kubernetes集群中利用etcd和grpc实现golang服务间通信 - Go语言中文社区

kubernetes集群中利用etcd和grpc实现golang服务间通信


kubernetes集群中利用etcd和grpc实现golang服务间通信

注:文中涉及工作环境相关的网址和IP已经被替换

1. 项目背景

  • 服务运行于docker容器中
  • 使用kubernetes管理容器
  • 服务有多个节点作为一个集群
  • 使用rest接口设置服务缓存中的信息
  • 需要将信息同步到集群中其他节点

2. 项目方案

  • 使用grpc做服务间通信
  • 从etcd中读取服务所有状态为running的节点信息,包括:podIp、status、hostIp、startedAt(启动时间)
  • 服务启动时选取运行时间最长的节点,调用grpc接口请求缓存的信息同步到本容器的服务中
  • 使用rest接口设置缓存的时候,遍历所有节点(不包括自身),调用grpc接口将信息同步到其他节点

方案特点:

  • 不需要借助额外的配置管理工具(如:zookeeper)
  • 不需要自行管理节点的配置信息(因为kubernetes的etcd中已经有完整的节点信息)
  • grpc开发、传输效率高,扩展性好
  • grpc使用http2.0方便后续提供rest接口

1. etcd简介

etcd 是用 golang 实现的一种 K-V 分布式存储系统,内部用raft协议做一致性校验,对外提供http的访问接口,最新版中提供了grpc的访问接口。
etcd主要用于:

  • 配置管理
  • 服务注册于发现
  • 选主
  • 应用调度
  • 分布式队列
  • 分布式锁

与etcd类似的还有zookeeper
这里 有一篇文章简单介绍了etcd和zookeeper的优缺点以及etcd的工作原理

2. kubernetes与etcd

前面介绍了etcd特别适合用于做集群服务的配置管理,kubernets 是用于docker容器编排的,也是用golang实现的,所以自然而然就采用etcd作为服务配置的存储方式了。这里 有一篇kubernets的架构介绍。

etcd在kubernetes中的最大作用是保存容器节点(pod)信息,包括:容器的服务名、状态、IP、版本以及其他信息

通过类似如下的命令可以获取到pod的信息

curl http://10.20.30.40:2379/v2/keys/registry/pods/default

etcd中保存的容器节点信息格式如下:

{
    "action": "get",
    "node": {
      "key": "/registry/pods/default",
      "dir": true,
      "nodes": [
        {
          "key": "/registry/pods/default/hello-web-29a74e26ea3c2138e1727f35a111f4c6-dknwh",
          "value": "{"kind":"Pod","apiVersion":"v1","metadata":{"name":"hello-web-29a74e26ea3c2138e1727f35a111f4c6-dknwh","generateName":"hello-web-29a74e26ea3c2138e1727f35a111f4c6-","namespace":"default","selfLink":"/api/v1/namespaces/default/pods/hello-web-29a74e26ea3c2138e1727f35a111f4c6-dknwh","uid":"09c45029-3fa0-11e7-a46c-00163e327954","creationTimestamp":"2017-05-23T10:10:24Z","labels":{"app":"hello","deployment":"bb6de7bfc7f357818a8c07faf3987d40","tier":"frontend"},"annotations":{"kubernetes.io/created-by":"{\"kind\":\"SerializedReference\",\"apiVersion\":\"v1\",\"reference\":{\"kind\":\"ReplicationController\",\"namespace\":\"default\",\"name\":\"hello-web-29a74e26ea3c2138e1727f35a111f4c6\",\"uid\":\"e42ce61a-3f9f-11e7-a46c-00163e327954\",\"apiVersion\":\"v1\",\"resourceVersion\":\"4361319\"}}\n"},"ownerReferences":[{"apiVersion":"v1","kind":"ReplicationController","name":"hello-web","uid":"32559b88-3fa0-11e7-a46c-00163e327954","controller":true}]},"spec":{"containers":[{"name":"hello-web","image":"docker.helloword.com/hello-web:f022d25","ports":[{"containerPort":8087,"protocol":"TCP"}],"env":[{"name":"SERVER","valueFrom":{"configMapKeyRef":{"name":"cluster-config","key":"external.ip"}}},{"name":"SERVER_PORT","valueFrom":{"configMapKeyRef":{"name":"hello-config","key":"hello.api.port"}}}],"resources":{"limits":{"cpu":"1","memory":"1Gi"},"requests":{"cpu":"100m","memory":"512Mi"}},"terminationMessagePath":"/dev/termination-log","imagePullPolicy":"IfNotPresent"}],"restartPolicy":"Always","terminationGracePeriodSeconds":30,"dnsPolicy":"ClusterFirst","nodeName":"10.30.58.179","securityContext":{},"imagePullSecrets":[{"name":"cn-registry"}]},"status":{"phase":"Running","conditions":[{"type":"Initialized","status":"True","lastProbeTime":null,"lastTransitionTime":"2017-05-23T10:10:24Z"},{"type":"Ready","status":"True","lastProbeTime":null,"lastTransitionTime":"2017-05-23T10:10:29Z"},{"type":"PodScheduled","status":"True","lastProbeTime":null,"lastTransitionTime":"2017-05-23T10:10:24Z"}],"hostIP":"10.30.58.179","podIP":"172.80.13.4","startTime":"2017-05-23T10:10:24Z","containerStatuses":[{"name":"hello-web","state":{"running":{"startedAt":"2017-05-23T10:10:29Z"}},"lastState":{},"ready":true,"restartCount":0,"image":"docker.helloword.com/hello-web:f022d25","imageID":"docker-pullable://docker.helloword.com/hello-web@sha256:f8e0460983b0d3f87733453b588469d8e225afbfc764da2ae55238cd524ef70a","containerID":"docker://78cd912de942f744a36bd51907562c5e670fb300ddc85267e3ec72572fdb5617"}]}}n",
          "modifiedIndex": 4361528,
          "createdIndex": 4361320
        }
	  ]
	}
}

其中value部分的json数据格式化后如下:

{
  "kind": "Pod",
  "apiVersion": "v1",
  "metadata": {
    "name": "hello-web-29a74e26ea3c2138e1727f35a111f4c6-dknwh",
    "generateName": "hello-web-29a74e26ea3c2138e1727f35a111f4c6-",
    "namespace": "default",
    "selfLink": "/api/v1/namespaces/default/pods/hello-web-29a74e26ea3c2138e1727f35a111f4c6-dknwh",
    "uid": "09c45029-3fa0-11e7-a46c-00163e327954",
    "creationTimestamp": "2017-05-23T10:10:24Z",
    "labels": {
      "app": "hello",
      "deployment": "bb6de7bfc7f357818a8c07faf3987d40",
      "tier": "frontend"
    },
    "annotations": {
      "kubernetes.io/created-by": "{"kind":"SerializedReference","apiVersion":"v1","reference":{"kind":"ReplicationController","namespace":"default","name":"hello-web-29a74e26ea3c2138e1727f35a111f4c6","uid":"e42ce61a-3f9f-11e7-a46c-00163e327954","apiVersion":"v1","resourceVersion":"4361319"}}n"
    },
    "ownerReferences": [
      {
        "apiVersion": "v1",
        "kind": "ReplicationController",
        "name": "hello-web",
        "uid": "32559b88-3fa0-11e7-a46c-00163e327954",
        "controller": true
      }
    ]
  },
  "spec": {
    "containers": [
      {
        "name": "hello-web",
        "image": "docker.helloword.com/hello-web:f022d25",
        "ports": [
          {
            "containerPort": 8087,
            "protocol": "TCP"
          }
        ],
        "env": [
          {
            "name": "SERVER",
            "valueFrom": {
              "configMapKeyRef": {
                "name": "cluster-config",
                "key": "external.ip"
              }
            }
          },
          {
            "name": "SERVER_PORT",
            "valueFrom": {
              "configMapKeyRef": {
                "name": "hello-config",
                "key": "hello.api.port"
              }
            }
          }
        ],
        "resources": {
          "limits": {
            "cpu": "1",
            "memory": "1Gi"
          },
          "requests": {
            "cpu": "100m",
            "memory": "512Mi"
          }
        },
        "terminationMessagePath": "/dev/termination-log",
        "imagePullPolicy": "IfNotPresent"
      }
    ],
    "restartPolicy": "Always",
    "terminationGracePeriodSeconds": 30,
    "dnsPolicy": "ClusterFirst",
    "nodeName": "10.30.58.179",
    "securityContext": {},
    "imagePullSecrets": [
      {
        "name": "cn-registry"
      }
    ]
  },
  "status": {
    "phase": "Running",
    "conditions": [
      {
        "type": "Initialized",
        "status": "True",
        "lastProbeTime": null,
        "lastTransitionTime": "2017-05-23T10:10:24Z"
      },
      {
        "type": "Ready",
        "status": "True",
        "lastProbeTime": null,
        "lastTransitionTime": "2017-05-23T10:10:29Z"
      },
      {
        "type": "PodScheduled",
        "status": "True",
        "lastProbeTime": null,
        "lastTransitionTime": "2017-05-23T10:10:24Z"
      }
    ],
    "hostIP": "10.30.58.179",
    "podIP": "172.80.13.4",
    "startTime": "2017-05-23T10:10:24Z",
    "containerStatuses": [
      {
        "name": "hello-web",
        "state": {
          "running": {
            "startedAt": "2017-05-23T10:10:29Z"
          }
        },
        "lastState": {},
        "ready": true,
        "restartCount": 0,
        "image": "docker.helloword.com/hello-web:f022d25",
        "imageID": "docker-pullable://docker.helloword.com/hello-web@sha256:f8e0460983b0d3f87733453b588469d8e225afbfc764da2ae55238cd524ef70a",
        "containerID": "docker://78cd912de942f744a36bd51907562c5e670fb300ddc85267e3ec72572fdb5617"
      }
    ]
  }
}

3. grpc简介

grpc是google实现的一种基于protobuf的远程服务调用框架,数据采用二进制传输,其传输协议是基于http2.0。

相比于其他各种rpc框架,grpc由于基于protobuf和http2.0,具有以下优点:

  • 通用性好,支持各种语言
  • 二进制传输,效率高
  • 扩展性好,只需要修改protobuf文件并重新生成代码

4. grpc开发环境搭建

4.1 protobuf环境

首先,去https://github.com/google/protobuf/releases/tag/v3.3.0 这个页面下载对应的protobuf编译器安装文件并安装好protoc

go get -u github.com/golang/protobuf
cd $GOPATH/src/github.com/golang/protobuf
# 如果有安装makefile,直接执行make install,如果没有则执行以下命令
go install ./proto ./jsonpb ./ptypes
go install ./protoc-gen-go

4.2 grpc环境

#安装grpc依赖库
go get -u google.golang.org/grpc
#安装grpc-go插件,用于将proto文件编译成grpc的golang代码
go get -u github.com/grpc/grpc-go
cd $GOPATH/src

mv github.com/grpc/grpc-go google.golang.org/grpc/grpc-go

遇到go get无法下载的包,也可以通过 http://gopm.io/ 或者 http://golangtc.com/download/package 进行下载

###5. 定义proto文件

syntax = "proto3"; //使用proto3版本

//用于java等语言的package配置
option java_multiple_files = true;
option java_package = "io.grpc.examples.hellorpc";
option java_outer_classname = "hellorpcProto";

//用于golang等语言的package配置
package hellorpc;

//定义服务接口,其中rpc关键字表示 rpc 接口,用于生成grpc接口代码
service Sync {
    rpc Get (SyncRequest) returns(SyncResponse) {}
    rpc Set (SyncRequest) returns(SyncResponse) {}
    rpc GetAll(SyncRequest)returns(SyncResponse) {}
}

//定义请求数据类型, repeated最终会转换成golang中的数组/切片
message SyncRequest {
    repeated SyncData data= 1;
}

//定义返回的数据类型
message SyncResponse {
    repeated SyncData data= 1;
}

//定义实体数据类型,用type字段表示请求的数据类型,用data字段保存请求的数据或者返回的数据
//map<string, string>最终会转换成golang中的map[string]string类型
message SyncData {
    int32 type = 1;
    map<string, string> data = 2;
}

编译proto文件

protoc --go_out=plugins=grpc:./hellorpc hellorpc.proto

其中–go_out用于指定go的proto编译插件以及插件参数
编译成功后,会在 hellorpc目录中生成 hellorpc.pb.go 文件,可以在其他go文件中通过 import “hello-api/hellorpc” 来使用文件中定义的接口

6. hellorpc.pb.go 文件分析

前面提到的 service Sync 部分会编译成如下两部分

type SyncClient interface {
	Get(ctx context.Context, in *SyncRequest, opts ...grpc.CallOption) (*SyncResponse, error)
	Set(ctx context.Context, in *SyncRequest, opts ...grpc.CallOption) (*SyncResponse, error)
	GetAll(ctx context.Context, in *SyncRequest, opts ...grpc.CallOption) (*SyncResponse, error)
}

type SyncServer interface {
	Get(context.Context, *SyncRequest) (*SyncResponse, error)
	Set(context.Context, *SyncRequest) (*SyncResponse, error)
	GetAll(context.Context, *SyncRequest) (*SyncResponse, error)
}

其中 SyncClient 的接口 在 hellorpc.pb.go 里面已经实现好了接口,直接调用即可,但SyncServer定义的接口是需要我们自己实现

7. 服务端代码实现(rtc_server.go)

//先定义server类型,并实现好SyncServer定义的接口
type server struct {}

const (
    HELLO_SYNC_REST_CLUSTER_INFO = iota
)

func (s *server)Get(ctx context.Context, in *hellorpc.SyncRequest) (*hellorpc.SyncResponse, error){
    var response = hellorpc.SyncResponse{Data: make([]*hellorpc.SyncData, 0, 10)}
    for i := 0; i < len(in.Data); i++{
        request := in.Data[i]
        switch request.Type {
        case hello_SYNC_REST_CLUSTER_INFO:
    		// get something from local cache and set to response
            break
        }
    }
    return &response, nil
}


func (s *server)Set(ctx context.Context, in *hellorpc.SyncRequest) (*hellorpc.SyncResponse, error){
    var response = hellorpc.SyncResponse{Data: make([]*hellorpc.SyncData, 0, 10)}
    for i := 0; i < len(in.Data); i++{
        request := in.Data[i]
        switch request.Type {
        case HELLO_SYNC_REST_CLUSTER_INFO:
    		// set something to local cache, and set the result to response
            break
        }
    }
    return &response, nil
}

func (s *server)GetAll(ctx context.Context, in *hellorpc.SyncRequest) (*hellorpc.SyncResponse, error){
    var response = hellorpc.SyncResponse{Data: make([]*hellorpc.SyncData, 0, 10)}
    for i := 0; i < len(in.Data); i++{
		request := in.Data[i]
        switch request.Type {
        case HELLO_SYNC_REST_CLUSTER_INFO:
    		// get all data from local cache, and set the result to response
            break
        }
    }
    return &response, nil
}

实现好接口后,我们需要将服务注册到grpc,这里我们实现一个名为StartSyncServer的函数来做这些事情

func StartSyncServer(address string) error{
    lis, err := net.Listen("tcp", address)
    if err != nil {
        beego.Debug("start sync server error: %v", err)
        return err
    }
    s := grpc.NewServer()
    hellorpc.RegisterSyncServer(s, &server{})

    //由于s.Serve方法是会一直阻塞住,所以我们需要起一个go routine来执行,在其停止后输出错误信息
    go func(){
        err := s.Serve(lis)
        beego.Debug("sync server stopped with error: %v", err)
    }()
    return nil
}

将StartSyncServer函数添加到模块的 init 函数中执行,我们服务端的代码就基本完成了

8. 客户端代码实现(rtc_client.go)

//先定义好客户端类型syncClient,这里我们利用继承的方式将hellorpc.SyncClient实现的方法继承过来
type syncClient struct{
	hellorpc.SyncClient
    conn *grpc.ClientConn
    address string
}

func OpenSyncClient(address string)(syncClient, error) {
	s := syncClient{}
    //grpc.WithInsecure用于关闭安全验证,因为我们是在docker内部环境里使用,不暴露在外网,就没有加安全认证了
    conn, err := grpc.Dial(address, grpc.WithInsecure(), grpc.WithTimeout(5*time.Second))
    if err != nil {
        fmt.Println("----open client error %v, conn: %v", err, conn)
        return s, err
    }
    s.conn = conn
    s.address = address
    s.client = hellorpc.NewSyncClient(conn)
    return s, nil
}

func CloseSyncClient(s *syncClient) {
    if s.conn != nil {
        s.conn.Close()
        s.conn = nil
        s.client = nil
    }
}

这样我们只需要编写 c, err := OpenSyncClient(address),既可通过 response, err := c.Get(context.Background(), request) 的方式调用hellorpc.SyncClient定义的方法

9. etcd客户端代码实现(部分功能)

根据etcd的返回值数据结构,我们定义一下两种类型的数据

//用于保存etcd的返回的数据
type EtcdData struct{
    Key             string
    Dir             bool
    Value           interface{}
    CreatedIndex    int32
    ModifiedIndex   int32
    Nodes           []EtcdData
}
//用于保存pod相关的数据
type PodData struct {
    Name            string
    PodIP           string
    HostIP          string
    Status          string
    UpdateTime      string
    Timestamp       int64
}

func newEtcdData() EtcdData{
    return EtcdData{Dir: false, Value: "", Key: "", Nodes: make([]EtcdData, 0, 100)}
}

接下来我们实现EtcdClient

//先定义好EtcdClient的数据结构
type EtcdClient struct{}
//用于解析etcd返回的数据
func parseEtcdData(dataIn map[string]interface{}, dataOut *EtcdData) error {
    if key, ok := dataIn["key"]; ok {
        dataOut.Key = key.(string)
    }
    if isDir, ok := dataIn["dir"]; ok {
        dataOut.Dir = isDir.(bool)
    }
    if value, ok := dataIn["value"]; ok {
        dataOut.Value = value
    }
    if createdIndex, ok := dataIn["createdIndex"]; ok {
        dataOut.CreatedIndex = int32(createdIndex.(float64))
    }
    if modifiedIndex, ok := dataIn["modifiedIndex"]; ok {
        dataOut.ModifiedIndex = int32(modifiedIndex.(float64))
    }
    if nodes, ok := dataIn["nodes"]; ok {
        var subnodes = nodes.([]interface{})
        for i := 0; i < len(subnodes); i++{
            node := subnodes[i].(map[string]interface{})
            var nodeData = newEtcdData()
            parseEtcdData(node, &nodeData)
            dataOut.Nodes = append(dataOut.Nodes,nodeData)
        }
    }
    return nil
}

//实现Get方法用于获取某个key的值
func (c *EtcdClient)Get(baseUrl, key string)(EtcdData, error){
	var url = baseUrl + key
    var res = newEtcdData()
    var result = make(map[string]interface{})
    resp, err := http.Get(url)
    if err == nil{
        out, err1 := ioutil.ReadAll(resp.Body)
        if err1 == nil{
            err2 := json.Unmarshal([]byte(out), &result)
            if err2 != nil{
                return res, err2
            }
            node := result["node"].(map[string]interface{})
            err = parseEtcdData(node, &res)
        }else{
            return res, err1
        }
    }
    return res, err
}

由于我们的服务是跑在docker里,由kubernetes进行服务编排,所以我们需要解析kubernetes在etcd中保存的数据

//用于解析pod的状态信息
func parsePodStatus(podStatus interface{}, podData *PodData){
    pod_status := podStatus.(map[string]interface{}
                            
                            版权声明:本文来源CSDN,感谢博主原创文章,遵循 CC 4.0 by-sa 版权协议,转载请附上原文出处链接和本声明。
原文链接:https://blog.csdn.net/letian0805/article/details/72862354
站方申明:本站部分内容来自社区用户分享,若涉及侵权,请联系站方删除。

0 条评论

请先 登录 后评论

官方社群

GO教程

猜你喜欢