Fix repeated login node deployments

Replace OHPC Lmod with system Lmod
Add opt mount and tests

playbook-svc-idm.yml

@@ -52,7 +52,7 @@
         ipaadmin_password: "{{ipa_password}}"
         defaultgroup: "{{idm_default_group}}"
         defaultshell: /bin/bash
-        homedirectory: /nfs/home
+        homedirectory: /home
 
     - name: Trim
       ansible.builtin.shell: fstrim -av

playbook-svc-login.yml

@@ -7,6 +7,10 @@
     - include_tasks: tasks/backup_or_restore_host_keys.yml
     - include_tasks: tasks/setup_unattended_security_updates.yml
 
+    - name: Setup missing Slurm dependencies
+      ansible.builtin.dnf:
+        name: ["/bin/mailx", "Lmod"]
+
     - name: Setup motd
       ansible.builtin.copy:
         content: "{{login_message_of_the_day}}"
@@ -30,6 +34,11 @@
         state: enabled
         permanent: true
 
+    - name: Remove stale CA from past IPA joins
+      ansible.builtin.file:
+        state: absent
+        path: /etc/ipa/ca.crt
+
   roles:
     - role: linux-system-roles.postfix
       postfix_conf: { relayhost: "{{postfix_smtp_relay}}" }
@@ -47,6 +56,7 @@
       openhpc_slurm_service_enabled: true
       openhpc_login_only_nodes: "{{inventory_hostname}}"
       openhpc_cluster_name: "{{mgmt_cluster_name}}"
+      openhpc_module_system_install: false
 
     - role: freeipa.ansible_freeipa.ipaclient
       state: present

playbook-svc-mgmt.yml

@@ -4,11 +4,12 @@
 
   pre_tasks:
     - include_tasks: tasks/setup_backup_dir.yml
+    - include_tasks: tasks/backup_or_restore_host_keys.yml
     - include_tasks: tasks/setup_unattended_security_updates.yml
 
     - name: Setup missing Slurm dependencies
       ansible.builtin.dnf:
-        name: ["/bin/mailx"]
+        name: ["/bin/mailx", "Lmod"]
 
   roles:
     - role: linux-system-roles.postfix
@@ -50,6 +51,7 @@
       openhpc_packages: []
       openhpc_slurm_partitions:
         - name: dummy # empty partition causes slurm to stall, we'll delete this partition later to make slurm start
+      openhpc_module_system_install: false
 
     - role: freeipa.ansible_freeipa.ipaclient
       state: present
@@ -280,15 +282,15 @@
         ipaadmin_password: "{{ipa_password}}"
         name: auto.rds1
         location: default
-        mount: "/nfs"
+        mount: "/-"
         state: present
 
     - name: Create rds1 mount keys
       freeipa.ansible_freeipa.ipaautomountkey:
         ipaadmin_password: "{{ipa_password}}"
         location: default
         mapname: auto.rds1
-        key: "{{item}}" # points to /nfs/{{item}}
+        key: "/{{item}}"
         info: "mgmt.{{domain}}:/mnt/rds1/{{item}}"
         state: present
       loop: "{{mgmt_exported_directories}}"

staging.rb

@@ -113,7 +113,7 @@ def self.write_inventory(pve_ip:, storage_pool:, extra_hosts:, host_common_hash:
       mgmt_compute_dhcp_start: '10.10.10.150',
       mgmt_compute_dhcp_end: '10.10.10.254',
       mgmt_webhook_port: '808',
-      mgmt_exported_directories: %w[home shared],
+      mgmt_exported_directories: %w[home shared opt],
       mgmt_cluster_name: 'staging'
     }
     login_node_vars = {

tasks/tests.yml

@@ -2,10 +2,21 @@
   ansible.builtin.shell: pwd
   register: dir
 
-- ansible.builtin.fail:
+- name: Assert home dir path is correct
+  ansible.builtin.fail:
     msg: "Bad home dir path"
   when:
-    - dir.rc != 0 or dir.stdout.strip() != "/nfs/home/" + ansible_user
+    - dir.rc != 0 or dir.stdout.strip() != "/home/" + ansible_user
+
+- name: Get df
+  ansible.builtin.shell: df
+  register: df
+
+- name: Assert mountpoint is correct
+  ansible.builtin.fail:
+    msg: "Missing mount in df"
+  when:
+    - df.rc != 0 or  'mgmt.{{domain}}:/mnt/rds1/home' not in df.stdout
 
 - name: Get srun node hostnames
   ansible.builtin.shell: srun --nodelist={{node}} hostname