Open Stack

Updates inline.

On 2/5/2019 8:35 AM, Matt Riedemann wrote:
> I'd like to propose some changes primarily to the CI jobs that run on 
> nova changes, but also impact cinder and tempest.
> 
> 1. Drop the nova-multiattach job and move test coverage to other jobs
> 
> This is actually an old thread [1] and I had started the work but got 
> hung up on a bug that was teased out of one of the tests when running in 
> the multi-node tempest-slow job [2]. For now I've added a conditional 
> skip on that test if running in a multi-node job. The open changes are 
> here [3].

Done: 
https://review.openstack.org/#/q/status:merged+topic:drop-multiattach-job

> 
> 2. Only run compute.api and scenario tests in nova-next job and run 
> under python3 only
> 
> The nova-next job is a place to test new or advanced nova features like 
> placement and cells v2 when those were still optional in Newton. It 
> currently runs with a few changes from the normal tempest-full job:
> 
> * configures service user tokens
> * configures nova console proxy to use TLS
> * disables the resource provider association refresh interval
> * it runs the post_test_hook which runs some commands like 
> archive_delete_rows, purge, and looks for leaked resource allocations [4]
> 
> Like tempest-full, it runs the non-slow tempest API tests concurrently 
> and then the scenario tests serially. I'm proposing that we:
> 
> a) change that job to only run tempest compute API tests and scenario 
> tests to cut down on the number of tests to run; since the job is really 
> only about testing nova features, we don't need to spend time running 
> glance/keystone/cinder/neutron tests which don't touch nova.

Proposed: https://review.openstack.org/#/c/636459/ (+2 from Stephen)

> 
> b) run it with python3 [5] which is the direction all jobs are moving 
> anyway

Done: https://review.openstack.org/#/c/634739/

> 
> 3. Drop the integrated-gate (py2) template jobs (from nova)
> 
> Nova currently runs with both the integrated-gate and 
> integrated-gate-py3 templates, which adds a set of tempest-full and 
> grenade jobs each to the check and gate pipelines. I don't think we need 
> to be gating on both py2 and py3 at this point when it comes to 
> tempest/grenade changes. Tempest changes are still gating on both so we 
> have coverage there against breaking changes, but I think anything 
> that's py2 specific would be caught in unit and functional tests (which 
> we're running on both py27 and py3*).

Proposed: https://review.openstack.org/#/c/634949/ (+2 from Stephen)

-- 

Thanks,

Matt