Skip to content

Inventory and Config

Scope: telling Ansible what machines exist and how to reach them. Nothing runs until this file is right, and every "Ansible can't see my server" problem lives here.

The inventory

An inventory lists your machines, usually grouped by what they're for. The simple INI format (from this guide's examples/):

[webservers]
web1 ansible_host=127.0.0.1 ansible_connection=local
web2 ansible_host=127.0.0.1 ansible_connection=local

[webservers:vars]
app_port=8080

Read it as: a group called webservers containing two hosts. Variables after a hostname apply to that host; the [group:vars] block applies to the whole group — app_port=8080 here becomes usable on both machines (see Variables and facts).

Two lab-isms to translate for real life:

  • ansible_host=127.0.0.1 would normally be the server's real address or DNS name (or omitted — the hostname itself is used).
  • ansible_connection=local runs tasks on the control machine itself, which is how this guide's two "servers" live on one box so you can reproduce everything. Real servers use the default connection: SSH. Same playbooks, same modules — only the transport differs.

A YAML inventory says the same thing with more structure; INI is fine until it isn't (nested groups are where YAML starts winning).

Check what Ansible sees

Before running anything real, ask:

ansible webservers --list-hosts

  hosts (2):
    web1
    web2

If a machine is missing here, no playbook will find it either. If groups are nested or overlapping, this command (and ansible-inventory --graph) is how you confirm the shape matches your mental model.

ansible.cfg

A small config file next to your inventory saves flags on every command:

[defaults]
inventory = inventory.ini
host_key_checking = False
retry_files_enabled = False
  • inventory — so you stop typing -i inventory.ini.
  • host_key_checking = False — don't stop to ask "are you sure you want to connect" for new hosts. Fine for a lab and a home fleet; on anything serious, pre-seed known_hosts instead and leave checking on. (Documented — the lab never makes a real SSH connection.)
  • retry_files_enabled = False — don't scatter .retry files after failed runs.

Ansible reads ansible.cfg from the current directory first, so one project folder = one self-contained setup: config, inventory, playbooks. That's the layout the examples in this guide use, and the one to copy.

The first contact

ansible webservers -m ping

web1 | SUCCESS => {
    "changed": false,
    "ping": "pong"
}

ping is an Ansible module, not the network command: it proves Ansible can connect, find Python on the target, and run code there. SUCCESS + pong from every host is the green light for everything after this chapter. If it fails instead, the Troubleshooting chapter has the usual suspects word for word.