Comment 6 for bug 1645979

Sorry for not being a little more thorough in the description. This is causing a problem for the playbooks that we run to deploy code that checks the status of the l3_agent(note: we are running the linuxbridge plugin). Our playbooks target the neutron_l3_agent and neutron_metadata group then deploys code that targets the hosts in said groups to see if the agents are running in these hosts. Since the compute node now is part of these groups, our code tries to see if the l3_agent is running on the compute node and fails because there is no l3 agent running on the compute node. I personally still feel like this is incorrect behavior. The compute nodes are not running the l3 or metadata agent, but are still part of that group. IMO that shouldn't be happening. Only containers/hosts that are running the l3 agent should be part of the neutron_l3_agent group, similarly for the containers/hosts running the metadata agent.