Spring Batch查找开始的工作执行

Question

I have a Spring Batch application with one job. Job executions order is critical to me, so JobLauncher configured as a single thread job pool. In other words if few executions were triggered, one execution will be running, and other gonna wait in the queue.

Job executions table will look like that:

        ╒═════╤═════════════════════╤════════════╤════════════╤═════╕
        │ ... │ START_TIME          │ END_TIME   │ STATUS     │ ... │
        ╞═════╪═════════════════════╪════════════╪════════════╪═════╡
        │ ... │ 2019-07-11 11:03:08 │ NULL       │ STARTED    │ ... │
        ├─────┼─────────────────────┼────────────┼────────────┼─────┤
        │ ... │ NULL                │ NULL       │ STARTING   │ ... │
        └─────┴─────────────────────┴────────────┴────────────┴─────┘

If JVM crashes, I need to recover from that state and do some audit, so I need to find all STARTED and STARTING executions.

Find the first one is not a problem, I can do it with JobExplorer :

Set<JobExecution> executions = jobExplorer.findRunningJobExecutions("jobName");

It will give me only STARTED execution with present START_TIME . But how to find all STARTING executions?

The only way I've found so far is to iterate over the job instances, then find STARTING execution for every instance. Something like this:

        jobExplorer.findJobInstancesByJobName("jobName", 0, 100)
                .forEach(jobInstance -> {
                    jobExplorer.getJobExecutions(jobInstance)
                            .stream()
                            .filter(execution -> STARTING.equals(execution.getStatus()))
                            .forEach(execution -> {
                                // do the job
                            });
                });

The problem here that you always doing a full scan with a limit (I specified 100). Are there any better way like jobExplorer.findStartingJobExecutions("jobName") in Spring Batch?

Answer 1

I see no public APIs to get a job execution by status. What you can do is (in pseudo code):

JobInstance jobInstance = jobInstanceDao.getJobInstance(jobName, jobParameters);
List<JobExecution> jobExecutions =  jobExplorer.findJobExecutions(jobInstance);
// iterate over job execution and filter by status

This will avoid loading all job instances (one hundred at a time).

That said, I think the JobRepository interface could add a method to get a job instance by name and parameters (similar to isJobInstanceExists ) to avoid using the jobInstanceDao as shown in my previous snippet.

EDIT:

I guess your initial requirement is to mark a job execution as failed to be able to restart it. For this, you can manually change the status of job execution (as well as its step executions) to FAILED and set its END_TIME to a non null value (This is explained here: https://docs.spring.io/spring-batch/4.1.x/reference/html/job.html#aborting-a-job ). With that, you will be able to restart failed executions.