Who says cat can't do link tracking? Stand up for me

background

Link tracking , We have a lot of options . Common are zipkin,pinpoint,skywalking,jaeger etc. .

It's basically based on Google 《Dapper A tracking system for large-scale distributed systems 》 This paper developed .

Today, let's talk about Cat How to implement link tracking in , Never used Cat You can check this article of mine 《 Stay up late ： One article for you to understand Cat Distributed monitoring 》 Get to know .

stay Cat It is convenient to see the total time consumption and business operation of each request , The time consumption of database operations . The call between services can also be monitored by means of embedding points .

Here's the picture , As you can see, one was initiated within the request RPC Call to ,callRPC The first record . Time consuming 11ms, But this RPC Where is the time spent inside the service , You can't view it directly here , You can only check it in another service , Not very convenient .

In detail, I drew a picture to illustrate the current problem ：

As can be seen from the above figure , A request passes through multiple services , Each service has a buried point for remote or local calls , In this way, the exceptions and performance indicators of the call can be monitored .

The next part is in Cat Let's take a look at the scenarios of these indicators ,Cat Data presentation in is presented in terms of project dimensions , So each service has its own monitoring data .

If I want to know about that request , Where is the slowest in the whole link , Where does it take , I have to go separately 4 This information can only be seen in the following services , Is not intuitive .

Realization way

As shown in the figure below :

From gateway to service , From service to service , We need to Trace Only when the information is transmitted can the whole link be concatenated . Only when they are connected together can they be in Cat View the time-consuming information of the whole link in the .

The effect that this article needs to realize is that it can be at the entrance of the request （ gateway ）, See all the services that this request passed through , Time spent in each service .

To connect the whole request , There must be a unique request identifier , Generally we call it traceId. The rest of the work is to pass the link related information layer by layer .

First, receive the request header information in the filter of each service , From gateway to service A, So service A This information needs to be received and passed on to the next service .

HTTP Request message tree construction ：

  

   
//  Building a remote message tree if(request.getHeader(CatConstantsExt.CAT_HTTP_HEADER_ROOT_MESSAGE_ID) != null){
   
        CatContext catContext = new CatContext();
   
        catContext.addProperty(Cat.Context.ROOT,request.getHeader(CatConstantsExt.CAT_HTTP_HEADER_ROOT_MESSAGE_ID));
   
        catContext.addProperty(Cat.Context.PARENT,request.getHeader(CatConstantsExt.CAT_HTTP_HEADER_PARENT_MESSAGE_ID));
   
        catContext.addProperty(Cat.Context.CHILD,request.getHeader(CatConstantsExt.CAT_HTTP_HEADER_CHILD_MESSAGE_ID));
   
        Cat.logRemoteCallServer(catContext);
   
}
  

When passing the message tree information to the next service, it depends on how you call it , If you use Feign perhaps RestTemplate You can use interceptors to deliver .

  

   
public class FeignRequestInterceptor implements RequestInterceptor {
   
    @Override
   
    public void apply(RequestTemplate template) {
   
        CatContext catContext = new CatContext();
   
        Cat.logRemoteCallClient(catContext,Cat.getManager().getDomain());
   
        template.header(CatConstantsExt.CAT_HTTP_HEADER_ROOT_MESSAGE_ID, catContext.getProperty(Cat.Context.ROOT));
   
        template.header(CatConstantsExt.CAT_HTTP_HEADER_PARENT_MESSAGE_ID, catContext.getProperty(Cat.Context.PARENT));
   
        template.header(CatConstantsExt.CAT_HTTP_HEADER_CHILD_MESSAGE_ID, catContext.getProperty(Cat.Context.CHILD));
   
    }
   
}
  

If I use theta Dubbo You can use Dubbo Of Filter To achieve the same effect .

The final effect is shown in the figure below , Called articles/newest Interface , Gateway forwards request to article-provider service ,article-provider Call... Again user-provider Of users/uid Interface to get user information , most important of all user-provider We can also see the time consuming of these operations here , Very convenient .

Complete source reference ： https://github.com/yinjihuan/kitty

About author ： Yin Jihuan , Simple technology enthusiasts ,《Spring Cloud Microservices - Full stack technology and case analysis 》, 《Spring Cloud Microservices introduction Actual combat and advanced 》 author , official account Ape world Originator . Personal wechat jihuan900 , Welcome to hook up with .

I have compiled a complete set of learning materials , Those who are interested can search through wechat 「 Ape world 」, Reply key 「 Learning materials 」 Get what I've sorted out Spring Cloud,Spring Cloud Alibaba,Sharding-JDBC Sub database and sub table , Task scheduling framework XXL-JOB,MongoDB, Reptiles and other related information .