Remove enable_rbac option and ensure Keystone project synchronization #19
Conversation
Remove deprecated configuration option [patrole].enable_rbac Reference: https: //docs.openstack.org/releasenotes/patrole/v0.4.0.html#deprecation-notes Change-Id: Icb575812d4a921d69f4cf472e6bad329043c62bd
This patch set ensures that projects are known to Tungsten Fabric before project-specific resources (such as virtual networks) are created. Projects are not automatically synced from Keystone to TF. The fqname-to-id call in resource_setup forces the synchronization. Reference: https: //bugs.launchpad.net/opencontrail/+bug/1685059 Change-Id: If3943572a5a05eab11c54ff4f4ad9350bce7cd23
| @@ -395,6 +392,9 @@ def setup_clients(cls): | |||
| @classmethod | |||
| def resource_setup(cls): | |||
| cls.tenant_name = cls.os_primary.credentials.tenant_name | |||
| cls.tenant_id = cls.fq_client.fqname_to_id(fq_name=['default-domain', | |||
There was a problem hiding this comment.
IMO usage of cls.get_tenant_by_name(tenant_name) instead of cls.fq_client is better.
There was a problem hiding this comment.
It doesn't look like get_tenant_by_name will make the necessary call to the TF API to make it pull the information from Keystone.
There was a problem hiding this comment.
Ok, I see, but what tests use the tenant_id variable? What is a reason of having this variable?
There was a problem hiding this comment.
The point is just to make the API call. If we don't do this, subsequent tests that rely on the creation of project-scoped resources (such as virtual-networks) fail.
The return value could be discarded or ignored instead, but it seemed to make sense to keep it around in case a future test wants it.
No description provided.