tfserving-deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: tfserving
  labels:
    app: tfserving
spec:
  selector:
    matchLabels:
      app: tfserving
  template:
    metadata:
      labels:
        app: tfserving
        version: v1
    spec:
      containers:
      - name: tfserving
        image: tensorflow/serving:latest
        imagePullPolicy: IfNotPresent
        args:
        - --port=8500
        - --rest_api_port=8501
        - --monitoring_config_file=s3://config/prometheus.config
        - --model_config_file=s3://config/models.config
        - --model_config_file_poll_wait_seconds=60
        env:
        - name: S3_ENDPOINT
          value: 'minio.default:9000'
        - name: AWS_ACCESS_KEY_ID
          value: 'minio'
        - name: AWS_SECRET_ACCESS_KEY
          value: 'minio123'
        - name: TF_CPP_MIN_LOG_LEVEL
          value: '3' 
        - name: S3_USE_HTTPS
          value: '0'
        - name: S3_USE_SSL
          value: '0' 
        ports:
        - containerPort: 8500
        - containerPort: 8501
        resources:
          limits:
            cpu: '100m'
          requests:
            cpu: '50m'
---
apiVersion: v1
kind: Service
metadata:
  name: tfserving
spec:
  type: LoadBalancer
  ports:
    - name: grpc-tf-serving
      port: 8500
      targetPort: 8500
    - name: http-tf-serving
      port: 8501
      targetPort: 8501
  selector:
    app: tfserving
---
apiVersion: autoscaling/v2beta1
kind: HorizontalPodAutoscaler
metadata:
  name: tfserving
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: tfserving
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      targetAverageUtilization: 50