Merge "B&R: Recover Ceph data on standard with storage nodes"

playbookconfig/src/playbooks/roles/bootstrap/validate-config/tasks/main.yml

@@ -658,10 +658,15 @@
          apiserver_oidc.username_claim is not defined)
 
 # Wipe ceph osds
+# Note that due to Ansible mishandling of boolean values via extra-vars we are
+# adding supplementary validation here.
+# See: https://github.com/ansible/ansible/issues/17193
 - name: Check for Ceph data wipe flag
   fail:
     msg: "wipe_ceph_osds is misconfigured. Valid value is either 'true' or 'false'"
-  when: (mode == "restore") and (not wipe_ceph_osds | type_debug == 'bool')
+  when: (mode == "restore") and (not wipe_ceph_osds | type_debug == 'bool') and
+        (wipe_ceph_osds != 'true') and
+        (wipe_ceph_osds != 'false')
 
 - block:
   - name: Wipe ceph osds

playbookconfig/src/playbooks/roles/recover-ceph-data/tasks/main.yml

@@ -9,26 +9,6 @@
 
 - name: Restore CEPH Monitor data
   block:
-    # Get system_mode after restore
-    - name: Retrieve system mode
-      shell: source /etc/platform/platform.conf; echo $system_mode
-      register: restore_system_mode_result
-
-    - name: Fail if system mode is not defined
-      fail:
-        msg: "system_mode is missing in /etc/platform/platform.conf"
-      when: restore_system_mode_result.stdout_lines|length == 0
-
-    - name: Set facts
-      set_fact:
-        restore_system_mode: "{{ restore_system_mode_result.stdout_lines[0] }}"
-
-    - name: Create flag file in /etc/platform to skip wiping OSDs
-      file:
-        path: "{{ skip_ceph_osds_wipe_flag }}"
-        state: touch
-      when: restore_system_mode != 'simplex'
-
     - name: Restore ceph.conf file
       command: >-
         tar -C / -xpf {{ target_backup_dir }}/{{ backup_filename }}

playbookconfig/src/playbooks/roles/recover-ceph-data/vars/main.yml

@@ -1,5 +1,4 @@
 ---
-skip_ceph_osds_wipe_flag: /etc/platform/.skip_ceph_osds_wipe
 ceph_mon_manifest_apply_log: /tmp/ceph_mon_apply_manifest.log
 ceph_wait_time: 60
 drbd_cephmon_res: /etc/drbd.d/drbd-cephmon.res

playbookconfig/src/playbooks/roles/restore-platform/prepare-env/tasks/main.yml

@@ -28,6 +28,16 @@
     failed_when: false
     register: search_result
 
+  # Note that due to Ansible mishandling of boolean values via extra-vars we are
+  # adding supplementary validation here.
+  # See: https://github.com/ansible/ansible/issues/17193
+  - name: Check for Ceph data wipe flag
+    fail:
+      msg: "wipe_ceph_osds is misconfigured. Valid value is either 'true' or 'false'"
+    when: (not wipe_ceph_osds | type_debug == 'bool') and
+          (wipe_ceph_osds != 'true') and
+          (wipe_ceph_osds != 'false')
+
   - block:
     - name: Extract override file from backup tarball
       shell: >-

playbookconfig/src/playbooks/roles/restore-platform/restore-more-data/tasks/main.yml

@@ -277,6 +277,14 @@
       psql -c "update i_host set administrative='locked', operational='disabled',
       availability='offline'" sysinv
     become_user: postgres
+    when: wipe_ceph_osds|bool
+
+  - name: Set all the hosts, except storage nodes to locked/disabled/offline state
+    shell: >-
+      psql -c "update i_host set administrative='locked', operational='disabled',
+      availability='offline' where personality!='storage'" sysinv
+    become_user: postgres
+    when: not wipe_ceph_osds|bool
 
   # Set platform-integ-apps to "uploaded" state, so that once ceph is up after
   # controller-0 is unlocked for the first time, the manifest will be applied.
@@ -347,8 +355,42 @@
 
   # Restore ceph-mon data
   - block:
-    - include_role:
-        name: recover-ceph-data
+    - block:
+      # Recover procedure for systems with storage nodes is different from
+      # that of systems with controller storage:
+      # - For controller storage we recover ceph-mon data by scanning OSDs.
+      # - For systems with storage nodes we get ceph-mon data from storage-0
+      #   ceph-mon that is already up and will not be reinstalled.
+      - name: Check if setup has storage nodes
+        shell: source /etc/platform/openrc; system host-list | awk '$6=="storage"{exit 1}'
+        register: storage_nodes
+        failed_when: false
+
+      # Get system_mode after restore and create flag file to skip wiping OSDs
+      - name: Retrieve system mode
+        shell: source /etc/platform/platform.conf; echo $system_mode
+        register: restore_system_mode_result
+
+      - name: Fail if system mode is not defined
+        fail:
+          msg: "system_mode is missing in /etc/platform/platform.conf"
+        when: restore_system_mode_result.stdout_lines|length == 0
+
+      - name: Set system mode fact
+        set_fact:
+          restore_system_mode: "{{ restore_system_mode_result.stdout_lines[0] }}"
+
+      - name: Create flag file in /etc/platform to skip wiping OSDs
+        file:
+          path: "{{ skip_ceph_osds_wipe_flag }}"
+          state: touch
+        when: restore_system_mode != 'simplex'
+
+      # Recover ceph data for systems with controller storage
+      - include_role:
+          name: recover-ceph-data
+        when: storage_nodes.rc == 0
+
       when: not wipe_ceph_osds|bool
 
     - name: Inform user that restore_platform is run successfully

playbookconfig/src/playbooks/roles/restore-platform/restore-more-data/vars/main.yml

@@ -5,6 +5,7 @@ extension_permdir: /opt/extension
 patch_vault_permdir: /opt/patch-vault
 ldap_permdir: /var/lib/openldap-data
 crushmap_file: crushmap.bin.backup
+skip_ceph_osds_wipe_flag: /etc/platform/.skip_ceph_osds_wipe
 
 # The staging area to process the data from the backup tarball
 # after bootstrap controller-0 is done during the restore.