传送门
分布式定时任务系列1:XXL-job安装
分布式定时任务系列2:XXL-job使用
分布式定时任务系列3:任务执行引擎设计
分布式定时任务系列4:任务执行引擎设计续
Java并发编程实战1:java中的阻塞队列
引子
这篇文章的主要目不是讨论XXL-job的使用,而是要通过它的任务线程实现机制来分析java中阻塞队列的应用!
而这一切要从上周某天,公司一个普通的下午说起。
当时一个同事要添加任务,就随口问了一句:“阻塞处理策略选啥?”,我心里一惊,以前没注意过这个地方,每次都是默认的,也就是单机串行。
就是下图这个:
这里面有3个选项:
众人先是侃侃讨论了半天,后面再去翻了源码来查看,发现了下面的blockingQueue
阻塞队列
什么是阻塞队列
关于java阻塞队列,可以看看java中的阻塞队列。里面例子用到的是有界队列ArrayBlockingQueue,XXL-job里面用的是无界队列LinkedBlockingQueue。不论它们的相似及区别点,最重要的都是拥有阻塞特性。
要分析XXL-job是怎么使用阻塞队列的,就从XXL-job的调度触发机制入手!
什么是XXL-job任务
前面在分布式定时任务系列2:XXL-job使用里面介绍过XXL-job的使用,注册一个任务也很简单,只要在业务代码方法中打上@XxlJob注解即可!下面是官方提供的demo:
/**
* 1、简单任务示例(Bean模式)
*/
@XxlJob("demoJobHandler")
public void demoJobHandler() throws Exception {
XxlJobHelper.log("XXL-JOB, Hello World.");
for (int i = 0; i < 5; i++) {
XxlJobHelper.log("beat at:" + i);
TimeUnit.SECONDS.sleep(2);
}
// default success
}
开发步骤
1、任务开发:在Spring Bean实例中,开发Job方法;
2、注解配置:为Job方法添加注解 "@XxlJob(value="自定义jobhandler名称", init = "JobHandler初始化方法", destroy = "JobHandler销毁方法")",注解value值对应的是调度中心新建任务的JobHandler属性的值。
3、执行日志:需要通过 "XxlJobHelper.log" 打印执行日志;
4、任务结果:默认任务结果为 "成功" 状态,不需要主动设置;如有诉求,比如设置任务结果为失败,可以通过 "XxlJobHelper.handleFail/handleSuccess" 自主设置任务结果;
任务注册
对于XXL-job来说,是怎么注册并管理这些任务的呢?
在程序启动的时候,xxl-job-core包中的XxlJobSpringExecutor类由于实现了接口SmartInitializingSingleton,在Bean实例化完成时会调用afterSingletonsInstantiated()方法:
@Override
public void afterSingletonsInstantiated() {
// init JobHandler Repository (for method)
initJobHandlerMethodRepository(applicationContext);
// refresh GlueFactory
GlueFactory.refreshInstance(1);
// super start
try {
super.start();
} catch (Exception e) {
throw new RuntimeException(e);
}
}
在initJobHandlerMethodRepository(applicationContext)方法中,现在看一下这个方法的代码:该方法会提取所有SpringBean的实例中方法上添加了XxlJob注解的方法,并进行注册!
private void initJobHandlerMethodRepository(ApplicationContext applicationContext) {
if (applicationContext == null) {
return;
}
// init job handler from method
String[] beanDefinitionNames = applicationContext.getBeanNamesForType(Object.class, false, true);
for (String beanDefinitionName : beanDefinitionNames) {
Object bean = applicationContext.getBean(beanDefinitionName);
Map<Method, XxlJob> annotatedMethods = null; // referred to :org.springframework.context.event.EventListenerMethodProcessor.processBean
try {
annotatedMethods = MethodIntrospector.selectMethods(bean.getClass(),
new MethodIntrospector.MetadataLookup<XxlJob>() {
@Override
public XxlJob inspect(Method method) {
return AnnotatedElementUtils.findMergedAnnotation(method, XxlJob.class);
}
});
} catch (Throwable ex) {
logger.error("xxl-job method-jobhandler resolve error for bean[" + beanDefinitionName + "].", ex);
}
if (annotatedMethods==null || annotatedMethods.isEmpty()) {
continue;
}
for (Map.Entry<Method, XxlJob> methodXxlJobEntry : annotatedMethods.entrySet()) {
Method executeMethod = methodXxlJobEntry.getKey();
XxlJob xxlJob = methodXxlJobEntry.getValue();
// regist
registJobHandler(xxlJob, bean, executeMethod);
}
}
// ---------------------- job handler repository ----------------------
private static ConcurrentMap<String, IJobHandler> jobHandlerRepository = new ConcurrentHashMap<String, IJobHandler>();
public static IJobHandler loadJobHandler(String name){
return jobHandlerRepository.get(name);
}
public static IJobHandler registJobHandler(String name, IJobHandler jobHandler){
logger.info(">>>>>>>>>>> xxl-job register jobhandler success, name:{}, jobHandler:{}", name, jobHandler);
return jobHandlerRepository.put(name, jobHandler);
}
注册的方法就是registJobHandler(name, new MethodJobHandler(bean, executeMethod, initMethod, destroyMethod)),会将XxlJob注解标记的方法,提取出对应的方法名,通过反射得到Method信息并实例为对应的MethodJobHandler,最后通过ConcurrentMap来管理,其中key是注任务名称,value为JobHandler实例。
看到这里,不知你有没有一个疑问:这里的JobHandler里面并没有阻塞队列BlockingQueue,它们是怎么关联起来的呢?所以这里就要讨论一下XXL-job的任务执行!
XXL-job的任务执行
在后台管理手动触发一下测试任务,并把代码打上debug断点:
首先会发送后台请求:
一路跟踪下去,会进入到XxlJobExecutor.runExecutor(TriggerParam triggerParam, String address)方法,最终是到ExecutorBizImpl.run(TriggerParam triggerParam)方法:
public ReturnT<String> run(TriggerParam triggerParam) {
// load old:jobHandler + jobThread
JobThread jobThread = XxlJobExecutor.loadJobThread(triggerParam.getJobId());
IJobHandler jobHandler = jobThread!=null?jobThread.getHandler():null;
String removeOldReason = null;
// valid:jobHandler + jobThread
GlueTypeEnum glueTypeEnum = GlueTypeEnum.match(triggerParam.getGlueType());
if (GlueTypeEnum.BEAN == glueTypeEnum) {
// new jobhandler
IJobHandler newJobHandler = XxlJobExecutor.loadJobHandler(triggerParam.getExecutorHandler());
// valid old jobThread
if (jobThread!=null && jobHandler != newJobHandler) {
// change handler, need kill old thread
removeOldReason = "change jobhandler or glue type, and terminate the old job thread.";
jobThread = null;
jobHandler = null;
}
// valid handler
if (jobHandler == null) {
jobHandler = newJobHandler;
if (jobHandler == null) {
return new ReturnT<String>(ReturnT.FAIL_CODE, "job handler [" + triggerParam.getExecutorHandler() + "] not found.");
}
}
} else if (GlueTypeEnum.GLUE_GROOVY == glueTypeEnum) {
// valid old jobThread
if (jobThread != null &&
!(jobThread.getHandler() instanceof GlueJobHandler
&& ((GlueJobHandler) jobThread.getHandler()).getGlueUpdatetime()==triggerParam.getGlueUpdatetime() )) {
// change handler or gluesource updated, need kill old thread
removeOldReason = "change job source or glue type, and terminate the old job thread.";
jobThread = null;
jobHandler = null;
}
// valid handler
if (jobHandler == null) {
try {
IJobHandler originJobHandler = GlueFactory.getInstance().loadNewInstance(triggerParam.getGlueSource());
jobHandler = new GlueJobHandler(originJobHandler, triggerParam.getGlueUpdatetime());
} catch (Exception e) {
logger.error(e.getMessage(), e);
return new ReturnT<String>(ReturnT.FAIL_CODE, e.getMessage());
}
}
} else if (glueTypeEnum!=null && glueTypeEnum.isScript()) {
// valid old jobThread
if (jobThread != null &&
!(jobThread.getHandler() instanceof ScriptJobHandler
&& ((ScriptJobHandler) jobThread.getHandler()).getGlueUpdatetime()==triggerParam.getGlueUpdatetime() )) {
// change script or gluesource updated, need kill old thread
removeOldReason = "change job source or glue type, and terminate the old job thread.";
jobThread = null;
jobHandler = null;
}
// valid handler
if (jobHandler == null) {
jobHandler = new ScriptJobHandler(triggerParam.getJobId(), triggerParam.getGlueUpdatetime(), triggerParam.getGlueSource(), GlueTypeEnum.match(triggerParam.getGlueType()));
}
} else {
return new ReturnT<String>(ReturnT.FAIL_CODE, "glueType[" + triggerParam.getGlueType() + "] is not valid.");
}
// executor block strategy
if (jobThread != null) {
ExecutorBlockStrategyEnum blockStrategy = ExecutorBlockStrategyEnum.match(triggerParam.getExecutorBlockStrategy(), null);
if (ExecutorBlockStrategyEnum.DISCARD_LATER == blockStrategy) {
// discard when running
if (jobThread.isRunningOrHasQueue()) {
return new ReturnT<String>(ReturnT.FAIL_CODE, "block strategy effect:"+ExecutorBlockStrategyEnum.DISCARD_LATER.getTitle());
}
} else if (ExecutorBlockStrategyEnum.COVER_EARLY == blockStrategy) {
// kill running jobThread
if (jobThread.isRunningOrHasQueue()) {
removeOldReason = "block strategy effect:" + ExecutorBlockStrategyEnum.COVER_EARLY.getTitle();
jobThread = null;
}
} else {
// just queue trigger
}
}
// replace thread (new or exists invalid)
if (jobThread == null) {
jobThread = XxlJobExecutor.registJobThread(triggerParam.getJobId(), jobHandler, removeOldReason);
}
// push data to queue
ReturnT<String> pushResult = jobThread.pushTriggerQueue(triggerParam);
return pushResult;
}
任务执行过程中,如果发现XXL-job对应的执行线程不存在,会创建一个新new Thread实例,并绑定前面的JobHandler实例,这样对于每一个任务都有一个单独的线程,也就将执行线程JobThread跟任务关联起来了:
public static JobThread registJobThread(int jobId, IJobHandler handler, String removeOldReason){
JobThread newJobThread = new JobThread(jobId, handler);
newJobThread.start();
logger.info(">>>>>>>>>>> xxl-job regist JobThread success, jobId:{}, handler:{}", new Object[]{jobId, handler});
JobThread oldJobThread = jobThreadRepository.put(jobId, newJobThread); // putIfAbsent | oh my god, map's put method return the old value!!!
if (oldJobThread != null) {
oldJobThread.toStop(removeOldReason);
oldJobThread.interrupt();
}
return newJobThread;
}
public JobThread(int jobId, IJobHandler handler) {
this.jobId = jobId;
this.handler = handler;
// 初始化阻塞队列
this.triggerQueue = new LinkedBlockingQueue<TriggerParam>();
this.triggerLogIdSet = Collections.synchronizedSet(new HashSet<Long>());
// assign job thread name
this.setName("xxl-job, JobThread-"+jobId+"-"+System.currentTimeMillis());
}
而执行任务的最后一步,就是将当前任务入队到刚才的执行线程的BlockingQueue中:
// push data to queue
ReturnT<String> pushResult = jobThread.pushTriggerQueue(triggerParam);
而这也就跟文章最前面的代码关联起来了:
/**
* new trigger to queue
*
* @param triggerParam
* @return
*/
public ReturnT<String> pushTriggerQueue(TriggerParam triggerParam) {
// avoid repeat
if (triggerLogIdSet.contains(triggerParam.getLogId())) {
logger.info(">>>>>>>>>>> repeate trigger job, logId:{}", triggerParam.getLogId());
return new ReturnT<String>(ReturnT.FAIL_CODE, "repeate trigger job, logId:" + triggerParam.getLogId());
}
triggerLogIdSet.add(triggerParam.getLogId());
triggerQueue.add(triggerParam);
return ReturnT.SUCCESS;
}
BlockingQueue的应用
消息入队
至此,就可以看出任务触发的大致逻辑了,用户在admin手工触发任务之后最终会放入到任务对应的队列之中:通过add方法将当前执行参数入队(FIFO),不阻塞队列,如果满了就抛出异常(这也是一个比较大的隐患,如果触了过于频繁,可能会导致OOM,而这也就是为什么要设计不同阻塞策略的原因)。
消息消费
而这些触发的任务怎么消费的呢?其实也是在上面的提到的执行线程JobThread:
- 当实例化一个线程的时候(com.xxl.job.core.executor.XxlJobExecutor类),会立即启动:调用线程的start()方法
- 启动之后会调用线程的run()方法(com.xxl.job.core.thread.JobThread类)
- 在执行线程中,会写一个while的无限循环来不停的从阻塞队列中取出任务
- 循环结束的条件就是上面的阻塞队列:如果是覆盖之前调度将会结束循环(com.xxl.job.core.biz.impl.ExecutorBizImpl类)