mgmt/configuration/playbooks/add-warden-to-nomad-cluster.yml

---
- name: Add Warden Server as Nomad Client to Cluster
  hosts: warden
  become: yes
  gather_facts: yes

  vars:
    nomad_plugin_dir: "/opt/nomad/plugins"
    nomad_datacenter: "dc1"
    nomad_region: "global"
    nomad_servers:
      - "100.117.106.136:4647"
      - "100.116.80.94:4647"
      - "100.97.62.111:4647"
      - "100.116.112.45:4647"
      - "100.84.197.26:4647"

  tasks:
    - name: 显示当前处理的节点
      debug:
        msg: "🔧 将 warden 服务器添加为 Nomad 客户端: {{ inventory_hostname }}"

    - name: 检查 Nomad 是否已安装
      shell: which nomad || echo "not_found"
      register: nomad_check
      changed_when: false

    - name: 下载并安装 Nomad
      block:
        - name: 下载 Nomad 1.10.5
          get_url:
            url: "https://releases.hashicorp.com/nomad/1.10.5/nomad_1.10.5_linux_amd64.zip"
            dest: "/tmp/nomad.zip"
            mode: '0644'

        - name: 解压并安装 Nomad
          unarchive:
            src: "/tmp/nomad.zip"
            dest: "/usr/local/bin/"
            remote_src: yes
            owner: root
            group: root
            mode: '0755'

        - name: 清理临时文件
          file:
            path: "/tmp/nomad.zip"
            state: absent
      when: nomad_check.stdout == "not_found"

    - name: 验证 Nomad 安装
      shell: nomad version
      register: nomad_version_output

    - name: 创建 Nomad 配置目录
      file:
        path: /etc/nomad.d
        state: directory
        owner: root
        group: root
        mode: '0755'

    - name: 创建 Nomad 数据目录
      file:
        path: /opt/nomad/data
        state: directory
        owner: nomad
        group: nomad
        mode: '0755'
      ignore_errors: yes

    - name: 创建 Nomad 插件目录
      file:
        path: "{{ nomad_plugin_dir }}"
        state: directory
        owner: nomad
        group: nomad
        mode: '0755'
      ignore_errors: yes

    - name: 获取服务器 IP 地址
      shell: |
        ip route get 1.1.1.1 | grep -oP 'src \K\S+'        
      register: server_ip_result
      changed_when: false

    - name: 设置服务器 IP 变量
      set_fact:
        server_ip: "{{ server_ip_result.stdout }}"

    - name: 停止 Nomad 服务（如果正在运行）
      systemd:
        name: nomad
        state: stopped
      ignore_errors: yes

    - name: 创建 Nomad 客户端配置文件
      copy:
        content: |
          # Nomad Client Configuration for warden
          datacenter = "{{ nomad_datacenter }}"
          data_dir = "/opt/nomad/data"
          log_level = "INFO"
          bind_addr = "{{ server_ip }}"
          
          server {
            enabled = false
          }
          
          client {
            enabled = true
            servers = [
              {% for server in nomad_servers %}"{{ server }}"{% if not loop.last %}, {% endif %}{% endfor %}
            ]
          }
          
          plugin_dir = "{{ nomad_plugin_dir }}"
          
          plugin "podman" {
            config {
              socket_path = "unix:///run/podman/podman.sock"
              volumes {
                enabled = true
              }
            }
          }
          
          consul {
            address = "127.0.0.1:8500"
          }          
        dest: /etc/nomad.d/nomad.hcl
        owner: root
        group: root
        mode: '0644'

    - name: 验证 Nomad 配置
      shell: nomad config validate /etc/nomad.d/nomad.hcl
      register: nomad_validate
      failed_when: nomad_validate.rc != 0

    - name: 创建 Nomad systemd 服务文件
      copy:
        content: |
          [Unit]
          Description=Nomad
          Documentation=https://www.nomadproject.io/docs/
          Wants=network-online.target
          After=network-online.target

          [Service]
          Type=notify
          User=root
          Group=root
          ExecStart=/usr/local/bin/nomad agent -config=/etc/nomad.d
          ExecReload=/bin/kill -HUP $MAINPID
          KillMode=process
          KillSignal=SIGINT
          TimeoutStopSec=5
          LimitNOFILE=65536
          LimitNPROC=32768
          Restart=on-failure
          RestartSec=2

          [Install]
          WantedBy=multi-user.target          
        dest: /etc/systemd/system/nomad.service
        mode: '0644'

    - name: 重新加载 systemd 配置
      systemd:
        daemon_reload: yes

    - name: 启动并启用 Nomad 服务
      systemd:
        name: nomad
        state: started
        enabled: yes

    - name: 等待 Nomad 服务启动
      wait_for:
        port: 4646
        host: "{{ server_ip }}"
        delay: 5
        timeout: 60

    - name: 检查 Nomad 客户端状态
      shell: nomad node status -self
      register: nomad_node_status
      retries: 5
      delay: 5
      until: nomad_node_status.rc == 0
      ignore_errors: yes

    - name: 显示 Nomad 客户端配置结果
      debug:
        msg: |
          ✅ warden 服务器已成功配置为 Nomad 客户端
          📦 Nomad 版本: {{ nomad_version_output.stdout.split('\n')[0] }}
          🌐 服务器 IP: {{ server_ip }}
          🏗️ 数据中心: {{ nomad_datacenter }}
          📊 客户端状态: {{ 'SUCCESS' if nomad_node_status.rc == 0 else 'PENDING' }}
          🚀 warden 现在是 Nomad 集群的一部分