K8s+Flask+TF2.0部署

最新推荐文章于 2024-05-17 17:42:47 发布

learn-to-live

最新推荐文章于 2024-05-17 17:42:47 发布

阅读量291

点赞数

文章标签： leetcode 算法动态规划

本文链接：https://blog.csdn.net/ineedstudytosurvive/article/details/119533223

版权

1.将模型导出，制作TF2.0+Flask服务。

2.将flask服务文件，加入docker容器内部，提交新images.

3.制作Dockerfile.将(2)中的镜像启动服务。

From tf2_flask:v1
WORKDIR /model_server/albert_predict
EXPOSE 8101
CMD ["python", "albert_predict.py"]

docker build -t new_images:new_tag .

3.利用K8S写出deployment和service文件。

Deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: flask-demo
  labels:
    app: flask-albert
spec:
  selector:
    matchLabels:
      app: flask-albert
  replicas: 4
  template:
    metadata:
      labels:
        app: flask-albert
    spec:
      containers:
        - name: flaskdemo
          image: tf2flask:v2
          imagePullPolicy: IfNotPresent
          ports:
            - containerPort: 8101

Service.yaml

apiVersion: v1
kind: Service
metadata:
  name: flask-demo
  labels:
    app: flask-albert
spec:
  ports:
  - port: 8101
    protocol: "TCP"
    targetPort: 8101
  selector:
    app: flask-albert
  type: LoadBalancer
  externalIPs:
  - 192.168.101.1

4.kubectl create -f deployment.yaml

kubectl create -f service.yaml

启动完成。

5.查看服务状态：

kubectl get pods