如何在远程Windows服务器而非Ansible本地机读取解析CSV文件？

原Playbook及背景 #

原本用于在localhost执行的Playbook，核心逻辑是从共享目录拉取CSV文件到本地，再读取处理数据：

- name: Fetch file
  fetch:
    src: \\10.14.2.130\shared\folder\Data\{{ filename }}
    dest: /var/lib/awx/projects/Windows/
    flat: yes
  delegate_to: 10.12.201.60
  become: yes
  become_method: runas
  become_user: SYSTEM

- name: Initialise data list
  set_fact:
    data_list: []
  delegate_to: localhost

- name: Read input file
  read_csv:
    path: "{{ filename }}"
    key: FirstName  
    fieldnames: FirstName,LastName,EmailAddress
    delimiter: ','
  register: userdata
  delegate_to: localhost

- name: Extract the list
  set_fact:
    data_list: "{{ data_list + [{ 'FirstName': item.FirstName, 'LastName': item.LastName, 'EmailAddress': item.EmailAddress }] }}"
  loop: "{{ userdata | community.general.json_query('dict.[*][0]') }}"

- name: Set fact for data_list
  set_fact: 
    data_list: "{{ data_list[1:] }}"

由于第一个fetch任务失败，无法将CSV文件拉取到localhost，于是修改Playbook，改为先把文件复制到Windows远程主机（10.12.201.60），再在该主机上读取处理。

修改后的Playbook及报错 #

- name: copy file
  win_copy:
    src: \\10.14.2.130\shared\folder\Data\{{ filename }}
    dest: C:/
    remote_src: true
  delegate_to: 10.12.201.60
  become: yes
  become_method: runas
  become_user: SYSTEM

- name: Initialise data list
  set_fact:
    data_list: []

- name: Read input file
  read_csv:
    path: C:\{{ filename }}
    key: FirstName  
    fieldnames: FirstName,LastName,EmailAddress
    delimiter: ','
  register: userdata
  delegate_to: 10.12.201.60

- name: Extract the list
  set_fact:
    data_list: "{{ data_list + [{ 'FirstName': item.FirstName, 'LastName': item.LastName, 'EmailAddress': item.EmailAddress }] }}"
  loop: "{{ userdata | community.general.json_query('dict.[*][0]') }}"

- name: Set fact for data_list
  set_fact: 
    data_list: "{{ data_list[1:] }}"

运行后出现报错：

fatal: [10.12.201.60]: FAILED! => {
    "msg": "The module read_csv was redirected to community.general.read_csv, which could not be loaded."
}

原因分析 #

read_csv模块属于community.general集合，是基于Python编写的Ansible模块。通过WinRM连接的Windows远程节点默认使用PowerShell执行模块，无法直接运行Python模块（除非在Windows上额外配置Python环境并适配Ansible连接），因此将read_csv任务委托到Windows主机时会加载失败。

解决方案 #

方案一：拉回CSV到localhost处理（推荐） #

将文件复制到Windows主机后，再拉回控制节点（localhost），利用控制节点的Python环境和已安装的community.general集合完成CSV读取和处理：

- name: copy file to Windows host
  win_copy:
    src: \\10.14.2.130\shared\folder\Data\{{ filename }}
    dest: C:/
    remote_src: true
  delegate_to: 10.12.201.60
  become: yes
  become_method: runas
  become_user: SYSTEM

- name: Fetch CSV file back to localhost
  fetch:
    src: C:\{{ filename }}
    dest: ./{{ filename }}
    flat: yes
  delegate_to: 10.12.201.60

- name: Initialise data list
  set_fact:
    data_list: []
  delegate_to: localhost

- name: Read input file on localhost
  read_csv:
    path: ./{{ filename }}
    key: FirstName  
    fieldnames: FirstName,LastName,EmailAddress
    delimiter: ','
  register: userdata
  delegate_to: localhost

- name: Extract the list
  set_fact:
    data_list: "{{ data_list + [{ 'FirstName': item.FirstName, 'LastName': item.LastName, 'EmailAddress': item.EmailAddress }] }}"
  loop: "{{ userdata | community.general.json_query('dict.[*][0]') }}"
  delegate_to: localhost

- name: Remove header row from data_list
  set_fact: 
    data_list: "{{ data_list[1:] }}"
  delegate_to: localhost

方案二：用PowerShell在Windows主机处理CSV #

如果必须在Windows主机上完成CSV处理，可以使用Windows原生的PowerShell命令替代read_csv模块，无需依赖Python：

- name: copy file to Windows host
  win_copy:
    src: \\10.14.2.130\shared\folder\Data\{{ filename }}
    dest: C:/
    remote_src: true
  delegate_to: 10.12.201.60
  become: yes
  become_method: runas
  become_user: SYSTEM

- name: Read CSV with PowerShell
  win_shell: |
    Import-Csv -Path "C:\{{ filename }}" -Header FirstName,LastName,EmailAddress -Delimiter ',' | ConvertTo-Json
  register: csv_output
  delegate_to: 10.12.201.60

- name: Parse CSV output to data_list
  set_fact:
    data_list: "{{ csv_output.stdout | from_json | list }}"

- name: Remove header row（若CSV本身自带表头，可删除此步骤）
  set_fact:
    data_list: "{{ data_list[1:] }}"

内容的提问来源于stack exchange，提问作者Blitzden