How can I employ a "truly" dynamic inventory with Ansible, which changes along the way when new hosts become reachable?

Question

I provision virtual machines dynamically and they become reachable along the way, whilst my playbook is already executing. How do I obtain a "truly" dynamic inventory?

By "truly" dynamic inventory I mean:

Dynamic inventories are always evaluated when Ansible starts, does not matter if the inventory is a immutable file or a script which dynamically discovers a bunch of IP addresses. I need to evaluate the inventory along the way, after tasks which provision virtual machines and before executing further other tasks.

Answer 1

Dynamic inventories always evaluated when Ansible starts, does not matter if the inventory is a immutable file or a script which dynamically discovers a bunch of IP addresses. In my use case, I need to evaluate the inventory along the way, after provisioning virtual machines and before executing further other tasks.

However, nothing prevents you from spliting one big playbook into smaller ones and breaking the logic so that a certain playbook has chance to evaluate the environment when it (the playbook) starts and finds all hosts you expect to find.

The first playbook would be responsible for provisioning, waiting for the virtual machines to be reachable and retrieving their host keys (see notes below). Then the last thing the first playbook does is kicking off the second playbook, which is responsible for all further tasks, like this:

- name: Kicks off the second playbook, passing the IP address of the recently provisioned virtual machine
  shell: "ansible-playbook -i '{{provisioned_host}},' second_playbook,yaml

This approach seems to work. This is a proof of concept:

file: first_playbook.yml

---
- name: This is the first playbook
  hosts: localhost

  tasks:

    - name: set facts -- obtain hostname
      shell: "cat /etc/hostname"
      register: hostname
     
    - name: debug
      debug:
        msg: "This is the first playbook: hostname is {{hostname.stdout}}"

    - name: set facts -- provisioned_ip
      set_fact:
        provisioned_host: "192.168.3.152"

    # see: ansible-doc -t inventory host_list
    - name: Kicks off the second playbook, passing the IP address of the recently provisioned virtual machine
      shell: "ansible-playbook -i '{{provisioned_host}},' {{playbook_dir}}/second_playbook.yml"

file: second_playbook.yml

---
- name: This is the second playbook
  hosts: all

  tasks:

    - name: set facts -- obtain hostname
      shell: "cat /etc/hostname"
      register: hostname
          
     
    - name: debug
      debug:
        msg: "This is the second playbook: hostname is {{hostname.stdout}}"

Now call first_playbook:

ansible-playbook -vv first_playbook.yml

---

Notes:

• Ansible needs to know the host key of the virtual machines to be contacted. So, before the second playbook starts, you have to retrieve the host keys and insert them into your ~/.ssh/known_hosts, otherwise the second playbook will fail to connect.