[openstack-dev] [TripleO] Easier way of trying TripleO

Robert Collins robertc at robertcollins.net
Thu Nov 21 18:55:31 UTC 2013


On 22 November 2013 05:55, James Slagle <james.slagle at gmail.com> wrote:
> On Tue, Nov 19, 2013 at 6:57 PM, Robert Collins

>> Right now I think we need to continue the two primary arcs we have:
>> CI/CD integration and a CD HA overcloud so that we are being tested,
>> and then work on making the artifacts from those tests available.
>
> Yes, understood.
>
> There are people focused on CI/CD work currently, and I don't think this effort
> would take away from that focus, other than the time it takes to do patch
> reviews, but I don't think that should be a reason not to do it.

It won't *help* with that work:
https://etherpad.openstack.org/p/tripleo-test-cluster - there is more
there that can be helped, and iteration 3 and four are not even
started. And we can't deploy this until we have the cloud to run the
nodepool instances in, which means:
 - the rebuild work (Joe and Roman are on that)
 - migrating our elements to split out state and config properly with
use-ephemeral
 - The HA spec
 - Heat rolling upgrades.

There is a /tonne/ of stuff that will drive TripleO directly towards
higher quality and a wider set of use cases. I'd really like to see
you putting time into one of those things, because they will all
support the virtuous circle of folk using it -> contributing to it.

> It'd be nice to have the images delivered as output from a well tested CD run,
> but I'd like to not delay until we have that.  I think we could make this
> available quicker than we could get to that point.

Sure. But if it's not tested, it's broken right? Our previous
experience from such images is that we had to say 'build your own'
because we didn't know if a given image would work or not.

> Plus, I think if we make this easier to try, we might get more community
> participation.  Right now, I don't think we're attracting people who want to
> try a test/development tripleo based deployment with devtest.  We're really
> only attracting people who want to contribute and develop on tripleo.  Which,
> may certainly be as designed at this point.  But I feel we have a lot of
> positive momentum coming out of summit, so it makes sense to me to try and give
> people something easier to try.

I totally get that, and I would love to see that too. However, I'm
pushing back for three reasons:
 - dib is quite reliable: the issues I see with folk that ask for
support are bad /contents/. And bad Fedora mirrors :(.
 - being able to use something that doesn't meet your needs isn't very
*ahem* useful.
 - we can't push in a meaningful way on a lot of things until the
current core stories are complete, and while folk will work on what
they want to work on, I want to encourage folk to collaborate on the
things the project /really needs/ right now.

> Given that, I think the next steps would be:
>  - add a bit more detail in a blueprint and get some feedback on that
>  - open up cards to track the work in the tripleo trello
>  - start the work on it :).  If it's just me working on it, I'm fine with
>    that.  I expect there may be 1 or 2 other folks that might work on it
>    as well, but these aren't folks that are looking at the CI/CD stories right
>    now.

If it gets /more/ people working on this instead of on the CI or  CD
stories, then I think it's actively harmful as opposed to not-helpful.
I really don't want to see that happen. Lets focus in as a team and
get the hard yards done for a stack we can recommend to anyone and
*then* spread out sideways into useful additional stories like this.

-Rob


-- 
Robert Collins <rbtcollins at hp.com>
Distinguished Technologist
HP Converged Cloud



More information about the OpenStack-dev mailing list