全链路spring cloud sleuth+zipkin

http://blog.csdn.net/qq_15138455/article/details/72956232

 

版权声明：@入江之鲸

1、About ZipKin

please google

2、 Demo Scene

3、 Result Display

4、Prepare

一、soft version

kafka：2.10-0.10.2.0
zokeeper：3.4.10
elasticsearch：5.2.2
jdk：1.8
spring boot：1.5.3.RELEASE
sprign cloud：Dalston.RELEASE
rabbit mq：3.6.9

二、install

kafka+zookeeper
elasticsearch
rabbit mq
mysql

三、create four spring cloud project

web-api、user-api、order-api、zipkin
ps:

why i will create zipkin project use spring boot by myself not use  zipkin.jar from http://zipkin.io/,actually，zipkin.jar is a spring boot project，check it's dependency lib you will find it din't use spring-cloud-sleuth-stream,but i will send trace info to kafka for zipkin server collector ,so i must use spring-cloud-sleuth-stream in my service

and the message send to kafka is a  sleuth.span object and use kafka default serialized，but zipkin.jar only receive zipkin.span and json or thrift encode,so it‘s not matching,That's the reason i create zipkin server

but if you use rabbit mq，that's no problem.

四、configuration

4.一、the service web-api、user-api、order-api config part like:
pom.xml

<dependency>
    <groupId>org.springframework.cloud</groupId>
    <artifactId>spring-cloud-sleuth-zipkin-stream</artifactId>
</dependency>
<dependency>
    <groupId>org.springframework.cloud</groupId>
    <artifactId>spring-cloud-starter-stream-kafka</artifactId>
</dependency>

application.properties

spring.sleuth.sampler.percentage=1.0
spring.cloud.stream.kafka.binder.brokers=10.20.1.11:9092,10.20.1.12:9092
spring.cloud.stream.kafka.binder.zkNodes=10.20.1.11:2181,10.20.1.12:2181

4.二、the zipkinconfig part like:

pom.xml

 

<!-- the  first one dependency below，In principle, there is no need，beause sleuth-zipkin-stream 1.5.3 will  Introduce zipkin version1.19 Automaticly，but 1.19 only support  elasticsearch version 2.X -->
<dependency>
    <groupId>io.zipkin.java</groupId>
    <artifactId>zipkin</artifactId>
    <version>1.24.0</version>
</dependency>
<dependency>
    <groupId>org.springframework.cloud</groupId>
    <artifactId>spring-cloud-sleuth-zipkin-stream</artifactId>
</dependency>
<dependency>
    <groupId>org.springframework.cloud</groupId>
    <artifactId>spring-cloud-starter-stream-kafka</artifactId>
</dependency>
<dependency>
    <groupId>io.zipkin.java</groupId>
    <artifactId>zipkin-autoconfigure-storage-elasticsearch-http</artifactId>
    <version>1.24.0</version>
    <optional>true</optional>
</dependency>

application.properties

 

 

#kafka config
spring.sleuth.enabled=false
spring.sleuth.sampler.percentage=1.0
spring.cloud.stream.kafka.binder.brokers=10.20.1.11:9092,10.20.1.12:9092
spring.cloud.stream.kafka.binder.zkNodes=10.20.1.11:2181,10.20.1.12:2181
#elasticsearch config
zipkin.storage.type=elasticsearch
zipkin.storage.elasticsearch.hosts=10.20.1.11:9200,10.20.1.12:9200
zipkin.storage.elasticsearch.cluster=elasticsearch
zipkin.storage.elasticsearch.index=zipkin
zipkin.storage.elasticsearch.index-shards=5
zipkin.storage.elasticsearch.index-replicas=1

ZipKin Server Startup class configuration

 

 

<span style="font-size:14px;">@SpringBootApplication
//@EnableZipkinServer //this is used by interface receive trace info
@EnableZipkinStreamServer //can be used kafka,rabbit
public class ZkingApplication {
    public static void main(String[] args) {
        SpringApplication.run(ZkingApplication.class, args);
    }
}</span>

 

5、Demo DownLoad

click me

by the way,spring cloud is a pretty boy,i like its combination of terseness and elegance

 

6、补充

若是kafka没有启动，spring boot会启动失败，这个异常处理设计的真是缺德

/** * 一、修改背景  * 因kafka节点没有启动 在spring boot启动时初始化outputBindingLifecycle、inputBindingLifecycle * 两个bean时候链接kafka失败，向外抛出了异常直到EmbeddedWebApplicationContext类 * 捕获处理，处理方式为：stopAndReleaseEmbeddedServletContainer()致使整个应用中止启动  * 二、修改方案 * 干预上面两个bean的初始化，在链接kafka异常时，将异常处理掉，不向上层抛出 * 三、修改步骤 * 3.一、使用自定义MyBindingLifecycle的bean将BindingServiceConfiguration中的两个bean初始化替换掉 * 3.二、在自定bean中启动线程MyBindingThread来控制两个bean的初始化 * 四、解决启动问题以后，实际上kafka仍是没有链接的，此时向kafka发送span时会失败，默认的处理方案是捕获到异常以后使用 *   handleError处理，再次发送新的span，这就致使循环发送 *   参见：ErrorHandlingTaskExecutor中的execute方法 *   catch (Throwable t)  *   { *        ErrorHandlingTaskExecutor.this.errorHandler.handleError(t); *     } * 五、解决方案 * 重写ErrorHandler的handleError方法 * 六、跟踪代码发现 * 跟踪发现ErrorHandler对线对象是在SourcePollingChannelAdapterFactoryBean初始化时候设置的 * spca.setErrorHandler(this.pollerMetadata.getErrorHandler()); * 进一步发现是在pollerMetadata对象中，因此须要在pollerMetadata对象初始化时候作修改 * 七、修改步骤 * 自定义MyPollerMetadata且须要@Configuration，重写handleError方法以下 * @author zhangdingxin、yangxi */