# 参考 Deployment：vLLM Qwen（作业17 · 实验 K8s 集群用）
# 对照现网：116.62.10.183 Docker 容器 vllm-qwen3.6-27b（H20，非 K8s）
#
# 应用：kubectl apply -f vllm-qwen-deployment.yaml -n ai-inference
# 发布检查清单 PR：python hw17_01_发布检查清单PR模板.py --deployment 本文件

apiVersion: apps/v1
kind: Deployment
metadata:
  name: vllm-qwen3-6-27b
  namespace: ai-inference
  labels:
    app: vllm-qwen
    model: qwen3.6-27b
    release-track: stable
  annotations:
    config/release-id: "20260827-001"
    config/owner: "platform-team"
    config/rollback-ref: "deployment/vllm-qwen3-6-27b@rev-previous"
spec:
  replicas: 1
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxSurge: 0
      maxUnavailable: 1
  selector:
    matchLabels:
      app: vllm-qwen
  template:
    metadata:
      labels:
        app: vllm-qwen
        model: qwen3.6-27b
      annotations:
        prometheus.io/scrape: "true"
        prometheus.io/port: "8000"
        prometheus.io/path: "/metrics"
    spec:
      # GPU 调度：需集群已装 NVIDIA GPU Operator + device plugin
      nodeSelector:
        nvidia.com/gpu.present: "true"
      tolerations:
        - key: nvidia.com/gpu
          operator: Exists
          effect: NoSchedule
      containers:
        - name: vllm
          image: vllm/vllm-openai:v0.8.5
          imagePullPolicy: IfNotPresent
          args:
            - --model
            - /models/Qwen3-6-27B-Instruct
            - --served-model-name
            - qwen3.6-27b
            - --host
            - "0.0.0.0"
            - --port
            - "8000"
            - --max-num-seqs
            - "8"
            - --gpu-memory-utilization
            - "0.75"
            - --trust-remote-code
          ports:
            - name: http
              containerPort: 8000
          resources:
            limits:
              nvidia.com/gpu: "1"
              memory: 64Gi
            requests:
              nvidia.com/gpu: "1"
              memory: 32Gi
          livenessProbe:
            httpGet:
              path: /health
              port: http
            initialDelaySeconds: 120
            periodSeconds: 30
            timeoutSeconds: 10
          readinessProbe:
            httpGet:
              path: /health
              port: http
            initialDelaySeconds: 60
            periodSeconds: 15
            timeoutSeconds: 5
          volumeMounts:
            - name: model-cache
              mountPath: /models
              readOnly: true
      volumes:
        - name: model-cache
          persistentVolumeClaim:
            claimName: qwen3-6-27b-model-pvc
---
apiVersion: v1
kind: Service
metadata:
  name: vllm-qwen
  namespace: ai-inference
  labels:
    app: vllm-qwen
spec:
  selector:
    app: vllm-qwen
  ports:
    - name: http
      port: 8000
      targetPort: http