背景
我们希望springboot默认的日志框架slf4j,在打印日志时,每行数据都有一个请求ID,这样就很方面追踪日志。并且可以在linux中用将所有该请求ID的日志写在单独的文件里(grep 你的RequestId xxx.log > 该请求ID的所有日志.log
)
(其实是有一些链路追踪的框架的,比如zipkin等,本文写这个是一个不想用这些框架的情况下的方案)
办法
slf4j有个MDC的类,是ThreadLocal的实现,保存在这里的变量会绑定到某个请求线程,于是在该请求的线程里的日志代码都可以使用设入的变量。例如:
MDC中设入了名为 RequestId 变量,则日志格式可以配置
%d{yyyy-MM-dd HH:mm:ss.SSS} [%X{RequestId}] [%thread] %-5level %logger{36} - %msg%n
其中 %X{参数名} 是引用的语法
详细的步骤
我们必须在springboot项目里定义一个拦截器,可以是javax.servlet.Filter,也可以是spring的AOP,反正就是要拦截所有请求并在MDC中设入变量。
设入MDC的变量最好在finally中进行clear,以避免不必要的麻烦。注意要用clear()方法,不要一个个key去remove,避免以后添加了新的变量后忘记remove
代码如下:
package com.wyf.test.requestidlog.mdc;
import org.slf4j.MDC;
import org.springframework.stereotype.Component;
import javax.servlet.*;
import javax.servlet.http.HttpServletRequest;
import javax.servlet.http.HttpServletResponse;
import java.io.IOException;
import java.util.UUID;
/**
* 在日志上下文增加参数,在日志中可以打印该参数,如RequestId,则日志配置:%d{yyyy-MM-dd HH:mm:ss.SSS} [%X{RequestId}] [%thread] %-5level %logger{36} - %msg%n
* 则 RequestId 会打印在中括号里,%X{参数名} 是格式
*
* 该bean自动被注册,默认的mapping是/*,参考文档
* https://docs.spring.io/spring-boot/docs/2.2.1.RELEASE/reference/htmlsingle/#howto-add-a-servlet-filter-or-listener-as-spring-bean (说明了只要是Filter/Servlet/*Listener的bean自动配置)
* https://docs.spring.io/spring-boot/docs/2.2.1.RELEASE/reference/htmlsingle/#boot-features-embedded-container-servlets-filters-listeners-beans (说明了Filters默认map to /*)
*
* 如果需要复杂点的配置,例如多个Filter的顺序、mapping路径的指定,可以在@Configuration类里使用@Bean来产生所需的bean,产生的过程即可指定顺序和mapping路径
*
* @author Stone
* @version V1.0.0
* @date 2019/12/10
*/
@Component
public class MDCFilter implements Filter {
/**
* 设置响应头,响应头名字(惯例是大写)
*/
private static final String REQUEST_ID_HEADER = "RequestId";
/**
* 设置到 MDC 里的key(惯例使用小驼峰,为了统一这里也用大写)
*/
private static final String REQUEST_ID = "RequestId";
@Override
public void init(FilterConfig filterConfig) throws ServletException {
}
@Override
public void doFilter(ServletRequest request, ServletResponse response, FilterChain chain) throws IOException, ServletException {
HttpServletRequest httpRequest = (HttpServletRequest) request;
HttpServletResponse httpResponse = (HttpServletResponse) response;
String requestId = httpRequest.getHeader(REQUEST_ID_HEADER);
if (requestId == null) {
requestId = UUID.randomUUID().toString();
}
// 还有更加准确的方法
String ip = getRequestSourceIp(httpRequest);
MDC.put(REQUEST_ID, requestId);
MDC.put("IP", ip);
httpResponse.setHeader(REQUEST_ID_HEADER, requestId);
try {
chain.doFilter(request, response);
} finally {
MDC.clear();
}
}
@Override
public void destroy() {
}
/**
* 获取请求来源的IP,可以适应前置部署有Nginx等反向代理软件等的情况. HTTP_CLIENT_IP 无法伪造,所以放在第一个
* http://blog.csdn.net/songylwq/article/details/7701718
*
* @param request
* @return
* @author Stone
*/
public static String getRequestSourceIp(HttpServletRequest request) {
String ip = request.getHeader("HTTP_CLIENT_IP");
if (ip == null || ip.length() == 0 || "unknown".equalsIgnoreCase(ip)) {
ip = request.getHeader("HTTP_X_FORWARDED_FOR");
}
if (ip == null || ip.length() == 0 || "unknown".equalsIgnoreCase(ip)) {
ip = request.getHeader("HTTP_X_FORWARDED");
}
if (ip == null || ip.length() == 0 || "unknown".equalsIgnoreCase(ip)) {
ip = request.getHeader("HTTP_FORWARDED");
}
if (ip == null || ip.length() == 0 || "unknown".equalsIgnoreCase(ip)) {
ip = request.getHeader("X-FORWARDED-FOR");
}
if (ip == null || ip.length() == 0 || "unknown".equalsIgnoreCase(ip)) {
ip = request.getHeader("Proxy-Client-IP");
}
if (ip == null || ip.length() == 0 || "unknown".equalsIgnoreCase(ip)) {
ip = request.getHeader("WL-Proxy-Client-IP");
}
if (ip == null || ip.length() == 0 || "unknown".equalsIgnoreCase(ip)) {
ip = request.getRemoteAddr();
}
return ip;
}
}
lockback-spring.xml配置如下:
<?xml version="1.0" encoding="UTF-8" ?>
<configuration>
<appender name="STDOUT" class="ch.qos.logback.core.ConsoleAppender">
<target>System.out</target>
<encoder>
<pattern>%d{yyyy-MM-dd HH:mm:ss.SSS} [%X{RequestId}] [%X{IP}] [%thread] %-5level %logger{36} - %msg%n</pattern>
</encoder>
</appender>
<root level="INFO">
<appender-ref ref="STDOUT"/>
</root>
</configuration>
上述代码配置了@Component,会被spring容器管理,其是javax.servlet.Filter的实现,所以默认拦截器能生效,拦截所有/* 的请求。
运行效果
2019-12-11 17:18:22.451 [fd4053e3-2a79-4277-a16b-2390f5b62ce3] [0:0:0:0:0:0:0:1] [http-nio-8080-exec-1] INFO c.w.t.r.controller.TestController - === log ===
补充
思考:
- 对于junit或main方法里打印的日志,能获取到RequestId的变量吗?
获取不到,因为不是http(s)请求,所以不会进入拦截器,故未往MDC中设入值 - 父线程能打印日志,其子线程能获取到吗?
打印不来,因为MDC是ThreadLocal的实现,父线程绑定了变量但是子线程没有。
这时候必须由父线程传入,下面演示的代码使用了 “装饰者模式” 的设计模式,重新写了Runnable和Callable,使用这个替代Runnable和Callable就能让子线程获取到父线程里的变量
- MyRunnable
package com.wyf.test.requestidlog.mdc;
import org.slf4j.MDC;
import java.util.HashMap;
import java.util.Map;
/**
* 封装Runnable接口,装饰者模式
*
* @author Stone
* @version V1.0.0
* @date 2019/12/10
*/
public class MyRunnable implements Runnable {
private Runnable runnable;
private Map<String, String> map;
/**
* 传入runnable类,并获取主线程中的所有参数(调用该构造器是父线程,所以这里的MDC.getXxxx获取到绑定在父线程的变量)
* @param runnable
*/
public MyRunnable(Runnable runnable) {
this.runnable = runnable;
this.map = MDC.getCopyOfContextMap();
}
@Override
public void run() {
// 可能是null,必须判断,追踪进代码可见null时会发生NPE
if (map != null) {
MDC.setContextMap(map);
}
try {
runnable.run();
} finally {
// 即使未执行setContextMap,这里可以clear不出错
MDC.clear();
}
}
}
- MyCallable
package com.wyf.test.requestidlog.mdc;
import org.slf4j.MDC;
import java.util.Map;
import java.util.concurrent.Callable;
/**
* 封装MyCallable接口,装饰者模式
*
* @author Stone
* @version V1.0.0
* @date 2019/12/10
*/
public class MyCallable<V> implements Callable<V> {
private Callable<V> callable;
private Map<String, String> map;
public MyCallable(Callable<V> callable) {
this.callable = callable;
this.map = MDC.getCopyOfContextMap();
}
@Override
public V call() throws Exception {
// 可能是null,必须判断,追踪进代码可见null时会发生NPE
if (map != null) {
MDC.setContextMap(map);
}
try {
return callable.call();
} finally {
// 即使未执行setContextMap,这里可以clear不出错
MDC.clear();
}
}
}
- 使用 MyRunnable 和 MyCallable
3.1 TestController
package com.wyf.test.requestidlog.controller;
import com.wyf.test.requestidlog.mdc.MyRunnable;
import lombok.extern.slf4j.Slf4j;
import org.springframework.web.bind.annotation.GetMapping;
import org.springframework.web.bind.annotation.RestController;
import java.util.Date;
/**
* @author Stone
* @version V1.0.0
* @date 2019/12/10
*/
@RestController
@Slf4j
public class TestController {
/**
* 能打印出RequestId
*
* @return
*/
@GetMapping("/log")
public String log() {
log.info("=== log ===");
return "log," + new Date();
}
/**
* 主线程能打印RequestId,子线程不能
*
* @return
*/
@GetMapping("/log/sub")
public String logSubThread() {
log.info("=== main thread ===");
new Thread(new Runnable() {
@Override
public void run() {
log.info("=== sub thread ===");
}
}).start();
return "log," + new Date();
}
/**
* 父线程、子线程都能打印出来
*
* @return
*/
@GetMapping("/log/sub2")
public String logSubThreadWithRequestId() {
log.info("=== main thread ===");
new Thread(new MyRunnable(new Runnable() {
@Override
public void run() {
log.info("=== sub thread ===");
}
})).start();
return "log," + new Date();
}
/**
* 即使多层,只要使用MyRunnable即可
*
* @return
*/
@GetMapping("/log/sub3")
public String logSubThreadWithRequestId2() {
log.info("=== main thread ===");
new Thread(new MyRunnable(new Runnable() {
@Override
public void run() {
log.info("=== sub thread ===");
new Thread(new MyRunnable(new Runnable() {
@Override
public void run() {
log.info("=== sub sub thread ===");
}
})).start();
}
})).start();
return "log," + new Date();
}
/**
* 因为是main方法,并未加载springboot的配置,包括log的配置。所以是按照默认的log的打印格式打印出来的,即并不带上 %X{RequestId} 的配置,
* 况且本身这个方法非http(s)请求故未在拦截器中设入RequestId
*
* @param args
*/
public static void main(String[] args) {
log.info("=== 打印不出RequestId ====");
}
@GetMapping("/")
public String hello() {
return "hello," + new Date();
}
}
3.2 CallableController
package com.wyf.test.requestidlog.controller;
import com.wyf.test.requestidlog.mdc.MyCallable;
import lombok.extern.slf4j.Slf4j;
import org.springframework.web.bind.annotation.GetMapping;
import org.springframework.web.bind.annotation.RestController;
import java.util.Date;
import java.util.concurrent.*;
/**
* @author Stone
* @version V1.0.0
* @date 2019/12/10
*/
@RestController
@Slf4j
public class CallableController {
/**
* 注意不要定义成1,否则 `/log/callable2` 里会死循环(外层使用了唯一的线程,在内存又需要等待这个线程可用)
*/
private ExecutorService threadPool = Executors.newFixedThreadPool(10);
@GetMapping("/log/callable")
public String logCallable() throws ExecutionException, InterruptedException {
log.info("=== main thread ===");
Future<String> future = threadPool.submit(new MyCallable<String>(new Callable<String>() {
@Override
public String call() throws Exception {
log.info("=== sub thread ===");
return "ok_from_sub";
}
}));
log.info("=== main wait for return value:" + future.get());
return "log," + new Date();
}
@GetMapping("/log/callable2")
public String logCallable2() throws ExecutionException, InterruptedException {
log.info("=== main thread ===");
Future<String> future = threadPool.submit(new MyCallable<String>(new Callable<String>() {
@Override
public String call() throws Exception {
log.info("=== sub thread ===");
Future<String> subsubFuture = threadPool.submit(new MyCallable<String>(new Callable<String>() {
@Override
public String call() throws Exception {
log.info("=== sub sub thread ===");
return "ok_from_sub_sub";
}
}));
String value = subsubFuture.get();
return value;
}
}));
log.info("=== main wait for return value:" + future.get());
return "log," + new Date();
}
}
补充
-
上述的拦截器见到一个细节:在响应中写了一个Header,即
httpResponse.setHeader(REQUEST_ID_HEADER, requestId);
这是一个好的实践,即使不看日志,从浏览器中也可以看到请求的RequestId,方便回到日志中进行搜索 -
建议使用AOP的方式拦截
这样即使是直接调用controller方法,也能够进行拦截并设置对应的参数到MDC,如果是javax.servlet.Filter 或者其他拦截器,只能拦截http(s)请求