Open Stack

On 02/13/2018 01:57 PM, Tom Barron wrote:
> Since python 2.7 will not be maintained past 2020 [1] it is a reasonable 
> conjecture that downstream distributions
> will drop support for python 2 between now and then, perhaps as early as 
> next year.

I'm not sure I agree.  I suspect python 2 support will not go quietly 
into that good night.  Personally I anticipate a lot of kicking and 
screaming right up to the end, especially from change averse enterprise 
users.

But that's neither here nor there.  I think we're all in agreement that 
python 3 support is needed. :-)

> In Pike, OpenStack projects, including TripleO, added python 3 unit 
> tests.  That effort was a good start, but likely we can agree that it is 
> *only* a start to gaining confidence that real life TripleO deployments 
> will "just work" running python 3.  As agreed in the TripleO community 
> meeting, this email is intended to kick off a discussion in advance of 
> PTG on what else needs to be done.
> 
> In this regard it is worth observing that TripleO currently only 
> supports CentOS deployments and CentOS won't have python 3 support until 
> RHEL does, which may be too late to test deploying with python3 before 
> support for python2 is dropped.  Fedora does have support for python 3 
> and for this reason RDO has decided [2] to begin work to run with 
> *stabilized* Fedora repositories in the Rocky cycle, aiming to be ready 
> on time to migrate to Python 3 and support its use in downstream and 
> upstream CI pipelines.

So that means we'll never have Python 3 on CentOS 7 and we need to start 
supporting Fedora again in order to do functional testing on py3? 
That's potentially messy.  My recollection of running TripleO CI on 
Fedora is that it was, to put it nicely, a maintenance headache.  Even 
with the "stabilized" repos from RDO, TripleO has a knack for hitting 
edge case bugs in a fast-moving distro like Fedora.  I guess it's not 
entirely clear to me what the exact plan is since there's some 
discussion of frozen snapshots and such, which might address the 
fast-moving part.

It also means more CI jobs, unless we're okay with dropping CentOS 
support for some scenarios and switching them to Fedora.  Given the 
amount of changes between CentOS 7 and current Fedora that's a pretty 
big gap in our testing.

I guess if RDO has chosen this path then we don't have much choice.  As 
far as next steps, the first thing that would need to be done is to get 
TripleO running on Fedora again.  I suggest starting with 
https://github.com/openstack/instack-undercloud/blob/3e702f3bdfea21c69dc8184e690f26e142a13bff/instack_undercloud/undercloud.py#L1377 
:-)

-Ben