infra/ansible/roles/raid/tasks/main.yml

---
- name: Install RAID management tools
  apt:
    name:
      - mdadm
      - smartmontools
      - parted
      - nvme-cli
      - lsscsi
    state: present
    update_cache: yes

- name: Gather disk information
  shell: |
    echo "=== Block devices ==="
    lsblk -d -o NAME,SIZE,TYPE,MODEL
    echo "=== NVMe devices ==="
    nvme list 2>/dev/null || echo "No NVMe devices found"
    echo "=== SCSI devices ==="
    lsscsi -s 2>/dev/null || echo "No SCSI devices found"
    echo "=== All disk devices ==="
    ls -la /dev/sd* /dev/nvme* /dev/vd* /dev/mmcblk* 2>/dev/null | grep -E "^b" || echo "Checking other paths..."
    echo "=== Detailed block device tree ==="
    lsblk -o NAME,SIZE,TYPE,FSTYPE,MOUNTPOINT,MODEL
  register: disk_info
  changed_when: false

- name: Display disk information
  debug:
    msg: "{{ disk_info.stdout_lines }}"

- name: Check if we should auto-detect RAID devices
  set_fact:
    auto_detect_devices: "{{ raid_devices is not defined or raid_devices | length == 0 }}"

- name: Auto-detect two largest unused disks
  when: auto_detect_devices
  block:
    - name: Find unused block devices
      shell: |
        lsblk -dpno NAME,SIZE,TYPE,MOUNTPOINT | grep -E "disk.*$" | grep -v -E "/(|/boot)" | sort -k2 -h -r | head -2 | awk '{print $1}'
      register: detected_devices
      changed_when: false

    - name: Set raid_devices from detected devices
      set_fact:
        raid_devices: "{{ detected_devices.stdout_lines }}"
      when: detected_devices.stdout_lines | length >= 2

    - name: Fail if not enough devices detected
      fail:
        msg: "Could not detect at least 2 unused block devices for RAID. Please specify raid_devices manually."
      when: detected_devices.stdout_lines | length < 2

- name: Display RAID devices to be used
  debug:
    msg: "Will use these devices for RAID0: {{ raid_devices }}"

- name: Check if RAID array already exists
  command: mdadm --detail /dev/md0
  register: raid_status
  failed_when: false
  changed_when: false

- name: Stop and remove existing RAID array if it exists
  block:
    - name: Stop RAID array
      command: mdadm --stop /dev/md0
      when: raid_status.rc == 0

    - name: Zero superblocks on disks
      command: mdadm --zero-superblock {{ item }}
      loop:
        - /dev/nvme0n1
        - /dev/nvme1n1
      when: raid_status.rc == 0
  when: force_raid_rebuild | default(false)

- name: Create partition table on first NVMe drive
  parted:
    device: /dev/nvme0n1
    number: 1
    state: present
    part_type: primary
    part_start: 0%
    part_end: 100%
    label: gpt
  when: raid_status.rc != 0

- name: Create partition table on second NVMe drive
  parted:
    device: /dev/nvme1n1
    number: 1
    state: present
    part_type: primary
    part_start: 0%
    part_end: 100%
    label: gpt
  when: raid_status.rc != 0

- name: Create RAID0 array
  command: >
    mdadm --create /dev/md0
    --level=0
    --raid-devices=2
    /dev/nvme0n1p1 /dev/nvme1n1p1
    --force
  when: raid_status.rc != 0
  register: raid_create

- name: Wait for RAID array to be ready
  command: mdadm --detail /dev/md0
  register: raid_ready
  until: raid_ready.rc == 0
  retries: 10
  delay: 5
  when: raid_create is changed

- name: Create mdadm configuration
  shell: mdadm --detail --scan >> /etc/mdadm/mdadm.conf
  when: raid_create is changed

- name: Update initramfs
  command: update-initramfs -u
  when: raid_create is changed

- name: Create filesystem on RAID array
  filesystem:
    fstype: ext4
    dev: /dev/md0
    opts: "-E lazy_itable_init=0,lazy_journal_init=0"
  when: raid_create is changed

- name: Create mount point for PostgreSQL data
  file:
    path: /mnt/pgdata
    state: directory
    mode: '0755'

- name: Mount RAID array
  mount:
    path: /mnt/pgdata
    src: /dev/md0
    fstype: ext4
    opts: noatime,nodiratime,nobarrier
    state: mounted

- name: Create PostgreSQL user and group
  block:
    - name: Create postgres group
      group:
        name: postgres
        state: present
        system: yes

    - name: Create postgres user
      user:
        name: postgres
        group: postgres
        home: /var/lib/postgresql
        shell: /bin/bash
        system: yes
        state: present

- name: Ensure PostgreSQL directories exist on RAID
  file:
    path: "{{ item }}"
    state: directory
    owner: postgres
    group: postgres
    mode: '0700'
  loop:
    - /mnt/pgdata
    - /mnt/pgdata/16
    - /mnt/pgdata/16/main
    - /mnt/pgdata/wal
    - /mnt/pgdata/archive
    - /mnt/pgdata/backups

- name: Set up RAID monitoring
  lineinfile:
    path: /etc/mdadm/mdadm.conf
    line: "MAILADDR root"
    insertafter: EOF

- name: Enable mdadm monitoring service
  systemd:
    name: mdmonitor
    enabled: yes
    state: started

- name: Create RAID monitoring script
  copy:
    content: |
      #!/bin/bash
      # RAID health check script
      RAID_STATUS=$(mdadm --detail /dev/md0 | grep "State :" | awk '{print $3}')
      if [ "$RAID_STATUS" != "clean" ] && [ "$RAID_STATUS" != "active" ]; then
          echo "RAID array /dev/md0 is in state: $RAID_STATUS"
          exit 1
      fi
      exit 0
    dest: /usr/local/bin/check-raid-health
    mode: '0755'

- name: Set up RAID health check cron job
  cron:
    name: "Check RAID health"
    minute: "*/5"
    job: "/usr/local/bin/check-raid-health || echo 'RAID health check failed' | logger -t raid-health"