We've decided that, like Workstation, we want to focus on deliverable,
complete OS environments for users instead of "parts". For Container
Cloud, that looks a lot like Fedora+OpenShiftOrigin, rather than just
Atomic Host.
I feel like I've not been out of the loop (aside from the recent FAD) but this comes as a bit of a surprise to me.
To be clear I don't see this as an "instead", but more of "in addition".
I think OpenShift is a *lot* more pretty face to be presenting by default than
just raw Kubernetes, and in addition we need to drive the integration
story between S2I and Fedora/RPM packaging.
This is a lot less scary, since I don't think moving the whole SIG to focus on a Fedora based delivery of Origin is all that interesting. I definitely agree that Origin is prettier than the current state, but that we should be looking to improve that state. It seems like "PaaS is the answer" comes from the existence of a PaaS, not from what users might want from an Atomic Host Cluster. S2I is great, but not the only way to build containers, and definitely drives the Fedora container story in an opinionated direction that also impacts people running Atomic Host Clusters (must have a registry and K8S implementation that understands the ImageStream extension) away from "general" use cases for K8S and Docker.
I love Origin, but I just don't see it as the "right" answer for "I need a better way to manage a Docker/Rocket/Kubernetes cluster". I absolutely think that Atomic Host as the execution environment for Origin is The Right Thing (TM) and we should be working together there. I think there's missing cluster management in both projects right now, but Cockpit and Commissaire are making big strides in that arena.
I can also see the potential argument that Atomic Host on it's own wasn't compelling enough for a full Edition. So, I hope if we do move toward an Atomic+Origin as primary deliverable, we don't lose focus on Atomic Host Clusters, since one can make the argument that while related there's two separate use cases.
my knee-jerk $0.02
- Matt M
On Wed, Jun 15, 2016 at 2:27 PM, Colin Walters <walters@xxxxxxxxxx> wrote:
On Wed, Jun 15, 2016, at 02:09 PM, Josh Boyer wrote:
>
> To me it is. As I said originally, something that might be
> self-evident to the WG likely isn't to someone that is following the
> IRC meetings, etc. It's a case of "we said we were going to do thing
> X on the list, suddenly on the list we're going to do thing Y
> instead." Please note that I did NOT say either thing X or thing Y
> were bad or negative. Just surprising.
To be clear I don't see this as an "instead", but more of "in addition".
I think OpenShift is a *lot* more pretty face to be presenting by default than
just raw Kubernetes, and in addition we need to drive the integration
story between S2I and Fedora/RPM packaging.
Currently a major disconnect with Atomic Host is that Kubernetes
is included, whereas many many people want to control the
version, or use OpenShift.
So we're looking at addressing this by dropping Kube out of the host
by default, and supporting installation via system containers (`atomic install --system`)
or package layering.
> Also, while not explicitly so, Atomic Host (more os-tree) is a
> foundation for the work that Workstation is looking at as well. Which
> is where some of my surprise comes from I guess. I thought we'd
> finally have some cohesion between the Editions, at least at a
> fundamentals level. That's still possible I guess.
Given the above then, WorkstationOstree is still entirely possible
and makes sense. I'm in fact running builds of it right now =)
The only crosscutting thing here was - does it make sense to
ship the Workstation as Docker images instead, and use Atomic Host?
I don't think that's quite technically feasible yet, though with the
system containers work we're potentially closer. That's a big
discussion.
_______________________________________________
cloud mailing list
cloud@xxxxxxxxxxxxxxxxxxxxxxx
https://lists.fedoraproject.org/admin/lists/cloud@xxxxxxxxxxxxxxxxxxxxxxx
_______________________________________________ cloud mailing list cloud@xxxxxxxxxxxxxxxxxxxxxxx https://lists.fedoraproject.org/admin/lists/cloud@xxxxxxxxxxxxxxxxxxxxxxx
