【xxl-job】源码分析 - 启动流程

【引言】

上篇博客对xxl-job分布式定时任务框架的架构做了整体介绍，本篇博客开始主要从源码入手，简单学习下xxl-job的实现原理。

【项目结构】

此项目主要分为三个模块：xxl-job-admin（管理中心），xxl-job-core（核心），xxl-job-executor-samples（示例）.

<modules>
		<module>xxl-job-core</module>
		<module>xxl-job-admin</module>
		<module>xxl-job-executor-samples</module>
</modules>

【启动流程】

从xxl-job-admin的spring 配置文件（applicationcontext-xxl-job-admin.xml）出发，主要配置了三部分：数据源，事务，基于quartz实现的调度任务，配置如下：

<!-- *************************** part 1 :for datasource ****************************-->
... 此处省略

<!-- ************************** part 2 :for tx ********************************* -->
... 此处省略

<!-- *********************** part 3 :for xxl-job scheduler ************************* -->

	<bean id="quartzScheduler" lazy-init="false" class="org.springframework.scheduling.quartz.SchedulerFactoryBean">
		<property name="dataSource" ref="dataSource" />
		<property name="autoStartup" value="true" />			<!--自动启动 -->
		<property name="startupDelay" value="20" />				<!--延时启动，应用启动成功后在启动 -->
		<property name="overwriteExistingJobs" value="true" />	<!--覆盖DB中JOB：true、以数据库中已经存在的为准：false -->
		<property name="applicationContextSchedulerContextKey"  value="applicationContextKey" />
		<property name="configLocation" value="classpath:quartz.properties"/>
	</bean>

	<bean id="xxlJobDynamicScheduler" class="com.xxl.job.admin.core.schedule.XxlJobDynamicScheduler" init-method="init" destroy-method="destroy" >
		<!-- (轻易不要变更“调度器名称”, 任务创建时会绑定该“调度器名称”) -->
		<property name="scheduler" ref="quartzScheduler"/>
		<property name="accessToken" value="${xxl.job.accessToken}" />
	</bean>

通过以上配置，可以看出启动的主要工作在XxlJobDynamicScheduler中的init()方法中实现，源码如下：

public void init() throws Exception {
        // admin registry monitor run
        // 启动自动注册线程， 获取类型为自动注册的执行器信息，完成机器的自动注册与发现
        JobRegistryMonitorHelper.getInstance().start();

        // admin monitor run
        // 启动失败日志监控线程
        JobFailMonitorHelper.getInstance().start();

        // admin-server(spring-mvc)
        NetComServerFactory.putService(AdminBiz.class, XxlJobDynamicScheduler.adminBiz);
        NetComServerFactory.setAccessToken(accessToken);

        // init i18n
        initI18n();

        // valid
        Assert.notNull(scheduler, "quartz scheduler is null");
        logger.info(">>>>>>>>> init xxl-job admin success.");
}

JobRegistryMonitorHelper.getInstance().start() 代码如下：

public void start(){
// 创建一个线程
		registryThread = new Thread(new Runnable() {
			@Override
			public void run() {
				while (!toStop) {
					try {
						// auto registry group
						// 获取类型为自动注册的执行器地址列表
						List<XxlJobGroup> groupList = XxlJobDynamicScheduler.xxlJobGroupDao.findByAddressType(0);
						if (CollectionUtils.isNotEmpty(groupList)) {

							// remove dead address (admin/executor)
							// 删除 90秒之内没有更新信息的注册机器， 90秒没有心跳信息返回，代表机器已经出现问题，故移除
							XxlJobDynamicScheduler.xxlJobRegistryDao.removeDead(RegistryConfig.DEAD_TIMEOUT);

							// fresh online address (admin/executor)
							HashMap<String, List<String>> appAddressMap = new HashMap<String, List<String>>();
							// 查询在90秒之内有过更新的机器列表
							List<XxlJobRegistry> list = XxlJobDynamicScheduler.xxlJobRegistryDao.findAll(RegistryConfig.DEAD_TIMEOUT);
							if (list != null) {
							// 循环注册机器列表，  根据执行器不同，将这些机器列表区分拿出来
								for (XxlJobRegistry item: list) {
								// 判断该机器注册信息RegistryGroup ，RegistType 是否是EXECUTOR , EXECUTOR 代表该机器是注册到执行器上面的
									if (RegistryConfig.RegistType.EXECUTOR.name().equals(item.getRegistryGroup())) {
									// 获取注册的执行器 KEY  （也就是执行器）
										String appName = item.getRegistryKey();
										List<String> registryList = appAddressMap.get(appName);
										if (registryList == null) {
											registryList = new ArrayList<String>();
										}

										if (!registryList.contains(item.getRegistryValue())) {
											registryList.add(item.getRegistryValue());
										}
										// 收集 机器信息，根据执行器做区分
										appAddressMap.put(appName, registryList);
									}
								}
							}

							// fresh group address
							//  遍历执行器列表
							for (XxlJobGroup group: groupList) {
							// 通过执行器的APP_NAME  拿出他下面的集群机器地址
								List<String> registryList = appAddressMap.get(group.getAppName());
								String addressListStr = null;
								if (CollectionUtils.isNotEmpty(registryList)) {
									Collections.sort(registryList);
									addressListStr = StringUtils.join(registryList, ",");
								}
								group.setAddressList(addressListStr);
								// 将 这个执行器的 集群机器地址列表，写入到数据库
								XxlJobDynamicScheduler.xxlJobGroupDao.update(group);
							}
						}
					} catch (Exception e) {
						logger.error("job registry instance error:{}", e);
					}
					try {
						TimeUnit.SECONDS.sleep(RegistryConfig.BEAT_TIMEOUT);
					} catch (InterruptedException e) {
						logger.error("job registry instance error:{}", e);
					}
				}
			}
		});
		registryThread.setDaemon(true);
		 //启动线程
		registryThread.start();
	}

JobFailMonitorHelper.getInstance().start() 代码如下：

public void start(){
 // 启动线程
		monitorThread = new Thread(new Runnable() {

			@Override
			public void run() {
				// monitor
				while (!toStop) {
					try {
						List<Integer> jobLogIdList = new ArrayList<Integer>();
						// 从队列中拿出所有可用的 jobLogIds
						int drainToNum = JobFailMonitorHelper.instance.queue.drainTo(jobLogIdList);

						if (CollectionUtils.isNotEmpty(jobLogIdList)) {
							for (Integer jobLogId : jobLogIdList) {
								if (jobLogId==null || jobLogId==0) {
									continue;
								}
								XxlJobLog log = XxlJobDynamicScheduler.xxlJobLogDao.load(jobLogId);
								if (log == null) {
									continue;
								}
								//任务触发成功， 但是JobHandle 还没有返回结果
								if (IJobHandler.SUCCESS.getCode() == log.getTriggerCode() && log.getHandleCode() == 0) {
									// job running
									//将 JobLogId 放入队列 ， 继续监控
									JobFailMonitorHelper.monitor(jobLogId);
									logger.info(">>>>>>>>>>> job monitor, job running, JobLogId:{}", jobLogId);
								} else if (IJobHandler.SUCCESS.getCode() == log.getHandleCode()) {
									// job success, pass
									logger.info(">>>>>>>>>>> job monitor, job success, JobLogId:{}", jobLogId);
								} else /*if (IJobHandler.FAIL.getCode() == log.getTriggerCode()
										|| IJobHandler.FAIL.getCode() == log.getHandleCode()
										|| IJobHandler.FAIL_RETRY.getCode() == log.getHandleCode() )*/ {

									// job fail,

									// 1、fail retry
									XxlJobInfo info = XxlJobDynamicScheduler.xxlJobInfoDao.loadById(log.getJobId());

									if (log.getExecutorFailRetryCount() > 0) {
										JobTriggerPoolHelper.trigger(log.getJobId(), (log.getExecutorFailRetryCount()-1), TriggerTypeEnum.RETRY);
										String retryMsg = "<br><br><span style=\"color:#F39C12;\" > >>>>>>>>>>>"+ I18nUtil.getString("jobconf_trigger_type_retry") +"<<<<<<<<<<< </span><br>";
										log.setTriggerMsg(log.getTriggerMsg() + retryMsg);
										XxlJobDynamicScheduler.xxlJobLogDao.updateTriggerInfo(log);
									}

									// 2、fail alarm
									failAlarm(info, log);
								 // 任务执行失败， 执行发送邮件等预警措施
									logger.info(">>>>>>>>>>> job monitor, job fail, JobLogId:{}", jobLogId);
								}/* else {
									JobFailMonitorHelper.monitor(jobLogId);
									logger.info(">>>>>>>>>>> job monitor, job status unknown, JobLogId:{}", jobLogId);
								}*/
							}
						}

						TimeUnit.SECONDS.sleep(10);
					} catch (Exception e) {
						logger.error("job monitor error:{}", e);
					}
				}

				// monitor all clear
				List<Integer> jobLogIdList = new ArrayList<Integer>();
				int drainToNum = getInstance().queue.drainTo(jobLogIdList);
				if (jobLogIdList!=null && jobLogIdList.size()>0) {
					for (Integer jobLogId: jobLogIdList) {
						XxlJobLog log = XxlJobDynamicScheduler.xxlJobLogDao.load(jobLogId);
						if (ReturnT.FAIL_CODE == log.getTriggerCode()|| ReturnT.FAIL_CODE==log.getHandleCode()) {
							// job fail,
							XxlJobInfo info = XxlJobDynamicScheduler.xxlJobInfoDao.loadById(log.getJobId());

							failAlarm(info, log);
							logger.info(">>>>>>>>>>> job monitor last, job fail, JobLogId:{}", jobLogId);
						}
					}
				}

			}
		});
		monitorThread.setDaemon(true);
		monitorThread.start();
	}

【总结】

通过以上分析，可以得到如下总结：

xxl-job在启动的时候，启动了两个线程，一是用来监控自动注册的机器，达到自动注册的目的；二是监控任务的执行状态，若失败，则发送邮件报警。