Spring cloud系列十四 分布式链路监控Spring Cloud Sleuth

1. 概述

Spring Cloud Sleuth实现对Spring cloud 分布式链路监控 
本文介绍了和Sleuth相关的内容,主要内容如下:

  • Spring Cloud Sleuth中的重要术语和意义:Span、Trance、Annotation
  • Zipkin中图形化展示分布式链接监控数据并说明字段意义
  • Spring Cloud集成Sleuth + Zipkin 的代码demo: Sleuth集成Zipkin, Zipkin数据持久化等

2. 术语

Span 
Span是基本的工作单元。Span包括一个64位的唯一ID,一个64位trace码,描述信息,时间戳事件,key-value 注解(tags),span处理者的ID(通常为IP)。 
最开始的初始Span称为根span,此span中span id和 trace id值相同。

Trance 
包含一系列的span,它们组成了一个树型结构

Annotation 
用于及时记录存在的事件。常用的Annotation如下

  • cs - Client Sent:客户端发送一个请求,表示span的开始
  • sr - Server Received:服务端接收请求并开始处理它。(sr-cs)等于网络的延迟
  • ss - Server Sent:服务端处理请求完成,开始返回结束给服务端。(ss-sr)表示服务端处理请求的时间
  • cr - Client Received:客户端完成接受返回结果,此时span结束。(cr-sr)表示客户端接收服务端数据的时间

如果一个服务的调用关系如下: 
这里写图片描述

那么此时将Span和Trace在一个系统中使用Zipkin注解的过程图形化: 
这里写图片描述

每个颜色的表明一个span(总计7个spans,从A到G),每个span有类似的信息

Trace Id = X
Span Id = D
Client Sent
  • 1
  • 2
  • 3

此span表示span的Trance Id是X,Span Id是D,同时它发送一个Client Sent事件

spans 的parent/child关系图形化如下: 
这里写图片描述

3. 在Zipkin中图形化展示分布式链接监控数据

3.1 spans在zipkin界面的信息解读

在Zipkin中展示了上图的跟踪信息,红框里是对上图调用span的跟踪 
这里写图片描述

但是你点击这个trace时,我们只看到4个span 
这里写图片描述

为什么两个界面显示的span数量不同,一个是7,一个4.

  • 1 个 spans:来自servier1的接口http:/start 被调用,分别是Server Received (SR) 和 Server Sent (SS) annotations.
  • 2 个 spans:来自 service1 调用service2 的 http:/foo 接口。service1 端有两个span,分别为 Client Sent (CS) 和 Client Received (CR) annotations。service2 端也有两个span,分别为Server Received (SR) 和Server Sent (SS) 。物理上他有2个span,但是从逻辑上说这个他们组成一个RPC调用的span。
  • 2个span:来自 service2 调用 service3 的 http:/bar 接口,service2 端有两个span,分别为Client Sent (CS) 和 Client Received (CR) annotations。service3 端也有两个span,分别为Server Received (SR) 和Server Sent (SS) 。物理上他有2个span,但是从逻辑上说它们都是同一个RPC调用的span。
  • 2个span:来自 service2 调用 service4 的 http:/bar 接口,service2 端有两个span,分别为Client Sent (CS) 和 Client Received (CR) annotations。service4 端也有两个span,分别为Server Received (SR) and Server Sent (SS) 。物理上他有2个span,但是从逻辑上说这个它们都是同一个RPC调用的span。

所以我们计算物理spans有7个:

  • 1个来自 http:/start 被请求
  • 2个来自 service1调用service2
  • 2个来自 service2调用service3
  • 2个来自 service2调用service4.

从逻辑上说,我们只看到4个span:

  • 1个来自 service1 的接口http:/start 被请求
  • 3个来自 服务之前的RCP接口调用

3.2. Zipkin可视化错误

如果调用链路中发生接口调用失败,zipkin会默认使用红色展示信息,如下图: 
这里写图片描述
点击红色的span,可以看到详细的失败信息: 
这里写图片描述

4. Spring Cloud集成Sleuth + Zipkin

本节演示在Spring Cloud中集成Sleuth并将链路监控数据传送到Zipkin,使用Zipkin展示数据,并配置mysql进行数据持久化

4.1. 相关工程

cloud-registration-center 
注册中心

cloud-service-zipkin 
提供测试服务接口,对外提供有两个接口: 
- 调用成功后,马上返回一个结果: http://127.0.0.1:17602//zipkin/simple 
- 调用成功后,服务sleep 5s后再返回: http://127.0.0.1:17602//zipkin/sleep

cloud-consumer-zipkin 
消费服务 
此服务会通过Feign调用cloud-service-zipkin里提供的两个接口(/zipkin/sleep和/zipkin/simple),我们访问如下URL会调用的对应的接口: 
http://127.0.0.1:17603//zipkin/simple 
http://127.0.0.1:17603//zipkin/sleep

以上3个服务的代码比较简单,这里代码略,可以自己看github里的代码。下文中只列出和Sleuth和Zipkin相关的配置。

cloud-dashboard-zipkin 
配置Zipkin服务,提供可视化链路监控 
Zipkin首页:http://127.0.0.1:17601/zipkin/

4.2. 在工程中使用sleuth+zipkin+http配置

在cloud-service-zipkin和cloud-consumer-zipkin中启动sleuth,sleuth会收集spans信息,并使用http异步地将spans 信息发送到Zipkin,然后Zipkin展示信息

cloud-service-zipkin和cloud-consumer-zipkin配置Sleuth+Zipkin 
2个工程中为了集成Sleuth和Zipkin,需要在普通工程基础上增加如下配置:

  • pom.xml增加sleuth和zipkin相关的jar
<!-- sleuth配置 -->
<dependency>
    <groupId>org.springframework.cloud</groupId>
    <artifactId>spring-cloud-starter-sleuth</artifactId>
</dependency>
<!-- 引入zipkin  -->
<dependency>
    <groupId>org.springframework.cloud</groupId>
    <artifactId>spring-cloud-starter-zipkin</artifactId>
</dependency>
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • application-zipkin.yml 
    Zipkin+Sleuth配置参数: 
    • spring.zipkin.baseUrl:指定cloud-dashboard-zipkin的服务地址,本例子中使用真实的IP。在新的版本spring-cloud-sleuth-core-1.3.0.RELEASE中,可以实现通过服务名称进行访问
    • spring.sleuth.sampler.percentage:设置采样率,为了测试设置100%采集
spring:
   zipkin:
     enabled: true
     # zipkkin dashboard的地址:通过真实IP地址访问
     baseUrl: http://localhost:17601/
     # 通过cloud-dashboard-zipkin注册到注册中心的服务名称访问,本版本(spring-cloud-sleuth-core-1.2.5.RELEASE)不支持,需要从spring-cloud-sleuth-core-1.3.0.RELEASE开始支持这个功能
     # 配置如下:
     # baseUrl: http://cloud-dashboard-zipkin/
   sleuth:
     sampler:
       #  默认值为0.1f,现在为了测试设置100%采集
       percentage: 1
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12

cloud-dashboard-zipkin 
配置zipkin服务

  • pom.xml
<!-- spring-cloud-starter-zipkin -->
<dependency>
    <groupId>org.springframework.cloud</groupId>
    <artifactId>spring-cloud-starter-zipkin</artifactId>
</dependency>
<!-- zipkin 界面-->
<dependency>
    <groupId>io.zipkin.java</groupId>
    <artifactId>zipkin-autoconfigure-ui</artifactId>
</dependency>
<!-- zipkin 服务类-->
<dependency>
    <groupId>io.zipkin.java</groupId>
    <artifactId>zipkin-server</artifactId>
</dependency>
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 配置参数 
    bootstrap-zipkin-http.yml
# port
server:
  port: 17601

spring:
  application:
    # 本服务注册到注册到服务器的名称, 这个名称就是后面调用服务时的服务标识符
    name: cloud-dashboard-zipkin
eureka:
  client:
    serviceUrl:
      # 服务器注册/获取服务器的zone
      defaultZone: http://127.0.0.1:10761/eureka/
      # defaultZone: http://192.168.21.3:10761/eureka/,http://192.168.21.4:10761/eureka/
  instance:
    prefer-ip-address: true
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • application-zipkin-http.yml 
    关闭本工程的推送到zipkin服务的功能
spring:
  zipkin:
    enabled: false
  • 1
  • 2
  • 3

启动类: 
@EnableZipkinServer:注解此类为Zipkin服务

@EnableEurekaClient // 配置本应用将使用服务注册和服务发现
@SpringBootApplication
@EnableZipkinServer // 启动Zipkin服务
public class ZipkinDashboardCloudApplication {

    public static void main(String[] args) {
        args = new String[1];
        args[0] = "--spring.profiles.active=zipkin-http";
        SpringApplication.run(ZipkinDashboardCloudApplication.class, args);
    }
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10

4.3. 测试

启动zipkin服务,提供可视化链路监控 
Zipkin访问地址: http://127.0.0.1:17601/zipkin/

我们演示链路跟踪,访问以下两个请求: 
http://127.0.0.1:17603/zipkin/simple :正常方法 
http://127.0.0.1:17603/zipkin/sleep :访问超时

进入zipkin,访问成功为蓝色,失败为红部。同时显示各个服务花费的时间 
这里写图片描述

点击蓝色记录,进入详细界面 
可知整个接口花费约19ms,cloud-zipkin-service的业务执行15ms,cloud-zipkin-consumer访问cloud-zipkin-service的 网络延迟为2ms,返回结果给cloud-zipkin-service的网络延迟是4s 
这里写图片描述

点击cloud-zipkin-service得到更精确的数据

这里写图片描述

4.4. Zipkin数据持久化

默认情况,zipkin存储记录到内存,如果服务重启,则所有记录丢失。为了保证持久性,zipkin支持Mysql、Elasticsearch、Cassandra存储。我们演示为zipkin配置MYSQL数据库,其它两种配置类似,本文略 
pom.xml为cloud-dashboard-zipkin引入新jar

 <!-- zipkin 存储到数据库需要引入此类 -->
<dependency>
    <groupId>io.zipkin.java</groupId>
    <artifactId>zipkin-autoconfigure-storage-mysql</artifactId>
</dependency>

<!--保存到数据库需要如下依赖-->
<dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-jdbc</artifactId>
</dependency>
<dependency>
    <groupId>mysql</groupId>
    <artifactId>mysql-connector-java</artifactId>
    <version>6.0.6</version>
</dependency>
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17

配置参数: 
application-zipkin-http.yml: 
配置启动时会根据classpath:/mysql.sql初始化数据库

spring:
  zipkin:
    enabled: false
  # 配置mysql
  datasource:
    schema: classpath:/mysql.sql
    # url: jdbc:mysql://127.0.0.1/test
    url: jdbc:mysql://127.0.0.1:3306/test?zeroDateTimeBehavior=convertToNull&serverTimezone=GMT%2b8
    username: root
    password: root
# Switch this on to create the schema on startup:
    initialize: true
    continueOnError: true
  sleuth:
    enabled: false
zipkin:
  storage:
    type: mysql
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18

重启服务,服务成功后,查看数据库,自动生成数据库 
这里写图片描述

如此数据库配置完毕,所有的spans信息存储到数据库中,重启服务,也不会丢失记录

4.5 在工程中使用sleuth+zipkin+ Spring Cloud Stream配置

本版本(spring-cloud-sleuth-core-1.2.5.RELEASE)支持集成spring-cloud-sleuth-stream,但是在新版本从spring-cloud-sleuth-core-1.3.0.RELEASE开始不支持spring-cloud-sleuth-stream。推荐直接使用通过集成RabbitMQ 或 Kafka。关于集成RabbitMQ 或 Kafka,关键是引入spring-rabbit 或 spring-kafka依赖包。默认的目的名称为zipkin. 
对于这个的使用demo。这里暂时略

5. 代码

以上的详细的代码见下面 
github代码,请尽量使用tag v0.11,不要使用master,因为我不能保证master代码一直不变

猜你喜欢

转载自blog.csdn.net/hanghangaidoudou/article/details/80926858
今日推荐