microdao-daarion/infrastructure/ansible/playbooks/k3s-install.yml

# DAARION Network - K3s Installation Playbook
# Lightweight Kubernetes cluster setup
---
# =============================================================================
# INSTALL K3S SERVER (MASTERS)
# =============================================================================
- name: Install K3s Server on Masters
  hosts: masters
  become: yes
  
  tasks:
    - name: Check if K3s is already installed
      stat:
        path: /etc/rancher/k3s/k3s.yaml
      register: k3s_installed
      
    - name: Download K3s installer
      get_url:
        url: https://get.k3s.io
        dest: /tmp/k3s-install.sh
        mode: '0755'
      when: not k3s_installed.stat.exists
        
    - name: Install K3s server
      shell: |
        INSTALL_K3S_VERSION={{ k3s_version }} \
        sh /tmp/k3s-install.sh server \
          --disable traefik \
          --disable servicelb \
          --write-kubeconfig-mode 644 \
          --tls-san {{ ansible_host }} \
          --tls-san {{ inventory_hostname }} \
          --node-label "datacenter={{ datacenter }}" \
          --node-label "node-role={{ node_role }}" \
          --cluster-cidr {{ daarion_network_cidr | default('10.42.0.0/16') }} \
          --service-cidr {{ daarion_service_cidr | default('10.43.0.0/16') }}
      args:
        creates: /etc/rancher/k3s/k3s.yaml
      register: k3s_install
        
    - name: Wait for K3s to be ready
      wait_for:
        port: 6443
        delay: 10
        timeout: 300
        
    - name: Wait for node to be ready
      shell: |
        export KUBECONFIG=/etc/rancher/k3s/k3s.yaml
        kubectl wait --for=condition=Ready node/{{ inventory_hostname }} --timeout=300s
      register: node_ready
      retries: 10
      delay: 10
      until: node_ready.rc == 0
        
    - name: Get K3s token
      slurp:
        src: /var/lib/rancher/k3s/server/node-token
      register: k3s_token_file
      
    - name: Save K3s token as fact
      set_fact:
        k3s_join_token: "{{ k3s_token_file.content | b64decode | trim }}"
        
    - name: Fetch kubeconfig
      fetch:
        src: /etc/rancher/k3s/k3s.yaml
        dest: "{{ playbook_dir }}/../kubeconfig/{{ inventory_hostname }}.yaml"
        flat: yes
        
    - name: Update kubeconfig with external IP
      delegate_to: localhost
      become: no
      replace:
        path: "{{ playbook_dir }}/../kubeconfig/{{ inventory_hostname }}.yaml"
        regexp: '127.0.0.1'
        replace: "{{ ansible_host }}"
        
    - name: Show K3s status
      shell: |
        export KUBECONFIG=/etc/rancher/k3s/k3s.yaml
        kubectl get nodes -o wide
      register: k3s_status
      changed_when: false
      
    - name: Display K3s status
      debug:
        var: k3s_status.stdout_lines

# =============================================================================
# INSTALL K3S AGENT (WORKERS)
# =============================================================================
- name: Install K3s Agent on Workers
  hosts: workers
  become: yes
  
  vars:
    k3s_master_host: "{{ hostvars[groups['masters'][0]].ansible_host }}"
    k3s_master_token: "{{ hostvars[groups['masters'][0]].k3s_join_token }}"
    
  tasks:
    - name: Check if K3s agent is already installed
      stat:
        path: /var/lib/rancher/k3s/agent
      register: k3s_agent_installed
      
    - name: Download K3s installer
      get_url:
        url: https://get.k3s.io
        dest: /tmp/k3s-install.sh
        mode: '0755'
      when: not k3s_agent_installed.stat.exists
        
    - name: Build node labels
      set_fact:
        node_labels: >-
          --node-label datacenter={{ datacenter }}
          --node-label node-role={{ node_role }}
          {% if gpu is defined and gpu %}
          --node-label gpu=true
          --node-label gpu-type={{ gpu_type | default('unknown') }}
          --node-label gpu-memory={{ gpu_memory | default('unknown') }}
          {% endif %}
          
    - name: Install K3s agent
      shell: |
        INSTALL_K3S_VERSION={{ k3s_version }} \
        K3S_URL=https://{{ k3s_master_host }}:6443 \
        K3S_TOKEN={{ k3s_master_token }} \
        sh /tmp/k3s-install.sh agent \
          {{ node_labels }}
      args:
        creates: /var/lib/rancher/k3s/agent
      register: k3s_agent_install
        
    - name: Wait for agent to connect
      pause:
        seconds: 30
      when: k3s_agent_install.changed

# =============================================================================
# VERIFY CLUSTER
# =============================================================================
- name: Verify K3s Cluster
  hosts: masters
  become: yes
  
  tasks:
    - name: Get cluster nodes
      shell: |
        export KUBECONFIG=/etc/rancher/k3s/k3s.yaml
        kubectl get nodes -o wide
      register: cluster_nodes
      changed_when: false
      
    - name: Display cluster nodes
      debug:
        var: cluster_nodes.stdout_lines
        
    - name: Get cluster info
      shell: |
        export KUBECONFIG=/etc/rancher/k3s/k3s.yaml
        kubectl cluster-info
      register: cluster_info
      changed_when: false
      
    - name: Display cluster info
      debug:
        var: cluster_info.stdout_lines
        
    - name: Create daarion namespace
      shell: |
        export KUBECONFIG=/etc/rancher/k3s/k3s.yaml
        kubectl create namespace daarion --dry-run=client -o yaml | kubectl apply -f -
      changed_when: false
      
    - name: Label GPU nodes
      shell: |
        export KUBECONFIG=/etc/rancher/k3s/k3s.yaml
        kubectl label nodes {{ item }} nvidia.com/gpu=true --overwrite
      loop: "{{ groups['gpu_nodes'] | default([]) }}"
      when: groups['gpu_nodes'] is defined
      ignore_errors: yes