Skip to main content

Schedule Java EE 7 Batch Jobs (Tech Tip #36)

Posted by arungupta on July 8, 2014 at 7:04 AM PDT

Java EE 7 added the capability to perform Batch jobs in a standard way using JSR 352.

<job id="myJob" xmlns="http://xmlns.jcp.org/xml/ns/javaee" version="1.0">
  <step id="myStep">
    <chunk item-count="3">
    <reader ref="myItemReader"/>
    <processor ref="myItemProcessor"/>
  <writer ref="myItemWriter"/>
</chunk>

This code fragment is the Job Specification Language defined as XML, a.k.a. Job XML. It defines a canonical job, with a single step, using item-oriented or chunk-oriented processing. A chunk can have a reader, optional processor, and a writer. Each of these elements are identified using the corresponding elements in the Job XML, and are CDI beans packaged in the archive.

This job can be easily started using:
BatchRuntime.getJobOperator().start("myJob", new Properties());
A typical question asked in different forums and conferences is how to schedule these jobs in a Java EE runtime. Batch 1.0 API itself does not offer anything to be schedule these jobs. However Java EE platform offers three different ways to schedule these jobs:

  1. Use the @javax.ejb.Schedule annotation in an EJB.
    Here is a sample code that will trigger the execution of batch job at 11:59:59 PM every day.
    @Singleton
    public class MyEJB {
      @Schedule(hour = "23", minute = "59", second = "59")
      public void myJob() {
        BatchRuntime.getJobOperator().start("myJob", new Properties());
      }
    }

    Of course, you can change the parameters of @Schedule to start the batch job at the desired time.
  2. Use ManagedScheduledExecutorService using javax.enterprise.concurrent.Trigger as shown:
    @Stateless
    public class MyStatelessEJB {
        @Resource
        ManagedScheduledExecutorService executor;

        public void runJob() {
            executor.schedule(new MyJob(), new Trigger() {

                public Date getNextRunTime(LastExecution lastExecutionInfo, Date taskScheduledTime) {
                    Calendar cal = Calendar.getInstance();
                    cal.setTime(taskScheduledTime);
                    cal.add(Calendar.DATE, 1);
                    return cal.getTime();
                }

                public boolean skipRun(LastExecution lastExecutionInfo, Date scheduledRunTime) {
                    return null == lastExecutionInfo;
                }

            });
        }

        public void cancelJob() {
            executor.shutdown();
        }
    }

    Call runJob to initiate job execution and cancelJob to terminate job execution. In this case, a new job is started a day later than the previous task. And its not started until previous one is terminated. You will need more error checks for proper execution.

    MyJob is very trivial:

    public class MyJob implements Runnable {

        public void run() {
            BatchRuntime.getJobOperator().start("myJob", new Properties());
        }

    }

    Of course, you can automatically schedule it by calling this code in @PostConstruct.
  3. A slight variation of second technique allows to run the job after a fixed delay as shown:
    public void runJob2() {
        executor.scheduleWithFixedDelay(new MyJob(), 2, 3, TimeUnit.HOURS);
    }

    The first task is executed 2 hours after the runJob2 method is called. And then with a 3 hours delay between subsequent execution.

This support is available to you within the Java EE platform. In addition, you can also invoke BatchRuntime.getJobOperator().start("myJob", new Properties()); from any of your Quartz-scheduled methods as well.

You can try all of this on WildFly.

And there are a ton of Java EE 7 samples at github.com/javaee-samples/javaee7-samples.

This particular sample is available at github.com/javaee-samples/javaee7-samples/tree/master/batch/scheduling.

How are you scheduling your Batch jobs ?