# 主模型：qwen3.6-27b（对齐现网逻辑名）；1 卡；挂载模型 PVC
apiVersion: apps/v1
kind: Deployment
metadata:
  name: vllm-qwen36-27b
  namespace: gpu-infer
  labels:
    app: vllm-qwen36-27b
    model: qwen3.6-27b
spec:
  replicas: 2
  selector:
    matchLabels:
      app: vllm-qwen36-27b
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxUnavailable: 0
      maxSurge: 1
  template:
    metadata:
      labels:
        app: vllm-qwen36-27b
        model: qwen3.6-27b
    spec:
      nodeSelector:
        nvidia.com/gpu.present: "true"
      tolerations:
        - key: nvidia.com/gpu
          operator: Equal
          value: "true"
          effect: NoSchedule
      containers:
        - name: vllm
          image: harbor.example.local/ai/vllm-openai:v0.8.5
          args:
            - --model
            - /models/qwen3.6-27b
            - --served-model-name
            - qwen3.6-27b
            - --host
            - 0.0.0.0
            - --port
            - "8000"
            - --tensor-parallel-size
            - "1"
          ports:
            - name: http
              containerPort: 8000
          env:
            - name: HUGGING_FACE_HUB_TOKEN
              valueFrom:
                secretKeyRef:
                  name: hf-token
                  key: token
                  optional: true
          resources:
            requests:
              cpu: "8"
              memory: 48Gi
              nvidia.com/gpu: "1"
            limits:
              cpu: "16"
              memory: 64Gi
              nvidia.com/gpu: "1"
          readinessProbe:
            httpGet:
              path: /health
              port: http
            initialDelaySeconds: 60
            periodSeconds: 10
          livenessProbe:
            httpGet:
              path: /health
              port: http
            initialDelaySeconds: 120
            periodSeconds: 20
          volumeMounts:
            - name: model
              mountPath: /models/qwen3.6-27b
              readOnly: true
      volumes:
        - name: model
          persistentVolumeClaim:
            claimName: pvc-qwen36-27b
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: pvc-qwen36-27b
  namespace: gpu-infer
spec:
  accessModes: ["ReadOnlyMany"]
  resources:
    requests:
      storage: 80Gi
  storageClassName: nfs-models