kubernetes 长连接负载不均衡 kubernetes grpc 负载均衡

grpc一般都是内部服务调用，在k8s集群中进行服务发现和负载均衡的方式我所知道的有三种：
一、直接service nodepod方式部署，缺点就是会占用宿主机port,服务多起来，团队大起来的时候，port端口使用混乱，一不小心就冲突，服务无法访问查都查不到原因
二、使用ngxin-ingress进行服务发现和负载均衡，缺点必须配置证书，只支持https访问
三、traefik-ingress进行服务发现和负载均衡（强烈推荐，生产环境已用）
本篇就讲一讲如何通过traefik-ingress 进行grpc负载均衡。

安装traefik-ingress环境
traefik-rbac.yaml：
 

## ServiceAccount
apiVersion: v1
kind: ServiceAccount
metadata:
  namespace: kube-system
  name: traefik-ingress-controller
---
## ClusterRole
kind: ClusterRole
apiVersion: /v1beta1
metadata:
  name: traefik-ingress-controller
rules:
  - apiGroups: [""]
    resources: ["services","endpoints","secrets"]
    verbs: ["get","list","watch"]
  - apiGroups: ["extensions",""]
    resources: ["ingresses","ingressclasses"]
    verbs: ["get","list","watch"]
  - apiGroups: ["extensions"]
    resources: ["ingresses/status"]
    verbs: ["update"]
  - apiGroups: [""]
    resources: ["middlewares"]
    verbs: ["get","list","watch"]
  - apiGroups: [""]
    resources: ["ingressroutes"]
    verbs: ["get","list","watch"]
  - apiGroups: [""]
    resources: ["ingressroutetcps"]
    verbs: ["get","list","watch"]
  - apiGroups: [""]
    resources: ["tlsoptions"]
    verbs: ["get","list","watch"]
  - apiGroups: [""]
    resources: ["traefikservices"]
    verbs: ["get","list","watch"]
---
## ClusterRoleBinding
kind: ClusterRoleBinding
apiVersion: /v1beta1
metadata:
  name: traefik-ingress-controller
roleRef:
  apiGroup: 
  kind: ClusterRole
  name: traefik-ingress-controller
subjects:
  - kind: ServiceAccount
    name: traefik-ingress-controller
    namespace: kube-system

traefik-config.yaml：
 

kind: ConfigMap
apiVersion: v1
metadata:
  name: traefik-config
data:
  traefik.yaml: |-
    ping:                       ## 启用 Ping
      entryPoint: "web"
    serversTransport:
      insecureSkipVerify: true  ## Traefik 忽略验证代理服务的 TLS 证书
    api:
      insecure: true            ## 允许 HTTP 方式访问 API
      dashboard: true           ## 启用 Dashboard
      debug: false              ## 启用 Debug 调试模式
    metrics:
      prometheus: ""            ## 配置 Prometheus 监控指标数据，并使用默认配置
    entryPoints:
      web:
        address: ":18080"          ## 配置 80 端口，并设置入口名称为 web
      websecure:
        address: ":18443"         ## 配置 443 端口，并设置入口名称为 websecure
      traefik:
        address: ":28080"         ## dashboard端口
    providers:
      kubernetesIngress: ""     ## 启动 Kubernetes Ingress 方式来配置路由规则
    log:
      level: INFO              ## 设置调试日志级别
      ##format: json              ## 设置调试日志格式
    accessLog:
      ##format: json              ## 设置访问调试日志格式
      bufferingSize: 100          ## 设置访问日志缓存行数

traefik-deploy.yaml:
 

apiVersion: apps/v1
kind: DaemonSet
metadata:
  name: traefik-ingress-controller
  labels:
    app: traefik
spec:
  selector:
    matchLabels:
      app: traefik
  template:
    metadata:
      name: traefik
      labels:
        app: traefik
    spec:
      dnsPolicy: ClusterFirstWithHostNet
      hostNetwork: true
      serviceAccountName: traefik-ingress-controller
      terminationGracePeriodSeconds: 1
      containers:
        - image: traefik:v2.3.7
          name: traefik-ingress-lb
          ports:
            - name: web
              containerPort: 18080
              hostPort: 18080
              protocol: TCP
            - name: websecure
              containerPort: 18443
              hostPort: 18443
              protocol: TCP
            - name: admin
              containerPort: 28080  ## Traefik Dashboard 端口
              hostPort: 28080
              protocol: TCP
          securityContext:
            allowPrivilegeEscalation: true
            capabilities:
              add:
              - NET_BIND_SERVICE
              drop:
              - ALL
            privileged: true
            readOnlyRootFilesystem: false
            runAsNonRoot: false
            runAsUser: 0
          args:
            - --configfile=/config/traefik.yaml
          volumeMounts:
            - mountPath: "/config"
              name: "config"
          readinessProbe:
            httpGet:
              path: /ping
              port: 18080
            initialDelaySeconds: 10
            periodSeconds: 5
      volumes:
        - name: config
          configMap:
            name: traefik-config 
      tolerations:              ## 设置容忍所有污点，防止节点被设置污点
        - operator: "Exists"
      nodeSelector:
        /worker: "true"
        /os: "linux"

一共三个部署文件：依次导入到集群中：traefik-rbac.yaml，traefik-config.yaml，traefik-deploy.yaml，
导入方式：rancher管理端或者kubectl (后面介绍如何用kubectl在集群中执行安装)

完成安装后可以查看traefik管理端：http://任何一个workerIP:28080/

准备测试程序：

打开vs2022新建项目：

 

 依旧添加一个健康检查控制器

 

 

 

安装GRPC依赖 Grpc.core和Grpc.Tools

添加google.protobu依赖

添加Grpc.AspNetCore依赖

 

 

syntax = "proto3";

package K8sGrpcDemo.Grpc;
import "google/protobuf/empty.proto";

service GrpcDemoService {
  rpc GetLocalIP (google.protobuf.Empty) returns (GetLocalIPResponse);
}

message GetLocalIPResponse {
	string localIP=1;//当前服务ip
}

 

 生成操作选择Protobuf compiler

 实现grpc服务：返回当前服务器ip

 

 Dockerfile

 .rancher-pipeline.yml

stages:
- name: build
  steps:
  - publishImageConfig:
      dockerfilePath: ./K8sGrpcDemo/Dockerfile
      buildContext: .
      tag: 192.168.21.4:8081/test/k8sgrpcdemo:${CICD_GIT_BRANCH}-${CICD_GIT_COMMIT}
      pushRemote: true
      registry: 192.168.21.4:8081
    env:
      PLUGIN_DEBUG: "true"
      PLUGIN_INSECURE: "true"
- name: deploy
  steps:
  - applyYamlConfig:
      path: ./deployment.yaml
timeout: 60
notification: {}

 deployment.yaml:（和上一篇nginx-ingress不一样的地方我注释了下）

apiVersion: v1
kind: Service
metadata:
  name: k8sgrpcdemo  
  namespace: default
  labels:
    app: k8sgrpcdemo
    service: k8sgrpcdemo
  annotations:
    /service.serversscheme: h2c    #ingress负载均衡
spec:
  ports:
  - port: 80
    name: http
  - port: 18080     #ingress代理grpc的端口
    targetPort: 8080  #容器暴露出来grpc端口
    protocol: TCP
    name: grpc
  selector:
    app: k8sgrpcdemo
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: k8sgrpcdemo
  namespace: default
  labels:
    app: k8sgrpcdemo
    version: v1
spec:
  replicas: 3   #三个副本
  selector:
    matchLabels:
      app: k8sgrpcdemo
      version: v1
  template:
    metadata:
      labels:
        app: k8sgrpcdemo
        version: v1
    spec:
      containers:
      - name: k8sgrpcdemo
        image: 192.168.21.4:8081/test/k8sgrpcdemo:${CICD_GIT_BRANCH}-${CICD_GIT_COMMIT}
        readinessProbe:
          httpGet:
            path: /health/status
            port: 80
          initialDelaySeconds: 10
          periodSeconds: 5
        ports:
        - containerPort: 80
      imagePullSecrets:
        - name: mydockerhub

健康检查走的是http 80端口，grpc对外是8080端口  所以应用需要同时监听两个端口。需要修改代码 Program.cs：

到此，项目已经准备完成。提交代码。然后到rancher流水线进行部署 （我已经在私有仓库创建好项目）

 

， 

 

准备ingress文件：k8sgrpcdemo_ingress.yaml
 

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: k8sgrpcdemo
  namespace: default
spec:
  rules:
  - host: www.k8sgrpcdemo.cn
    http:
      paths:
      - path: /                #所有www.k8sgrpcdemo.cn:80过来的http请求都会打到服务名为k8sgrpcdemo 的80上 其实这里走的是ingress-nginx
        backend:
          serviceName: k8sgrpcdemo
          servicePort: 80
      - path: /K8sGrpcDemoProto   #grpc proto 包名 所有www.k8sgrpcdemo.cn:18080的请求都会打到k8sgrpcdemo 的8080上
        backend:
          serviceName: k8sgrpcdemo
          servicePort: 18080  #ingress-traefik代理端口

导入ingress:

 

 

我本机hosts文件(C:\Windows\System32\drivers\etc\hosts)配置的：192.168.21.233  www.k8sgrpcdemo.cn，也可以是其他两台。因为三台worker都是有ingress-traefik的。

 测试grpc访问：使用的grpc测试工具：BloomRPC