Ansible剧本角色与任务执行异常处理方案咨询

Alright, let's tackle this Ansible problem step by step. Your requirement is straightforward but needs careful handling of error flow—here's how you can implement it cleanly:

Core Approach #

We'll split your playbook into two distinct plays to separate role execution from the mandatory post-task:

1. The first play runs all your roles, configured to halt immediately if any task fails (skipping all remaining roles).
2. The second play runs your specific task on system_a regardless of whether the first play succeeded or failed.

Playbook Implementation #

Here's the full playbook code with clear explanations:

# Play 1: Execute all roles, halt on any failure
- name: Run designated roles on system_a and system_b
  hosts: system_a, system_b
  # Critical setting: If ANY task fails on ANY host, stop the entire play immediately
  # This skips all leftover roles (for both system_a and system_b)
  any_errors_fatal: yes
  tasks:
    # Roles restricted to system_a
    - name: Execute role1 on system_a hosts
      import_role:
        name: role1
      when: inventory_hostname in groups['system_a']

    - name: Execute role2 on system_a hosts
      import_role:
        name: role2
      when: inventory_hostname in groups['system_a']

    # Roles restricted to system_b
    - name: Execute role3 on system_b hosts
      import_role:
        name: role3
      when: inventory_hostname in groups['system_b']

    - name: Execute role4 on system_b hosts
      import_role:
        name: role4
      when: inventory_hostname in groups['system_b']

    - name: Execute role5 on system_b hosts
      import_role:
        name: role5
      when: inventory_hostname in groups['system_b']

    - name: Execute role6 on system_b hosts
      import_role:
        name: role6
      when: inventory_hostname in groups['system_b']

# Play 2: Mandatory task for system_a (runs no matter what)
- name: Run required specific task on system_a
  hosts: system_a
  tasks:
    - name: Your specific task here
      # Replace this with your actual task (e.g., log cleanup, service restart, alert)
      debug:
        msg: "Executing the mandatory post-task on system_a"

Key Details Breakdown #

• any_errors_fatal: yes: This is the linchpin of the solution. It ensures that if any task in the first play fails (on any host), Ansible stops executing all remaining tasks/roles immediately—exactly what you need to skip leftover roles when something goes wrong.
• Host targeting with when: Each role is locked to its intended host group, so roles only run on the correct machines and don't waste cycles on unrelated hosts.
• Separate play for the specific task: By isolating this task in its own play, it will always execute after the first play finishes—whether the first play succeeded fully or failed halfway through. This guarantees your system_a task runs no matter the outcome of role execution.

Alternative: Single-Play with Block/Rescue #

If you prefer to keep all logic in one play, you can use Ansible's block/rescue structure. This works similarly but keeps everything contained:

- name: Handle roles and post-task in a single play
  hosts: system_a, system_b
  tasks:
    - block:
        # All role executions go here
        - import_role: name=role1
          when: inventory_hostname in groups['system_a']
        - import_role: name=role2
          when: inventory_hostname in groups['system_a']
        - import_role: name=role3
          when: inventory_hostname in groups['system_b']
        - import_role: name=role4
          when: inventory_hostname in groups['system_b']
        - import_role: name=role5
          when: inventory_hostname in groups['system_b']
        - import_role: name=role6
          when: inventory_hostname in groups['system_b']
      rescue:
        # Optional: Mark play failure for logging/alerting purposes
        - name: Flag play as failed
          set_fact:
            play_failed: true
          delegate_to: localhost
          run_once: true
    # Mandatory task for system_a (runs regardless of block success/failure)
    - name: Execute specific task on system_a
      debug:
        msg: "Running the required task on system_a"
      when: inventory_hostname in groups['system_a']

Either approach will meet your requirements—pick the one that fits your existing playbook structure best!

内容的提问来源于stack exchange，提问作者Ankur Gupta