实验目的

• 判断Liveness 探测 对副本数变化
• 判断Readiness 探测 对副本数变化
• 判断Readiness 探测 再次健康时是否从新加入svc

Liveness 实验

# cat centos.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  creationTimestamp: null
  labels:
    app: centos-test
  name: centos-test
spec:
  replicas: 5
  selector:
    matchLabels:
      app: centos-test
  strategy: {}
  template:
    metadata:
      creationTimestamp: null
      labels:
        app: centos-test
    spec:
      containers:
      - image: centos:7
        name: centos
        args:
        - tail
        - -f
        - /etc/hosts
        livenessProbe:
          httpGet:
            path: /
            port: 80
          initialDelaySeconds: 60
          periodSeconds: 3
          
# kubectl exec -it POD_NAME -- python -m SimpleHTTPServer 80

实验小结

副本数变化没有任何变化,pod名字也没有变化.

Readiness 实验

# cat centos.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  creationTimestamp: null
  labels:
    app: centos-test
  name: centos-test
spec:
  replicas: 5
  selector:
    matchLabels:
      app: centos-test
  strategy: {}
  template:
    metadata:
      creationTimestamp: null
      labels:
        app: centos-test
    spec:
      containers:
      - image: centos:7
        name: centos
        args:
        - tail
        - -f
        - /etc/hosts
        readinessProbe:
          httpGet:
            path: /
            port: 80
          initialDelaySeconds: 60
          periodSeconds: 3
---
apiVersion: v1
kind: Service
metadata:
  creationTimestamp: null
  labels:
    app: centos-test
  name: centos-svc
spec:
  ports:
  - port: 80
    protocol: TCP
    targetPort: 80
  selector:
    app: centos-test
  type: NodePort
  
# kubectl exec -it POD_NAME -- python -m SimpleHTTPServer 80

实验小结

副本数变化没有任何变化,健康状态下ep可以挂载Pod IP
健康健康,不健康从ep摘除,健康从ep挂载

整体总结

• Liveness 探测 用于应用本身导致的错误重启可以快速解决
• Readiness 探测 用于自身是否可以正常接收流量，和非本身问题导致的错误，等待后端问题解决就可以恢复