Spring Batch JSR-352重新启动了错误的分区



我使用Spring batch JSR-352实现创建了一个简单的批处理。批处理使用PartitionMapper为每个区块注入一个单独的属性。根据JSR-352规范(当partitionsOverride=False时),当其中一个分区块失败并且批处理重新启动时,只有失败的分区块才应该重新启动。

例如,如果我们有3个分区:分区0、分区1和分区2。如果partition1和partition2失败,则批处理应仅使用各自的批处理属性重新启动partition1或partition2。

然而,我注意到,当使用Spring Batch JSR-352实现(最新版本3.0.3.Release)时,重新启动批处理将重新启动partition0和partition1,而不是partition1和partition2。因此,它正确地检测到两个分区发生了故障,但它错误地重新启动了前(两个)分区,而不是应该重新启动的故障分区。

这是SpringBatch实现中的一个错误,还是我遗漏了什么?

参见JSR-352文档第10.8.5节:http://download.oracle.com/otn-pub/jcp/batch-1_0_revA-mrel-spec/JSR_352-v1.0_Rev_a-Maintenance_Release.pdf

这是我使用过的代码:

/META-INF/批处理作业/sampleBatch.xml

<?xml version="1.0" encoding="UTF-8"?>
<job id="sampleBatch" version="1.0" xmlns="http://xmlns.jcp.org/xml/ns/javaee" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
 xsi:schemaLocation="http://xmlns.jcp.org/xml/ns/javaee http://xmlns.jcp.org/xml/ns/javaee/jobXML_1_0.xsd">
<step id="sampleStep">
    <chunk item-count="100">
        <reader ref="com.springapp.batch.SampleReader">
            <properties>
                <property name="sample" value="#{partitionPlan['sample']}"/>
            </properties>
        </reader>
        <writer ref="com.springapp.batch.SampleWriter">
            <properties>
                <property name="sample" value="#{partitionPlan['sample']}"/>
            </properties>
        </writer>
    </chunk>
    <partition>
        <mapper ref="com.springapp.batch.SamplePartitionMapper"/>
    </partition>
</step>
</job>

com.springpapp.batch.SamplePartitionMapper:

package com.springapp.batch;
import java.util.Properties;
import javax.batch.api.partition.PartitionMapper;
import javax.batch.api.partition.PartitionPlan;
import javax.batch.api.partition.PartitionPlanImpl;
public class SamplePartitionMapper implements PartitionMapper {
@Override
public PartitionPlan mapPartitions() throws Exception {
    final PartitionPlan partitionPlan = new PartitionPlanImpl();
    int size = 3;
    Properties[] partitionProps = new Properties[size];
    for (int i=0; i<size; i++) {
        final Properties properties = new Properties();
        properties.put("sample", ""+i);
        partitionProps[i] = properties;
        System.out.println("mapPartitions: " + i);
    }
    partitionPlan.setThreads(1);
    partitionPlan.setPartitions(partitionProps.length);
    partitionPlan.setPartitionProperties(partitionProps);
    return partitionPlan;
}
}

com.springpapp.batch.SampleReader:

public class SampleReader extends AbstractItemReader {
@Inject
@BatchProperty
private String sample;
Iterator<Integer> iter;
@Override
public void open(Serializable checkpoint) throws Exception {
    System.out.println("open for reading sample: " + sample);
    ArrayList list = new ArrayList<Integer>();
    for(int i=0; i<Integer.parseInt(sample); i++) {
        list.add(new Integer(i));
    }
    iter = list.iterator();
}
@Override
public Integer readItem() throws Exception {
    if(iter.hasNext())
        return iter.next();
    else
        return null;
}
}

com.springpapp.batch.SampleWriter:

public class SampleWriter extends AbstractItemWriter {
@Inject
@BatchProperty
private String sample;
@Override
public void writeItems(List<Object> items) throws Exception {
    System.out.println("writeItems sample: " + sample);
    if(sample.equals("1")) {
        throw new Exception("FAIL PARTITION 1");
    }
    if(sample.equals("2")) {
        throw new Exception("FAIL PARTITION 2");
    }
    for (Object itemObj : items) {
        Integer item = (Integer) itemObj;
        System.out.println(item);
    }
}
}

TestJob测试运行程序:

package com.springapp.batch;
import java.util.Properties;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.TimeoutException;
import javax.batch.operations.JobOperator;
import javax.batch.runtime.BatchRuntime;
import javax.batch.runtime.JobExecution;
import javax.inject.Inject;
import junit.framework.Assert;
import org.junit.Test;
//@RunWith(SpringJUnit4ClassRunner.class)
//@ContextConfiguration("classpath:spring-config.xml")
public class AppTests {
@Test
public void testJob() throws Exception {
    JobOperator jobOperator = BatchRuntime.getJobOperator();
    long jobExecution = jobOperator.start("sampleBatch", new Properties());
    int attempt = 0;
    while (true) {
        JobExecution execution = jobOperator.getJobExecution(jobExecution);
        if (execution.getEndTime() != null) {
            //check status
            if( "FAILED".equals(execution.getExitStatus()) && attempt < 3 ) {
                attempt++;
                System.out.println("Batch failed, trying to restart (attempt " + attempt + ")..");
                jobExecution = jobOperator.restart(jobExecution,  new Properties());
                continue;
            }
            System.out.println("Batch ended with status: " + execution.getExitStatus());
            break;
        }
    }
    Assert.assertEquals("COMPLETED", jobOperator.getJobExecution(jobExecution).getExitStatus());
}
}

看看这个,它看起来像一个bug。我已经记录了Jira问题BATCH-2364来跟踪它。

最新更新