Description of problem: When running tempest in our CI and using ceph, the nova resize tests are disabled due to the compute_feature_enabled.resize flag not being enabled in the tempest.conf file, which depends on shared storage. As the tempest config tool can't discover by itself the kind of object storage that is running on the system, we'll need to rely on director to provide the setting when creating the deployer-input-conf file. Version-Release number of selected component (if applicable): How reproducible: Run CI on a downstream tempest with ceph enabled. nova resize tests would be skipped due to the skip decorator: @testtools.skipUnless(CONF.compute_feature_enabled.resize Expected results: For them to be run. Additional info:
This bug did not make the OSP 8.0 release. It is being deferred to OSP 10.
I gave a quick look at how this could be implemented; from what I can tell, it would be trivial to add a condition in tripleoclient to set the additional value in the deployer input file based on the number of CephStorage nodes. Yet, given we have composable roles now, that won't cover scenarios where Ceph is either external or deployed on custom roles. To behave correctly in those scenarios we'd need to introspect the list of services per-role and the Heat registry from tripleoclient ... which is more time consuming.
This indeed can be set in the python-tripleoclient since we are using the information contained in tempest-deployer-input.conf to generate the tempest.conf using the python-tempestconf
Upstream we are using here: https://github.com/openstack/tripleo-quickstart-extras/blob/master/roles/validate-tempest/templates/configure-tempest.sh.j2#L103
It looks like we could start addressing this in tripleoclient now that we emit EnabledServices as a stack output
*** Bug 1388972 has been marked as a duplicate of this bug. ***