Open Stack

Greetings,

I'd like to propose that we start removing multinode ( 2 node ) based
scenario jobs from the upstream that still exist in queens and rocky.   We
will continue to support a basic multinode deployment  [1 ] across all the
branches. From stable/stein on we only maintain one multinode job, and the
scenarios have all been migrated to single node deployments.

The loss of coverage will be augmented by the RDO promotion pipeline [2]
that should be verifying all the upstream check/gate jobs across the
branches.   Issues will be caught periodically vs. on a per patch basis.
Issues will be reported, debugged and sent to the appropriate teams.

I am not proposing we do this overnight or in one giant patch, but over
time start to remove the upstream coverage.  We also need to ensure that
the upstream version is properly covered in RDO software factory.

WHY???
Two node deployments in upstream infrastructure are very unreliable and
often timeout.  Generally devolving into noise and wasted resources and
time spent by developers trying to land patches  [3].

Comments / Questions??


[1] tripleo-ci-centos-[7,8]-containers-multinode
[2]
https://review.rdoproject.org/zuul/builds?pipeline=openstack-periodic-wednesday-weekend
[3] https://bugs.launchpad.net/tripleo/+bug/1879565
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20200519/31c1c6d3/attachment.html>