services之Endpoints

Services: 将运行在一组pods上的应用程序公开为网络服务的抽象方法。

1. 为什么需要Services

​ 若我们使用 Deployment 来管理 Pod ， 那么该Deployment可以动态的 创建 和 销毁 Pod , 而每个Pod都有自己的IP地址 ， 此时就需要一种服务， 能够帮助我们找到集群对应的 Pod 机器的IP地址,这就需要services。

2. 什么是endpoints

​ endpoints 资源就是暴露一个服务的IP地址和端口的列表, 只要服务中的 Pod 集合发生改变，endpoints 就会被更新。

​ 默认在创建Servers 时，若编写了标签选择器selector，则会自动创建 endpoints ,否则不会创建。

3. 案例

3.1 创建 deployment

若有deployment资源(该镜像是简单的dns服务器)，定义为如下

apiVersion: apps/v1
kind: Deployment

metadata:
  name: sampledns
  namespace: default

spec:
  replicas: 3

  selector:
    matchLabels:
      app: dns

  template:
    metadata:
      labels:
        app: dns

    spec:
      containers: 
        - name: sampledns2
          image: docker.io/2859413527/sample-dns

          env:
          - name: redisHost
            value: '192.168.1.5:6379'

          - name: proxyDNSHost
            value: '8.8.8.8'

          - name: dnsOnlineStatus
            value: 'false'
        
          ports:
          - name: tcpdns
            containerPort: 53
            protocol: TCP
        
          - name: udpdns
            containerPort: 53
            protocol: UDP

          - name: web
            containerPort: 5001
            protocol: TCP

我们查看pod的状态信息

# kubectl get pods -o wide  -l app=dns
NAME                         READY   STATUS    RESTARTS   AGE     IP             NODE    NOMINATED NODE   READINESS GATES
sampledns-77c56c9fb9-6twzl   1/1     Running   0          2m22s   10.244.2.205   node2   <none>           <none>
sampledns-77c56c9fb9-g97jw   1/1     Running   0          106s    10.244.1.195   node1   <none>           <none>
sampledns-77c56c9fb9-l7j5x   1/1     Running   0          2m5s    10.244.1.194   node1   <none>           <none>
# 

3.2 创建 services

创建带 endpoints 的 services

apiVersion: v1
kind: Service

metadata:
  name: dnsservices
  namespace: default

spec:
  selector:
    app: dns

  type: NodePort

  ports:
  - name: tcpdns
    port: 53
    targetPort: 53
    protocol: TCP

  - name: udpdns
    port: 53
    targetPort: 53
    nodePort: 53
    protocol: UDP

    
  - name: web
    port: 5001
    targetPort: 5001
    nodePort: 5001
    protocol: TCP

创建好后，查看 services

# kubectl get services dnsservices -o wide
NAME          TYPE       CLUSTER-IP     EXTERNAL-IP   PORT(S)                             AGE   SELECTOR
dnsservices   NodePort   10.1.146.200   <none>        53:53/TCP,53:53/UDP,5001:5001/TCP   8s    app=dns
# 

3.3 观察endpoints

再次查看 endpoints

# kubectl describe endpoints dnsservices        
Name:         dnsservices
Namespace:    default
Labels:       <none>
Annotations:  endpoints.kubernetes.io/last-change-trigger-time: 2022-01-28T10:20:13Z
Subsets:
  Addresses:          10.244.1.194,10.244.1.195,10.244.2.205
  NotReadyAddresses:  <none>
  Ports:
    Name    Port  Protocol
    ----    ----  --------
    tcpdns  53    TCP
    udpdns  53    UDP
    web     5001  TCP

Events:  <none>
#

3.4 重建pods并且查看endpoints

# kubectl delete pod sampledns-77c56c9fb9-6twzl
pod "sampledns-77c56c9fb9-6twzl" deleted
# 

由于使用的 deployment 来定义的 Pod ， 所以，该 Pod 会被重建，如下

# kubectl get pods -o wide
NAME                         READY   STATUS    RESTARTS   AGE     IP             NODE    NOMINATED NODE   READINESS GATES
sampledns-77c56c9fb9-g97jw   1/1     Running   0          8m39s   10.244.1.195   node1   <none>           <none>
sampledns-77c56c9fb9-gcqcd   1/1     Running   0          28s     10.244.2.206   node2   <none>           <none>
sampledns-77c56c9fb9-l7j5x   1/1     Running   0          8m58s   10.244.1.194   node1   <none>           <none>
#

再次查看 endpoints

# kubectl describe endpoints dnsservices
Name:         dnsservices
Namespace:    default
Labels:       <none>
Annotations:  endpoints.kubernetes.io/last-change-trigger-time: 2022-01-28T10:24:31Z
Subsets:
  Addresses:          10.244.1.194,10.244.1.195,10.244.2.206
  NotReadyAddresses:  <none>
  Ports:
    Name    Port  Protocol
    ----    ----  --------
    tcpdns  53    TCP
    udpdns  53    UDP
    web     5001  TCP

Events:  <none>
# 

可以看到 endpoints Address 已经被更新掉了