tidy up cluster.yml files

environments/arcus/inventory/cluster.yml

@@ -1,23 +1,22 @@
 all:
   vars:
     # required:
-    cluster_name: templatetf # FIXME: needs defining for each arcus cluster!
+    cluster_name: "{{ lookup('env', 'CI_CLUSTER_NAME'}}"
     image_name: openhpc-221118-1422.qcow2 # https://github.com/stackhpc/slurm_image_builder/pull/12
     flavor: vm.alaska.cpu.general.small
     key_pair: slurm-app-ci
     network_name: WCDC-iLab-60
     subnet_name: WCDC-iLab-60
+    openhpc_slurm_partitions:
+      - name: small
+      - name: extra
 
     # overrides:
     vnic_type: direct
     block_device_prefix: sd
     home_volume_size: 10
     state_volume_size: 10
 
-    openhpc_slurm_partitions:
-      - name: small
-      - name: extra
-
 # compute must contain child groups matching openhpc_slurm_partitions
 compute:
   children:

environments/common/inventory/cluster.yml

@@ -1,7 +1,7 @@
-# yaml-format ansible inventory with default hostvars
-# This provides defaults when no actual hosts are available, i.e. during provisining or image build. It should be overriden/extended
-# using an environment's own cluster.yml file.
+# Yaml-format ansible inventory with default hostvars. It should be overriden/extended by an environment's own cluster.yml file.
 # NB: This is the lowest-possible priority - it will be overridden by inventory files in other inventories, 'all' group_vars, etc.
+# Host and group vars defined here are available even when no hosts have been provisioned (e.g. during provisioning or image build).
+
 all:
   vars:
     # Must be provided in an environment:
@@ -12,25 +12,28 @@ all:
     network_name: "{{ undef(hint='network_name must be defined') }}"
     subnet_name: "{{ undef(hint='subnet_name must be defined') }}"
 
-    # tf meta:
+    # top-level domain for cluster (FQDN is nodename.clustername.tld):
+    tld: invalid
+
+    # terraform lifecycle:
     tf_ignore_changes:
       - image_name
       - user_data
 
+    # openstack volume options - note hostvars too:
     state_volume_size: 150 # GB
     home_volume_size: 100 # GB
+    block_device_prefix: vd
     volumes: []
 
-    # tf openstack_networking_port_v2:binding
+    # terrraform openstack_networking_port_v2:binding
     vnic_profile: {}
     vnic_type: normal
 
     security_groups: ["default"]
-
-    openhpc_cluster_name: "{{ cluster_name }}" # FIXME move to defaults?
-    block_device_prefix: vd
-    tld: invalid
 
+    # this default works when there are no child groups in the 'compute' group
+    # and no partition customisation is required
     openhpc_slurm_partitions:
       - name: "compute"
 

environments/common/inventory/group_vars/all/openhpc.yml

@@ -3,6 +3,7 @@
 # See: https://github.com/stackhpc/ansible-role-openhpc
 # for variable definitions
 
+openhpc_cluster_name: "{{ cluster_name }}"
 openhpc_enable:
   control: "{{ inventory_hostname in groups['control'] }}"
   batch: "{{ inventory_hostname in groups['compute'] }}"