如何从Spring Boot Batch应用中终止AWS Batch作业

首先明确：AWS Batch作业执行完成（成功或失败）后，默认会自动停止对应的ECS容器，大多数场景下无需手动干预。如果你的业务需求必须通过Spring Batch代码主动触发终止操作，可参考以下方案：

方案1：通过Spring Batch作业监听器调用AWS SDK #

步骤1：引入AWS SDK依赖 #

在项目构建文件（如Maven pom.xml）中添加AWS Batch和ECS的SDK依赖：

<dependency>
    <groupId>software.amazon.awssdk</groupId>
    <artifactId>batch</artifactId>
</dependency>
<dependency>
    <groupId>software.amazon.awssdk</groupId>
    <artifactId>ecs</artifactId>
</dependency>

步骤2：实现作业完成监听器 #

创建Spring Batch的JobExecutionListener，在作业成功完成时调用AWS API终止作业或容器：

@Component
public class JobCompletionTerminationListener implements JobExecutionListener {

    private final BatchClient batchClient;
    private final EcsClient ecsClient;

    public JobCompletionTerminationListener(BatchClient batchClient, EcsClient ecsClient) {
        this.batchClient = batchClient;
        this.ecsClient = ecsClient;
    }

    @Override
    public void afterJob(JobExecution jobExecution) {
        if (BatchStatus.COMPLETED.equals(jobExecution.getStatus())) {
            // 从作业参数中获取AWS Batch作业ID（提交作业时需传入该参数）
            String awsBatchJobId = jobExecution.getJobParameters().getString("awsBatchJobId");
            
            // 终止AWS Batch作业
            TerminateJobRequest terminateJobReq = TerminateJobRequest.builder()
                    .jobId(awsBatchJobId)
                    .reason("Job completed successfully, initiating termination")
                    .build();
            batchClient.terminateJob(terminateJobReq);

            // （可选）直接终止对应的ECS容器
            DescribeJobsRequest describeJobsReq = DescribeJobsRequest.builder()
                    .jobs(awsBatchJobId)
                    .build();
            Job batchJob = batchClient.describeJobs(describeJobsReq).jobs().get(0);
            String ecsTaskArn = batchJob.container().taskArn();
            
            // 解析任务ARN获取集群名称和任务ID
            String[] arnSegments = ecsTaskArn.split("/");
            String clusterName = arnSegments[1];
            String taskId = arnSegments[2];
            
            StopTaskRequest stopTaskReq = StopTaskRequest.builder()
                    .cluster(clusterName)
                    .task(taskId)
                    .reason("Job completed, stopping associated ECS task")
                    .build();
            ecsClient.stopTask(stopTaskReq);
        }
    }
}

步骤3：注册监听器到作业 #

将监听器绑定到你的Spring Batch作业定义：

@Bean
public Job batchProcessingJob(JobBuilderFactory jobBuilders, Step processingStep,
                              JobCompletionTerminationListener terminationListener) {
    return jobBuilders.get("batchProcessingJob")
            .listener(terminationListener)
            .flow(processingStep)
            .end()
            .build();
}

方案2：确认AWS Batch自动终止配置 #

如果不需要代码干预，检查以下配置确保容器自动终止：

• 作业定义中，容器的essential属性设为true：主进程（Spring Batch应用）退出后，容器会自动停止
• 计算环境配置：按需/Spot计算环境会在无作业时自动缩减实例，避免资源浪费

权限说明 #

确保运行Spring Batch的容器拥有对应的IAM权限，需包含以下权限：

• batch:TerminateJob
• batch:DescribeJobs
• ecs:StopTask

内容的提问来源于stack exchange，提问作者Shuvan Sarkar