Sleuth brief introduction

Sleuth yes Spring Cloud One of the components of , It's for Spring Cloud A distributed tracking solution is implemented , compatible Zipkin,HTrace And other log based tracking systems , for example ELK（Elasticsearch 、Logstash、 Kibana）.

Related terms

Sleuth Introduced many Dapper Terms in ：

• Span : Basic unit of work . Whether it's sending a RPC Or to RPC Sending a response is a Span. every last Span Through one 64 position ID To uniquely identify , And through another 64 position ID Yes Span Where Trace Unique identification .
  Span Be able to start and stop , They keep tracking their time information , When you create a Span, You have to stop it at some point in the future . Tips ： Start a Trace The initialization Span It's called Root Span , its Span ID and Trace Id identical .

• Trace By a series of Span A tree structure of . for example , If you want to perform a distributed big data storage operation , This Trace It may be up to you PUT Request to form .

• Annotation： Used to record the existence of an event in time . By introducing Brave library , We don't have to set up a series of special events , So that Zipkin Be able to know who the client and server are 、 Where did the request start 、 Where does it end . For the purpose of learning , Let's list these events here ：

  cs （Client Sent） - Client initiates a request , This annotation indicates a Span The beginning of .
  sr （Server Received） - The server receives the request and begins to process it , If you use sr Timestamp minus cs The timestamp shows how many network delays there are .
  ss（Server Sent）- Comment request processing completed ( The response was sent to the client ), If you use ss Timestamp minus sr
  The time stamp can get the time spent by the server processing the request .
  cr（Client Received）- Heralded a Span The end of , The client successfully received the response from the server , If you use cr Timestamp minus cs
  The time stamp can get the whole time that the client needs to get the response from the server .

The following figure shows the... In a system Span and Trace It looks like ：

Use Sleuth+zipkin

Introduce dependencies

The premise is that SpringCloud Of Demo

  <dependency>
		 <groupId>org.springframework.cloud</groupId>
       <artifactId>spring-cloud-starter-sleuth</artifactId>
       <scope>compile</scope>
 </dependency>
  <dependency>
      <groupId>org.springframework.cloud</groupId>
      <artifactId>spring-cloud-starter-zipkin</artifactId>
  </dependency>

To configure

#  Collection type 
spring.zipkin.sender.type=web
# zipkin The address of the server 
spring.zipkin.base-url=http://localhost:9411/
#  Set sampling collection to 100%, The default is 0.1, namely 10%
spring.sleuth.sampler.probability=1

Be careful ：

spring-cloud-sleuth-stream Expired and incompatible with these targets . If you want to use RabbitMQ perhaps Kafka replace HTTP , We need to introduce spring-rabbit perhaps spring-kafka rely on . The default target name is zipkin . If you're using Kafka , You have to set the corresponding spring.zipkin.sender.type attribute

Download and enable zipkin

Download address ：

https://search.maven.org/remote_content?g=io.zipkin.java&a=zipkin-server&v=LATEST&c=exec

Start with the following command Zipkin：

java -jar zipkin-server-2.12.2-exec.jar

visit http://localhost:9411/zipkin

Start the service and test

View the log after sending the message ：

Sleuth take Trace Id and Span Id Add to Slf4J MDC（Mapped Diagnostic Context） And printed in the log , such , You can extract any given from the log aggregator Trace perhaps Span All the logs of .

Next , Focus on the explanation of [appname,traceId,spanId,exportable] The meaning of each part ：

• appname： The name of the application that logs , namely spring.application.name Value ;
• traceId：Sleuth Unique data generated for a request link ID, One Trace Can contain more than one Span;
• spanId： The basic unit of work of the request link , Represents the occurrence of a specific operation , for example ： Send a Http request ;
• exportable： Do you need to export logs to Zipkin;

Sleuth It provides an abstraction of common distributed link tracking data models ：Trace、Span、Annotation And key value Annotation.Spring-Cloud-Sleuth Although based on htrace, But with Zipkin（dapper） Also compatible .

Sleuth Record time information to help with delay analysis . By using sleuth, You can find out the cause of the delay in your application .

When spring-cloud-sleuth-zipkin Included in classpath In the middle of the day , Generate and collect applications zipkin Compatible tracking records . By default , Will pass HTTP Send them to the local host （ port 9411） Upper Zipkin The server . You can set spring.zipkin.baseurl To configure the address of the service .

If you rely on spring-rabbit, Then the application will send the tracking record to Rabbit MQ agent , instead of HTTP.

If you rely on spring-kafka, And set up spring.zipkin.sender.type:kafka, Then the application will send the tracking record to Kafka Agency, not HTTP.

Be careful

• If you're using Zipkin, Please set spring.sleuth.sampler.probability To configure the export Span Probability （ The default value is ：0.1, namely 10%）. otherwise , You may think that Sleuth It doesn't work , Because it omits some Span.

• If you're using SLF4J,Trace and Span The tracking record of will be recorded to by default MDC, So users of the log can immediately see . But if you use other logging systems , You also need to set the print format of the log to see the same results ：

logging.pattern.level = %5p [${spring.zipkin.service.name:${spring.application.name:-}},%X{X-B3-TraceId:-},%X{X-B3-SpanId:-},%X{X-Span-Export:-}]

View the results after the request is sent ：

Reference blog ：https://blog.csdn.net/pengjunlee/article/details/87797969