When periodic tasks run the calls to nova don't go to the endpoint specified in nova_compute_url
When taskmanager.Manager sets up a admin_context to use for periodic tasks it passes in nova_proxy_admin_tenant_name for the tenant attribute for the context. (A standard openstack context uses tenant_id for the tenant attribute, so this is confusing)
That admin context is used to create a nova client in common.remote
It creates the client correctly but then when it overrides the management_url it sets it to <nova_compute_url>/<nova_proxy_admin_tenant_name>
This is the wrong URL, it should be the tenant ID not the name. Nova seems to then ignore this value and in our case uses the default compute endpoint gotten from keystone.
Fix proposed to branch: master /review. openstack. org/111679
Review: https:/