k3s (D6): Dienst-VM k3s per tofu, Rollen k3s-server/k3s-agent/k3s-apps, Manifeste, NEX-Sprachdienste als k3s-Umzugstest

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

platform/proxmox/ansible/roles/k3s-apps/defaults/main.yml

@@ -0,0 +1,6 @@
+# Grundausstattung im k3s-Cluster (von ops per kubectl): NVIDIA-Runtime und Device-Plugin, Portainer-Agent.
+k3s_manifest_dir: "{{ playbook_dir }}/../../../../deploy/k8s"
+k3s_base_manifests:
+  - runtimeclass-nvidia.yaml
+  - nvidia-device-plugin.yaml
+  - portainer-agent.yaml

platform/proxmox/ansible/roles/k3s-apps/tasks/main.yml

@@ -0,0 +1,28 @@
+- name: Grund-Manifeste anwenden
+  ansible.builtin.command: kubectl apply -f {{ k3s_manifest_dir }}/{{ item }}
+  loop: "{{ k3s_base_manifests }}"
+  register: k3s_apply
+  changed_when: "'configured' in k3s_apply.stdout or 'created' in k3s_apply.stdout"
+
+- name: Warten, bis jede GPU-VM ihre GPUs meldet (nvidia.com/gpu)
+  ansible.builtin.command: >
+    kubectl get node {{ item }} -o jsonpath='{.status.allocatable.nvidia\.com/gpu}'
+  loop: "{{ groups['k3s_agents'] }}"
+  register: k3s_gpus
+  until: (k3s_gpus.stdout | default('0') | int) > 0
+  retries: 24
+  delay: 5
+  changed_when: false
+
+- name: Warten, bis der Portainer-Agent läuft
+  ansible.builtin.command: kubectl -n portainer rollout status deployment/portainer-agent --timeout=180s
+  changed_when: false
+
+- name: Übersicht
+  ansible.builtin.command: kubectl get nodes -o custom-columns=NAME:.metadata.name,STATUS:.status.conditions[-1].type,GPU:.status.allocatable.nvidia\.com/gpu,VERSION:.status.nodeInfo.kubeletVersion
+  register: k3s_nodes
+  changed_when: false
+
+- name: Knoten
+  ansible.builtin.debug:
+    msg: "{{ k3s_nodes.stdout_lines }}"