> > OpenStack Development Mailing List (not for usage questions)
> > Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:
> unsubscribe
> > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
> __________________________________________________________________________
>
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
> __________________________________________________________________________
>
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20170111/845e7d0a/attachment.html>


More information about the OpenStack-dev mailing list