roles/nfs_stability/tasks/main.yml

@@ -0,0 +1,128 @@
+---
+- name: Ensure this role only targets the three mergerfs media hosts
+  ansible.builtin.assert:
+    that:
+      - inventory_hostname in ['emby-sascha', 'emby-chris', 'arrapps']
+    fail_msg: "nfs_stability must not run on {{ inventory_hostname }}"
+
+- name: Stop and disable the destructive legacy retry timer first
+  ansible.builtin.systemd:
+    name: remount-nfs.timer
+    state: stopped
+    enabled: false
+  failed_when: false
+
+- name: Stop a currently running destructive legacy retry service
+  ansible.builtin.systemd:
+    name: remount-nfs.service
+    state: stopped
+  failed_when: false
+
+- name: Mark mergerfs as a network-dependent mount without remounting it live
+  ansible.builtin.replace:
+    path: /etc/systemd/system/mnt-media.mount
+    regexp: '^Options=(?![^\n]*_netdev)(.*)$'
+    replace: 'Options=_netdev,\1'
+    backup: true
+  register: mergerfs_unit
+
+- name: Install non-destructive one-shot NFS recovery helper
+  ansible.builtin.copy:
+    dest: /usr/local/sbin/remount-nfs-safe
+    owner: root
+    group: root
+    mode: '0755'
+    content: |
+      #!/bin/bash
+      set -u
+      is_nfs() {
+        findmnt -rn --target "$1" -t nfs,nfs4 -o TARGET | grep -Fxq "$1"
+      }
+      is_media() {
+        findmnt -rn --target /mnt/media -t fuse.mergerfs -o TARGET | grep -Fxq /mnt/media
+      }
+      for attempt in 1 2 3; do
+        missing=0
+        for entry in 'mnt-nas.mount:/mnt/nas' 'mnt-qnap1.mount:/mnt/qnap1' 'mnt-qnap2.mount:/mnt/qnap2'; do
+          unit="${entry%%:*}"
+          target="${entry#*:}"
+          if ! is_nfs "$target"; then
+            systemctl start "$unit" || true
+          fi
+          is_nfs "$target" || missing=1
+        done
+        if [ "$missing" -eq 0 ]; then
+          if ! is_media; then
+            systemctl start mnt-media.mount || true
+          fi
+          is_media && exit 0
+        fi
+        sleep 10
+      done
+      echo 'NFS recovery failed without restarting any active mount' >&2
+      exit 1
+
+- name: Install safe one-shot NFS recovery service
+  ansible.builtin.copy:
+    dest: /etc/systemd/system/remount-nfs.service
+    owner: root
+    group: root
+    mode: '0644'
+    backup: true
+    content: |
+      [Unit]
+      Description=Safely start missing NFS mounts once after boot
+      After=network-online.target
+      Wants=network-online.target
+
+      [Service]
+      Type=oneshot
+      ExecStart=/usr/local/sbin/remount-nfs-safe
+
+- name: Install once-per-boot NFS recovery timer
+  ansible.builtin.copy:
+    dest: /etc/systemd/system/remount-nfs.timer
+    owner: root
+    group: root
+    mode: '0644'
+    backup: true
+    content: |
+      [Unit]
+      Description=One-shot NFS mount validation after boot
+
+      [Timer]
+      OnBootSec=45
+      AccuracySec=5
+      Persistent=false
+      Unit=remount-nfs.service
+
+      [Install]
+      WantedBy=timers.target
+
+- name: Reload systemd and enable the one-shot boot timer
+  ansible.builtin.systemd:
+    daemon_reload: true
+    name: remount-nfs.timer
+    enabled: true
+    state: started
+
+- name: Verify systemd unit dependency graph
+  ansible.builtin.command:
+    argv:
+      - systemd-analyze
+      - verify
+      - /etc/systemd/system/mnt-media.mount
+      - /etc/systemd/system/remount-nfs.service
+      - /etc/systemd/system/remount-nfs.timer
+  changed_when: false
+
+- name: Verify all active storage mounts without restarting them
+  ansible.builtin.shell: |
+    set -e
+    for target in /mnt/nas /mnt/qnap1 /mnt/qnap2; do
+      findmnt -rn --target "$target" -t nfs,nfs4 -o TARGET | grep -Fxq "$target"
+    done
+    findmnt -rn --target /mnt/media -t fuse.mergerfs -o TARGET | grep -Fxq /mnt/media
+  args:
+    executable: /bin/bash
+  changed_when: false