前言-为什么要添加id?
1.新建项目
<dependency><groupId>org.springframework.boot</groupId><artifactId>spring-boot-starter-web</artifactId>
</dependency>
<!--lombok-->
<dependency><groupId>org.projectlombok</groupId><artifactId>lombok</artifactId><optional>true</optional>
</dependency>
除了spring的web依赖包外,添加一个lombok,方便编码。
2.编写controller方法
@RestController
public class TestController {@GetMapping("hello")public String hello(){ log.info("this is hello 1");log.info("this is hello 2");log.info("this is hello 3");return "hello";}}
这个时候一个简单的rest接口已经完成了,我们可以先看看日志的效果。
2019-10-23 10:45:10.917 INFO 1146 --- [nio-8080-exec-1] c.e.demo.controller.TestController : this is hello 1
2019-10-23 10:45:10.917 INFO 1146 --- [nio-8080-exec-1] c.e.demo.controller.TestController : this is hello 2
2019-10-23 10:45:10.917 INFO 1146 --- [nio-8080-exec-1] c.e.demo.controller.TestController : this is hello 3
从这个日志中,我们根本无法区分单独的请求。如果同一时间有2到3个请求过来的话,那么你还能分得清哪个对哪个么?所以这就是今天要改造的地方。
改造思路
其实要改造的话其实很简单,我们可以在每个controller入口处,生成唯一的uuid,并传递下去。这样的话缺点就是对代码干扰太大,每个方法都要多加一个参数。
那么我们能不能把这个参数存在一个统一的地方,需要打印日志的时候,直接去取呢?大家应该可以想到了,用ThreadLocal类。其实到这儿思路已经对了,不过日志框架也想到了这个问题,他们已经封装好了现成的功能,就是日志框架中的MDC
1.首先将日志的id添加进MDC中
@Component
public class TraceIdInterceptor extends HandlerInterceptorAdapter {@Overridepublic boolean preHandle(HttpServletRequest request, HttpServletResponse response, Object handler) throws Exception {String traceId = getTraceId(request); MDC.put("traceId", traceId);//将traceId添加进响应头response.addHeader("traceId",traceId);return true;}private String getTraceId(HttpServletRequest request){return String.format("%s - %s",request.getRequestURI(), UUID.randomUUID());}
}
@Component
public class GlobalWebMvcConfigurer implements WebMvcConfigurer {@Autowiredprivate TraceIdInterceptor traceIdInterceptor;@Override/*traceId 拦截器需要在最开始执行*/public void addInterceptors(InterceptorRegistry registry) {registry.addInterceptor(traceIdInterceptor).order(0);}
}
我们使用一下Spring的拦截器功能。在请求开始之前,将请求id添加进MDC。
2.修改日志的配置文件
新建一个logback-spring.xml文件添加如下内容
<?xml version="1.0" encoding="UTF-8"?>
<configuration><property name="LOG_PATTERN"value="%d{yyyy-MM-dd} %d{HH:mm:ss.SSS} [%-5level] [%X{traceId}] [%thread] %logger{36} %F.%L %msg%n"></property><appender name="STDOUT" class="ch.qos.logback.core.ConsoleAppender"><encoder><pattern>${LOG_PATTERN}</pattern></encoder><!-- 控制台打印INFO及以上级别的日志 --><filter class="ch.qos.logback.classic.filter.ThresholdFilter"><level>INFO</level></filter></appender><root><appender-ref ref="STDOUT"/></root>
</configuration>
在日志的格式LOG_PATTERN中,添加%X{traceId} ,这样日志在打印的时候便会去MDC中取出traceid,放在这儿。现在我们可以看看效果。
2019-10-23 11:02:46.649 [INFO ] [/hello - 79ec561a-ef5e-4dc5-91cc-b2143fa3dbd3] [http-nio-8080-exec-1] c.e.demo.controller.TestController TestController.java.18 this is hello 1
2019-10-23 11:02:46.649 [INFO ] [/hello - 79ec561a-ef5e-4dc5-91cc-b2143fa3dbd3] [http-nio-8080-exec-1] c.e.demo.controller.TestController TestController.java.19 this is hello 2
2019-10-23 11:02:46.650 [INFO ] [/hello - 79ec561a-ef5e-4dc5-91cc-b2143fa3dbd3] [http-nio-8080-exec-1] c.e.demo.controller.TestController TestController.java.20 this is hello 3
2019-10-23 11:02:47.612 [INFO ] [/hello - 2ac7450f-40f5-441c-8e10-9b937c006484] [http-nio-8080-exec-2] c.e.demo.controller.TestController TestController.java.18 this is hello 1
2019-10-23 11:02:47.612 [INFO ] [/hello - 2ac7450f-40f5-441c-8e10-9b937c006484] [http-nio-8080-exec-2] c.e.demo.controller.TestController TestController.java.19 this is hello 2
2019-10-23 11:02:47.612 [INFO ] [/hello - 2ac7450f-40f5-441c-8e10-9b937c006484] [http-nio-8080-exec-2] c.e.demo.controller.TestController TestController.java.20 this is hello 3
可以看到,两个请求通过traceId可以很清楚的区分开了。这样我们在排查问题的时候,可以通过响应头里面的traceId,直接查找到相关日志,非常方便。
进阶版traceId
之前说过traceId的实现思路是通过ThreadLocal来实现的。使用ThreadLocal有一个前提就是一个请求进来始终是一个线程在处理。如果用到spring中的异步方法,traceId就会失效了。
我们可以做个实验
//编写一个service类
public class Service { @Async public void run(){ log.info("this is service run!");}}//在hello方法中调用service的run方法@GetMapping("hello")
public String hello(){ log.info("this is hello 1");log.info("this is hello 2");log.info("this is hello 3");service.run(); return "hello";
}
可以看的打印出来的日志
2019-10-23 11:12:23.265 [INFO ] [/hello - 96b31833-e8e6-46c5-8459-d423309d1488] [http-nio-8080-exec-1] c.e.demo.controller.TestController TestController.java.22 this is hello 1
2019-10-23 11:12:23.266 [INFO ] [/hello - 96b31833-e8e6-46c5-8459-d423309d1488] [http-nio-8080-exec-1] c.e.demo.controller.TestController TestController.java.23 this is hello 2
2019-10-23 11:12:23.267 [INFO ] [/hello - 96b31833-e8e6-46c5-8459-d423309d1488] [http-nio-8080-exec-1] c.e.demo.controller.TestController TestController.java.24 this is hello 3
2019-10-23 11:12:23.278 [INFO ] [] [task-1] com.example.demo.controller.Service Service.java.17 this is service run!
其实也很简单,只要对异步线程池跑的对象稍作封装即可。
@EnableAsync
@Configuration
public class AsyncConfiguration {@Autowiredprivate AppConfig config;@Bean("async")public Executor taskExecutor() {ThreadPoolTaskExecutor executor = new ThreadPoolTaskExecutor();//设置核心线程数量executor.setCorePoolSize(config.getAsync().getCorePoolSize());//设置最大线程数量executor.setMaxPoolSize(config.getAsync().getMaxPoolSize());//设置队列最大长度executor.setQueueCapacity(config.getAsync().getQueueCapacity());//设置线程空闲时间executor.setKeepAliveSeconds(config.getAsync().getKeepLiveSeconds());//设置线程前缀executor.setThreadNamePrefix("async-");//设置拒绝策略executor.setRejectedExecutionHandler(new ThreadPoolExecutor.CallerRunsPolicy());//设置线程装饰器executor.setTaskDecorator(runnable -> ThreadMdcUtils.wrapAsync(runnable, MDC.getCopyOfContextMap()));return executor;}
}
public class ThreadMdcUtils {public static Runnable wrapAsync(Runnable task, Map<String,String> context){return () -> {if(context==null){MDC.clear();}else {MDC.setContextMap(context);}if(MDC.get(Constant.TraceId)==null){MDC.put(Constant.TraceId,UUIDUtils.randomUUID());}try {task.run();}finally {MDC.clear();}};}
}
这样的话,即便调用异步方法,也能获得统一的日志id。