How can I have a Task always executed on localhost at the end of a Playbook even if the hosts all fail?

[Larry Fast]

I have a diagnostic task that dumps the Ansible dictionaries at the end of a run. However if the hosts fail the Playbook stops and my task doesn't get executed. But of course that's when it's most important.  How can I arrange that this task always runs after everything else?

For what it's worth, here's my dump_vars code. The real goal is to have a data dump at the end of each Playbook. Another means to this end would also be a valuable solution.

... Other Playbook code ...

# Last Play

- hosts: localhost

  gather_facts: no

  tasks:

    - name: dump all vars

      local_action: template src=dump_vars.j2 dest=./dump_vars.out

      tags: [testing1]

      ignore_errors: yes

      changed_when: False

cat dump_vars.j2

{

  "aaa_export_info1": "Export of all Ansible 1.x Dicts as a single JSON dict",

  "aaa_export_info2": "This file combines the dicts into a single JSON record",

  "aaa_export_info3": "Note that indentation is a bit off",

  "vars":         {{ vars        | remove_sensitive_data | to_nice_json }},

  "environment":  {{ environment | remove_sensitive_data | to_nice_json }},

  "group_names":  {{ group_names | remove_sensitive_data | to_nice_json }},

  "groups":       {{ groups      | remove_sensitive_data | to_nice_json }},

  "hostvars":     {{ hostvars    | remove_sensitive_data | to_nice_json }}

}

[Giovanni Tirloni]

On Mon, 30 Mar 2015 22:07 -0700, Larry Fast <lfas...@gmail.com> wrote:
> I have a diagnostic task that dumps the Ansible dictionaries at the end
> of
> a run. However if the hosts fail the Playbook stops and my task doesn't
> get
> executed. But of course that's when it's most important. How can I
> arrange
> that this task always runs after everything else?

You can have a handler for that and start ansible with
'--force-handlers'.

v2 is coming with try/except/finally [1]

[1] -
http://www.slideshare.net/jimi-c/whats-new-in-v2-ansiblefest-london-2015/10

Giovanni

[Larry Fast]

I want to be sure I understand the scope of the stop-on-failure behavior.  Is this correct: If a single Task fails for all selected servers, execution of the whole Playbook stops.  The exception is that all handlers will run IF force-handlers is true.

For clarity this implies that a Playbook managing 100 servers will stop if any Task fails for all-selected-servers even if that Task only applies to perhaps a single server.

[Brian Coca]

you can just have a play that acts on localhost and uses that task, it
will run independent of previous play failures.



--
Brian Coca

[Larry Fast]

Thanks Brian,

I think I tried that.  The following cut down example stops after the first Play fails for all selected hosts.  It did not execute the localhosts Play.  I'm all ears if you're suggesting something different or can show me what I've done wrong.

- hosts: some_hosts

  tasks:

     - fail: always fail to test the operation

       when: true

- hosts: localhost

  dump_vars