Merge pull request #798 from stackhpc/fix/caas-upgrade

Fix .caas secrets not persisting post-reimage and skip tofu vars validation for .caas

Author: wtripp180901

ansible/roles/persist_openhpc_secrets/tasks/main.yml

@@ -15,6 +15,14 @@
     - "{{ appliances_state_dir }}/ansible.facts.d"
     - "/etc/ansible/facts.d"
 
+- name: Symlink to persisted facts if present
+  ansible.builtin.file:
+    state: link
+    src: "{{ appliances_state_dir }}/ansible.facts.d/openhpc_secrets.fact"
+    dest: /etc/ansible/facts.d/openhpc_secrets.fact
+    owner: root
+  when: openhpc_secrets_stat.stat.exists
+
 - name: Load existing OpenHPC secrets if present
   ansible.builtin.setup:
     filter: ansible_local

ansible/validate.yml

@@ -77,6 +77,7 @@
           - cluster_home_volume is defined
           - cluster_compute_groups is defined
         fail_msg: "One or more expected variables are missing: is OpenTofu inventory template up to date?"
+      when: appliances_validate_tofu_vars | bool
 
 - name: Ensure control node is in inventory
   hosts: all

environments/.caas/inventory/group_vars/all/defaults.yml

@@ -0,0 +1 @@
+appliances_validate_tofu_vars: false

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

@@ -8,6 +8,7 @@ appliances_cockpit_state: absent # RHEL cockpit installed but not enabled in gen
 # appliances_state_dir: # define an absolute path here to use for persistent state: NB: This is defined as /var/lib/state in inventory by the default Terraform
 appliances_mode: configure
 appliances_pulp_url: https://ark.stackhpc.com
+appliances_validate_tofu_vars: true
 
 # Address(ip/dns) for internal communication between services. This is
 # normally traffic you do no want to expose to users.