[Pulp-dev] [CI] [Installer] We should not test ansible-pulp against multiple versions of Python on the mgmt node

[Mike DePaulo]

Hi CI & Installer teams,

I talked with James Tanner (who works on Ansible Core), and his advice was
that we should not bother to test our ansible-pulp (a collection of roles)
against multiple versions of Python on the management node (the Travis
host.)

We still need to test a variety of Linux distros (with their bundled python
versions) for managed nodes, which we do (molecule creates containers for
multiple distros.)

This is because we meet the following criteria:
1. We do not write any python code, "action plugins" that run on the
management node (and none for the managed node either, like role-bundled
modules.)
2. We use few "action plugins" on the management node (the "package" module
includes one to decide whether to use dnf/yum/apt/etc), and those we do use
are probably all well tested by Ansible's first-party CI. It's difficult to
enumerate those we use, and not worth enumerating. It's not like we use
community-supported action plugins.

If there are no objections, I will create a low redmine tasks to make our
Travis PR approach (static test case on Python 2.7 only, upgrade test case
on 3.6 only, dynamic test case on 3.7 only) replace our complete 3x3 matrix
approach of cron jobs.

If in the future we do not test every version of Python on the management
node, we should not be overly concerned either.

-Mike

-- 

Mike DePaulo

He / Him / His

Service Reliability Engineer, Pulp

Red Hat <https://www.redhat.com/>

IM: mikedep333

GPG: 51745404
<https://www.redhat.com/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/pulp-dev/attachments/20200214/a07fdcb7/attachment.htm>