全链路追踪竟然如此简单? bytebuddy搭建全链路追踪的demo 附代码

大家好，我是烤鸭：
   最近一直在研究全链路追踪，比如cat、skywalking、zipkin等。
   发现 skywalking 是基于bytebuddy 实现的，想自己试着写一下demo。
   demo的git地址,感兴趣的可以自己试下。代码在idea中可以跑,至于其他场景需要自己研究(比如用cmd或者linux可能会报NoClassDefDoundError)。

demo地址:(仅实现了http方式的链路，有需要的可以自己补充，比如dubbo或者其他rpc方式的拦截)

https://gitee.com/fireduck_admin/link-trace-demo
环境：
JDK 8

1. 设计目标

监控接口(方法)耗时和链路关系(http请求),对比aop方式,zipkin和cat 是基于拦截的形式。

2. bytebuddy

bytebuddy网上资料虽然不多,但是api比较简单,看看基本就会了。我也不介绍了。具体想看的去官网看下吧。
https://bytebuddy.net/

3. 接口耗时伪代码说明

由于测试链路，我们需要一个agent项目和demo项目(用于请求转发)。
agent项目创建拦截和写具体的拦截逻辑,这里以拦截 Spring的service注解为例。这是在agent项目里的。

public static void premain(String agentArgs, Instrumentation inst) {System.out.println("==============Client=============== premain =============start============");AgentBuilder.Transformer transformerService = new AgentBuilder.Transformer() {@Overridepublic DynamicType.Builder<?> transform(DynamicType.Builder<?> builder, TypeDescription typeDescription, ClassLoader classLoader, JavaModule module) {return builder.method(ElementMatchers.<MethodDescription>any()) // 拦截任意方法.intercept(MethodDelegation.to(MyServiceAdvice.class)); // 委托}};// 拦截 ServiceAgentBuilder agentBuilder = agentBuilder.type(ElementMatchers.isAnnotatedWith(ElementMatchers.named("org.springframework.stereotype.Service"))) // 指定需要拦截的类.transform(transformerService);// 注入 instagentBuilder.installOn(inst);System.out.println("================Client============ premain ================finish===========");}

demo项目启动的时候需要在idea配置vm参数。不知道怎么配的看图。

 -javaagent:\xx\xx\target\link-trace-demo-agent-1.0-SNAPSHOT.jar

拦截效果如图，这样就实现了接口(方法)调用的耗时统计。

4. 全链路伪代码说明

其实自从google在2010年提出了dapper论文后,后续的链路追踪基本都是按照这个思路来实现的，我这就是简易版。
   agent拦截 controller 注解跟上面的service类似，就不贴代码了。
   这里我们需要一个span对象,当前的请求信息记录在span对象(主要是谁调的你)，并且放到threadlocal的调用堆栈中，这样当前的请求和线程就绑定了(方便单个服务内的流转，比如controller调service)。
   这里截图可以看下只单独拦截了web,由于没有上游的信息,所以生成的新的span，seq为1。如图。

模拟下拦截web后调用web方法的链路信息。如图所示。pid(parentid)指的是上个链路的id，这样就可以获取到整个调用链的完整信息（出入参、时间、方法等）

再多链路，比如web-service-web或者更多服务的自己试下吧，思路就是这样的。

5. aop个agent的对比

只是单纯的统计aop方式和bytebuddy两种方式。aop底层有接口使用jdk 代理，无接口使用cglib(底层asm)。而bytebuddy底层也是 asm。
先放一张官方的对比图。

至于到底快不快，我试下。先链路的代码先注释，单纯调下接口试试。（测试方法在test包下）
先看下单次的：

   单独调用controller的时候，bytebuddy明显快的，几乎没有损耗。
   调用web+serivce的时候时间差不多。
   下面单独调用两种情况,500次的平均值：
   仅调用Controller：
   AOP方式： 5.7 ms,主要损耗在首次调用。
   [329, 12, 9, 6, 7, 5, 6, 5, 6, 7, 9, 8, 6, 5, 8, 8, 9, 22, 7, 11, 7, 5, 7, 6, 6, 6, 5, 6, 7, 10, 9, 9, 6, 8, 8, 13, 20, 10, 38, 7, 9, 8, 7, 10, 12, 9, 7, 7, 8, 9, 11, 11, 6, 7, 8, 5, 5, 5, 4, 5, 4, 3, 3, 4, 4, 4, 6, 4, 4, 4, 5, 6, 6, 7, 7, 6, 7, 5, 5, 8, 6, 7, 4, 7, 7, 6, 5, 5, 4, 4, 3, 4, 4, 4, 4, 5, 9, 4, 5, 5, 7, 4, 5, 11, 7, 7, 6, 9, 7, 22, 8, 14, 8, 4, 3, 4, 3, 3, 3, 4, 3, 4, 4, 5, 4, 5, 5, 11, 4, 4, 4, 4, 4, 7, 5, 8, 8, 7, 6, 6, 6, 7, 7, 4, 5, 3, 4, 4, 3, 4, 3, 4, 3, 4, 4, 5, 4, 4, 4, 4, 6, 10, 4, 6, 8, 7, 11, 12, 5, 8, 7, 6, 5, 4, 4, 5, 4, 5, 4, 4, 4, 4, 4, 4, 4, 4, 8, 4, 5, 8, 6, 12, 4, 8, 5, 8, 6, 7, 6, 9, 3, 4, 6, 4, 3, 4, 3, 3, 3, 3, 3, 3, 4, 5, 5, 4, 3, 3, 3, 4, 4, 5, 4, 6, 6, 5, 4, 7, 4, 9, 4, 5, 6, 5, 8, 5, 6, 4, 4, 4, 3, 3, 4, 4, 2, 3, 3, 3, 2, 3, 3, 3, 4, 3, 4, 5, 4, 3, 3, 3, 4, 4, 5, 4, 5, 4, 7, 4, 6, 5, 5, 6, 3, 3, 4, 4, 5, 5, 5, 4, 4, 3, 3, 3, 3, 3, 2, 3, 3, 3, 3, 2, 3, 3, 3, 2, 2, 3, 3, 3, 3, 2, 3, 6, 4, 4, 4, 4, 4, 7, 3, 4, 6, 3, 5, 7, 5, 6, 4, 6, 7, 5, 4, 7, 6, 3, 3, 3, 4, 3, 3, 2, 3, 3, 3, 3, 2, 3, 3, 4, 4, 3, 4, 3, 4, 5, 5, 4, 10, 12, 7, 7, 5, 10, 4, 6, 6, 5, 4, 6, 6, 5, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 3, 3, 3, 4, 13, 6, 11, 4, 6, 4, 4, 6, 4, 4, 5, 4, 4, 6, 4, 3, 4, 3, 4, 3, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 2, 3, 3, 2, 2, 2, 3, 2, 3, 4, 3, 4, 5, 4, 4, 6, 5, 4, 6, 4, 4, 6, 5, 4, 7, 3, 4, 4, 5, 3, 4, 4, 3, 3, 12, 5, 3, 3, 3, 4, 3, 2, 2, 2, 3, 2, 3, 3, 4, 3, 4, 4, 2, 3, 5, 4, 3, 7, 4, 4, 5, 3, 12, 21, 15, 16, 7, 2, 3, 2, 3, 2, 3, 3, 3, 3, 3, 4, 2, 2, 3, 3, 2, 2, 2, 2, 2]
   avg = OptionalDouble[5.706]
   AGENT方式： 5.4 ms,主要损耗在首次调用。
   [251, 9, 11, 12, 9, 9, 9, 9, 8, 11, 13, 6, 5, 5, 5, 6, 5, 5, 4, 6, 6, 6, 4, 10, 7, 6, 8, 7, 8, 6, 7, 7, 6, 6, 7, 6, 8, 5, 6, 5, 4, 4, 5, 5, 4, 4, 4, 4, 3, 4, 3, 5, 4, 3, 4, 4, 5, 5, 6, 5, 7, 6, 11, 10, 7, 9, 6, 5, 7, 5, 8, 6, 10, 7, 7, 7, 5, 5, 5, 5, 5, 5, 7, 5, 9, 19, 10, 7, 4, 3, 4, 4, 4, 3, 3, 4, 4, 6, 5, 4, 3, 5, 6, 4, 4, 5, 9, 10, 5, 3, 4, 5, 7, 4, 5, 10, 5, 8, 5, 11, 6, 6, 10, 4, 4, 5, 5, 7, 5, 4, 4, 7, 4, 4, 4, 5, 6, 4, 4, 5, 6, 6, 4, 6, 4, 4, 5, 8, 4, 5, 5, 4, 4, 5, 4, 4, 4, 7, 7, 10, 4, 4, 4, 3, 4, 3, 3, 4, 5, 4, 5, 8, 4, 5, 6, 7, 4, 4, 6, 5, 4, 7, 4, 4, 5, 5, 8, 5, 4, 6, 4, 9, 4, 5, 3, 3, 3, 4, 4, 3, 3, 4, 4, 4, 6, 5, 12, 6, 5, 5, 6, 8, 7, 14, 8, 5, 6, 6, 5, 5, 10, 5, 6, 5, 4, 4, 3, 4, 4, 4, 4, 5, 4, 6, 7, 4, 8, 5, 6, 4, 5, 7, 4, 6, 5, 5, 6, 3, 4, 3, 3, 4, 4, 3, 3, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 3, 4, 4, 4, 5, 3, 5, 7, 3, 4, 5, 4, 4, 7, 4, 5, 6, 5, 8, 4, 4, 5, 4, 3, 3, 3, 3, 4, 2, 3, 6, 4, 3, 3, 3, 3, 4, 4, 3, 3, 4, 4, 5, 8, 5, 6, 5, 10, 3, 4, 14, 5, 7, 3, 7, 3, 3, 8, 3, 4, 5, 3, 4, 4, 3, 3, 4, 3, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 3, 4, 6, 6, 6, 7, 5, 7, 5, 5, 4, 4, 4, 3, 5, 5, 6, 4, 6, 3, 2, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 5, 4, 3, 3, 4, 4, 5, 8, 4, 6, 3, 5, 5, 4, 6, 8, 5, 3, 3, 4, 6, 7, 5, 8, 4, 4, 4, 5, 6, 5, 4, 3, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 3, 3, 4, 7, 3, 7, 4, 6, 6, 3, 5, 7, 4, 5, 3, 5, 5, 4, 4, 5, 3, 3, 2, 5, 3, 26, 17, 8, 20, 9, 10, 3, 3, 4, 6, 3, 5, 6, 3, 5, 5, 4, 4, 5, 4, 4, 5, 3, 3, 3, 3, 2, 3, 2, 2, 2, 2, 2, 3, 2, 2, 2, 3, 3, 2, 2, 2, 2, 3, 2, 2, 2, 3, 3, 3, 2]
   avg = OptionalDouble[5.438]

结论是在仅有cglib代理的时候(单独调用controller)，耗时差不多,bytebuddy要稍快一些。
   调用Controller+Service：
   AOP方式： 5.7 ms,主要损耗在首次调用。
   [328, 14, 8, 12, 11, 9, 8, 7, 8, 8, 13, 6, 6, 6, 7, 7, 6, 6, 8, 5, 7, 8, 11, 11, 9, 8, 9, 5, 12, 8, 8, 5, 5, 8, 6, 9, 6, 8, 6, 4, 4, 5, 5, 4, 5, 7, 7, 11, 6, 9, 5, 8, 10, 5, 10, 6, 6, 6, 7, 6, 5, 4, 4, 4, 4, 3, 5, 4, 5, 5, 5, 5, 6, 7, 9, 5, 8, 6, 7, 7, 6, 4, 5, 7, 5, 8, 5, 5, 5, 4, 5, 4, 5, 4, 6, 9, 8, 9, 6, 6, 5, 4, 7, 9, 4, 4, 10, 8, 25, 30, 29, 4, 5, 6, 6, 7, 4, 6, 5, 5, 6, 6, 9, 6, 10, 12, 8, 13, 8, 6, 6, 4, 5, 4, 3, 5, 3, 3, 3, 4, 3, 3, 4, 3, 3, 3, 3, 4, 4, 4, 6, 5, 5, 6, 4, 6, 4, 4, 6, 5, 4, 5, 6, 6, 5, 5, 5, 4, 4, 4, 6, 12, 7, 5, 6, 5, 4, 5, 4, 6, 9, 8, 10, 4, 3, 6, 4, 4, 4, 4, 11, 8, 5, 4, 14, 6, 7, 5, 6, 6, 5, 5, 9, 4, 6, 8, 5, 5, 5, 6, 3, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 2, 2, 3, 4, 4, 6, 14, 8, 6, 6, 4, 4, 9, 4, 6, 5, 6, 4, 4, 5, 6, 4, 4, 3, 4, 4, 4, 3, 5, 4, 3, 3, 5, 15, 11, 11, 6, 9, 5, 6, 5, 5, 7, 5, 8, 6, 7, 5, 4, 5, 5, 7, 4, 4, 4, 5, 5, 4, 4, 5, 9, 4, 6, 5, 4, 7, 5, 7, 6, 9, 7, 5, 7, 5, 5, 5, 5, 6, 6, 5, 6, 4, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 5, 3, 5, 3, 5, 5, 8, 5, 5, 6, 8, 5, 5, 4, 4, 6, 4, 4, 5, 4, 4, 4, 3, 4, 3, 4, 4, 3, 3, 3, 3, 2, 3, 3, 3, 4, 4, 3, 7, 4, 4, 8, 4, 7, 4, 3, 8, 4, 4, 6, 5, 3, 7, 3, 4, 5, 3, 3, 2, 3, 2, 3, 3, 3, 2, 3, 3, 3, 4, 7, 4, 6, 3, 3, 3, 3, 3, 3, 3, 4, 3, 3, 5, 4, 4, 4, 6, 4, 7, 4, 5, 6, 4, 4, 6, 5, 7, 4, 3, 6, 3, 3, 3, 3, 3, 2, 3, 2, 3, 3, 3, 3, 3, 3, 4, 3, 3, 2, 3, 4, 4, 4, 4, 4, 5, 4, 5, 3, 5, 4, 7, 4, 31, 5, 4, 5, 6, 5, 6, 5, 4, 5, 8, 4, 4, 4, 4, 4, 7, 4, 4, 4, 4, 4, 5, 4, 3, 8, 4, 32, 13, 37, 28, 5, 4, 3, 3, 3, 3, 2, 2, 3, 3, 3, 5, 6, 3, 3, 3, 4, 5, 3, 3, 3, 4]
   avg = OptionalDouble[6.186]
   AGENT方式： 5.89 ms,主要损耗在首次调用。
   [306, 16, 7, 9, 10, 11, 9, 7, 8, 11, 18, 11, 8, 9, 11, 9, 6, 6, 6, 11, 7, 8, 10, 9, 15, 9, 7, 12, 22, 8, 6, 6, 5, 5, 5, 5, 4, 4, 5, 5, 5, 4, 5, 4, 8, 7, 11, 15, 17, 8, 9, 5, 6, 8, 7, 6, 9, 5, 4, 5, 5, 4, 4, 5, 4, 5, 4, 5, 4, 4, 6, 6, 10, 6, 6, 8, 5, 8, 6, 10, 6, 9, 7, 7, 6, 8, 6, 5, 6, 5, 5, 6, 7, 7, 6, 8, 6, 10, 8, 9, 6, 4, 5, 8, 4, 5, 6, 5, 33, 19, 19, 6, 5, 4, 6, 5, 6, 5, 7, 6, 6, 7, 7, 5, 6, 6, 5, 8, 5, 4, 4, 5, 4, 3, 3, 4, 3, 4, 3, 3, 3, 3, 3, 3, 4, 3, 5, 4, 4, 6, 6, 5, 7, 4, 5, 6, 4, 6, 4, 7, 6, 3, 6, 3, 3, 3, 4, 3, 4, 3, 5, 3, 4, 5, 4, 4, 3, 3, 4, 3, 8, 4, 11, 5, 6, 6, 9, 4, 5, 7, 5, 16, 6, 4, 6, 4, 4, 5, 3, 4, 2, 3, 3, 4, 4, 4, 4, 4, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 5, 6, 5, 6, 4, 5, 9, 5, 8, 4, 4, 5, 4, 8, 7, 8, 5, 5, 4, 5, 5, 7, 4, 6, 4, 4, 6, 4, 4, 4, 5, 4, 4, 5, 8, 18, 12, 5, 7, 5, 5, 5, 4, 4, 7, 5, 3, 3, 4, 3, 4, 3, 3, 3, 4, 3, 4, 3, 4, 3, 9, 4, 4, 4, 3, 3, 3, 5, 5, 4, 4, 5, 4, 5, 6, 4, 3, 4, 6, 4, 5, 7, 3, 5, 4, 4, 6, 4, 3, 4, 3, 3, 3, 3, 3, 2, 2, 3, 2, 3, 4, 3, 3, 3, 3, 3, 3, 4, 4, 4, 3, 7, 4, 5, 4, 4, 4, 4, 5, 4, 6, 4, 5, 7, 5, 6, 5, 3, 4, 4, 4, 3, 3, 4, 3, 3, 3, 3, 3, 3, 3, 2, 3, 2, 2, 3, 3, 3, 2, 3, 4, 4, 3, 4, 2, 6, 4, 4, 5, 3, 3, 5, 4, 5, 7, 7, 5, 12, 4, 5, 4, 3, 4, 5, 5, 3, 4, 4, 3, 3, 3, 3, 3, 3, 4, 3, 3, 3, 3, 3, 4, 4, 4, 6, 5, 6, 5, 5, 6, 4, 5, 6, 4, 7, 5, 5, 7, 6, 6, 4, 4, 5, 4, 5, 3, 3, 5, 4, 4, 4, 3, 4, 4, 4, 3, 3, 3, 4, 4, 3, 4, 4, 5, 4, 4, 4, 5, 4, 6, 4, 4, 6, 4, 4, 7, 4, 4, 6, 2, 2, 5, 3, 4, 4, 3, 3, 18, 7, 9, 10, 21, 28, 20, 12, 5, 4, 4, 3, 3, 5, 2, 3, 3, 2, 3, 2, 2, 2, 3, 3, 3]
   avg = OptionalDouble[5.896]
   结论是在仅有cglib代理+jdk proxy的时候(调用controller+service)，耗时差不多,bytebuddy要稍快一些。

事实证明，确实这样，尤其是我只是测试一个简单接口，如果链路长的时候这个差距会更加明显。

6. 结论

事实上链路追踪的框架已经很很多了，选一款适合自己的就好，如果业务个性化需求比较多，自己开发也是一个不错的选择。你也看到了，自己写一个也没有那么复杂。如果选用开源框架的话，我推荐 skywalking，社区生态整体都挺好的，而且也方便二次开发，网上文章和文档也挺多的，就不多介绍了。

推荐几篇 javaagent 的文章：
https://www.jianshu.com/p/5c62b71fd882

https://www.jianshu.com/p/b72f66da679f

https://www.jianshu.com/p/7b2072513819