This message was deleted.

You can (if you want to deploy it on an RKE cluster... If you have another method to deploy clusters, that might make more sense to use) I had this template: (the rest of the playbook might be relevant for the logic here as well)

nodes:
{% for host in groups['rancher'] %}
  - address: {{ hostvars[host]['inventory_hostname'] }}
{% if hostvars[host]['internal_ip'] is defined %}
    internal_address: {{ hostvars[host]['internal_ip'] }}
{% endif %}
    user: {{ hostvars[host]['rke_user'] | default(ansible_user) }}
    role: [{{ hostvars[host]['rke_roles'] | default('etcd,controlplane,worker') }}]
{% endfor %}

# TODO: Allow settings
services:
  etcd:
    snapshot: true
    creation: 6h
    retention: 24h

ssh_agent_auth: true

(Starting from a role for K3s or another cluster deployment method might be faster) (I never did upgrades of the RKE cluster with Ansible and used a kubeadm deployed cluster instead when the OS running the RKE cluster reached EoL) (You are probably less likely to have a role lying around for deploying a cluster if you use Rancher to deploy cluster, which we don't - it lacked features (e.g. AWS clusters could not autoscale) and was extremely fragile last time we tried it (~2.2))