Spring Batch after JVM crash

Question

How to restart Jobs after a JVM crash?

I was running a lot of Jobs implemented in Spring Batch framework, when my JVM crashed or the system failed. How can I restart these Jobs after failure?

Answer 1

You need to mark the "running" jobs as failed before restarting them, like this:

List<String> jobs = jobExplorer.getJobNames();
for (String job : jobs) {
    Set<JobExecution> runningJobs = jobExplorer.findRunningJobExecutions(job);

    for (JobExecution runningJob : runningJobs) {
        try {
            runningJob.setStatus(BatchStatus.FAILED);
            runningJob.setEndTime(new Date());
            jobRepository.update(runningJob);
            jobOperator.restart(runningJob.getId());
        } catch (Exception e) {
            LOG.error(e.getMessage(), e);
        }
    }
}

Answer 2

Basically, you can do as follows:

1. Configure a JobExplorer factory bean in your application context:

2. Configure a JobOperator bean in your applictaion context

3. Query the jobExplorer for distinct job names: jobExplorer.getJobNames()

4. For each job from step (3), query the jobExplorer for unfinished jobs: jobExplorer.findRunningJobExecutions(String jobName)

5. For each JobExecution from step (4) invoke:jobOperator.restart(jobExecution.getJobId())

6. Make sure to call this procedure during boot sequence, before any other job is launched

Technically it is possible to merge steps 3+4 for something like findRunningJobExecutions() by overriding JobExecutionDao, but the current API doesn't support it.

For help in aforementioned Spring bean configuration, consult the reference documentation