场景预设

1. role 的默认变量文件里，有不同类型的变量，K/V、Dict、List 都有。
2. 目标 host 同时归属于几个分组（groups）

期望结果

可以在 group_vars 里给某些组自定义一些变量，可以与 role 默认变量合并，优先使用自定义变量。

举例

role 默认变量 defaults/main.yml

  # mysql 配置，只包含基本设置，优化参数请自行决定。
  pxc_custom_cnf:
    mysqld:
      ssl_ca: "/cert/ca.pem"
      ssl_cert: "/cert/server-cert.pem"
      ssl_key: "/cert/server-key.pem"
      bind_address: "{
  { global_bond_ip }}"
      wsrep_node_address: "{
  { global_bond_ip }}"
      wsrep_node_name: "{
  { ansible_hostname }}"
      server_id: "{
  { global_bond_ip | ipaddr('int') }}"
      wsrep_cluster_address: "gcomm://{
  { gcomm_nodes }}"
      wsrep_node_incoming_address: "{
  { global_bond_ip }}:{
  { combined_vars.pxc_port | default(pxc_port) }}"
      port: "{
  { combined_vars.pxc_port | default(pxc_port) }}"
      skip_name_resolve: "on"
    client:
      ssl_ca: "/cert/ca.pem"
      ssl_cert: "/cert/client-cert.pem"
      ssl_key: "/cert/client-key.pem"
    sst:
      encrypt: "4"
      ssl_ca: "/cert/ca.pem"
      ssl_cert: "/cert/server-cert.pem"
      ssl_key: "/cert/server-key.pem"
    xtrabackup:
      host: "{
  { global_bond_ip }}"
      user: "{
  { mysql_backup_username }}"
      password: "{
  { mysql_backup_password }}"
      port: "{
  { combined_vars.pxc_port | default(pxc_port) }}"
      target_dir: "/backup"

自定义变量 group_vars/pxc.yml

# mysql 配置
pxc_custom_cnf:
  mysqld:
# 自定义配置，替代 role 对应的默认值，并合并其余默认值。
    character_set_server: "utf8mb4"
    collation_server: "utf8mb4_unicode_ci"
    max_connections: "1000"
    open_files_limit: "200000"
    pxc_strict_mode: "PERMISSIVE"
    skip_name_resolve: "on"
    expire_logs_days: "7"
    max_prepared_stmt_count: "65528"
    innodb_buffer_pool_instances: "16"
    innodb_buffer_pool_chunk_size: "256M"

合并后的变量内容

"pxc_custom_cnf": {
    "client": {
        "ssl_ca": "/cert/ca.pem",
        "ssl_cert": "/cert/client-cert.pem",
        "ssl_key": "/cert/client-key.pem"
    },
    "mysqld": {
        "bind_address": "192.168.56.16",
        "character_set_server": "utf8mb4",
        "collation_server": "utf8mb4_unicode_ci",
        "expire_logs_days": "7",
        "innodb_buffer_pool_chunk_size": "256M",
        "innodb_buffer_pool_instances": "16",
        "max_connections": "1000",
        "max_prepared_stmt_count": "65528",
        "open_files_limit": "200000",
        "port": "13306",
        "pxc_strict_mode": "PERMISSIVE",
        "server_id": "3232249872",
        "skip_name_resolve": "on",
        "ssl_ca": "/cert/ca.pem",
        "ssl_cert": "/cert/server-cert.pem",
        "ssl_key": "/cert/server-key.pem",
        "wsrep_cluster_address": "gcomm://192.168.56.17,192.168.56.18",
        "wsrep_node_address": "192.168.56.16",
        "wsrep_node_incoming_address": "192.168.56.16:13306",
        "wsrep_node_name": "192-168-56-16"
    },
    "sst": {
        "encrypt": "4",
        "ssl_ca": "/cert/ca.pem",
        "ssl_cert": "/cert/server-cert.pem",
        "ssl_key": "/cert/server-key.pem"
    },
    "xtrabackup": {
        "host": "192.168.56.16",
        "password": "jNeYsi0ATJkyDTh1vXs4",
        "port": "13306",
        "target_dir": "/backup",
        "user": "bkuser"
    }
}

playbook 代码

首先，在需要合并变量的 role 里，调用合并变量的 tasks

- name: Pre tasks
  block:
    - name: Combine role vars and group vars
      import_role:
        name: pre-tasks
        tasks_from: vars_combine
  tags: always

完整代码：https://gitee.com/bottlelee/haibinlee-ansible-roles/blob/develop/percona-xtradb-cluster/tasks/main.yml

roles/pre-tasks/vars_combine.yml 内容

- name: Combine default vars and custom vars
  block:
    - name: "Include role's defaults vars"
      ansible.builtin.include_vars:
        dir: "{
  { item }}/defaults"
        name: "temp_role_vars"
      loop: "{
  { ansible_parent_role_paths }}"

    - name: Assembling group vars files
      assemble:
        src: "{
  { inventory_dir }}/group_vars"
        dest: "{
  { inventory_dir }}/assembled_vars/{
  { group_names | hash('md5') }}.yml"
        regexp: "({
  { group_names | join('|') }})"
      delegate_to: localhost
      become: false

    - name: "Include group custom vars"
      ansible.builtin.include_vars:
        file: "{
  { inventory_dir }}/assembled_vars/{
  { group_names | hash('md5') }}.yml"
        name: "temp_group_vars"

    - name: Combine vars from role and groups
      set_fact:
        combined_vars: "{
  { role_vars | combine(group_vars, recursive=true) }}"
      vars:
        role_vars: "{
  { lookup('vars', 'temp_role_vars') }}"
        group_vars: "{
  { lookup('vars', 'temp_group_vars') }}"

  rescue:
    - name: "Create {
  { inventory_dir }}/assembled_vars"
      file:
        path: "{
  { inventory_dir }}/assembled_vars"
        state: directory
      delegate_to: localhost
      become: false

    - include_tasks: vars_combine.yml

完整代码：https://gitee.com/bottlelee/haibinlee-ansible-roles/blob/develop/pre-tasks/tasks/vars_combine.yml

变量调用

合并的变量，以 combined_vars 开头，那么变量的使用就从原来的 var_a 变成 combined_vars.var_a 即可。

这种方法，方便了使用者用最少的代码量，自定义所需的变量。尤其是你遇到 failed to combine variables, expected dicts but got a 'NoneType' and a 'dict' 这样的报错。