In quickstart, there is a template https://git.openstack.org/cgit/openstack/tripleo-quickstart-extras/tree/roles/overcloud-deploy/templates/config-download.yaml.j2 , and a flag deploy_steps_ansible_workflow, that configures port mappings for deployed servers, yet does this too opinionated and consumable with only CI multinode cases.
The flexible approach should follow the deployed server docs https://docs.openstack.org/tripleo-docs/latest/install/advanced_deployment/deployed_server.html#network-configuration and allow users, say RDOcloud to meet their case, which is to transform :
DeployedServerPortMap:
controller0-ctlplane:
fixed_ips:
- ip_address:
into configurable:
DeployedServerPortMap:
<host-cloud-nova-server-hostname>-ctlplane:
fixed_ips:
- ip_address: <host-cloud-ctlplane-subnet-IP>
Note that, 'tripleo-ansible-inventory --list' command shows overcloud group only when there is an undercloud stack called "overcloud" exists.
If it's called something different, you have to export TRIPLEO_PLAN_NAME=foo. The issue is that the quickstart's config-download.yaml template expects the overcloud stack to already present, while it's being created later, when overcloud deploy commands executes.
For config download to work the deploy of the heat stack must already be complete, that's not that quickstart's deploy_steps_ansible_workflow follows yet.
This chicken and egg problem for deployed servers should be resolved as well.
Added UX tag as this issue might block openstack clouds users using deployed servers and trying external deployments performed with configs download mechanism.