Question

某些作业比我们的调度程序间隔运行的时间更长。跳过这些工作是非常安全的。如何检测具有相同名称和参数（运行时间除外）的作业当前是否正在运行？

Quartz将Batch作业运行为：

import org.springframework.batch.core.JobParametersBuilder;

@Override
protected void executeInternal(org.quartz.JobExecutionContext context) {
    Map<String, Object> jobDataMap = context.getMergedJobDataMap();
    String jobName = (String) jobDataMap.get("job.name");

    JobParametersBuilder builder = new JobParametersBuilder();
    builder.addString("company", jobDataMap.get("company"));
    builder.addDate("run.date", new Date());

    try {
        jobLauncher.run(jobLocator.getJob(jobName), jobParametersBuilder.toJobParameters());
    }
    catch (JobExecutionException e) {
        log.error("Could not execute job.", e);
    }
}

我需要查找特定"job.name" / "company"（无论"run.date"）的任何执行是否正在运行。

我可以使用针对BATCH_表运行的普通SQL来实现这一点。

检查应该在Tasklet中运行，以便JobRepository可以自动装配。是否可以仅使用JobRepository或其他Spring Batch bean找到它？

Answer 1

您可以从识别中排除“run.date”。

使用公开JobParameter(Date parameter, boolean identifying)

来自JobParameter的文档。

识别标志用于指示参数是否是用作识别作业实例的一部分。

所以只需添加不带builder.addDate("run.date", new Date());的参数并运行作业。

Answer 2

我的解决方案基于JobExplorer.findRunningJobExecutions()（找到最新未完成的工作）：

/**
 * If there are jobs with same parameters, that started later then some number we can run new job again.
 *
 * {@link JobExplorer#findRunningJobExecutions(String)} selects jobs with BATCH_JOB_EXECUTION.END_TIME is null.
 *
 * @return  null if there is no jobs with same parameters, otherwise the latest running.
 */
public JobExecution getLatestRunningJob(JobExplorer jobExplorer, String jobName, String agency, String branch) {
    long now = new Date().getTime();
    long minDiff = Long.MAX_VALUE;
    JobExecution latestExecution = null;
    Set<JobExecution> runningJobs = jobExplorer.findRunningJobExecutions(jobName);
    for (JobExecution execution : Optional.ofNullable(runningJobs).orElse(Collections.emptySet())) {
        JobParameters params = execution.getJobParameters();
        // log.warn("agencyName {}", params.getString("agencyName"));
        // log.warn("branchName {}", params.getString("branchName"));
        if ( ! agency.equals(params.getString("agencyName")))
            continue;
        if ( ! branch.equals(params.getString("branchName")))
            continue;
        // log.warn("create {}", execution.getCreateTime());
        // log.warn("start {}", execution.getStartTime());

        long diff = now - execution.getCreateTime().getTime();
        if (diff < 0) {
            log.warn("Impossible, new job executed before old! Old JobExecution id: {}", execution.getJobId());
            continue;
        }
        if (diff < minDiff) {
            minDiff = diff;
            latestExecution = execution;
        }
    }
    return latestExecution;
}

和决策者：

/**
 * @param execution  not null
 * @param limitMinutes  how long job should be runnning to ignore
 */
public boolean canIgnoreLateJob(JobExecution execution, long limitMinutes) {
    long now = new Date().getTime();
    long diffMinutes = TimeUnit.MINUTES.convert(now - execution.getCreateTime().getTime(), TimeUnit.MILLISECONDS);
    if (diffMinutes < limitMinutes) {
        log.warn("Recent JobExecution {} is still not finished, can't run new job, diff (minute): {}",
                execution.getJobId(), diffMinutes);
        return false;
    } else {
        log.warn("Very old JobExecution {} is still not finished, can run new job, diff (minute): {}",
                execution.getJobId(), diffMinutes);
        return true;
    }
}

此解决方案的缺点是JobExplorer解析JobExecution中的每个字段（包括所有步骤）。所以表现并不完美。但解决方案是可靠的。

如果具有相同参数（但不同的日期/时间）的作业正在进行中，则Spring批处理跳过作业

2 个答案: