整合 Elastic-Job(支持动态任务)
最近公司的项目需要用到分布式任务调度,在结合xxl-job,elastic-job,quartz多款开源框架后决定使用当当的Elastic-job。不知道大家有没有这样的需求,就是动态任务。之前比较了xxl-job和elastic-job发现,都只是支持注解或者配置以及后台添加现有的任务,不支持动态添加。比如:类似订单半小时后自动取消的场景。
xxl-job理论上来说是可以支持的,但是需要高度整合admin端的程序,然后开放对应的接口才可以给其他服务调用,这样本质直接改源码对后期的升级十分不便,最后放弃了xxl-job。elastic-job在移交Apache后的版本规划中,有提到API的开放,但是目前还没有稳定版,所以只能使用之前的2.1.5的版本来做。在Github搜了很多整合方案,最后决定选择下面的来实现。
因为底层本质上还是用elastic-job的东西。下面引入elastic-job坐标
<dependency> <groupId>com.dangdang</groupId> <artifactId>elastic-job-lite-spring</artifactId> <version>2.1.5</version> </dependency> <dependency> <groupId>com.dangdang</groupId> <artifactId>elastic-job-lite-lifecycle</artifactId> <version>2.1.5</version> </dependency>
这是Github找到的集成方案,把代码clone下来,然后修改一下。github地址:https://github.com/xjzrc/elastic-job-lite-spring-boot-starter。因为要做的是动态的,所以这里没有直接使用maven坐标引入,直接将源码全部接入项目来使用并修改成自己想要的,这样比较灵活。项目采用的是SpringBoot,为简介代码,使用了Lombok。
先引入Zookeeper配置类
/** * Zookeeper配置 * @author Nil * @date 2020/9/14 11:24 */@Setter@Getterpublic class ZookeeperRegistryProperties { /** * 连接Zookeeper服务器的列表 * 包括IP地址和端口号 * 多个地址用逗号分隔 * 如: host1:2181,host2:2181 */ private String serverLists; /** * Zookeeper的命名空间 */ private String namespace; /** * 等待重试的间隔时间的初始值 * 单位:毫秒 */ private int baseSleepTimeMilliseconds = 1000; /** * 等待重试的间隔时间的最大值 * 单位:毫秒 */ private int maxSleepTimeMilliseconds = 3000; /** * 最大重试次数 */ private int maxRetries = 3; /** * 连接超时时间 * 单位:毫秒 */ private int connectionTimeoutMilliseconds = 15000; /** * 会话超时时间 * 单位:毫秒 */ private int sessionTimeoutMilliseconds = 60000; /** * 连接Zookeeper的权限令牌 * 缺省为不需要权限验 */ private String digest;}
引入Elastic-Job配置类,这个类主要是为了可以支持配置式的任务
/** * 任务配置类 * @author Nil * @date 2020/9/14 17:25 */@Getter@Setter@RefreshScope@ConfigurationProperties(prefix = "spring.elasticjob")public class ElasticJobProperties { /** * 注册中心 */ private ZookeeperRegistryProperties zookeeper; /** * 简单作业配置(String为任务名) */ private Map<String, SimpleConfiguration> simples = new LinkedHashMap<>(16); /** * 流式作业配置(String为任务名) */ private Map<String, DataflowConfiguration> dataflows = new LinkedHashMap<>(16); /** * 脚本作业配置(String为任务名) */ private Map<String, ScriptConfiguration> scripts = new LinkedHashMap<>(16); @Getter @Setter public static class SimpleConfiguration extends JobConfiguration { /** * 作业类型 */ private final JobType jobType = JobType.SIMPLE; } @Getter @Setter public static class ScriptConfiguration extends JobConfiguration { /** * 作业类型 */ private final JobType jobType = JobType.SCRIPT; /** * 脚本型作业执行命令行 */ private String scriptCommandLine; } @Getter @Setter public static class DataflowConfiguration extends JobConfiguration { /** * 作业类型 */ private final JobType jobType = JobType.DATAFLOW; /** * 是否流式处理数据 * 如果流式处理数据, 则fetchData不返回空结果将持续执行作业 * 如果非流式处理数据, 则处理数据完成后作业结束 */ private boolean streamingProcess = false; } @Getter @Setter public static class JobConfiguration { /** * 作业实现类,需实现ElasticJob接口 */ private String jobClass; /** * 注册中心Bean的引用,需引用reg:zookeeper的声明 */ private String registryCenterRef = ElasticJobAutoConfiguration.DEFAULT_REGISTRY_CENTER_NAME; /** * cron表达式,用于控制作业触发时间 */ private String cron; /** * 作业分片总数 */ private int shardingTotalCount = 1; /** * 分片序列号和参数用等号分隔,多个键值对用逗号分隔 * 分片序列号从0开始,不可大于或等于作业分片总数 * 如:0=a,1=b,2=c */ private String shardingItemParameters = "0=A"; /** * 作业实例主键,同IP可运行实例主键不同, 但名称相同的多个作业实例 */ private String jobInstanceId; /** * 作业自定义参数 * 作业自定义参数,可通过传递该参数为作业调度的业务方法传参,用于实现带参数的作业 * 例:每次获取的数据量、作业实例从数据库读取的主键等 */ private String jobParameter; /** * 监控作业运行时状态 * 每次作业执行时间和间隔时间均非常短的情况,建议不监控作业运行时状态以提升效率。 * 因为是瞬时状态,所以无必要监控。请用户自行增加数据堆积监控。并且不能保证数据重复选取,应在作业中实现幂等性。 * 每次作业执行时间和间隔时间均较长的情况,建议监控作业运行时状态,可保证数据不会重复选取。 */ private boolean monitorExecution = true; /** * 作业监控端口 * 建议配置作业监控端口, 方便开发者dump作业信息。 * 使用方法: echo “dump” | nc 127.0.0.1 9888 */ private int monitorPort = -1; /** * 最大允许的本机与注册中心的时间误差秒数 * 如果时间误差超过配置秒数则作业启动时将抛异常 * 配置为-1表示不校验时间误差 */ private int maxTimeDiffSeconds = -1; /** * 是否开启失效转移 */ private boolean failover = false; /** * 是否开启错过任务重新执行 */ private boolean misfire = true; /** * 作业分片策略实现类全路径 * 默认使用平均分配策略 * 详情参见:作业分片策略http://elasticjob.io/docs/elastic-job-lite/02-guide/job-sharding-strategy */ private String jobShardingStrategyClass; /** * 作业描述信息 */ private String description; /** * 作业是否禁止启动 * 可用于部署作业时,先禁止启动,部署结束后统一启动 */ private boolean disabled = false; /** * 本地配置是否可覆盖注册中心配置 * 如果可覆盖,每次启动作业都以本地配置为准 */ private boolean overwrite = true; /** * 扩展异常处理类 */ private String jobExceptionHandler = DefaultJobExceptionHandler.class.getCanonicalName(); /** * 扩展作业处理线程池类 */ private String executorServiceHandler = DefaultExecutorServiceHandler.class.getCanonicalName(); /** * 修复作业服务器不一致状态服务调度间隔时间,配置为小于1的任意值表示不执行修复 * 单位:分钟 */ private int reconcileIntervalMinutes = 10; /** * 作业事件追踪的数据源Bean引用 */ private String eventTraceRdbDataSource = "dataSource"; /** * 监听器 */ private Listener listener; @Getter @Setter public static class Listener { /** * 每台作业节点均执行的监听 * 若作业处理作业服务器的文件,处理完成后删除文件,可考虑使用每个节点均执行清理任务。 * 此类型任务实现简单,且无需考虑全局分布式任务是否完成,请尽量使用此类型监听器。 * * <p>注意:类必须继承com.dangdang.ddframe.job.lite.api.listener.ElasticJobListener</p> */ String listenerClass; /** * 分布式场景中仅单一节点执行的监听 * 若作业处理数据库数据,处理完成后只需一个节点完成数据清理任务即可。 * 此类型任务处理复杂,需同步分布式环境下作业的状态同步,提供了超时设置来避免作业不同步导致的死锁,请谨慎使用。 * * <p>注意:类必须继承com.dangdang.ddframe.job.lite.api.listener.AbstractDistributeOnceElasticJobListener</p> */ String distributedListenerClass; /** * 最后一个作业执行前的执行方法的超时时间 * 单位:毫秒 */ Long startedTimeoutMilliseconds = Long.MAX_VALUE; /** * 最后一个作业执行后的执行方法的超时时间 * 单位:毫秒 */ Long completedTimeoutMilliseconds = Long.MAX_VALUE; } }}
创建自动配置类
/** * elastic-job配置 * @author Nil * @date 2020/9/14 11:24 */@Configuration@RequiredArgsConstructor@EnableConfigurationProperties(ElasticJobProperties.class)public class ElasticJobAutoConfiguration { public static final String DEFAULT_REGISTRY_CENTER_NAME = "elasticJobRegistryCenter"; private final ElasticJobProperties elasticJobProperties; /** * 初始化Zookeeper * @return */ @Bean(name = DEFAULT_REGISTRY_CENTER_NAME, initMethod = "init") @ConditionalOnMissingBean public ZookeeperRegistryCenter regCenter() { ZookeeperRegistryProperties zookeeperRegistryProperties = elasticJobProperties.getZookeeper(); ZookeeperConfiguration zookeeperConfiguration = new ZookeeperConfiguration(zookeeperRegistryProperties.getServerLists(), zookeeperRegistryProperties.getNamespace()); zookeeperConfiguration.setBaseSleepTimeMilliseconds(zookeeperRegistryProperties.getBaseSleepTimeMilliseconds()); zookeeperConfiguration.setConnectionTimeoutMilliseconds(zookeeperRegistryProperties.getConnectionTimeoutMilliseconds()); zookeeperConfiguration.setMaxSleepTimeMilliseconds(zookeeperRegistryProperties.getMaxSleepTimeMilliseconds()); zookeeperConfiguration.setSessionTimeoutMilliseconds(zookeeperRegistryProperties.getSessionTimeoutMilliseconds()); zookeeperConfiguration.setMaxRetries(zookeeperRegistryProperties.getMaxRetries()); zookeeperConfiguration.setDigest(zookeeperRegistryProperties.getDigest()); return new ZookeeperRegistryCenter(zookeeperConfiguration); } /** * 初始化简单任务 * @return */ @Bean(initMethod = "init") @ConditionalOnMissingBean public SimpleJobInitialization simpleJobInitialization() { return new SimpleJobInitialization(elasticJobProperties.getSimples()); } /** * 流式任务初始化 * @return */ @Bean(initMethod = "init") @ConditionalOnMissingBean public DataflowJobInitialization dataflowJobInitialization() { return new DataflowJobInitialization(elasticJobProperties.getDataflows()); } /** * 脚本任务初始化 * @return */ @Bean(initMethod = "init") @ConditionalOnMissingBean @ConditionalOnBean(ZookeeperRegistryCenter.class) public ScriptJobInitialization scriptJobInitialization() { return new ScriptJobInitialization(elasticJobProperties.getScripts()); } /** * 动态任务初始化 * @return */ @Bean(initMethod = "init") @ConditionalOnMissingBean public DynamicJobInitialization dynamicJobInitialization() { return new DynamicJobInitialization(this.regCenter()); }}
AbstractJobInitialization核心基类,主要用作初始化任务的操作
/** * 任务初始化抽象类 * @author Nil * @date 2020/9/14 19:24 */public abstract class AbstractJobInitialization implements ApplicationContextAware { protected ApplicationContext applicationContext; @Override public void setApplicationContext(ApplicationContext applicationContext) { this.applicationContext = applicationContext; } /** * 初始化任务 * @param jobName 任务名 * @param jobType 任务类型 * @param configuration 配置 */ protected void initJob(String jobName, JobType jobType, ElasticJobProperties.JobConfiguration configuration) { //向spring容器中注册作业任务 ElasticJob elasticJob = registerElasticJob(jobName, configuration.getJobClass(), jobType); //获取注册中心 ZookeeperRegistryCenter regCenter = getZookeeperRegistryCenter(configuration.getRegistryCenterRef()); //构建核心配置 JobCoreConfiguration jobCoreConfiguration = getJobCoreConfiguration(jobName, configuration); //获取作业类型配置 JobTypeConfiguration jobTypeConfiguration = getJobTypeConfiguration(jobName, jobType, jobCoreConfiguration); //获取Lite作业配置 LiteJobConfiguration liteJobConfiguration = getLiteJobConfiguration(jobTypeConfiguration, configuration); //获取作业事件追踪的数据源配置 JobEventRdbConfiguration jobEventRdbConfiguration = getJobEventRdbConfiguration(configuration.getEventTraceRdbDataSource()); //获取作业监听器 ElasticJobListener[] elasticJobListeners = creatElasticJobListeners(configuration.getListener()); //注册作业 if (null == jobEventRdbConfiguration) { new SpringJobScheduler(elasticJob, regCenter, liteJobConfiguration, elasticJobListeners).init(); } else { new SpringJobScheduler(elasticJob, regCenter, liteJobConfiguration, jobEventRdbConfiguration, elasticJobListeners).init(); } } /** * 获取作业类型配置 * @param jobName 任务名称 * @param jobType 任务类型 * @param jobCoreConfiguration 任务核心配置 * @return JobTypeConfiguration */ public abstract JobTypeConfiguration getJobTypeConfiguration(String jobName, JobType jobType, JobCoreConfiguration jobCoreConfiguration); /** * 获取作业任务实例 * @param jobName 任务名称 * @param jobType 任务类型 * @param strClass 任务类全路径 * @return ElasticJob */ private ElasticJob registerElasticJob(String jobName, String strClass, JobType jobType) { switch (jobType) { case SIMPLE: return registerBean(jobName, strClass, SimpleJob.class); case DATAFLOW: return registerBean(jobName, strClass, DataflowJob.class); default: return null; } } /** * 获取注册中心 * @param registryCenterRef 注册中心引用 * @return ZookeeperRegistryCenter */ protected ZookeeperRegistryCenter getZookeeperRegistryCenter(String registryCenterRef) { if (StringUtils.isBlank(registryCenterRef)) { registryCenterRef = ElasticJobAutoConfiguration.DEFAULT_REGISTRY_CENTER_NAME; } if (!applicationContext.containsBean(registryCenterRef)) { throw new ServiceException("not exist ZookeeperRegistryCenter [" + registryCenterRef + "] !"); } return applicationContext.getBean(registryCenterRef, ZookeeperRegistryCenter.class); } /** * 获取作业事件追踪的数据源配置 * @param eventTraceRdbDataSource 作业事件追踪的数据源Bean引用 * @return JobEventRdbConfiguration */ private JobEventRdbConfiguration getJobEventRdbConfiguration(String eventTraceRdbDataSource) { if (StringUtils.isBlank(eventTraceRdbDataSource)) { return null; } if (!applicationContext.containsBean(eventTraceRdbDataSource)) { throw new ServiceException("not exist datasource [" + eventTraceRdbDataSource + "] !"); } DataSource dataSource = (DataSource) applicationContext.getBean(eventTraceRdbDataSource); return new JobEventRdbConfiguration(dataSource); } /** * 构建Lite作业 * @param jobTypeConfiguration 任务类型 * @param jobConfiguration 任务配置 * @return LiteJobConfiguration */ private LiteJobConfiguration getLiteJobConfiguration(JobTypeConfiguration jobTypeConfiguration, ElasticJobProperties.JobConfiguration jobConfiguration) { //构建Lite作业 return LiteJobConfiguration.newBuilder(Objects.requireNonNull(jobTypeConfiguration)) .monitorExecution(jobConfiguration.isMonitorExecution()) .monitorPort(jobConfiguration.getMonitorPort()) .maxTimeDiffSeconds(jobConfiguration.getMaxTimeDiffSeconds()) .jobShardingStrategyClass(jobConfiguration.getJobShardingStrategyClass()) .reconcileIntervalMinutes(jobConfiguration.getReconcileIntervalMinutes()) .disabled(jobConfiguration.isDisabled()) .overwrite(jobConfiguration.isOverwrite()).build(); } /** * 构建任务核心配置 * @param jobName 任务执行名称 * @param jobConfiguration 任务配置 * @return JobCoreConfiguration */ protected JobCoreConfiguration getJobCoreConfiguration(String jobName, ElasticJobProperties.JobConfiguration jobConfiguration) { JobCoreConfiguration.Builder builder = JobCoreConfiguration.newBuilder(jobName, jobConfiguration.getCron(), jobConfiguration.getShardingTotalCount()) .shardingItemParameters(jobConfiguration.getShardingItemParameters()) .jobParameter(jobConfiguration.getJobParameter()) .failover(jobConfiguration.isFailover()) .misfire(jobConfiguration.isMisfire()) .description(jobConfiguration.getDescription()); if (StringUtils.isNotBlank(jobConfiguration.getJobExceptionHandler())) { builder.jobProperties(JobProperties.JobPropertiesEnum.JOB_EXCEPTION_HANDLER.getKey(), jobConfiguration.getJobExceptionHandler()); } if (StringUtils.isNotBlank(jobConfiguration.getExecutorServiceHandler())) { builder.jobProperties(JobProperties.JobPropertiesEnum.EXECUTOR_SERVICE_HANDLER.getKey(), jobConfiguration.getExecutorServiceHandler()); } return builder.build(); } /** * 获取监听器 * @param listener 监听器配置 * @return ElasticJobListener[] */ private ElasticJobListener[] creatElasticJobListeners(ElasticJobProperties.JobConfiguration.Listener listener) { if (null == listener) { return new ElasticJobListener[0]; } List<ElasticJobListener> elasticJobListeners = new ArrayList<>(16); //注册每台作业节点均执行的监听 ElasticJobListener elasticJobListener = registerBean(listener.getListenerClass(), listener.getListenerClass(), ElasticJobListener.class); if (null != elasticJobListener) { elasticJobListeners.add(elasticJobListener); } //注册分布式监听者 AbstractDistributeOnceElasticJobListener distributedListener = registerBean(listener.getDistributedListenerClass(), listener.getDistributedListenerClass(), AbstractDistributeOnceElasticJobListener.class, listener.getStartedTimeoutMilliseconds(), listener.getCompletedTimeoutMilliseconds()); if (null != distributedListener) { elasticJobListeners.add(distributedListener); } if (CollUtil.isEmpty(elasticJobListeners)) { return new ElasticJobListener[0]; } //集合转数组 ElasticJobListener[] elasticJobListenerArray = new ElasticJobListener[elasticJobListeners.size()]; for (int i = 0; i < elasticJobListeners.size(); i++) { elasticJobListenerArray[i] = elasticJobListeners.get(i); } return elasticJobListenerArray; } /** * 向spring容器中注册bean * @param beanName bean名字 * @param strClass 类全路径 * @param tClass 类类型 * @param constructorArgValue 构造函数参数 * @param <T> 泛型 * @return T */ protected <T> T registerBean(String beanName, String strClass, Class<T> tClass, Object... constructorArgValue) { //判断是否配置了监听者 if (StringUtils.isBlank(strClass)) { return null; } if (StringUtils.isBlank(beanName)) { beanName = strClass; } //判断监听者是否已经在spring容器中存在 if (applicationContext.containsBean(beanName)) { return applicationContext.getBean(beanName, tClass); } //不存在则创建并注册到Spring容器中 BeanDefinitionBuilder beanDefinitionBuilder = BeanDefinitionBuilder.rootBeanDefinition(strClass); beanDefinitionBuilder.setScope(BeanDefinition.SCOPE_PROTOTYPE); //设置参数 for (Object argValue : constructorArgValue) { beanDefinitionBuilder.addConstructorArgValue(argValue); } getDefaultListableBeanFactory().registerBeanDefinition(beanName, beanDefinitionBuilder.getBeanDefinition()); return applicationContext.getBean(beanName, tClass); } /** * 获取beanFactory * @return DefaultListableBeanFactory */ private DefaultListableBeanFactory getDefaultListableBeanFactory() { return (DefaultListableBeanFactory) ((ConfigurableApplicationContext) applicationContext).getBeanFactory(); }}
分别初始化简单任务,流式任务,脚本任务
/** * 流式任务初始化 * @author Nil * @date 2020/9/14 19:15 */public class DataflowJobInitialization extends AbstractJobInitialization { private Map<String, ElasticJobProperties.DataflowConfiguration> dataflowConfigurationMap; public DataflowJobInitialization(Map<String, ElasticJobProperties.DataflowConfiguration> dataflowConfigurationMap) { this.dataflowConfigurationMap = dataflowConfigurationMap; } public void init() { if (CollUtil.isNotEmpty(dataflowConfigurationMap)) { dataflowConfigurationMap.forEach((k, v) -> { ElasticJobProperties.DataflowConfiguration configuration = dataflowConfigurationMap.get(k); super.initJob(k, configuration.getJobType(), configuration); }); } } @Override public JobTypeConfiguration getJobTypeConfiguration(String jobName, JobType jobType, JobCoreConfiguration jobCoreConfiguration) { ElasticJobProperties.DataflowConfiguration configuration = dataflowConfigurationMap.get(jobName); return new DataflowJobConfiguration(jobCoreConfiguration, configuration.getJobClass(), configuration.isStreamingProcess()); }}
/** * 简单任务初始化 * @author Nil * @date 2020/9/14 19:14 */public class SimpleJobInitialization extends AbstractJobInitialization { private Map<String, ElasticJobProperties.SimpleConfiguration> simpleConfigurationMap; public SimpleJobInitialization(Map<String, ElasticJobProperties.SimpleConfiguration> simpleConfigurationMap) { this.simpleConfigurationMap = simpleConfigurationMap; } public void init() { if (CollUtil.isNotEmpty(simpleConfigurationMap)) { simpleConfigurationMap.forEach((k, v) -> { ElasticJobProperties.SimpleConfiguration configuration = simpleConfigurationMap.get(k); super.initJob(k, configuration.getJobType(), configuration); }); } } @Override public JobTypeConfiguration getJobTypeConfiguration(String jobName, JobType jobType, JobCoreConfiguration jobCoreConfiguration) { ElasticJobProperties.SimpleConfiguration configuration = simpleConfigurationMap.get(jobName); return new SimpleJobConfiguration(jobCoreConfiguration, configuration.getJobClass()); }}
/** * 脚本任务初始化 * @author Nil * @date 2020/9/14 20:45 */public class ScriptJobInitialization extends AbstractJobInitialization { private Map<String, ElasticJobProperties.ScriptConfiguration> scriptConfigurationMap; public ScriptJobInitialization(final Map<String, ElasticJobProperties.ScriptConfiguration> scriptConfigurationMap) { this.scriptConfigurationMap = scriptConfigurationMap; } public void init() { if (CollUtil.isNotEmpty(scriptConfigurationMap)) { scriptConfigurationMap.forEach((k, v) -> { ElasticJobProperties.ScriptConfiguration configuration = scriptConfigurationMap.get(k); super.initJob(k, configuration.getJobType(), configuration); }); } } @Override public JobTypeConfiguration getJobTypeConfiguration(String jobName, JobType jobType, JobCoreConfiguration jobCoreConfiguration) { ElasticJobProperties.ScriptConfiguration configuration = scriptConfigurationMap.get(jobName); return new ScriptJobConfiguration(jobCoreConfiguration, configuration.getScriptCommandLine()); }}
下面就是核心部分,用来做动态任务的添加修改和删除,同时支持自动启动初始化任务,就可以满足大部分场景了。代码如下
/** * 动态任务初始化(支持简单、流式任务) * @author Nil * @date 2020/9/14 19:22 */@Slf4jpublic class DynamicJobInitialization extends AbstractJobInitialization { private JobStatisticsAPI jobStatisticsAPI; private JobSettingsAPI jobSettingsAPI; public DynamicJobInitialization(ZookeeperRegistryCenter zookeeperRegistryCenter) { this.jobStatisticsAPI = new JobStatisticsAPIImpl(zookeeperRegistryCenter); this.jobSettingsAPI = new JobSettingsAPIImpl(zookeeperRegistryCenter); } public void init() { Collection<JobBriefInfo> allJob = jobStatisticsAPI.getAllJobsBriefInfo(); if (CollUtil.isNotEmpty(allJob)) { allJob.forEach(jobInfo -> { // 已下线的任务 if (JobBriefInfo.JobStatus.CRASHED.equals(jobInfo.getStatus())) { try { Date currentDate = new Date(); CronExpression cronExpression = new CronExpression(jobInfo.getCron()); Date nextValidTimeAfter = cronExpression.getNextValidTimeAfter(currentDate); // 表达式还生效的任务 if (ObjectUtil.isNotNull(nextValidTimeAfter)) { this.initJobHandler(jobInfo.getJobName()); } } catch (ParseException e) { log.error(e.getMessage(), e); } } }); } } /** * 初始化任务操作 * @param jobName 任务名 */ private void initJobHandler(String jobName) { try { JobSettings jobSetting = jobSettingsAPI.getJobSettings(jobName); if (ObjectUtil.isNotNull(jobSetting)) { String jobCode = StrUtil.subBefore(jobSetting.getJobName(), StrUtil.UNDERLINE, false); JobClassEnum jobClassEnum = JobClassEnum.convert(jobCode); if (ObjectUtil.isNotNull(jobClassEnum)) { ElasticJobProperties.JobConfiguration configuration = new ElasticJobProperties.JobConfiguration(); configuration.setCron(jobSetting.getCron()); configuration.setJobParameter(jobSetting.getJobParameter()); configuration.setShardingTotalCount(jobSetting.getShardingTotalCount()); configuration.setDescription(jobSetting.getDescription()); configuration.setShardingItemParameters(jobSetting.getShardingItemParameters()); configuration.setJobClass(jobClassEnum.getClazz().getCanonicalName()); super.initJob(jobName, JobType.valueOf(jobSetting.getJobType()), configuration); } } } catch (Exception e) { log.error("初始化任务操作失败: {}", e.getMessage(), e); } } /** * 保存/更新任务 * @param job * @param jobClass */ public void addOrUpdateJob(Job job, Class<? extends ElasticJob> jobClass) { ElasticJobProperties.JobConfiguration configuration = new ElasticJobProperties.JobConfiguration(); configuration.setCron(job.getCron()); configuration.setJobParameter(job.getJobParameter()); configuration.setShardingTotalCount(job.getShardingTotalCount()); configuration.setShardingItemParameters(job.getShardingItemParameters()); configuration.setJobClass(jobClass.getCanonicalName()); super.initJob(job.getJobName(), JobType.valueOf(job.getJobType()), configuration); } @Override public JobTypeConfiguration getJobTypeConfiguration(String jobName, JobType jobType, JobCoreConfiguration jobCoreConfiguration) { String jobCode = StrUtil.subBefore(jobName, StrUtil.UNDERLINE, false); JobClassEnum jobClassEnum = JobClassEnum.convert(jobCode); if (ObjectUtil.isNotNull(jobClassEnum)) { if (JobType.SIMPLE.equals(jobType)) { return new SimpleJobConfiguration(jobCoreConfiguration, jobClassEnum.getClazz().getCanonicalName()); } else if (JobType.DATAFLOW.equals(jobType)) { return new DataflowJobConfiguration(jobCoreConfiguration, jobClassEnum.getClazz().getCanonicalName(), false); } } return null; }}
在ElasticJobAutoConfiguration类中初始化动态任务
/** * 动态任务初始化 * @return */ @Bean(initMethod = "init") @ConditionalOnMissingBean public DynamicJobInitialization dynamicJobInitialization() { return new DynamicJobInitialization(this.regCenter()); }
为什么是这样的实现?我发现每次重新发布服务后,现在的未执行的任务都会变成“已下线”,这可能跟Zookeeper有关,需要重新初始化才行,对于注解和配置式的,会自动初始化,但是动态添加的不会自动初始化。所以必须自己初始化,之前有个思路是自己建张表来维护定时,每次启动时进行初始化,但是这样太麻烦,后来实现使用elastic-job现有的API来实现,即启动时,遍历Zookeeper已有的节点,然后判断Cron表达式是否过期,如果还没有过期,则重新初始化任务,初始化时配置设置了会覆盖原来的配置,所以不会有影响。然后外层可以通过MQ来新增任务,在通过服务调用去指定对应的定时逻辑即可。
(不知道大家有没有更好的实现方案,可以初始化动态任务的)
而配置式的,可以直接在配置文件指定并实现即可
spring: elasticjob: #注册中心配置 zookeeper: server-lists: 127.0.0.1:6181 namespace: elastic-job-spring-boot-stater-demo #简单作业配置 simples: #spring简单作业示例配置 spring-simple-job: #配置简单作业,必须实现com.dangdang.ddframe.job.api.simple.SimpleJob job-class: com.zen.spring.boot.demo.elasticjob.job.SpringSimpleJob cron: 0/2 * * * * ? sharding-total-count: 3 sharding-item-parameters: 0=Beijing,1=Shanghai,2=Guangzhou #配置监听器 listener: #配置每台作业节点均执行的监听,必须实现com.dangdang.ddframe.job.lite.api.listener.ElasticJobListener listener-class: com.zen.spring.boot.demo.elasticjob.listener.MyElasticJobListener #流式作业配置 dataflows: #spring简单作业示例配置 spring-dataflow-job: #配置简单作业,必须实现com.dangdang.ddframe.job.api.dataflow.DataflowJob<T> job-class: com.zen.spring.boot.demo.elasticjob.job.SpringDataflowJob cron: 0/2 * * * * ? sharding-total-count: 3 sharding-item-parameters: 0=Beijing,1=Shanghai,2=Guangzhou streaming-process: true #配置监听器 listener: #配置分布式场景中仅单一节点执行的监听,必须实现com.dangdang.ddframe.job.lite.api.listener.AbstractDistributeOnceElasticJobListener distributed-listener-class: com.zen.spring.boot.demo.elasticjob.listener.MyDistributeElasticJobListener started-timeout-milliseconds: 5000 completed-timeout-milliseconds: 10000 #脚本作业配置 scripts: #脚本作业示例配置 script-job: cron: 0/2 * * * * ? sharding-total-count: 3 sharding-item-parameters: 0=Beijing,1=Shanghai,2=Guangzhou script-command-line: youPath/spring-boot-starter-demo/elastic-job-spring-boot-starter-demo/src/main/resources/script/demo.bat
以上整合基本可以满足现在的使用,如果大家有更好的整合方案也可以一起交流一下。比较期待移交Apache后的3的版本,这样可以有更多API的支持,而不用自己造轮子。
版权声明: 本文为 InfoQ 作者【Nil】的原创文章。
原文链接:【http://xie.infoq.cn/article/0b0cd01fe77d4da058bd7bc87】。文章转载请联系作者。
Nil
非宁静无以致远 2020.09.17 加入
还未添加个人简介
评论