业务复杂的微服务架构中,往往服务之间的调用关系比较难梳理,一次http请求中,可能涉及到多个服务的调用(eg: service A -> service B -> service C...),如果想分析各服务间的调用关系,以及各服务的响应耗时,找出有性能瓶颈的服务,这时zipkin就派上用场,它是Twitter公司开源的一个tracing系统,官网地址为: http://zipkin.io/ , spring cloud可以跟它无疑集成。
使用步骤:
一、微服务方
1.1 添加依赖jar包
compile ‘org.springframework.cloud:spring-cloud-starter-bus-kafka‘ compile ‘org.springframework.cloud:spring-cloud-starter-sleuth‘ compile ‘org.springframework.cloud:spring-cloud-sleuth-stream‘
注:为了实现tracing数据埋点与采集的解耦,spring cloud引入了message bus(消息总线)的概念,微服务无需关心tracing系统在哪,长什么样,只要向bus总线上扔消息就行,所以引入了bus-kafka以及sleuth-stream。
1.2 application.yml配置
spring: ... cloud: bus: enabled: true stream: default-binder: kafka kafka: binder: brokers: 10.0.1.2,10.0.1.3,10.0.1.4 //kafaka的服务器集群列表 zkNodes: 10.0.1.5,10.0.1.6,10.0.1.7 //zk的服务器集群列表 defaultZkPort: 2181 //zk的端口 defaultBrokerPort: 9092 //kafka的broker端口 ... sleuth: sampler: percentage: 0.2 //采样率 0.2为20%
上面2项配置好就行了,代码不用任何修改,真正的代码零侵入
二、zipkin-server
zipkin从kafka上接收过来数据后,有4种保存方式:in-memory(保存在内存中)、mysql、cassandra、elasticsearch
个人开发调试的话,推荐用in-memory模式,其它环境不要使用!(注:因为随着收集的数据越来越多,都放在内存中 很容易造成OOM)
2.1 mysql 存储
2.1.1 主要jar包依赖
dependencies { ... 关键是下面几个 compile ‘org.springframework.cloud:spring-cloud-starter-sleuth‘ compile ‘org.springframework.cloud:spring-cloud-sleuth-zipkin-stream‘ compile ‘org.springframework.cloud:spring-cloud-starter-bus-kafka‘ compile ‘io.zipkin.java:zipkin-server‘ compile ‘io.zipkin.java:zipkin-autoconfigure-ui‘ compile ‘io.zipkin.java:zipkin-autoconfigure-storage-mysql‘ #mysql的存储 ... 下面几个是spring-boot/cloud的常规项 compile ‘org.springframework.boot:spring-boot-starter-actuator‘ compile ‘org.springframework.boot:spring-boot-starter-web‘ compile ‘org.springframework.boot:spring-boot-starter-security‘ compile ‘log4j:log4j:1.2.17‘ //zipkin的storage jar包,依赖低版本的log4j compile ‘org.apache.logging.log4j:log4j-slf4j-impl:2.8.2‘ compile ‘mysql:mysql-connector-java:6.0.5‘ }
2.1.2 application.yml配置
spring: application: name: zipkin-server datasource: //指定mysql数据源 schema: classpath:/mysql.sql url: jdbc:mysql://192.168.1.2:3306/zipkin?autoReconnect=true&useSSL=false username: root password: *** driver-class-name: com.mysql.cj.jdbc.Driver initialize: true continue-on-error: true sleuth: enabled: false cloud: bus: enabled: true ... stream: default-binder: kafka kafka: binder: brokers: ${kafka.brokers} zkNodes: ${kafka.zkNodes} defaultZkPort: ${kafka.zkPort} defaultBrokerPort: ${kafka.brokerPort} zipkin: storage: type: mysql //配置成mysql存储
2.1.3 main入口代码
@SpringBootApplication(exclude = { MybatisAutoConfiguration.class, RedisAutoConfiguration.class, RedisRepositoriesAutoConfiguration.class}) @EnableZipkinStreamServer public class ZipkinServer { public static void main(String[] args) { SpringApplication.run(ZipkinServer.class, args); } }
注:如果你的项目中依赖了redis,mybatis等其它包,可以参考上面的写法,排除掉这些自动配置,否则的话,不用加那一堆exclude。
2.2 cassandra
2.2.1 依赖jar包
注:cassandra和elasticsearch下,可能会遇到zipkin中的dependencies面板无数据,详情见github上的讨论:https://github.com/openzipkin/zipkin-dependencies/issues/22
compile ‘org.springframework.boot:spring-boot-starter-data-cassandra‘ compile(‘io.zipkin.java:zipkin-autoconfigure-storage-cassandra3:1.29.3‘) { exclude group: "com.datastax.cassandra", module: "cassandra-driver-core" } compile ‘com.datastax.cassandra:cassandra-driver-core:3.1.1‘ compile ‘com.datastax.cassandra:cassandra-driver-mapping:3.1.1‘
2.2.2 application.yml
spring: data: cassandra: contact-points: localhost port: 9042 keyspace-name: zipkin3 ... zipkin: storage: type: cassandra3
2.3 elasticsearch
2.3.1 依赖jar包
compile ‘io.zipkin.dependencies:zipkin-dependencies-elasticsearch:1.7.2‘ compile ‘io.zipkin.java:zipkin-autoconfigure-storage-elasticsearch-http:1.29.2‘
2.3.2 application.yml
zipkin: storage: type: elasticsearch elasticsearch: cluster: elasticsearch hosts: http://localhost:9200 index: zipkin index-shards: 5 index-replicas: 1