Spring Boot优雅关闭服务

在之前，我们要重新部署Spring Boot应用时，都是用简单粗暴的kill -9 PID命令，这种方式的弊端很明显，假如应用内还有未处理完的业务，突然把应用给kill掉，这显然是出问题的。

现在好了，Spring Boot提供了优雅关闭服务的功能，它支持目前Jetty, Reactor Netty, Tomcat, Undertow这四种容器，使用该功能很简单，只需要在配置文件中配置server.shutdown(关闭方式)和spring.lifecycle.timeout-per-shutdown-phase(超时时间)这两个参数即可，server.shutdown有两个可选值:immediate(立即停止，默认值)和graceful(优雅停止)，spring.lifecycle.timeout-per-shutdown-phase的默认值为30s,我们可以根据自己的业务耗时情况来设置时间。我们先到yml配置好参数：

server:
  shutdown: graceful
spring:
  lifecycle:
    timeout-per-shutdown-phase: 20s

然后再写个简单的http接口

@RestController
@SpringBootApplication
public class GracefulApplication {

    public static void main(String[] args) {
        SpringApplication.run(GracefulApplication.class, args);
    }

    @RequestMapping("health")
    public String health() {
        return "health";
    }

}

这里我用的是最新的Spring Boot 2.6.1,其默认的tomcat版本为Apache Tomcat/9.0.55,此时，我们打包到Linux环境下，启动后执行kill -2 PID可以看到以下日志

2021-12-03 11:37:25.282  INFO 638081 --- [ionShutdownHook] o.s.b.w.e.tomcat.GracefulShutdown        : Commencing graceful shutdown. Waiting for active requests to complete
2021-12-03 11:37:25.290  INFO 638081 --- [tomcat-shutdown] o.s.b.w.e.tomcat.GracefulShutdown        : Graceful shutdown complete

因为我们服务没有需要处理的请求，所以关闭得很快，我们在接口中等待10s

public String health() {
        try {
            Thread.sleep(10000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        return "health";
    }

我们再次执行kill -2 PID会看到有一定的时间间隔

2021-12-03 11:44:44.239  INFO 647083 --- [ionShutdownHook] o.s.b.w.e.tomcat.GracefulShutdown        : Commencing graceful shutdown. Waiting for active requests to complete
2021-12-03 11:44:51.911  INFO 647083 --- [tomcat-shutdown] o.s.b.w.e.tomcat.GracefulShutdown        : Graceful shutdown complete

在官网有提到,在kill -2之后，服务端不会再接受新请求，但是Undertow比较例外，它会接受请求，但是会返回503错误 ，我们把容器改为Undertow来看一下

<dependency>
			<groupId>org.springframework.boot</groupId>
			<artifactId>spring-boot-starter-web</artifactId>
			<exclusions>
				<!-- Exclude the Tomcat dependency -->
				<exclusion>
					<groupId>org.springframework.boot</groupId>
					<artifactId>spring-boot-starter-tomcat</artifactId>
				</exclusion>
			</exclusions>
		</dependency>
		<dependency>
			<groupId>org.springframework.boot</groupId>
			<artifactId>spring-boot-starter-undertow</artifactId>
		</dependency>

然后启动起来

2021-12-03 11:58:46.767  INFO 673474 --- [           main] io.undertow                              : starting server: Undertow - 2.2.12.Final
2021-12-03 11:58:46.783  INFO 673474 --- [           main] org.xnio                                 : XNIO version 3.8.4.Final
2021-12-03 11:58:46.795  INFO 673474 --- [           main] org.xnio.nio                             : XNIO NIO Implementation Version 3.8.4.Final
2021-12-03 11:58:46.979  INFO 673474 --- [           main] org.jboss.threads                        : JBoss Threads version 3.1.0.Final
2021-12-03 11:58:47.089  INFO 673474 --- [           main] o.s.b.w.e.undertow.UndertowWebServer     : Undertow started on port(s) 8888 (http)
2021-12-03 11:58:47.108  INFO 673474 --- [           main] c.example.graceful.GracefulApplication   : Started GracefulApplication in 3.076 seconds (JVM running for 3.652)

但是在我们测试过程中并没返回503的错误，而是直接拒绝了请求

经过一番测试后，我们发现，如果在接收到kill -2 的信号时，会先打印Commencing graceful shutdown. Waiting for active requests to complete，如果在spring.lifecycle.timeout-per-shutdown-phase设置的时间内处理完所有的请求，那么最后会成功打印Graceful shutdown complete,如果超过spring.lifecycle.timeout-per-shutdown-phase的时间则会作出以下提示

2021-12-03 13:07:15.215  INFO 796937 --- [ionShutdownHook] o.s.b.w.e.undertow.UndertowWebServer     : Commencing graceful shutdown. Waiting for active requests to complete
2021-12-03 13:07:20.216  INFO 796937 --- [ionShutdownHook] o.s.c.support.DefaultLifecycleProcessor  : Failed to shut down 1 bean with phase value 2147483647 within timeout of 5000ms: [webServerGracefulShutdown]
2021-12-03 13:07:20.217  INFO 796937 --- [ionShutdownHook] o.s.b.w.e.undertow.UndertowWebServer     : Graceful shutdown aborted with one or more requests still active
2021-12-03 13:07:20.217  INFO 796937 --- [ionShutdownHook] io.undertow                              : stopping server: Undertow - 2.2.12.Final

好了，接下我们再大胆尝试，在等待服务处理完毕的这段时间内，我们的还能重新启动应用吗？这里，先写一段简单的脚本

#!/bin/bash
set -m

APP_NAME="graceful-0.0.1-SNAPSHOT.jar"

PID=$(ps -ef |grep $APP_NAME | grep -v $0 |grep -v grep |awk '{print $2}')
echo "PID=${PID}"
if [ "${PID}" != "" ]; then
     echo "ok"
     kill -15 ${PID}
     echo ${PID}
fi
sleep 1
nohup java -jar $APP_NAME 2>&1 &
PID=$(echo $!)
echo "NEW PID=$!"

运行该脚本，先找到原有进程->发送中断请求->启动新程序

可以看到，再程序未退出前，端口依然占用，此时无法启动该端口的新应用。