写在前面:如有问题,以你为准,
目前24年应届生,各位大佬轻喷,部分资料与图片来自网络
内容较长,页面右上角目录方便跳转
Service 介绍 架构
- 在kubernetes中,Pod是应用程序的载体,我们可以通过Pod的IP来访问应用程序,但是Pod的IP地址不是固定的,这就意味着不方便直接采用Pod的IP对服务进行访问。
- Pod IP 会随着Pod的重建产生变化
- Pod IP 仅仅是集群内可见的虚拟IP,外部无法访问
- 为了解决这个问题,kubernetes提供了Service资源,Service会对提供同一个服务的多个Pod进行聚合,并且提供一个统一的入口地址,通过访问Service的入口地址就能访问到后面的Pod服务。
- 架构如图
注:通过labels标签选择器来将service与后端Pod进行绑定
- Service在很多情况下只是一个概念,真正起作用的其实是kube-proxy服务进程,每个Node节点上都运行了一个kube-proxy的服务进程。当创建Service的时候会通过API Server向etcd写入创建的Service的信息,而kube-proxy会基于监听的机制发现这种Service的变化,然后它会将最新的Service信息转换为对应的。
- 访问规则即流量负载分发 使用ipvs实现,其也是实现LVS的核心
# 10.97.97.97:80 是service提供的访问入口(VIP)# 当访问这个入口的时候,可以发现后面有三个pod的服务在等待调用,# kube-proxy会基于rr(轮询)的策略,将请求分发到其中一个pod上去# 这个规则会同时在集群内的所有节点上都生成,所以在任何一个节点上访问都可以。[root@k8s-node1 ~]# ipvsadm -LnIP Virtual Server version 1.2.1 (size=4096)Prot LocalAddress:Port Scheduler Flags-> RemoteAddress:Port Forward Weight ActiveConn InActConnTCP 10.97.97.97:80 rr-> 10.244.1.39:80 Masq 1 0 0-> 10.244.1.40:80 Masq 1 0 0-> 10.244.2.33:80 Masq 1 0 0
kube-proxy(service 工作原理)
是逐步进化的,从userspace 到 iptables 再到 ipvs
-
- userspace 模式:
- userspace模式下,kube-proxy会为每一个Service创建一个监听端口,发向Cluster IP的请求被iptables规则重定向到kube-proxy监听的端口上,kube-proxy根据LB算法(负载均衡算法)选择一个提供服务的Pod并和其建立连接,以便将请求转发到Pod上。
- 该模式下,kube-proxy充当了一个四层负载均衡器的角色。由于kube-proxy运行在userspace中,在进行转发处理的时候会增加内核和用户空间之间的数据拷贝,虽然比较稳定,但是效率非常低下。
- userspace 模式:
-
- iptables 模式:
- iptables模式下,kube-proxy为Service后端的每个Pod创建对应的iptables规则,直接将发向Cluster IP的请求重定向到一个Pod的IP上。
- 该模式下kube-proxy不承担四层负载均衡器的角色,只负责创建iptables规则。该模式的优点在于较userspace模式效率更高,但是不能提供灵活的LB策略,当后端Pod不可用的时候无法进行重试。
- iptables 模式:
- 开启ipvs(必须安装ipvs内核模块,否则会降级为iptables)
ipvs
开启 查看 ipvs
# 更改模式 vim 下输入/mode[root@master ~]# kubectl edit cm kube-proxy -n kube-systemkind: KubeProxyConfigurationmetricsBindAddress: ""mode: "ipvs"nodePortAddresses: null# 查看与重新创建每个node 对应的 kube-proxy(三个node) [root@master ~]# kubectl get pod -l k8s-app=kube-proxy -n kube-systemNAME READY STATUS RESTARTS AGEkube-proxy-b8mzd 1/1 Running 0 13dkube-proxy-g6q8z 1/1 Running 4 (23h ago) 13dkube-proxy-trzb9 1/1 Running 3 (23h ago) 12d[root@master ~]# kubectl delete pod -l k8s-app=kube-proxy -n kube-systempod "kube-proxy-b8mzd" deletedpod "kube-proxy-g6q8z" deletedpod "kube-proxy-trzb9" deleted[root@master ~]# kubectl get pod -l k8s-app=kube-proxy -n kube-systemNAME READY STATUS RESTARTS AGEkube-proxy-cppq6 1/1 Running 0 8skube-proxy-cw8xn 1/1 Running 0 9skube-proxy-nlnpl 1/1 Running 0 9s
[root@master ~]# ipvsadm -LnIP Virtual Server version 1.2.1 (size=4096)Prot LocalAddress:Port Scheduler Flags-> RemoteAddress:Port Forward Weight ActiveConn InActConnTCP 10.96.0.1:443 rr-> 192.168.100.53:6443 Masq 1 1 0 TCP 10.96.0.10:53 rr-> 10.244.219.65:53 Masq 1 0 0 -> 10.244.219.69:53 Masq 1 0 0 TCP 10.96.0.10:9153 rr-> 10.244.219.65:9153 Masq 1 0 0 -> 10.244.219.69:9153 Masq 1 0 0 TCP 10.98.145.184:9094 rr-> 10.244.219.67:9094 Masq 1 0 0 TCP 10.98.165.172:443 rr-> 192.168.100.51:4443 Masq 1 0 0 -> 192.168.100.52:4443 Masq 1 0 0 TCP 10.109.241.243:5473 rr-> 192.168.100.51:5473 Masq 1 0 0 -> 192.168.100.52:5473 Masq 1 0 0 TCP 10.111.111.114:443 rr-> 10.244.219.66:5443 Masq 1 0 0 -> 10.244.219.68:5443 Masq 1 0 0 UDP 10.96.0.10:53 rr-> 10.244.219.65:53 Masq 1 0 0 -> 10.244.219.69:53 Masq 1 0 0
负载分发策略
- 对Service的访问被分发到了后端的Pod上去,目前kubernetes提供了两种负载分发策略:
- 如果不定义,默认使用kube-proxy的策略,比如随机、轮询等。
- 基于客户端地址的会话保持模式,即来自同一个客户端发起的所有请求都会转发到固定的一个Pod上,这对于传统基于Session的认证项目来说很友好,此模式可以在spec中添加sessionAffinity: ClientIP选项。
- 查看ipvs的映射规则,rr表示轮询:
[root@master k8s]# ipvsadm -LnIP Virtual Server version 1.2.1 (size=4096)Prot LocalAddress:Port Scheduler Flags-> RemoteAddress:Port Forward Weight ActiveConn InActConnTCP 10.97.97.97:80 rr # vip:端口 rr(轮询策略)-> 10.244.104.10:80 Masq 1 0 1 -> 10.244.166.129:80 Masq 1 0 1 -> 10.244.166.191:80 Masq 1 0 2 [root@master k8s]# curl 10.97.97.97IP: 10.244.166.191 pod-1 node1[root@master k8s]# curl 10.97.97.97IP: 10.244.166.129 pod-2 node1[root@master k8s]# curl 10.97.97.97IP: 10.244.104.10 pod-3 node2
Endpoint
- Endpoint是kubernetes中的一个资源对象,存储在etcd中,用来记录一个service对应的所有Pod的访问地址,它是根据service配置文件中的selector描述产生的。
- 一个service由一组Pod组成,这些Pod通过Endpoints暴露出来,Endpoints是实现实际服务的端点集合。换言之,service和Pod之间的联系是通过Endpoints实现的,也就是通过labels 进行 selector 实现service 与Pod之间的绑定
[root@master k8s]# kubectl describe service -n study service-clusteripName: service-clusteripNamespace: studyLabels: <none>Annotations: <none>Selector: app=nginx-podType: ClusterIPIP Family Policy: SingleStackIP Families: IPv4IP: 10.97.97.97IPs: 10.97.97.97Port: <unset> 80/TCPTargetPort: 80/TCPEndpoints: 10.244.1.39:80,10.244.1.40:80 #连接的podSession Affinity: NoneEvents: <none>
[root@master k8s]# kubectl get endpoints -n study -o wideNAME ENDPOINTS AGEservice-clusterip 10.244.1.39:80,10.244.1.40:80 5m47s
域名
- 当我们创建一个 Service 的时候,Kubernetes 会创建一个相应的 DNS 条目。
- 该条目的形式是
<namespace-name>.svc.cluster.local,这意味着如果容器中只使用
<服务名称>,它将被解析到本地名称空间的服务器。这对于跨多个名字空间(如开发、测试和生产) 使用相同的配置非常有用。如果你希望跨名字空间访问,则需要使用完全限定域名(FQDN)
自带 service
[root@master cks]# kubectl get svcNAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGEkubernetes ClusterIP 10.96.0.1 <none> 443/TCP 278d
kubernetes 用于给 pod 访问 kube apiserver
Service 类型(实操)
命令行实操(expose)
集群内部访问
# 创建 deploy[root@master k8s]# kubectl create deployment nginx -n default --image=nginx:1.8 --replicas=2deployment.apps/nginx created# 暴露端口,其实就是创建 service[root@master k8s]# kubectl expose deploy nginx --name=nginx --type=ClusterIP --port=80 --target-port=80 -n defaultservice/nginx exposed[root@master k8s]# kubectl get svc -n defaultNAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGEkubernetes ClusterIP 10.96.0.1 <none> 443/TCP 2dnginx ClusterIP 10.105.198.94 <none> 80/TCP 22s[root@master k8s]# curl 10.105.198.94 #集群内部地址[root@master k8s]# kubectl delete svc nginx -n defaultservice "nginx" deleted
集群外部访问
[root@master k8s]# kubectl expose deploy nginx --name=nginx --type=NodePort --port=80 --target-port=80 -n defaultservice/nginx exposed[root@master k8s]# kubectl get svc -n defaultNAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGEkubernetes ClusterIP 10.96.0.1 <none> 443/TCP 2dnginx NodePort 10.105.143.21 <none> 80:31627/TCP 7s# 80:31627/TCP 中 31627 是master的ip地址端口,# 会将master ip:31627 请求转发到 10.105.143.21[root@master k8s]# curl 192.168.100.53:31627[root@master k8s]# kubectl delete svc nginx -n defaultservice "nginx" deleted
yaml 整体解析
spec.type:
ClusterIP:默认值,它是kubernetes系统自动分配的虚拟IP,只能在集群内部访问。
NodePort:将Service通过指定的Node(集群节点上)上的端口暴露给外部,通过此方法,就可以在集群外部访问服务。
LoadBalancer:使用外接负载均衡器完成到服务的负载分发,注意此模式需要外部云环境的支持。
ExternalName:把集群外部的服务引入集群内部,直接使用,可以实现pod访问外部域名地址
sessionAffinity:ClientIP 同一个ip都全部请求去同一个Pod上(会话保持模式)
NodePort 的缺点是会占用很多集群机器的端口,那么当集群服务变多的时候,这个缺点就愈发明显。
LoadBalancer 的缺点是每个Service都需要一个LB,浪费,麻烦,并且需要kubernetes之外的设备的支持
apiVersion: v1 # 版本kind: Service # 类型metadata: # 元数据name: # 资源名称namespace: # 命名空间spec:selector: # 标签选择器,用于确定当前Service代理那些Podapp: nginxtype: NodePort # Service的类型,指定Service的访问方式clusterIP: # 虚拟服务的IP地址sessionAffinity: # session亲和性,支持ClientIP、None两个选项,默认值为None(不开启)ports: # 端口信息- port: 8080 # Service 开放端口protocol: TCP # 协议targetPort : # 转发到 Pod 的端口nodePort: # 主机 开放端口
环境准备
创建deployment控制器,注意labels为 app=nginx-pod
---apiVersion: apps/v1kind: Namespacemetadata:name: study---apiVersion: apps/v1kind: Deploymentmetadata:name: service-environment-deploymentnamespace: studyspec:replicas: 3selector:matchLabels:app: nginx-podtemplate:metadata:labels:app: nginx-podspec:containers:- name: nginximage: nginx:1.17.1ports:- containerPort: 80 # 容器开放端口
[root@master k8s]# kubectl apply -f controller.yamlnamespace/study createddeployment.apps/service-environment-deployment created[root@master k8s]# kubectl get deploy -n studyNAME READY UP-TO-DATE AVAILABLE AGEservice-environment-deployment 3/3 3 3 21s[root@master k8s]# kubectl get pod -n study -o wide --show-labelsNAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES LABELSservice-environment-deployment-6bb9d9f778-5k2x6 1/1 Running 0 57s 10.244.166.191 node1 <none> <none> app=nginx-pod,pod-template-hash=6bb9d9f778service-environment-deployment-6bb9d9f778-8bt7j 1/1 Running 0 57s 10.244.166.129 node1 <none> <none> app=nginx-pod,pod-template-hash=6bb9d9f778service-environment-deployment-6bb9d9f778-8hcm2 1/1 Running 0 57s 10.244.104.10 node2 <none> <none> app=nginx-pod,pod-template-hash=6bb9d9f778
[root@master k8s]# kubectl exec -it -n study service-environment-deployment-6bb9d9f778-5k2x6 -c nginx /bin/shecho " IP: 10.244.166.191 pod-1 node1" > /usr/share/nginx/html/index.html[root@master k8s]# kubectl exec -it -n study service-environment-deployment-6bb9d9f778-8bt7j -c nginx /bin/shecho " IP: 10.244.166.129 pod-2 node1" > /usr/share/nginx/html/index.html
[root@master k8s]# kubectl exec -it -n study service-environment-deployment-6bb9d9f778-8hcm2 -c nginx /bin/shecho " IP: 10.244.104.10 pod-3 node2" > /usr/share/nginx/html/index.html
[root@master k8s]# curl 10.244.166.191IP: 10.244.166.191 pod-1 node1[root@master k8s]# curl 10.244.166.129IP: 10.244.166.129 pod-2 node1[root@master k8s]# curl 10.244.104.10IP: 10.244.104.10 pod-3 node2
ClusterIP 类型
apiVersion: v1kind: Servicemetadata:name: service-clusteripnamespace: studyspec:selector:app: nginx-pod # Pod的标签clusterIP: 10.97.97.97 # service的IP地址,如果不写,默认会生成一个type: ClusterIPports:- port: 80 # Service的端口protocol: TCP # 协议targetPort: 80 # Pod的端口
[root@master k8s]# kubectl apply -f service.yamlservice/service-clusterip created[root@master k8s]# kubectl get service -n studyNAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGEservice-clusterip ClusterIP 10.97.97.97 <none> 80/TCP 10s[root@master k8s]# kubectl describe service -n study service-clusteripName: service-clusteripNamespace: studyLabels: <none>Annotations: <none>Selector: app=nginx-podType: ClusterIPIP Family Policy: SingleStackIP Families: IPv4IP: 10.97.97.97IPs: 10.97.97.97Port: <unset> 80/TCPTargetPort: 80/TCPEndpoints: 10.244.104.10:80,10.244.166.129:80,10.244.166.191:80 #连接的podSession Affinity: NoneEvents: <none>[root@master k8s]# ipvsadm -LnIP Virtual Server version 1.2.1 (size=4096)Prot LocalAddress:Port Scheduler Flags-> RemoteAddress:Port Forward Weight ActiveConn InActConn
TCP 10.97.97.97:80 rr #vip:端口 rr(轮询策略)-> 10.244.104.10:80 Masq 1 0 1 -> 10.244.166.129:80 Masq 1 0 1 -> 10.244.166.191:80 Masq 1 0 2
# 由下面可以看到是轮询策略[root@master k8s]# curl 10.97.97.97IP: 10.244.166.191 pod-1 node1[root@master k8s]# curl 10.97.97.97IP: 10.244.166.129 pod-2 node1[root@master k8s]# curl 10.97.97.97IP: 10.244.104.10 pod-3 node2
[root@master k8s]# kubectl delete -f service.yamlservice "service-clusterip" deleted
HeadLiness 类型
在某些场景中,开发人员可能不想使用Service提供的负载均衡功能,而希望自己来控制负载均衡策略,针对这种情况,kubernetes提供了HeadLinesss Service,这类Service不会分配Cluster IP,如果想要访问Service,只能通过Service的域名进行访问
一般用于实现 StatefulSet(常用来部署RabbitMQ集群、Zookeeper集群、MySQL集群、Eureka集群等)
apiVersion: v1kind: Servicemetadata:name: service-headlinessnamespace: studyspec:selector:app: nginx-podclusterIP: None # 将clusterIP设置为None,即可创建headliness Servicetype: ClusterIPports:- port: 80 # Service的端口targetPort: 80 # Pod的端口
[root@master k8s]# kubectl apply -f service.yamlservice/service-headliness created[root@master k8s]# kubectl get svc -n studyNAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGEservice-headliness ClusterIP None <none> 80/TCP 8s[root@master k8s]# kubectl describe svc service-headliness -n studyName: service-headlinessNamespace: studyLabels: <none>Annotations: <none>Selector: app=nginx-podType: ClusterIPIP Family Policy: SingleStackIP Families: IPv4IP: NoneIPs: NonePort: <unset> 80/TCPTargetPort: 80/TCPEndpoints: 10.244.104.10:80,10.244.166.129:80,10.244.166.191:80Session Affinity: NoneEvents: <none>
[root@master k8s]# kubectl get pod -n studyNAME READY STATUS RESTARTS AGEservice-environment-deployment-6bb9d9f778-5k2x6 1/1 Running 0 45mservice-environment-deployment-6bb9d9f778-8bt7j 1/1 Running 0 45mservice-environment-deployment-6bb9d9f778-8hcm2 1/1 Running 0 45m[root@master k8s]# kubectl exec -it -n study service-environment-deployment-6bb9d9f778-5k2x6 -c nginx /bin/shkubectl exec [POD] [COMMAND] is DEPRECATED and will be removed in a future version. Use kubectl exec [POD] -- [COMMAND] instead.# cat /etc/resolv.confnameserver 10.96.0.10search study.svc.cluster.local svc.cluster.local cluster.localoptions ndots:5
[root@master k8s]# dig @10.96.0.10 service-headliness.study.svc.cluster.local; <<>> DiG 9.11.4-P2-RedHat-9.11.4-26.P2.el8 <<>> @10.96.0.10 service-clusterip.study.svc.cluster.local; (1 server found);; global options: +cmd;; Got answer:;; WARNING: .local is reserved for Multicast DNS;; You are currently testing what happens when an mDNS query is leaked to DNS;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 6641;; flags: qr aa rd; QUERY: 1, ANSWER: 3, AUTHORITY: 0, ADDITIONAL: 1;; WARNING: recursion requested but not available
;; OPT PSEUDOSECTION:; EDNS: version: 0, flags:; udp: 4096; COOKIE: 2ea4525b90a0ac1b (echoed);; QUESTION SECTION:;service-headliness.study.svc.cluster.local. IN A;; ANSWER SECTION:service-headliness.study.svc.cluster.local. 30 IN A 10.244.104.10service-headliness.study.svc.cluster.local. 30 IN A 10.244.166.191service-headliness.study.svc.cluster.local. 30 IN A 10.244.166.129;; Query time: 18 msec;; SERVER: 10.96.0.10#53(10.96.0.10);; WHEN: Wed Feb 15 09:43:58 EST 2023;; MSG SIZE rcvd: 253
NodePort 类型
NodePort的工作原理就是将Service的端口映射到Node的一个端口上,然后就可以通过
apiVersion: v1kind: Servicemetadata:name: service-nodeportnamespace: studyspec:selector:app: nginx-podtype: NodePort # Service类型为NodePort,实现集群外部访问ports:- port: 80 # Service的端口targetPort: 80 # Pod的端口nodePort: 30002# 指定绑定的node的端口(默认取值范围是30000~32767),如果不指定,会默认分配
[root@master k8s]# kubectl apply -f service.yamlservice/service-nodeport created[root@master k8s]# kubectl get svc -n studyNAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGEservice-nodeport NodePort 10.100.46.87 <none> 80:30002/TCP 48s[root@master k8s]# ifconfig | grep inetinet6 fe80::ecee:eeff:feee:eeee prefixlen 64 scopeid 0x20<link>inet6 fe80::ecee:eeff:feee:eeee prefixlen 64 scopeid 0x20<link>inet6 fe80::ecee:eeff:feee:eeee prefixlen 64 scopeid 0x20<link>inet6 fe80::ecee:eeff:feee:eeee prefixlen 64 scopeid 0x20<link>inet6 fe80::ecee:eeff:feee:eeee prefixlen 64 scopeid 0x20<link>inet 192.168.100.53 netmask 255.255.255.0 broadcast 192.168.100.255inet6 fe80::5523:b3a4:8bc9:b40f prefixlen 64 scopeid 0x20<link>inet6 fe80::522c:e0a0:2c74:37c6 prefixlen 64 scopeid 0x20<link>inet6 fe80::dec0:9c00:3416:2561 prefixlen 64 scopeid 0x20<link>inet 127.0.0.1 netmask 255.0.0.0inet6 ::1 prefixlen 128 scopeid 0x10<host>[root@master k8s]# ipvsadm -LnIP Virtual Server version 1.2.1 (size=4096)Prot LocalAddress:Port Scheduler Flags-> RemoteAddress:Port Forward Weight ActiveConn InActConnTCP 192.168.100.53:30002 rr-> 10.244.104.10:80 Masq 1 0 0 -> 10.244.166.129:80 Masq 1 0 0 -> 10.244.166.191:80 Masq 1 0 0 # 另开一台同一网络的linux或windows进行访问,地址为 masterip 192.168.100.53[root@ip-15 ~]# curl 192.168.100.53:30002IP: 10.244.166.191 pod-1 node1[root@ip-15 ~]# curl 192.168.100.53:30002IP: 10.244.166.129 pod-2 node1[root@ip-15 ~]# curl 192.168.100.53:30002IP: 10.244.104.10 pod-3 node2
扩展,访问集群中你的节点也是可以的如 node1 node2
[root@node1 ~]# ipvsadm -LnIP Virtual Server version 1.2.1 (size=4096)Prot LocalAddress:Port Scheduler Flags-> RemoteAddress:Port Forward Weight ActiveConn InActConnTCP 192.168.100.51:30002 rr-> 10.244.104.10:80 Masq 1 0 0 -> 10.244.166.129:80 Masq 1 0 0 -> 10.244.166.191:80 Masq 1 0 0 [root@ip-15 ~]# curl 192.168.100.51:30002IP: 10.244.166.191 pod-1 node1[root@ip-15 ~]# curl 192.168.100.51:30002IP: 10.244.166.129 pod-2 node1[root@ip-15 ~]# curl 192.168.100.51:30002IP: 10.244.104.10 pod-3 node2
LoadBalancer 类型
LoadBalancer和NodePort很相似,目的都是向外部暴露一个端口,区别在于LoadBalancer会在集群的外部再来做一个负载均衡设备,而这个设备需要外部环境的支持,外部服务发送到这个设备上的请求,会被设备负载之后转发到集群中
ExternalName 类型
ExternalName类型的Service用于引入集群外部的服务,它通过externalName属性指定一个服务的地址,然后在集群内部访问此Service就可以访问到外部的服务了,访问外部域名地址
apiVersion: v1kind: Servicemetadata:name: service-externalnamenamespace: studyspec:type: ExternalName # Service类型为ExternalNameexternalName: www.baidu.com # 改成IP地址也可以
[root@master k8s]# kubectl get svc -n studyNAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGEservice-externalname ExternalName <none> www.baidu.com <none> 55s[root@master k8s]# dig @10.96.0.10 service-externalname.study.svc.cluster.local; <<>> DiG 9.11.4-P2-RedHat-9.11.4-26.P2.el8 <<>> @10.96.0.10 service-externalname.study.svc.cluster.local; (1 server found);; global options: +cmd;; Got answer:;; WARNING: .local is reserved for Multicast DNS;; You are currently testing what happens when an mDNS query is leaked to DNS;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 2116;; flags: qr aa rd; QUERY: 1, ANSWER: 4, AUTHORITY: 0, ADDITIONAL: 1;; WARNING: recursion requested but not available;; OPT PSEUDOSECTION:; EDNS: version: 0, flags:; udp: 4096; COOKIE: 144210b36fc272dd (echoed);; QUESTION SECTION:;service-externalname.study.svc.cluster.local. IN A;; ANSWER SECTION:service-externalname.study.svc.cluster.local. 30 IN CNAME www.baidu.com.www.baidu.com.30INCNAMEwww.a.shifen.com.www.a.shifen.com.30INA14.215.177.38www.a.shifen.com.30INA14.215.177.39;; Query time: 14 msec;; SERVER: 10.96.0.10#53(10.96.0.10);; WHEN: Wed Feb 15 09:57:33 EST 2023;; MSG SIZE rcvd: 263
实现会话保持(持久连接)
其实就是这个参数 sessionAffinity: ClientIP
apiVersion: v1kind: Servicemetadata:name: service-clusteripnamespace: studyspec:sessionAffinity: ClientIP # 实现保持会话,如果不开启则填 Noneselector:app: nginx-pod # Pod的标签clusterIP: 10.97.97.97 # service的IP地址,如果不写,默认会生成一个type: ClusterIPports:- port: 80 # Service的端口protocol: TCP # 协议targetPort: 80 # Pod的端口
[root@master k8s]# kubectl get svc -n studyNAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGEservice-clusterip ClusterIP 10.97.97.97 <none> 80/TCP 14s[root@master k8s]# kubectl describe svc -n studyName: service-clusteripNamespace: studyLabels: <none>Annotations: <none>Selector: app=nginx-podType: ClusterIPIP Family Policy: SingleStackIP Families: IPv4IP: 10.97.97.97IPs: 10.97.97.97Port: <unset> 80/TCPTargetPort: 80/TCPEndpoints: 10.244.104.10:80,10.244.166.129:80,10.244.166.191:80Session Affinity: ClientIP # 设置为这个就是保持会话Events: <none>[root@master k8s]# ipvsadm -LnIP Virtual Server version 1.2.1 (size=4096)Prot LocalAddress:Port Scheduler Flags-> RemoteAddress:Port Forward Weight ActiveConn InActConnTCP 10.97.97.97:80 rr persistent(保持会话的标志) 10800(保持会话的时间,单位秒)-> 10.244.104.10:80 Masq 1 0 0 -> 10.244.166.129:80 Masq 1 0 0 -> 10.244.166.191:80 Masq 1 0 0