Support custom Nova Compute Ironic host names

This fixes an issue where it is not possible to customise the `host`
config option in the Nova Compute Ironic config file without breaking
detection of the service.

This is a backwards compatible fix, which allows a user to set the
`host` config option using Ansible host or group vars.

Other reasons for not using the default host setting of
`{{ ansible_hostname }}-ironic` are covered in [1].

[1] https://specs.openstack.org/openstack/nova-specs/specs/2024.1/approved/ironic-shards.html#migrate-from-peer-list-to-shard-key.

Closes-Bug: #2056571
Change-Id: I9b562f6a5722f21b7dbec2a4d53a46a57c829155

ansible/roles/nova-cell/tasks/wait_discover_computes.yml

@@ -79,7 +79,7 @@
     # configure for [DEFAULT] host in nova.conf.
     ironic_compute_service_hosts: >-
       {{ ironic_computes_in_batch |
-         map('extract', hostvars, ['ansible_facts', 'hostname']) |
+         map('extract', hostvars) | json_query('[].nova_compute_ironic_custom_host || [].ansible_facts.hostname') |
          map('regex_replace', '^(.*)$', '\1-ironic') |
          list }}
     expected_compute_service_hosts: "{{ virt_compute_service_hosts + ironic_compute_service_hosts }}"

ansible/roles/nova-cell/templates/nova.conf.j2

@@ -9,7 +9,7 @@ state_path = /var/lib/nova
 allow_resize_to_same_host = true
 
 {% if service_name == "nova-compute-ironic" %}
-host={{ ansible_facts.hostname }}-ironic
+host={{ nova_compute_ironic_custom_host | default(ansible_facts.hostname) }}-ironic
 log_file = /var/log/kolla/nova/nova-compute-ironic.log
 compute_driver = ironic.IronicDriver
 ram_allocation_ratio = 1.0

doc/source/reference/bare-metal/ironic-guide.rst

@@ -148,6 +148,49 @@ variable ``ironic_enable_keystone_integration`` to ``"yes"``
 
     ironic_enable_keystone_integration: "yes"
 
+Avoiding problems with high availability
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. note::
+
+    This section assumes that you have not yet deployed the Nova Compute
+    Ironic service. If you have already deployed multiple instances of the
+    service and have one or more baremetal nodes registered, the following
+    operations are non-trivial. You will likely have to use the `nova-manage`
+    command (or pre-Caracal edit the DB) to ensure that all Ironic nodes
+    are registered with a single Nova Compute Ironic instance. This is
+    an advanced subject and is not covered here. Stop now if you don't
+    know what you are doing.
+
+Nova Compute Ironic HA is known to be unstable. Pending a better solution,
+a workaround is to avoid the feature by running a single Nova Compute Ironic
+instance. For example:
+
+.. code-block:: diff
+
+  - [nova-compute-ironic:children]
+  - nova
+  + [nova-compute-ironic]
+  + controller1
+
+If you choose to do this, it is helpful to pin the service host name
+to a 'synthetic' constant. This means that if you need to re-deploy the
+service to another host, the Ironic nodes will automatically use the new
+service instance. Otherwise you will need to manually move active Ironic nodes
+to the new service, with either the `nova-manage` CLI, or pre-Caracal, by
+editing the Nova database.
+
+The config option to pin the host name is `nova_compute_ironic_custom_host`
+and must be set as a group or host var. Note that, unless you know what you
+are doing, you must not change or set this option if you have already deployed
+Ironic nodes.
+
+This config option is also useful for Ironic Shards. Whilst these are not
+explicitly supported by Kolla Ansible, some further information can be found
+`here <https://specs.openstack.org/openstack/nova-specs/specs/2024.1/approved/ironic-shards.html>`__.
+
+Note that Ironic HA is not affected, and continues to work as normal.
+
 Deployment
 ~~~~~~~~~~
 Run the deploy as usual:

releasenotes/notes/fix-setting-nova-compute-ironic-host-option-a7a3f6ae095f5201.yaml

@@ -0,0 +1,6 @@
+---
+fixes:
+  - |
+    Fixes the detection of the Nova Compute Ironic service when a custom `host`
+    option is set in the service config file.
+    See `LP#2056571 <https://launchpad.net/bugs/2056571>`__