Golang rpc获取错误的结构变量

Question

1. 我在端口号8000和8100上注册了两个RPC服务，它们是 分别是server1和server2。
2. 然后我在server1端创建了一个RPC连接，连接的目的地是server2。
3. 连接成功建立后，我调用该连接进行远程调用。
4. 远程调用成功，但是当我获取远程调用的被叫方端口号时，有时获取到server1的端口号，有时获取到server2的端口号。在我的理解中，远程调用的被调用方是serve2，程序应该返回server2的端口号。
5. 是什么原因造成的？

我的代码:

package main

import (
    "log"
    "net"
    "net/rpc"
    "sync"
)

type RPCServer struct {
    port    string
    clients map[string]rpc.Client
}

type MessageArgs struct {
    Sender  string
    Message string
}

type MessageReply struct {
    Receive bool
}

func (rs *RPCServer) SendMessage(args *MessageArgs, reply *MessageReply) error {
    log.Printf("current rpc server [%s] Receive Message [%s] from Sender[%s]", rs.port, args.Message, args.Sender)
    reply.Receive = true
    return nil
}

func (rs *RPCServer) server(wg *sync.WaitGroup) {
    l, e := net.Listen("tcp", rs.port)
    if e != nil {
        log.Fatal("listen error: ", e)
    } else {
        log.Printf("server start successfully on port %s\n", rs.port)
    }
    rpc.Register(rs)

    // wait for all server construct but not wait for loop
    wg.Done()

    for {
        conn, err := l.Accept()
        if err != nil {
            log.Fatal("accept error: ", err)
        }

        go rpc.ServeConn(conn)
    }
}

func (rs *RPCServer) connect(port string) {
    client, err := rpc.Dial("tcp", port)
    if err != nil {
        log.Fatal("dialing: ", err)
    } else {
        log.Printf("[%s] server connect other server [%s] successfully.\n", rs.port, port)
    }
    rs.clients[port] = *client
}

func main() {
    servers := make([]RPCServer, 2)
    ports := [2]string{":8000", ":8100"}
    for i := range servers {
        servers[i].clients = make(map[string]rpc.Client)
        servers[i].port = ports[i]
    }

    // start all rpc server
    var wg sync.WaitGroup
    for i := 0; i < 2; i++ {
        wg.Add(1)
        go servers[i].server(&wg)
    }
    wg.Wait()

    servers[0].connect(":8100")

    args := &MessageArgs{Message: "Hello", Sender: ":8000"}
    reply := &MessageReply{}
    clientCall := servers[0].clients[":8100"]
    err := clientCall.Call("RPCServer.SendMessage", args, reply)
    if err != nil {
        log.Fatal(err)
    }
}

有时程序的输出是

2021/07/07 14:39:10 server start successfully on port :8000
2021/07/07 14:39:10 server start successfully on port :8100
2021/07/07 14:39:10 [:8000] server connect other server [:8100] successfully.
2021/07/07 14:39:10 current rpc server [:8000] Receive Message [Hello] from Sender[:8000]

有时程序的输出是

2021/07/07 14:39:16 server start successfully on port :8100
2021/07/07 14:39:16 server start successfully on port :8000
2021/07/07 14:39:16 [:8000] server connect other server [:8100] successfully.
2021/07/07 14:39:16 current rpc server [:8100] Receive Message [Hello] from Sender[:8000]

Answer 1

您正在调用rpc.Register(rs)两次:

1. 向您的代码添加错误检查:

    err := rpc.Register(rs)
    if err != nil {
        log.Println(err)
    }

你会看到:

rpc: service already defined: RPCServer

由于您在这里运行两个 goroutine:

for i := 0; i < 2; i++ {
    wg.Add(1)
    go servers[i].server(&wg)
}

然后他们都会使用以下方法注册一个rpc服务器:

rpc.Register(rs)

所以其中一个偶然是真正注册的服务器。 由于运行 goroutine 的时间未知，因此每次运行都会变化

只需仅注册一台服务器。 例如:

    go servers[1].server(&wg)

---

代码审查说明:

rs.clients[port] = *client

包含sync.Mutex和:

A Mutex must not be copied after first use.