Ansible handler - not called from rescue block

Question

Consider the following (reduced for brevity) playbook with a block and rescue block

- name: deploy block
  block:
  - name: debug meta
    debug: var=meta

  
  - name: create/update configmap with new data
    k8s:
      state: present
      namespace: "{{ namespace }}"
      definition: "{{ lookup('template', 'configmap.tpl.yml') }}"
      kind: ConfigMap
    notify:
      - successfull_deployment_slack_handler
  rescue:
    - name: Something went wrong handler
      debug:
        msg: "Something has failed in the playbook"
      notify: failure_deployment_slack_handler
- meta: flush_handlers

Testing this in the happy path works and calls the handlers as expected.
But when I fail the tasks in my testing I do see the expected debug message, but the actual handler isn't called (I've traded it for a debug message to verify).

Yes, I've tried adding meta: flush_handlers.

How can I make this work?

Ansible version: 2.9.9

score 2 · Accepted Answer · answered Jul 28 '20 at 17:57

The 'error' in your slim version of the playbook is that debug will never report to ansible as changed and because of that, the handler wont trigger.

In this playbook where i've replaced your rescue debug with a shell, the handler is called

---
- hosts: all
  handlers:
    - name: handler
      debug:
        msg: handler
  tasks:
    - name: deploy block
      block:
      - name: debug meta
        debug:
          msg: in the block

      - fail:
          msg: failing
      rescue:
        - name: Something went wrong handler
          shell: date
          notify: handler

results in

TASK [debug meta] **************************************************************************************************************************************************************************************************************************************************************
ok: [localhost] => {
    "msg": "in the block"
}

TASK [fail] ********************************************************************************************************************************************************************************************************************************************************************
fatal: [localhost]: FAILED! => {"changed": false, "msg": "failing"}

TASK [Something went wrong handler] ********************************************************************************************************************************************************************************************************************************************
changed: [localhost]

RUNNING HANDLER [handler] ******************************************************************************************************************************************************************************************************************************************************
ok: [localhost] => {
    "msg": "handler"
}

β.εηοιτ.βε · Answer 2 · 2020-07-28T18:07:56.677

Handlers are called only upon a change, the task meta: flush_handlers will execute the handlers only if there were handlers to be flushed in the first place, which is not your case.

As we’ve mentioned, modules should be idempotent and can relay when they have made a change on the remote system. Playbooks recognize this and have a basic event system that can be used to respond to change.

^{Source: https://docs.ansible.com/ansible/latest/user_guide/playbooks_intro.html#handlers-running-operations-on-change, emphasis, mine}

In order to make it work, you will have to create a changed state, somehow. That somehow could be by using changed_when:

- name: deploy block
  block:
  - name: debug meta
    debug: var=meta

  
  - name: create/update configmap with new data
    k8s:
      state: present
      namespace: "{{ namespace }}"
      definition: "{{ lookup('template', 'configmap.tpl.yml') }}"
      kind: ConfigMap
    notify:
      - successfull_deployment_slack_handler
  rescue:
    - name: Something went wrong handler
      debug:
        msg: "Something has failed in the playbook"
      changed_when: true
      notify: failure_deployment_slack_handler
- meta: flush_handlers

_{Related: https://stackoverflow.com/a/62183353/2123530}

Ansible handler - not called from rescue block

2 Answers2