kubernetes 集群内的 C# gRPC 无法将 DNS 名称解析为其他服务

C# gRPC inside kubernetes cluster cannot resolve DNS name to other service

我调试了几个小时都没有效果。我要么知道如何 google 它要么就是它不寻常的问题。

我在 google 云上有 Kubernetes Engine 集群。为了测试它,我部署了简单的应用程序: https://github.com/iniside/uSpark 它由三个控制台应用程序组成:

他们都在使用 .NET Core 2.1、gRPC 1.12 和 C#。 Kubernetes 引擎是 1.10

我已经尝试查看服务 dns 是否在集群内解析。 我也试过在创建频道时直接通过它的IP调用服务。

其中任何一个以:

结尾
Grpc.Core.RpcException: 'Status(StatusCode=Unavailable, Detail="Name resolution failure")'

GreeterServer 正在尝试与 Greeter.Backend:

对话
public override Task<HelloReply> SayHelloAgain(HelloRequest request, ServerCallContext context)
{
    Channel channel = new Channel("grpc-greeter-backend.default.svc.cluster.local", 9000, ChannelCredentials.Insecure);
    var backendClient = new GreeterBackend.GreeterBackendClient(channel);

    var reply = backendClient.SayHelloFromBackend(new BackendHelloRequest { Name = "iniside" });
    channel.ShutdownAsync().Wait(); //not neeed to wait, but easier to debug now.
    return Task.FromResult<HelloReply>(new HelloReply { Message = "Hello Again " + request.Name + " " + reply.Message });
}

这是我的后端代码:

class GreeterBackendImpl : GreeterBackend.GreeterBackendBase
{
    public override Task<BackendHelloReply> SayHelloFromBackend(BackendHelloRequest request, ServerCallContext context)
    {
        return Task.FromResult<BackendHelloReply>(new BackendHelloReply { Message = "Hello " + request.Name + " From Backend"});
    }
}

class Program
{
    static void Main(string[] args)
    {
        Console.WriteLine("Hello World!");

        Server server = new Server
        {
            Services = { GreeterBackend.BindService(new GreeterBackendImpl()) },
            Ports = { new ServerPort("127.0.0.1", 9000, ServerCredentials.Insecure) }
        };

        server.Start();

        Console.WriteLine("Gretter server is linstening on port 50051");
        Console.WriteLine("Press any key to stop server");

        int read = Console.Read();
        while (read < 0)
        {

        }
    }
}

前端配置:

apiVersion: v1
kind: Service
metadata:
  name: esp-grpc-greeter
spec:
  ports:
  # Port that accepts gRPC and JSON/HTTP2 requests over HTTP.
  - port: 80
    targetPort: 9100
    protocol: TCP
    name: http2
  selector:
    app: esp-grpc-greeter
  type: LoadBalancer
---
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: esp-grpc-greeter
spec:
  replicas: 1
  template:
    metadata:
      labels:
        app: esp-grpc-greeter
    spec:
      containers:
      - name: esp
        image: gcr.io/endpoints-release/endpoints-runtime:1
        args: [
          "--http2_port=9100",
          "--service=hellohorld3.endpoints.sa-game-206414.cloud.goog",
          "--rollout_strategy=managed",
          "--backend=grpc://127.0.0.1:9000"
        ]
        ports:
          - containerPort: 9100
      - name: greeter
        image: eu.gcr.io/sa-game-206414/greeter-service:v1
        ports:
          - containerPort: 8000

后端配置:

apiVersion: v1
kind: Service
metadata:
  name: grpc-greeter-backend
spec:
  ports:
  # Port that accepts gRPC and JSON/HTTP2 requests over HTTP.
  - port: 8080
    targetPort: 9000
    protocol: TCP
    name: http2
  - port: 9000
    targetPort: 9000
    protocol: TCP
    name: http2900
  selector:
    app: grpc-greeter-backend
  type: ClusterIP
---
apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: grpc-greeter-backend
spec:
  replicas: 1
  template:
    metadata:
      labels:
        app: grpc-greeter-backend
    spec:
      containers:
      - name: greeter-backend
        image: eu.gcr.io/sa-game-206414/greeter-backend:b1
        ports:
          - containerPort: 9000

我解决了问题,它是两倍。首先,正如@spender 所建议的,我确实将后端和前端的 gRPC 地址更改为 0.0.0.0。

但是更改没有被集群接收。因为一直以来它都在拉错图像。

每次我进行更改时,我都会删除带有旧标签的旧图像(本地和 GCR 上的)。使用相同的远程存储库和标签创建新图像并将它们推送到 GCR。

由于某些原因,图片没有更新。