hh

roles/vllm/README.md

@@ -0,0 +1,48 @@
+# vLLM
+
+Installs vLLM via Helm for GPU-enabled inference workloads.
+
+## Requirements
+
+- Kubernetes cluster with NVIDIA GPU workers
+- NVIDIA device plugin deployed
+- NGC API key for downloading models (stored in secret `vllm-ngc-secret`)
+
+## Role Variables
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `vllm_chart_version` | `0.1.0` | vLLM Helm chart version |
+| `vllm_namespace` | `vllm-system` | Kubernetes namespace |
+| `vllm_num_replicas` | `2` | Number of replicas for HA |
+| `vllm_image.repository` | `vllm/vllm-openai` | Container image repository |
+| `vllm_image.tag` | `latest` | Container image tag |
+| `vllm_image.pull_policy` | `IfNotPresent` | Image pull policy |
+| `vllm_resources.limits.nvidia.com/gpu` | `1` | GPU limit per pod |
+| `vllm_resources.requests.nvidia.com/gpu` | `1` | GPU request per pod |
+| `vllm_env` | `[]` | Environment variables (e.g., NGC_API_KEY) |
+| `vllm_extra_args` | `[]` | Additional CLI arguments |
+| `vllm_kubeconfig` | `/home/{{ ansible_user }}/.kube/config` | Path to kubeconfig |
+
+## Example Playbook
+
+```yaml
+- hosts: k8s_manager
+  become: false
+  roles:
+    - vllm
+```
+
+## Prerequisites
+
+- Create secret with NGC API key:
+  ```bash
+  kubectl create secret generic vllm-ngc-secret \
+    --from-literal=api-key=your-ngc-api-key \
+    -n vllm-system
+  ```
+
+- Ensure GPU workers are labeled:
+  ```bash
+  kubectl label nodes <node> nvidia.com/gpu.present=true
+  ```

roles/vllm/defaults/main.yml

@@ -0,0 +1,28 @@
+---
+vllm_chart_version: "0.1.0"
+vllm_namespace: vllm-system
+
+vllm_deployment_mode: "distributed"
+vllm_num_replicas: 2
+
+vllm_image:
+  repository: "vllm/vllm-openai"
+  tag: "latest"
+  pull_policy: "IfNotPresent"
+
+vllm_resources:
+  limits:
+    nvidia.com/gpu: "1"
+  requests:
+    nvidia.com/gpu: "1"
+
+vllm_env:
+  - name: "NGC_API_KEY"
+    valueFrom:
+      secretKeyRef:
+        name: "vllm-ngc-secret"
+        key: "api-key"
+
+vllm_extra_args: []
+
+vllm_kubeconfig: "/home/{{ ansible_user }}/.kube/config"

roles/vllm/meta/main.yml

@@ -0,0 +1,12 @@
+---
+galaxy_info:
+  role_name: vllm
+  author: ops
+  description: Installs vLLM via Helm for GPU-enabled inference workloads
+  license: MIT
+  min_ansible_version: "2.14"
+  platforms:
+    - name: EL
+      versions:
+        - "9"
+  dependencies: []

roles/vllm/tasks/deploy.yml

@@ -0,0 +1,36 @@
+---
+- name: Helm | Add vLLM chart repository
+  ansible.builtin.command: helm repo add vllm https://vllm-project.github.io/charts
+  environment:
+    KUBECONFIG: "{{ vllm_kubeconfig }}"
+    PATH: "/usr/local/bin:/usr/bin:/bin"
+  register: _helm_repo_add
+  changed_when: "'already exists' not in _helm_repo_add.stdout"
+  failed_when: _helm_repo_add.rc != 0 and 'already exists' not in _helm_repo_add.stdout
+
+- name: Helm | Update chart repositories
+  ansible.builtin.command: helm repo update
+  environment:
+    KUBECONFIG: "{{ vllm_kubeconfig }}"
+    PATH: "/usr/local/bin:/usr/bin:/bin"
+  changed_when: false
+
+- name: vLLM | Render values
+  ansible.builtin.template:
+    src: vllm-values.yml.j2
+    dest: /tmp/vllm-values.yml
+    mode: "0600"
+
+- name: vLLM | Install or upgrade
+  ansible.builtin.command: >
+    helm upgrade --install vllm vllm/vllm
+    --namespace {{ vllm_namespace }}
+    --version {{ vllm_chart_version }}
+    --values /tmp/vllm-values.yml
+    --timeout 15m0s
+    --wait
+  environment:
+    KUBECONFIG: "{{ vllm_kubeconfig }}"
+    PATH: "/usr/local/bin:/usr/bin:/bin"
+  register: _helm_install
+  changed_when: "'STATUS: deployed' in _helm_install.stdout or 'has been upgraded' in _helm_install.stdout"

roles/vllm/tasks/main.yml

@@ -0,0 +1,6 @@
+---
+- name: Create vLLM namespace
+  ansible.builtin.include_tasks: namespace.yml
+
+- name: Deploy vLLM via Helm
+  ansible.builtin.include_tasks: deploy.yml

roles/vllm/tasks/namespace.yml

@@ -0,0 +1,10 @@
+---
+- name: Create namespace
+  ansible.builtin.kubectl:
+    name: "{{ vllm_namespace }}"
+    api_version: "v1"
+    kind: "Namespace"
+    state: "present"
+  environment:
+    KUBECONFIG: "{{ vllm_kubeconfig }}"
+    PATH: "/usr/local/bin:/usr/bin:/bin"

roles/vllm/templates/vllm-values.yml.j2

@@ -0,0 +1,16 @@
+replicaCount: {{ vllm_num_replicas }}
+
+image:
+  repository: {{ vllm_image.repository }}
+  tag: {{ vllm_image.tag }}
+  pullPolicy: {{ vllm_image.pull_policy }}
+
+resources:
+  limits:
+    nvidia.com/gpu: {{ vllm_resources.limits.nvidia.com/gpu | default('1') }}
+  requests:
+    nvidia.com/gpu: {{ vllm_resources.requests.nvidia.com/gpu | default('1') }}
+
+env: {{ vllm_env | default([]) | to_json }}
+
+extraArgs: {{ vllm_extra_args | default([]) | to_json }}