Quartz JobDataMap数据存储存在的坑

场景

现在有这么一个场景，我们要统计一个job执行了多少次，然后打印出来. 这里我们使用JobDataMap展示一个错误案例。

错误实现

less 复制代码

@Slf4j
class JobCountTest {
    @Test
    @SneakyThrows
    @SuppressWarnings("all")
    void run() {
        StdSchedulerFactory stdSchedulerFactory = new StdSchedulerFactory();
        Scheduler scheduler = stdSchedulerFactory.getScheduler();
        JobDetail job = newJob(CounterJob.class)
                .withIdentity("job1", "group1")
                .build();

        // 一秒后执行，三秒一次
        Integer repeatCount = 6;
        Integer intervalSeconds = 3;
        SimpleTrigger trigger = newTrigger()
                .withIdentity("trigger1", "group1")
                .startAt(Date.from(Instant.now().plusSeconds(1)))
                .withSchedule(simpleSchedule()
                        .withIntervalInSeconds(intervalSeconds)
                        .withRepeatCount(repeatCount))
                .build();

        // 初始化
        job.getJobDataMap().put("counter", 0);
        scheduler.scheduleJob(job, trigger);
        scheduler.start();

        Thread.sleep(repeatCount * intervalSeconds * 1000 + 1);
        scheduler.shutdown();
        assertEquals(repeatCount, job.getJobDataMap().getInt("counter"));
    }
}

less 复制代码

@Slf4j
@PersistJobDataAfterExecution
@DisallowConcurrentExecution
public class CounterJob implements Job {

    @Override
    public void execute(JobExecutionContext context) {
        jobDataMap.put("counter", context.getJobDetail().getJobDataMap().getInt("counter") + 1);
        log.info("--- job: " + context.getJobDetail().getKey() + " executed.");
    }
}

结果

分析

经过简单的debug会发现counter在CounerJob的execute方法执行中确实叠加了，但是当所有job执行结束时发现counter的累积值消失了，并且发现在test中获取的jobDataMap和CounterJob利用context获取的jobDataMap不是同一个对象。所以经过初步分析怀疑jobDataMap对象可能是拷贝对象。进入源码发现

jobDataMap的维护非常简单，并不是拷贝，那我们只能怀疑它的封装对象jobDetail是拷贝了，从这里看起

我们的数据是存储在内存的，所以应该是这个RAMJobStore

最后这里印证了我们的猜想，scheduleJob时引用的JobDetail和执行时引用的JobDetail实例并不是同一个而是clone对象，所以他们的jobDataMap自然也就不一样, 使用context获取的jobDataMap和test中获取的jobDataMap更不是同一个对象！

正确代码

由于quartz的job是多线程的，我们只要有办法做到线程共享数据就行，这里我们用最简单的方式实现，静态变量。

less 复制代码

@Slf4j
@PersistJobDataAfterExecution
@DisallowConcurrentExecution
public class CounterJob implements Job {
   public static Integer counter;

   @Override
   public void execute(JobExecutionContext context) {
       JobDataMap jobDataMap = context.getJobDetail().getJobDataMap();
       counter = jobDataMap.getInt("counter") + 1;
       jobDataMap.put("counter", counter);
       log.info("--- job: " + context.getJobDetail().getKey() + " executed.");
   }
}

less 复制代码

@Slf4j
class JobCountTest {
    @Test
    @SneakyThrows
    @SuppressWarnings("all")
    void run() {
        StdSchedulerFactory stdSchedulerFactory = new StdSchedulerFactory();
        Scheduler scheduler = stdSchedulerFactory.getScheduler();
        JobDetail job = newJob(CounterJob.class)
                .withIdentity("job1", "group1")
                .build();

        // 一秒后执行，三秒一次
        Integer repeatCount = 6;
        Integer intervalSeconds = 3;
        SimpleTrigger trigger = newTrigger()
                .withIdentity("trigger1", "group1")
                .startAt(Date.from(Instant.now().plusSeconds(1)))
                .withSchedule(simpleSchedule()
                        .withIntervalInSeconds(intervalSeconds)
                        .withRepeatCount(repeatCount))
                .build();

        // 初始化
        job.getJobDataMap().put("counter", 0);
        scheduler.scheduleJob(job, trigger);
        scheduler.start();

        Thread.sleep(repeatCount * intervalSeconds * 1000 + 1);
        scheduler.shutdown();
//        assertEquals(repeatCount, job.getJobDataMap().getInt("counter"));
        assertEquals(repeatCount, CounterJob.counter);
    }
}