问题
Background
I'm using Spring Batch 2.1.8, and run jobs by CommandLineJobRunner. Such as:
java org.springframework.batch.core.launch.support.CommandLineJobRunner classpath:launchContext.xml theJobId
Problem
At some condition such as a server crash, running job could be interrupted.
But the interrupted job left a STARTED status in the Spring Batch Meta-Data tables, and can't be run again.
org.springframework.batch.core.repository.JobExecutionAlreadyRunningException: A job execution for this job is already running
I can think of two solutions:
Solution1
Add a new job parameter and change it everytime to make it a "new" job for Spring Batch. Such as:
java org.springframework.batch.core.launch.support.CommandLineJobRunner classpath:launchContext.xml theJobId times=0
And when need to rerun it, do clear all corresponding output data, count up times once, and then rerun the job.
Solution2
Change the Spring Batch Meta-Data tables manually.
To update the status to make the job restartable. Such as:
UPDATE BATCH_JOB_EXECUTION SET END_TIME = SYSTIMESTAMP, STATUS = 'FAILED', EXIT_CODE = 'FAILOVER' WHERE JOB_EXECUTION_ID =
(SELECT MAX(JOB_EXECUTION_ID) FROM BATCH_JOB_EXECUTION WHERE JOB_INSTANCE_ID =
(SELECT MAX(JOB_INSTANCE_ID) FROM BATCH_JOB_INSTANCE WHERE JOB_NAME = 'XXX'));
I've tried it and it seems works well.
Question
Is Solution2 a bad idea? Are there any traps?
Thanks in advance. And any other solutions are appreciated.
回答1:
Solution 2 is the accepted approach right now. The API does not provide a way to fix this scenario. There have been requests in the past for the framework to clean up automatically, but 99% of the time, a human decision is needed to determine if cleanup is truly required.
My only note for option 2 would be to check the BATCH_STEP_EXECUTION table as well to see what state the last executed step was left in.
回答2:
I created a specific spring bean for this which is triggered on a container refresh (which happens on app (re)start too).
It searches for 'running' jobs, marks them 'FAILED' and restarts them.
import java.util.Date;
import java.util.List;
import java.util.Set;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.batch.core.BatchStatus;
import org.springframework.batch.core.JobExecution;
import org.springframework.batch.core.explore.JobExplorer;
import org.springframework.batch.core.launch.JobOperator;
import org.springframework.batch.core.repository.JobRepository;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.context.ApplicationListener;
import org.springframework.context.event.ContextRefreshedEvent;
import org.springframework.stereotype.Component;
@Component
public class BatchJobRestarter implements ApplicationListener<ContextRefreshedEvent> {
private static final Logger LOGGER = LoggerFactory.getLogger(BatchJobRestarter.class);
@Autowired
private JobExplorer jobExplorer;
@Autowired
JobRepository jobRepository;
@Autowired
JobOperator jobOperator;
@Override
public void onApplicationEvent(ContextRefreshedEvent contextRefreshedEvent) {
LOGGER.info("Container restart: restarting 'running' batch jobs");
List<String> jobs = jobExplorer.getJobNames();
for (String job : jobs) {
Set<JobExecution> runningJobs = jobExplorer.findRunningJobExecutions(job);
for (JobExecution runningJob : runningJobs) {
try {
LOGGER.info("Restarting job {} with parameters {}", runningJob.getJobInstance().getJobName(), runningJob.getJobParameters().toString());
runningJob.setStatus(BatchStatus.FAILED);
runningJob.setEndTime(new Date());
jobRepository.update(runningJob);
jobOperator.restart(runningJob.getId());
} catch (Exception e) {
LOGGER.error(e.getMessage(), e);
}
}
}
}
}
Steef
来源:https://stackoverflow.com/questions/22775233/is-it-a-bad-idea-to-change-the-spring-batch-meta-data-tables-manually