Ansible Inventory and Configuration

Ansible Configuration

Every bare-metal node in this homelab is configured through version-controlled Ansible playbooks. Nothing is typed into a node by hand.

Step 1: Learn the repository layout

The automation structure in ansible/ manages the bare-metal lifecycle:

• ansible/inventory/hosts.yaml: Lists control plane nodes and worker nodes with hostnames, SSH users, and IP addresses.
• ansible/group_vars/all.yaml: Defines global cluster variables including pinned K3s versions, TLS SANs, and storage paths.
• ansible/site.yaml: The master playbook defining the sequence of roles applied to hosts.
• ansible/roles/: Modular tasks for OS preparation (common), Tailscale node settings (tailscale), GPU setup (nvidia), K3s server (k3s_server), and ArgoCD bootstrap (argocd). Each role has a matching tag in ansible/site.yaml, so you can run one role with --tags.

Step 2: Configure host inventory

Inspect and update ansible/inventory/hosts.yaml with your node information:

k3s_cluster:
  children:
    k3s_servers:
      hosts:
        legion:
          ansible_host: 100.66.139.118
          ansible_user: sudhanva
          k3s_node_ip: "100.66.139.118"
          k3s_external_ip: "100.66.139.118"

Step 3: Configure cluster variables

Review ansible/group_vars/all.yaml to ensure cluster parameters match your network and storage layout:

k3s_version: "v1.36.4+k3s1"
local_storage_path: "/home/k3s-storage"
tls_sans:
  - "100.66.139.118"
  - "10.0.0.133"
  - "legion"
  - "legion.ainu-herring.ts.net"
k3s_kubelet_args:
  - "kube-reserved=cpu=250m,memory=512Mi"
  - "system-reserved=cpu=250m,memory=512Mi"
  - "eviction-hard=memory.available<500Mi,nodefs.available<10%,imagefs.available<10%"

k3s_node_ip and k3s_external_ip are both the node’s Tailscale IP. The Tailscale IP stays the same across LAN, DHCP, and interface changes (Ethernet or Wi-Fi), so the API server, kubelet, and host-network pods always have a reachable address. k3s_kubelet_args reserves CPU and memory for K3s and the OS so workloads cannot starve the control plane. See Node Networking and DNS.

Step 4: Verify Ansible connectivity

Test connectivity from your workstation to the target node:

ansible all -i ansible/inventory/hosts.yaml -m ping

If the ping reports SUCCESS, your automation environment is ready. Proceed to K3s Bootstrap to execute the provisioning playbook.