一步一步教你写kubernetes sidecar

本文分享自华为云社区《一步一步教你写kubernetes sidecar》,作者: 张俭。

什么是sidecar?

sidecar,直译为边车。 如上图所示,边车就是加装在摩托车旁来达到拓展功能的目的,比如行驶更加稳定,可以拉更多的人和货物,坐在边车上的人可以给驾驶员指路等。边车模式通过给应用服务加装一个"边车"来达到控制和逻辑的分离的目的。

对于微服务来讲,我们可以用边车模式来做诸如 日志收集、服务注册、服务发现、限流、鉴权等不需要业务服务实现的控制面板能力。通常和边车模式比较的就是像spring-cloud那样的sdk模式,像上面提到的这些能力都通过sdk实现。

这两种实现模式各有优劣,sidecar模式会引入额外的性能损耗以及延时,但传统的sdk模式会让代码变得臃肿并且升级复杂,控制面能力和业务面能力不能分开升级。

本文的代码已经上传到gitee

sidecar 实现原理

介绍了sidecar的诸多功能,但是,sidecar是如何做到这些能力的呢?

原来,在kubernetes中,一个pod是部署的最小单元,但一个pod里面,允许运行多个container(容器),多个container(容器)之间共享存储卷和网络栈。这样子,我们就可以多container来做sidecar,或者init-container(初始化容器)来调整挂载卷的权限

日志收集sidecar

日志收集sidecar的原理是利用多个container间可以共用挂载卷的原理实现的,通过将应用程序的日志路径挂出,用另一个程序访问路径下的日志来实现日志收集,这里用cat来替代了日志收集,部署yaml模板如下

复制代码
apiVersion: v1
kind: Pod
metadata:
  name: webserver
spec:
  volumes:
    - name: shared-logs
      emptyDir: {}

  containers:
    - name: nginx
      image: ttbb/nginx:mate
      volumeMounts:
        - name: shared-logs
          mountPath: /opt/sh/openresty/nginx/logs

    - name: sidecar-container
      image: ttbb/base
      command: ["sh","-c","while true; do cat /opt/sh/openresty/nginx/logs/nginx.pid; sleep 30; done"]
      volumeMounts:
        - name: shared-logs
          mountPath: /opt/sh/openresty/nginx/logs

使用kubectl create -f 创建pod,通过kubectl logs命令就可以看到sidecar-container打印的日志输出

复制代码
kubectl logs webserver sidecar-container

转发请求sidecar

这一节我们来实现,一个给应用程序转发请求的sidecar,应用程序代码如下

复制代码
use std::io::prelude::*;
use std::net::{TcpListener, TcpStream};

fn main() {
    let listener = TcpListener::bind("127.0.0.1:7878").unwrap();

    for stream in listener.incoming() {
        let stream = stream.unwrap();

        handle_connection(stream);
    }
    println!("Hello, world!");
}

fn handle_connection(mut stream: TcpStream) {
    let mut buffer = [0; 1024];

    stream.read(&mut buffer).unwrap();

    let contents = "Hello";

    let response = format!(
        "HTTP/1.1 200 OK\r\nContent-Length: {}\r\n\r\n{}",
        contents.len(),
        contents
    );

    println!("receive a request!");
    stream.write(response.as_bytes()).unwrap();
    stream.flush().unwrap();
}

我们再来写一个sidecar,它会每15秒向应用程序发出请求

复制代码
use std::thread;
use std::time::Duration;

fn main() {
    loop {
        thread::sleep(Duration::from_secs(15));
        let response = reqwest::blocking::get("http://localhost:7878").unwrap();
        println!("{}", response.text().unwrap())
    }
}

通过仓库下的intput/build.sh脚本构造镜像,运行yaml如下

复制代码
apiVersion: v1
kind: Pod
metadata:
  name: webserver
spec:
  containers:
    - name: input-server
      image: sidecar-examples:input-http-server

    - name: input-sidecar
      image: sidecar-examples:sidecar-input

通过查看kubectl logs input input-http-server可以看到input-http-server收到了请求

复制代码
receive a request!
receive a request!

拦截请求sidecar

应用程序代码,它会每15s向localhost发出请求

复制代码
package com.shoothzj.sidecar

import akka.actor.typed.ActorSystem
import akka.actor.typed.scaladsl.Behaviors
import akka.http.scaladsl.Http
import akka.http.scaladsl.model._

import scala.concurrent.{ExecutionContextExecutor, Future}
import scala.util.{Failure, Success}

object HttpClient {
    def main(args: Array[String]): Unit = {
        while (true) {
            Thread.sleep(15_000L)
            implicit val system: ActorSystem[Nothing] = ActorSystem(Behaviors.empty, "SingleRequest")
            // needed for the future flatMap/onComplete in the end
            implicit val executionContext: ExecutionContextExecutor = system.executionContext

            val responseFuture: Future[HttpResponse] = Http().singleRequest(HttpRequest(uri = "http://localhost:7979/hello"))

            responseFuture
                    .onComplete {
                        case Success(res) => println(res)
                        case Failure(_) => sys.error("something wrong")
                    }
        }
    }
}

我们再来写一个sidecar,它会拦截http请求并打印日志

复制代码
package com.shoothzj.sidecar

import akka.actor.typed.ActorSystem
import akka.actor.typed.scaladsl.Behaviors
import akka.http.scaladsl.Http
import akka.http.scaladsl.model._
import akka.http.scaladsl.server.Directives._

import scala.concurrent.ExecutionContextExecutor
import scala.io.StdIn

object HttpServer {

    def main(args: Array[String]): Unit = {

        implicit val system: ActorSystem[Nothing] = ActorSystem(Behaviors.empty, "my-system")
        // needed for the future flatMap/onComplete in the end
        implicit val executionContext: ExecutionContextExecutor = system.executionContext

        val route =
            path("hello") {
                get {
                    println("receive a request")
                    complete(HttpEntity(ContentTypes.`text/html(UTF-8)`, "<h1>Say hello to akka-http</h1>"))
                }
            }

        val bindingFuture = Http().newServerAt("localhost", 7979).bind(route)
        while (true) {
            Thread.sleep(15_000L)
        }
    }
}

通过仓库下的output/build.sh脚本构造镜像,运行yaml如下

复制代码
apiVersion: v1
kind: Pod
metadata:
  name: output
spec:
  volumes:
    - name: shared-logs
      emptyDir: {}

  containers:
    - name: output-workload
      image: sidecar-examples:output-workload
      imagePullPolicy: Never

    - name: sidecar-output
      image: sidecar-examples:sidecar-output
      imagePullPolicy: Never

通过查看kubectl logs output output-workload可以看到output-sidecar收到了请求

复制代码
HttpResponse(200 OK,List(Server: akka-http/10.2.9, Date: Tue, 29 Mar 2022 00:15:47 GMT),HttpEntity.Strict(text/html; charset=UTF-8,31 bytes total),HttpProtocol(HTTP/1.1))
HttpResponse(200 OK,List(Server: akka-http/10.2.9, Date: Tue, 29 Mar 2022 00:16:02 GMT),HttpEntity.Strict(text/html; charset=UTF-8,31 bytes total),HttpProtocol(HTTP/1.1))
HttpResponse(200 OK,List(Server: akka-http/10.2.9, Date: Tue, 29 Mar 2022 00:16:17 GMT),HttpEntity.Strict(text/html; charset=UTF-8,31 bytes total),HttpProtocol(HTTP/1.1))
HttpResponse(200 OK,List(Server: akka-http/10.2.9, Date: Tue, 29 Mar 2022 00:16:32 GMT),HttpEntity.Strict(text/html; charset=UTF-8,31 bytes total),HttpProtocol(HTTP/1.1))
HttpResponse(200 OK,List(Server: akka-http/10.2.9, Date: Tue, 29 Mar 2022 00:16:47 GMT),HttpEntity.Strict(text/html; charset=UTF-8,31 bytes total),HttpProtocol(HTTP/1.1))
HttpResponse(200 OK,List(Server: akka-http/10.2.9, Date: Tue, 29 Mar 2022 00:17:02 GMT),HttpEntity.Strict(text/html; charset=UTF-8,31 bytes total),HttpProtocol(HTTP/1.1))
HttpResponse(200 OK,List(Server: akka-http/10.2.9, Date: Tue, 29 Mar 2022 00:17:17 GMT),HttpEntity.Strict(text/html; charset=UTF-8,31 bytes total),HttpProtocol(HTTP/1.1))
HttpResponse(200 OK,List(Server: akka-http/10.2.9, Date: Tue, 29 Mar 2022 00:17:32 GMT),HttpEntity.Strict(text/html; charset=

点击关注,第一时间了解华为云新鲜技术~

相关推荐
运维全栈笔记13 小时前
K8S部署Redis高可用全攻略:1主2从3哨兵架构实战
redis·docker·云原生·容器·架构·kubernetes·bootstrap
尘世壹俗人13 小时前
使用K8s部署模型
kubernetes
AI木马人15 小时前
9.人工智能实战:GPU 服务如何上 Kubernetes?从单机部署到 K8s + NVIDIA Device Plugin + HPA 的生产级改造
人工智能·容器·kubernetes
码点滴19 小时前
告别显存焦虑:PagedAttention 如何将大模型吞吐量提升 4 倍?
人工智能·架构·kubernetes·大模型·pagedattention
键盘鼓手苏苏20 小时前
Kubernetes 容器安全最佳实践
云原生·kubernetes·k8
键盘鼓手苏苏20 小时前
Kubernetes 安全最佳实践
云原生·kubernetes·k8
木雷坞2 天前
K8s GPU 推理服务 ImagePullBackOff 排查与预热
云原生·容器·kubernetes·gpu算力
吴爃2 天前
Spring Boot 项目在 K8S 中的打包、部署与运维发布实践
运维·spring boot·kubernetes
The Straggling Crow2 天前
Monitoring 2026-04-30
kubernetes
AOwhisky2 天前
Kubernetes调度与服务暴露:从“定时任务”到“服务发现”的完全指南
linux·运维·云原生·容器·kubernetes·服务发现