Ansible 2.4.2.0跨Play访问变量失败排查求助

嘿，我来帮你搞定这个Ansible跨Play访问变量的问题！你用Ansible 2.4.2.0写的Playbook里，第一个Play在all-compute-host主机上用heat-admin执行hostname命令并注册了变量，第二个Play想在localhost通过hostvars访问这个变量，但一直报变量未定义的错误，试过好几种写法都不行对吧？

先看看你原来的Playbook代码：

$ cat test.yml
- name: Finding Master VMs
  hosts: all-compute-host
  remote_user: heat-admin
  tasks:
    - name: Getting master VM's hostname
      shell: hostname
      register: hostname_output
- name: Access in different play
  hosts: localhost
  connection: local
  tasks:
    - name: Testing vars
      debug: var='{{ hostvars[item]['hostname_output']['stdout'] }}'
      with_items: groups['all-compute-host']

报错信息是这样的：

TASK [Testing vars] *******************************************************************************************************************
fatal: [localhost]: FAILED! => {"msg": "The task includes an option with an undefined variable. The error was: u\"hostvars['groups['all-compute-host']']\" is undefined\n\nThe error appears to have been in '/tmp/test.yml': line 18, column 5, but may\nbe elsewhere in the file depending on the exact syntax problem.\n\nThe offending line appears to be:\n\n tasks:\n - name: Testing vars\n ^ here\n\nexception type: <class 'ansible.errors.AnsibleUndefinedVariable'>\nexception: u\"hostvars['groups['all-compute-host']']\" is undefined"} to retry, use: --limit @/tmp/test.retry

问题出在哪？ #

仔细看报错信息里的hostvars['groups['all-compute-host']']，这说明Ansible把groups['all-compute-host']当成了一个字符串键，而不是去引用组变量，根本原因有两个：

1. with_items里的groups['all-compute-host']没有用{{ }}包裹，Ansible不会把它当成变量解析，而是直接作为字符串处理，导致循环时item的值就是groups['all-compute-host']，自然hostvars里没有这个主机。
2. debug的var参数不需要用{{ }}包裹变量，加了之后反而会引发引号嵌套的解析问题。

正确的解决方案 #

把第二个Play里的debug任务改成下面这样：

- name: Access in different play
  hosts: localhost
  connection: local
  tasks:
    - name: Testing vars
      debug:
        var: hostvars[item].hostname_output.stdout
      with_items: "{{ groups['all-compute-host'] }}"

或者用方括号的写法也可以，效果一样：

debug:
  var: hostvars[item]['hostname_output']['stdout']

为什么之前的写法都不行？ #

我来给你逐个分析：

• debug: var=hostvars[item]['hostname_output']['stdout']：with_items没加{{ }}，item还是字符串，无法正确匹配主机。
• debug: var=hostvars.item.hostname_output.stdout：动态变量名不能用点语法，得用hostvars[item]这种方括号形式才能解析item变量。
• debug: var=hostvars.all-compute-host.hostname_output.stdout：hostvars的键是单个主机的inventory名称，不是组名，直接用组名访问肯定找不到。
• debug: var=hostvars['all-compute-host']['hostname_output']['stdout']：同上，组名不能作为hostvars的键，得遍历组里的每个主机。
• debug: var='hostvars.compute01.hostname_output.stdout'：如果compute01是组里的单个主机，这个写法是对的，但只能访问这一台，没法遍历整个组。

修正后的完整Playbook #

- name: Finding Master VMs
  hosts: all-compute-host
  remote_user: heat-admin
  tasks:
    - name: Getting master VM's hostname
      shell: hostname
      register: hostname_output

- name: Access in different play
  hosts: localhost
  connection: local
  tasks:
    - name: Testing vars
      debug:
        var: hostvars[item].hostname_output.stdout
      with_items: "{{ groups['all-compute-host'] }}"

这样修改后，Ansible会正确遍历all-compute-host组里的每台主机，然后通过hostvars[item]获取对应主机上注册的hostname_output变量，就能正常输出每台主机的 hostname 了。

内容的提问来源于stack exchange，提问作者veerendra2