Fix management address lookup for metal hosts in some deployments

Git history indicates that when Nova runs on metal hosts its
address was identified via the management bridge and an IP lookup.
When the dynamic-address-fact was added this link was broken
and metal hosts have been using the fallback of ansible_host
defined in dynamic-address-fact

Note that originally the 'management_bridge' host var was relied
upon, but this change pulls this bridge from the
'container_networks.container_address.bridge'

Change-Id: I9aaca445d135b574753722d6b8de28e4f99c427b
This commit is contained in:
Andrew Bonney 2020-07-15 13:30:29 +01:00
parent 8fa414e1f8
commit fc1b6f3bf6

View File

@ -69,7 +69,7 @@
- name: Determine management bridge IP address
include_tasks: ../common-tasks/dynamic-address-fact.yml
vars:
network_address: "management_address"
network_address: "container_address"
tags:
- always
@ -142,7 +142,7 @@
roles:
- role: "os_nova"
nova_management_address: "{{ management_address }}"
nova_management_address: "{{ container_address }}"
nova_cinder_rbd_inuse: "{{ hostvars['localhost']['nova_cinder_rbd_inuse'] | default(False) }}"
- role: "system_crontab_coordination"