spine leaf undercloud installation missing ctlplane_ip exception error in master

[Michele Baldessari]

Since the change "131ff2b4b831 - (2021-01-19 09:17:01 +0100) Re-factor inventory.py use setdefault pervasively" https://review.opendev.org/c/openstack/tripleo-common/+/766109 undercloud.conf
fails to install with the following error:

PLAY [Load global variables] *************************************************** ^[[0;32m2021-01-23 20:57:15.255413 | 5254005e-d6c8-3f49-eab0-00000000000e | OK | include_vars | undercloud^[[0m ^[[1;30m2021-01-23 20:57:15.256753 | 5254005e-d6c8-3f49-eab0-00000000000e | TIMING | include_vars | undercloud | 0:00:02.635333 | 0.04s^[[0m ^[[0;32m2021-01-23 20:57:15.267562 | 5254005e-d6c8-3f49-eab0-00000000000e | OK | include_vars | undercloud-0^[[0m ^[[1;30m2021-01-23 20:57:15.268023 | 5254005e-d6c8-3f49-eab0-00000000000e | TIMING | include_vars | undercloud-0 | 0:00:02.646619 | 0.03s^[[0m ^M PLAY [Render all_nodes data as group_vars for overcloud] *********************** 2021-01-23 20:57:15.301062 | 5254005e-d6c8-3f49-eab0-000000000011 | TASK | Render all_nodes data as group_vars for overcloud Process Process-5:2:^M Traceback (most recent call last):^M File "/usr/lib64/python3.6/multiprocessing/process.py", line 258, in _bootstrap^M self.run()^M File "/usr/lib64/python3.6/multiprocessing/process.py", line 93, in run^M self._target(*self._args, **self._kwargs)^M File "/usr/share/ansible/plugins/action/tripleo_all_nodes_data.py", line 84, in compute_service^M service_hosts))^M File "/usr/share/ansible/plugins/action/tripleo_all_nodes_data.py", line 83, in ^M map(lambda host: self.h_vars[host][service_network + '_ip'],^M File "/usr/lib/python3.6/site-packages/ansible/vars/hostvars.py", line 129, in __getitem__^M foo = templar.template(self._vars[var], fail_on_undefined=False, static_vars=STATIC_VARS)^M KeyError: 'ctlplane_ip'^M

Reverting that change makes the installation work correctly.

My undercloud.conf is
[DEFAULT]
undercloud_ntp_servers=ntp.corp
overcloud_domain_name = bgp.ftw
container_images_file=/home/stack/containers-prepare-parameter.yaml
undercloud_timezone = UTC
undercloud_hostname = undercloud-0.bgp.ftw
undercloud_nameservers = 10.19.42.41,10.11.5.19

local_interface = eth2
local_ip = 192.168.4.2/24
undercloud_public_host = 192.168.4.200
undercloud_admin_host = 192.168.4.201
enable_routed_networks = true
subnets=r1,r2,r3,r4
local_subnet=r4

[r1]
cidr = 192.168.1.0/24
dhcp_start = 192.168.1.160
dhcp_end = 192.168.1.170
inspection_iprange = 192.168.1.171,192.168.1.175
gateway = 192.168.1.1
masquerade = False
[r2]
cidr = 192.168.2.0/24
dhcp_start = 192.168.2.160
dhcp_end = 192.168.2.170
inspection_iprange = 192.168.2.171,192.168.2.175
gateway = 192.168.2.1
masquerade = False
[r3]
cidr = 192.168.3.0/24
dhcp_start = 192.168.3.160
dhcp_end = 192.168.3.170
inspection_iprange = 192.168.3.171,192.168.3.175
gateway = 192.168.3.1
masquerade = False
[r4]
cidr = 192.168.4.0/24
dhcp_start = 192.168.4.160
dhcp_end = 192.168.4.170
inspection_iprange = 192.168.4.171,192.168.4.175
gateway = 192.168.4.1
masquerade = False

[Harald Jensås (harald-jensas)]

I reproduced this and did some debugging.
Prior to the patch the 'Undercloud' entry added at the beginning of the list() function was actually completely overwritten when a role names 'Undercloud' is present in the heat stack.

After re-factoring we add to it instead, and end up with two hosts in the "Undercloud" group.

{'Undercloud': {
    'hosts': {
       'undercloud': {},
       'undercloud-0': { ...

We probably need to special case for the 'Undercloud' role in the _inventory_from_heat_outputs() method to ensure it overwrites instead of appending.

[Harald Jensås (harald-jensas)]

fix proposed: https://review.opendev.org/c/openstack/tripleo-common/+/772971

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/tripleo-common 15.0.0 release.