优雅停机
目前Spring Boot已经发展到了2.3.4.RELEASE,伴随着2.3版本的到来,优雅停机机制也更加完善了。
目前版本的Spring Boot 优雅停机支持Jetty, Reactor Netty, Tomcat和 Undertow 以及反应式和基于 Servlet 的 web 应用程序都支持优雅停机功能。
优雅停机的目的:
如果没有优雅停机,服务器此时直接直接关闭(kill -9),那么就会导致当前正在容器内运行的业务直接失败,在某些特殊的场景下产生脏数据。
增加了优雅停机配置后:
在服务器执行关闭(kill -2)时,会预留一点时间使容器内部业务线程执行完毕,此时容器也不允许新的请求进入。新请求的处理方式跟web服务器有关,Reactor Netty、 Tomcat将停止接入请求,Undertow的处理方式是返回503.
新版配置
YAML配置
新版本配置非常简单,server.shutdown=graceful 就搞定了(注意,优雅停机配置需要配合Tomcat 9.0.33(含)以上版本)
server:
port: 6080
shutdown: graceful #开启优雅停机
spring:
lifecycle:
timeout-per-shutdown-phase: 20s #设置缓冲时间 默认30s
server:
port: 6080
shutdown: graceful #开启优雅停机
spring:
lifecycle:
timeout-per-shutdown-phase: 20s #设置缓冲时间 默认30s
在设置了缓冲参数timeout-per-shutdown-phase 后,在规定时间内如果线程无法执行完毕则会被强制停机。
下面我们来看下停机时,加了优雅停日志和不加的区别:
//未加优雅停机配置
Disconnected from the target VM, address: '127.0.0.1:49754', transport: 'socket'
Process finished with exit code 130 (interrupted by signal 2: SIGINT)
//未加优雅停机配置
Disconnected from the target VM, address: '127.0.0.1:49754', transport: 'socket'
Process finished with exit code 130 (interrupted by signal 2: SIGINT)
加了优雅停机配置后,可明显发现的日志 Waiting for active requests to cpmplete,此时容器将在ShutdownHook执行完毕后停止。
关闭方式
1、 一定不要使用kill -9 操作,使用kill -2 来关闭容器。这样才会触发java内部ShutdownHook操作,kill -9不会触发ShutdownHook。
2、可以使用端点监控 POST 请求 /actuator/shutdown 来执行优雅关机。
添加ShutdownHook
通过上面的日志我们发现Druid执行了自己的ShutdownHook,那么我们也来添加下ShutdownHook,有几种简单的方式:
1、实现DisposableBean接口,实现destroy方法
@Slf4j
@Service
public class DefaultDataStore implements DisposableBean {
private final ExecutorService executorService = new ThreadPoolExecutor(OSUtil.getAvailableProcessors(), OSUtil.getAvailableProcessors() + 1, 1, TimeUnit.MINUTES, new ArrayBlockingQueue<>(200), new DefaultThreadFactory("UploadVideo"));
@Override
public void destroy() throws Exception {
log.info("准备优雅停止应用使用 DisposableBean");
executorService.shutdown();
}
}
@Slf4j
@Service
public class DefaultDataStore implements DisposableBean {
private final ExecutorService executorService = new ThreadPoolExecutor(OSUtil.getAvailableProcessors(), OSUtil.getAvailableProcessors() + 1, 1, TimeUnit.MINUTES, new ArrayBlockingQueue<>(200), new DefaultThreadFactory("UploadVideo"));
@Override
public void destroy() throws Exception {
log.info("准备优雅停止应用使用 DisposableBean");
executorService.shutdown();
}
}
2、使用@PreDestroy注解
@Slf4j
@Service
public class DefaultDataStore {
private final ExecutorService executorService = new ThreadPoolExecutor(OSUtil.getAvailableProcessors(), OSUtil.getAvailableProcessors() + 1, 1, TimeUnit.MINUTES, new ArrayBlockingQueue<>(200), new DefaultThreadFactory("UploadVideo"));
@PreDestroy
public void shutdown() {
log.info("准备优雅停止应用 @PreDestroy");
executorService.shutdown();
}
}
@Slf4j
@Service
public class DefaultDataStore {
private final ExecutorService executorService = new ThreadPoolExecutor(OSUtil.getAvailableProcessors(), OSUtil.getAvailableProcessors() + 1, 1, TimeUnit.MINUTES, new ArrayBlockingQueue<>(200), new DefaultThreadFactory("UploadVideo"));
@PreDestroy
public void shutdown() {
log.info("准备优雅停止应用 @PreDestroy");
executorService.shutdown();
}
}
这里注意,@PreDestroy 比 DisposableBean 先执行
关闭原理
1、使用kill pid关闭,源码很简单,大家可以看下GracefulShutdown
private void doShutdown(GracefulShutdownCallback callback) {
List connectors = getConnectors();
connectors.forEach(this::close);try {for (Container host : this.tomcat.getEngine().findChildren()) {for (Container context : host.findChildren()) {while (isActive(context)) {if (this.aborted) {
logger.info("Graceful shutdown aborted with one or more requests still active");
callback.shutdownComplete(GracefulShutdownResult.REQUESTS_ACTIVE);return;
}
Thread.sleep(50);
}
}
}
}catch (InterruptedException ex) {
Thread.currentThread().interrupt();
}
logger.info("Graceful shutdown complete");
callback.shutdownComplete(GracefulShutdownResult.IDLE);
}
private void doShutdown(GracefulShutdownCallback callback) {
List connectors = getConnectors();
connectors.forEach(this::close);try {for (Container host : this.tomcat.getEngine().findChildren()) {for (Container context : host.findChildren()) {while (isActive(context)) {if (this.aborted) {
logger.info("Graceful shutdown aborted with one or more requests still active");
callback.shutdownComplete(GracefulShutdownResult.REQUESTS_ACTIVE);return;
}
Thread.sleep(50);
}
}
}
}catch (InterruptedException ex) {
Thread.currentThread().interrupt();
}
logger.info("Graceful shutdown complete");
callback.shutdownComplete(GracefulShutdownResult.IDLE);
}
2、使用端点监控 POST 请求 /actuator/shutdown关闭
因为actuator 都使用了SPI的扩展方式,所以我们看下AutoConfiguration,可以看到关键点就是ShutdownEndpoint
@Configuration(
proxyBeanMethods = false
)
@ConditionalOnAvailableEndpoint(
endpoint = ShutdownEndpoint.class
)
public class ShutdownEndpointAutoConfiguration {
public ShutdownEndpointAutoConfiguration() {
}
@Bean(
destroyMethod = ""
)
@ConditionalOnMissingBean
public ShutdownEndpoint shutdownEndpoint() {
return new ShutdownEndpoint();
}
}
@Configuration(
proxyBeanMethods = false
)
@ConditionalOnAvailableEndpoint(
endpoint = ShutdownEndpoint.class
)
public class ShutdownEndpointAutoConfiguration {
public ShutdownEndpointAutoConfiguration() {
}
@Bean(
destroyMethod = ""
)
@ConditionalOnMissingBean
public ShutdownEndpoint shutdownEndpoint() {
return new ShutdownEndpoint();
}
}
ShutdownEndpoint,为了节省篇幅只留了一点重要的
@Endpoint(
id = "shutdown",
enableByDefault = false
)
public class ShutdownEndpoint implements ApplicationContextAware {
@WriteOperation
public Map shutdown() {
if (this.context == null) {
return NO_CONTEXT_MESSAGE;
} else {
boolean var6 = false;
Map var1;
try {
var6 = true;
var1 = SHUTDOWN_MESSAGE;
var6 = false;
} finally {
if (var6) {
Thread thread = new Thread(this::performShutdown);
thread.setContextClassLoader(this.getClass().getClassLoader());
thread.start();
}
}
Thread thread = new Thread(this::performShutdown);
thread.setContextClassLoader(this.getClass().getClassLoader());
thread.start();
return var1;
}
}
private void performShutdown() {
try {
Thread.sleep(500L);
} catch (InterruptedException var2) {
Thread.currentThread().interrupt();
}
this.context.close(); //这里才是核心
}
}
@Endpoint(
id = "shutdown",
enableByDefault = false
)
public class ShutdownEndpoint implements ApplicationContextAware {
@WriteOperation
public Map shutdown() {
if (this.context == null) {
return NO_CONTEXT_MESSAGE;
} else {
boolean var6 = false;
Map var1;
try {
var6 = true;
var1 = SHUTDOWN_MESSAGE;
var6 = false;
} finally {
if (var6) {
Thread thread = new Thread(this::performShutdown);
thread.setContextClassLoader(this.getClass().getClassLoader());
thread.start();
}
}
Thread thread = new Thread(this::performShutdown);
thread.setContextClassLoader(this.getClass().getClassLoader());
thread.start();
return var1;
}
}
private void performShutdown() {
try {
Thread.sleep(500L);
} catch (InterruptedException var2) {
Thread.currentThread().interrupt();
}
this.context.close(); //这里才是核心
}
}
在调用了 this.context.close() ,其实就是AbstractApplicationContext 的close() 方法 (重点是其中的doClose())
/**
* Close this application context, destroying all beans in its bean factory.
*
/**
* Close this application context, destroying all beans in its bean factory.
*
Delegates to {@code doClose()} for the actual closing procedure.
* Also removes a JVM shutdown hook, if registered, as it's not needed anymore.
* @see #doClose()
* @see #registerShutdownHook()
*/
@Override
public void close() {
synchronized (this.startupShutdownMonitor) {
doClose(); //重点:销毁bean 并执行jvm shutdown hook
// If we registered a JVM shutdown hook, we don't need it anymore now:
// We've already explicitly closed the context.
if (this.shutdownHook != null) {
try {
Runtime.getRuntime().removeShutdownHook(this.shutdownHook);
}
catch (IllegalStateException ex) {
// ignore - VM is already shutting down
}
}
}
}
后记
到这里,关于单机
版本的Spring Boot优雅停机就说完了。为什么说单机
?因为大家也能发现,在关闭时,其实只是保证了服务端内部线程执行完毕,调用方的状态是没关注的。
不论是Dubbo还是Cloud 的分布式服务框架,需要关注的是怎么能在服务停止前,先将提供者在注册中心进行反注册,然后在停止服务提供者,这样才能保证业务系统不会产生各种503、timeout等现象。
好在当前Spring Boot 结合Kubernetes已经帮我们搞定了这一点,也就是Spring Boot 2.3版本新功能Liveness(存活状态) 和Readiness(就绪状态)
简单的提下这两个状态:
- Liveness(存活状态):Liveness 状态来查看内部情况可以理解为health check,如果Liveness失败就就意味着应用处于故障状态并且目前无法恢复,这种情况就重启吧。此时Kubernetes如果存活探测失败将杀死Container。
- Readiness(就绪状态):用来告诉应用是否已经准备好接受客户端请求,如果Readiness未就绪那么k8s就不能路由流量过来。