loading tasks file based on ansible_fqdn value

hi,

I want to take different sligthly routes in a playbook based on the particular remote. All remotes are in the same hosts group.
I am trying to do this by setting a fact based on the ansible_fqdn value.
Based on that fact ('keycloak' in my example here) I am including different tasks files.

these 2 options do not work though:

There’s a problem with your “when:” conditionals.

First of all, throw out the second, because the strings that are “when:” expressions are already being evaluated in a Jinja2 context, so the mustaches shouldn’t be there.

In the first option, on this line:

  when: ansible_fqdn == VM-0213.step.zrz.internal

the “ansible_fqdn” part is okay, but the expression on the right-hand side “==” should be a string. Instead it’s the value of the variable “VM” minus the octal number 0213’s missing attribute “step”'s missing attribute “zrz”'s missing attribute “internal”.

So, yeah, no wonder that doesn’t work.

Try this instead:

  when: ansible_fqdn == "VM-0213.step.zrz.internal"

and let us know how you get on.

Or, better yet in my opinion, set the default value for “level” in group_vars/<name_of_group>.yml, then for hosts which need a different “level” set that in an appropriately named host_vars file, like host_vars/VM-0213.step.zrz.internal.yml.

This let’s you expresses facts about hosts as data rather than as results of expressions in set_fact or other tasks, task which you no longer need to run.

when: ansible_fqdn == “VM-0213.step.zrz.internal”

does not create any error in the set_fact TASK. It does not set anything either unfortunately. Which, thinking of it, seems logical as we just compare the string ansible_fqdn with another string and not the value of the variale {{ ansible_fqdn }}. so that seems to be a dead end.

the second option seems too complex for my current scenario. I am trying to gather all variables into a singe place. And having to create a host_vars/.yml file for each host on each execution contradicts that a bit too much.

I could actually manage this with:

  • ansible.builtin.set_fact:
    level: keycloak
    when "‘VM-0312.step.zrz.internal’ in inventory_hostname’

- ansible.builtin.set_fact:
    level: keycloak
  when: ansible_fqdn == VM-0312.step.zrz.internal

Make your live easier and put the logic into the *vars*. For example,

  > cat group_vars/all/level.yml
  level_dict:
    VM-0311.step.zrz.internal: passkey
    VM-0312.step.zrz.internal: keycloak
    VM-0313.step.zrz.internal: falsekey
  level: "{{ level_dict[ansible_fqdn] }}"

Given the inventory for testing

  > cat hosts
  host1 ansible_fqdn=VM-0311.step.zrz.internal
  host2 ansible_fqdn=VM-0312.step.zrz.internal
  host3 ansible_fqdn=VM-0313.step.zrz.internal

use *set_fact" if you want to 'instantiate' the variable (put the
variable into the *hostvars*). You can omit *set_fact* if you don't
need hostvars.*.level. For example, the play

  - hosts: all
    tasks:
      - set_fact:
          level: "{{ level }}"
      - debug:
          var: level

gives (abridged)

  ok: [host1] =>
    level: passkey
  ok: [host2] =>
    level: keycloak
  ok: [shot3] =>
    level: falsekey

As a result, the code is cleaner. Also the concentration of data into
a single point of failure (a dictionary and assignment isolated in
*vars*) makes the code more robust.

This indicates that your inventory_hostname and your ansible_fqdn are not the same, which can certainly be the case.

But Vladimir Botka is correct, and so is your intuition, that you should have all these variables in the same place. However, that place shouldn’t be in your tasks:.

Rule of thumb: If you find you have to edit your task: sections / files to add, remove, or rename hosts, you’re doing it wrong.