Unusal behaviour with set_fact and add_host

ansible version: 2:11.2
jinja version: 3.0.3
python version: 3.6.8

Hello all

I have a peculiar problem with a relatively simple playbook I have written. It basically checks 2 servers to determine which one has more than 1 IP and uses that to set a simp;e fact called “is_leader” for each node.

This fact is then to be used to build an in-memory inventory using module add_host to iterate through the servers based on the value of that fact. See below:

  • name: Get Server Recycling Order
    hosts: “{{ my_servers | default (‘localhost’) }}”
    become: yes
    become_method: sudo
    become_user: root

tasks:

  • name: Set Patching Sequence
    tags: always
    block:

  • name: Get IP addresses of eth0 using nmcli command
    command: nmcli -g ip4.address conn show “System eth0”
    register: nmcli_output
    changed_when: false

  • name: Get nmcli output
    debug:
    var: nmcli_output

  • name: Set is_leader variable based on number of IPs
    set_fact:
    is_leader: “{{ (nmcli_output.stdout.split(‘|’) | length) > 1 }}”

  • name: Display if server is ACTIVE (1) or PASSIVE (0)
    debug:
    var: is_leader

  • name: Debug is_leader value
    debug:
    msg: “is_leader value for {{ inventory_hostname }} is {{ is_leader }}”

  • name: Add standby host
    add_host:
    name: “{{ inventory_hostname }}”
    groups: nodes
    when: not is_leader

  • name: Add primary host
    add_host:
    name: “{{ inventory_hostname }}”
    groups: nodes
    when: is_leader

#- name: Execute tasks on servers in a specified order

hosts: nodes

gather_facts: no

serial: 1

tasks:

- name: Run a command in a specified order

command: echo “Running on {{ inventory_hostname }}”

tags: always

The problem is that when this runs, the “is_leader” fact is ignored for when the condition is false and the standby node is not added to the inventory. The relevant extract from the execution is below:

TASK [Set is_leader variable based on number of IPs] **************************************************************************************************************************
Tuesday 23 April 2024 19:46:31 +0000 (0:00:00.079) 0:00:16.815 *********
ok: [node-1] => {
“ansible_facts”: {
“is_leader”: true
},
“changed”: false
}
ok: [node-2] => {
“ansible_facts”: {
“is_leader”: false
},
“changed”: false
}

TASK [Display if server is ACTIVE (1) or PASSIVE (0)] **************************************************************************************************************************
Tuesday 23 April 2024 19:46:31 +0000 (0:00:00.083) 0:00:16.898 *********
ok: [node-1] => {
“is_leader”: true
}
ok: [node-2] => {
“is_leader”: false
}

TASK [Debug is_leader value] **************************************************************************************************************************************************
Tuesday 23 April 2024 19:46:31 +0000 (0:00:00.072) 0:00:16.971 *********
ok: [node-1] => {}

MSG:

is_leader value for node-1 is True
ok: [node-2] => {}

MSG:

is_leader value for node-2 is False

TASK [Re-confirm is_leader before adding to group] ****************************************************************************************************************************
Tuesday 23 April 2024 19:46:31 +0000 (0:00:00.074) 0:00:17.045 *********
skipping: [node-1] => {}
ok: [node-2] => {}

MSG:

Preparing to add node-2 to stby group based on is_leader = False

TASK [Add standby proxysql node] ***********************************************************************************************************************************************
Tuesday 23 April 2024 19:46:31 +0000 (0:00:00.088) 0:00:17.134 *********
skipping: [node-1] => {
“changed”: false,
“skip_reason”: “Conditional result was False”
}

TASK [Re-confirm is_leader before adding to group] ****************************************************************************************************************************
Tuesday 23 April 2024 19:46:31 +0000 (0:00:00.036) 0:00:17.170 *********
ok: [node-1] => {}

MSG:

Preparing to add node-1 to prim group based on is_leader = True
skipping: [node-2] => {}

TASK [Add primary proxysql node] ***********************************************************************************************************************************************
Tuesday 23 April 2024 19:46:31 +0000 (0:00:00.059) 0:00:17.230 *********
changed: [node-1] => {
“add_host”: {
“groups”: [
“nodes_prim”
],
“host_name”: “node-1”,
“host_vars”: {}
},
“changed”: true
}

PLAY RECAP *********************************************************************************************************************************************************************
localhost : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
node-1 : ok=8 changed=1 unreachable=0 failed=0 skipped=2 rescued=0 ignored=0
node-2 : ok=7 changed=0 unreachable=0 failed=0 skipped=1 rescued=0 ignored=0

I’m not sure what I’m missing regarding this behaviour as it seems like it should simply work?

Thanks

Hard to say. The job log you show doesn’t match the tasks in the playbook you posted. They seem to be the same down to the “Debug is_leader value” task, but after that the task names in the log don’t match those in the playbook.

In the playbook you posted, task “Add standby host” an task “Add primary host” both add the hosts to the same group. That may be a source of your problem. But as we don’t see the output for those tasks …, hmm.

All that said, I don’t see why this one task wouldn’t work for your purposes:

    - name: Add hosts to groups
      ansible.builtin.add_host:
        name: "{{ inventory_hostname }}"
        groups:
          - nodes
          - "{{ 'nodes_prim' if is_active | d(false) else 'nodes_stby' }}"

Testing if my email is still deleted without any code

I’ve been staring at this for quite a while. I was able to reproduce your issue, and I now understand what’s going on.

Rather than trying to duplicate the “single vs. multiple ips on an interface”, I just ran a playbook against two hosts (tango and cloister), and did

    - name: Set is_leader for tango and cloister
      ansible.builtin.set_fact:
        is_leader: '{{ true if inventory_hostname == "tango" else false }}'

After that, I could play with ansible.builtin.add_host. And that’s where the confusion starts. Everything works “as expected” except for theses add_host tasks. So I went back and read the document on add_host very closely. I had not appreciated this bit about the bypass_host_loop property:

Forces a ‘global’ task that does not execute per host, this bypasses per host templating and serial, throttle and other loop considerations

Conditionals will work as if run_once is being used, variables used will be from the first available host

This action will not work normally outside of lockstep strategies

So you can’t run ansible.builtin.add_host on all your hosts the same way you would, well, almost everything else. If you want to loop over all your hosts (which is the default behavior for most other modules), then you have to do it explicitly, like this:

    - name: Display is_leader for all hosts
      ansible.builtin.debug:
        var: is_leader

    - name: Show groups BEFORE add_host
      ansible.builtin.debug:
        msg: "{{ groups }}"

    - name: Add hosts to either the nodes_prim or nodes_stby group
      **ansible.builtin.add_host**:
        name: "**{{ item }}**"
        groups:
          - "{{ 'nodes_prim' if **hostvars[item].is_leader** else 'notes_stby' }}"
      **loop: "{{ ansible_play_hosts }}"**

    - name: Show groups AFTER add_host
      ansible.builtin.debug:
        msg: "{{ groups }}"