Open Stack

I think you’re giving a great example of my point that we’re not yet at the stage where we can say, “Any tool should be able to deploy kolla containers”.  Right?

From: Pete Birley <pete at port.direct>
Reply-To: "OpenStack Development Mailing List (not for usage questions)" <openstack-dev at lists.openstack.org>
Date: Thursday, January 5, 2017 at 9:06 PM
To: "OpenStack Development Mailing List (not for usage questions)" <openstack-dev at lists.openstack.org>
Subject: Re: [openstack-dev] [tc][kolla] Adding new deliverables

I'll reply to Britts comments, and then duck out, unless explicitly asked back, as I don't want to (totally) railroad this conversation:

The Kolla containers entry-point is a great example of how the field have moved on. While it was initially required, in the Kkubernetes world the Kolla ABI is actually more of a hindrance than help, as it makes the containers much more of a 'black-box' to use. In the other Openstack on Kubernetes projects I contribute to, and my own independent work, in we actually just define the entry point to the container directly in the k8s manifest and make no use of Kolla's entry point and config mechanisms, either running another 'init' container to build and bind mount the configuration (Harbor), or use Kubernetes configmap object to achieve the same result (Openstack Helm). It would be perfectly possible for Kolla Ansible (and indeed Salt) to take a similar approach - meaning that rather maintaining an ABI that works for all platforms, Kolla would be free to just ensure that the required binaries were present in images.

I agree that this cannot happen overnight, but think that when appropriate we should take stock of where we are and how to plot a course that lets all of our projects flourish without competing for resources, or being so entwined that we become technically paralyzed and overloaded.

Sorry, Sam and Michal! You can have your thread back now :)

On Fri, Jan 6, 2017 at 1:17 AM, Britt Houser (bhouser) <bhouser at cisco.com<mailto:bhouser at cisco.com>> wrote:
I think both Pete and Steve make great points and it should be our long term vision.  However, I lean more with Michael that we should make that a separate discussion, and it’s probably better done further down the road.  Yes, Kolla containers have come a long way, and the ABI has been stable for awhile, but the vast majority of that “for awhile” was with a single deployment tool: ansible.  Now we have kolla-k8s and kolla-salt.  Neither one is fully featured yet as ansible, which to me means I don’t think we can say for sure that ABI won’t need to change as we try to support many deployment tools.  (Someone remind me, didn’t kolla-mesos change the ABI?)  Anyway, the point is I don’t think we’re at a point of maturity to be certain the ABI won’t need changing.  When we have 2-3 deployment tools with enough feature parity to say, “Any tool should be able to deploy kolla containers” then I think it make sense to have that discussion.  I just don’t think we’re there yet.  And until the point, changes to the ABI will be quite painful if each project is in outside of the kolla umbrella, IMHO.

Thx,
britt

From: Pete Birley <pete at port.direct>
Reply-To: "OpenStack Development Mailing List (not for usage questions)" <openstack-dev at lists.openstack.org<mailto:openstack-dev at lists.openstack.org>>
Date: Thursday, January 5, 2017 at 6:47 PM
To: "OpenStack Development Mailing List (not for usage questions)" <openstack-dev at lists.openstack.org<mailto:openstack-dev at lists.openstack.org>>
Subject: Re: [openstack-dev] [tc][kolla] Adding new deliverables

Also coming from the perspective of a Kolla-Kubernetes contributor, I am worried about the scope that Kolla is extending itself to.

Moving from a single repo to multiple repo's has made the situation much better, but by operating under a single umbrella I feel that we may potentially be significantly limiting the potential for each deliverable. Alex Schultz, Steve and Sam raise some good points here.

The interdependency between the projects is causing issues, the current reliance that Kolla-Kubernetes has on Kolla-ansible is both undesirable and unsustainable in my opinion. This is both because it limits the flexibility that we have as Kolla-Kubernetes developers, but also because it places burden and rigidity on Kolla-Ansible. This will ultimately prevent both projects from being able to take advantage of the capabilities offered to them by the deployment mechanism they use.

Like Steve, I don't think the addition of Kolla-aSlt should affect me, and as a result don't feel I should have any say in the project. That said, I'd really like to see it happen in one form or another - as having a wide variety of complementary projects and tooling for OpenStack deployment can only be a good thing for the community if correctly managed.

When Kolla started it was very experimental, containers (In their modern form) were a relatively new construct, and it took on the audacious task of trying to package and deploy OpenStack using the tooling that was available at the time. I really feel that this effort has succeeded admirably, and conversations like this are a result of that. Kolla is one of the most active projects in OpenStack, with two deployment mechanisms being developed currently, and hopefully to increase soon with a salt based deployment and potentially even more on the horizon.

With this in mind, I return to my original point and wonder if we may be better moving from our current structure of Kolla-deploy(x) to deploy(x)-Kolla and redefine the governance of these deliverables, turning them into freestanding projects. I think this would offer several potential advantages, as it would allow teams to form tighter bonds with the tools and communities they use (ie Kubernetes/Helm, Ansible or Salt). This would also make it easier for these projects to use upstream components where available (eg Ceph, RabbitMQ, and MariaDB) which are (and should be) in many cases better than the artifacts we can produce. To this end, I have been working with the Ceph community to get their Kubernetes Helm implementation to the point where we can use it for our own work, and would love to see more of this. It benefits not only us by offloading support to the upstream project, but gives them a vested interest in supporting us and also helps provide better quality tooling for the entire open source ecosystem.

This should also allow Kolla itself to become much more streamlined, and focused simply on producing docker containers for consumption by the community, and make the artifacts produced potentially much less opinionated and more attractive to other projects. And being honest, I have a real desire for this activity to eventually be taken on by the relevant OpenStack projects themselves - and would love to see Kolla help develop a framework that would allow projects to take ownership of the containerisation of their output.

Sorry for such a long email - but this seems like a good opportunity to raise some of these issues that have been on my mind. In summary, if it doesn't affect me then I wish a Salt based Kolla deployment the best of success and hope to see the project prosper so that we as OpenStack developers can all share from the increased experience and opportunities growing the community offers.


On Thu, Jan 5, 2017 at 9:43 PM, Steve Wilkerson <wilkers.steve at gmail.com<mailto:wilkers.steve at gmail.com>> wrote:
There are some interesting points in this topic.  I agree entirely with Sam Yaple.  It does not make sense to me to have kolla-ansible and kolla-kubernetes cores involved with the introduction of a new deliverable under the kolla umbrella.  A new deliverable (read: project, really) should not rely on a separate project to ratify its existence.  I feel this is dangerous.  I also feel looking at the different deployment methodologies scoped under the kolla project as competition or rivalry is folly.  I'm honestly a bit concerned about how broad the scope of the project kolla has become.  I think the conversation of separating the deployment projects from the kolla umbrella is a conversation worth having at some point.

The repo split was a step in the right direction, but currently the deliverables (4, if kolla-salt becomes a thing) are sharing a single PTL, a single IRC channel, and a single IRC weekly meeting.  This has the potential of introducing a significant amount of overhead for the overarching project as a whole.  What happens if kolla-puppet becomes a thing?  What if kolla-mesos was still about?  I think we can all agree this gets out of hand quickly.

Yes, people are religious about the tools they use, and deployment tools are no different.  I think scoping them all under the same umbrella project is a mistake in the long term.  The folks that want to focus on Ansible should be able to focus wholly on Ansible with like-minded folks, same for Salt, same for whatever.  Having them remain together for the sake of sharing a name isn't sustainable in the long term -- let each do what they do well.  As far as being able to talk and share experiences in deployments or whatever, let's not act as if IRC channels have walls we can't reach across.  As part of the kolla-kubernetes community, it's imperative that I can reach across the gap to work with people in the Helm and Kubernetes community.  If the deployment tools existed separately, there's nothing stopping them from asking either.

But in regards to the question, if kolla-salt is to be a thing, I think the PTL and the kolla team proper can decide that.  As a contributor for kolla-kubernetes, it does not and should not affect me.

On Thu, Jan 5, 2017 at 3:14 PM, Doug Hellmann <doug at doughellmann.com<mailto:doug at doughellmann.com>> wrote:
Excerpts from Michał Jastrzębski's message of 2017-01-05 11:45:49 -0800:
> I think total separation of projects would require much larger
> discussion in community. Currently we agreed on having kolla-ansible
> and kolla-k8s to be deliverables under kolla umbrella from historical
> reasons. Also I don't agree that there is "little or no overlap" in
> teams, in fact there is ton of overlap, just not 100%. Many
> contributors (myself included) jump between deliverables today.

OK, that's good to know. It wasn't clear from some of the initial
messages in this thread, which seemed to imply otherwise.

Doug

__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe<http://OpenStack-dev-request@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://OpenStack-dev-request@lists.openstack.org?subject:unsubscribe>
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev



--

[rt.direct]<https://port.direct>


Pete Birley / Director
pete at port.direct<mailto:pete at port.direct> / +447446862551<tel:+44%207446%20862551>

PORT.DIRECT
United Kingdom
https://port.direct




This e-mail message may contain confidential or legally privileged information and is intended only for the use of the intended recipient(s). Any unauthorized disclosure, dissemination, distribution, copying or the taking of any action in reliance on the information herein is prohibited. E-mails are not secure and cannot be guaranteed to be error free as they can be intercepted, amended, or contain viruses. Anyone who communicates with us by e-mail is deemed to have accepted these risks. Port.direct is not responsible for errors or omissions in this message and denies any responsibility for any damage arising from the use of e-mail. Any opinion and other statement contained in this message and any attachment are solely those of the author and do not necessarily represent those of the company.



__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe<http://OpenStack-dev-request@lists.openstack.org?subject:unsubscribe>
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev



--

[ort.direct]<https://port.direct>


Pete Birley / Director
pete at port.direct<mailto:pete at port.direct> / +447446862551

PORT.DIRECT
United Kingdom
https://port.direct




This e-mail message may contain confidential or legally privileged information and is intended only for the use of the intended recipient(s). Any unauthorized disclosure, dissemination, distribution, copying or the taking of any action in reliance on the information herein is prohibited. E-mails are not secure and cannot be guaranteed to be error free as they can be intercepted, amended, or contain viruses. Anyone who communicates with us by e-mail is deemed to have accepted these risks. Port.direct is not responsible for errors or omissions in this message and denies any responsibility for any damage arising from the use of e-mail. Any opinion and other statement contained in this message and any attachment are solely those of the author and do not necessarily represent those of the company.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20170106/56ac28d2/attachment.html>