如何为Hadoop的Map-reduce作业设置配置

Question

假设我想为MR作业设置这些配置:

mapred.map.tasks
mapred.reduce.tasks
mapred.tasktracker.map.tasks.maximum
mapred.tasktracker.reduce.tasks.maximum
mapred.reduce.slowstart.completed.maps

有哪些可能的方法可以设置这个?

我可以在mapred-site.xml中设置。但这将适用于我运行的所有作业。

如果我想为单个作业设置这些，这是否有效:

conf.set("mapred.tasktracker.map.tasks.maximum", 10)

(我没有在任何地方看到这种类型)

或仅通过命令行参数

如-D mapred.tasktracker.map.tasks.maximum=10

(这似乎是更常见的用法)

Answer 1

两种方式都是有效的，您可以在开始作业之前以任何方式编辑配置

Answer 2

解决方案1:创建一个BaseJob类:

public abstract class BaseJob extends Configured implements Tool {
// method to set the configuration for the job and the mapper and the reducer classes
protected Job setupJob(Transformation transformation, final Configuration conf) throws Exception {
    //Get the job object from the global configuration
    Job job = new Job(conf);
    //Set the transformation specific details
    if(transformation.getMapperClass() != null)
    job.setMapperClass(transformation.getMapperClass());
    if(transformation.getReducerClass() != null)
    job.setReducerClass(transformation.getReducerClass());
    if(transformation.getMapOutputKeyClass() != null)
    job.setMapOutputKeyClass(transformation.getMapOutputKeyClass());
    if(transformation.getMapOutputValueClass() != null)
    job.setMapOutputValueClass(transformation.getMapOutputValueClass());
    if(transformation.getPartitionerClass() != null)
    job.setPartitionerClass(transformation.getPartitionerClass());
    if(transformation.getSortComparatorClass() != null)
    job.setSortComparatorClass(transformation.getSortComparatorClass());
    if(transformation.getGroupingComparator() != null)
    job.setGroupingComparatorClass(transformation.getGroupingComparator());
    if(transformation.getInputFormatClass() != null)
    job.setInputFormatClass(transformation.getInputFormatClass());
    if(transformation.getOutputKeyClass() != null)
    job.setOutputKeyClass(transformation.getOutputKeyClass());
    if(transformation.getOutputValueClass() != null)
    job.setOutputValueClass(transformation.getOutputValueClass());
    if(transformation.getJarByClass() != null)
    job.setJarByClass(transformation.getJarByClass());
    return job;
}
protected abstract class Transformation {
    public abstract Class<?> getJarByClass();
    public abstract Class<? extends Mapper> getMapperClass();
    public abstract Class<? extends Reducer> getCombinerClass();
    public abstract Class<? extends Reducer> getReducerClass();
    public abstract Class<?> getOutputKeyClass();
    public abstract Class<?> getOutputValueClass();
    public abstract Class<?> getMapOutputKeyClass();
    public abstract Class<?> getMapOutputValueClass();
    public abstract Class<? extends Partitioner> getPartitionerClass();
    public abstract Class<? extends WritableComparator> getSortComparatorClass();
    public abstract Class<? extends WritableComparator> getGroupingComparator();
    public abstract Class<? extends InputFormat<?,?>> getInputFormatClass();
    public abstract Class<? extends OutputFormat<?,?>> getOutputFormatClass();
}

}

然后编写MyTransformationJob类并设置配置

 public class MyTransformationJob extends BaseJob {
           private Job getJobConf(final Configuration conf) throws Exception {

        Transformation tranformation = new Transformation() {
            @Override
            public Class<? extends Reducer> getCombinerClass() {
                return null;
            }
            @Override
            public Class<?> getJarByClass() {
                return MyTransformationJob .class;
            }
            @Override
            public Class<? extends Mapper> getMapperClass() {
                return MyMapper.class;
            }
            @Override
            public Class<?> getOutputKeyClass() {
                return Text.class;
            }
            @Override
            public Class<?> getOutputValueClass() {
                return NullWritable.class;
            }
            @Override
            public Class<? extends Reducer> getReducerClass() {
                if(StringUtils.equals(jobParams[3], "header")){
                    return HeaderReducer.class;
                }
                return ValuesReducer.class;
            }
            @Override
            public Class<?> getMapOutputKeyClass() {
                return Text.class;
            }
            @Override
            public Class<?> getMapOutputValueClass() {
                return LinkedMapWritable.class;
            }
            @Override
            public Class<? extends Partitioner> getPartitionerClass() {
                return StationKeyPartitioner.class;
            }
            @Override
            public Class<? extends WritableComparator> getSortComparatorClass() {
                return StationKeySortComparator.class;
            }
            @Override
            public Class<? extends WritableComparator> getGroupingComparator() {
                return UniqueIdGroupingComparator.class;
            }
            @Override
            public Class<? extends InputFormat<?,?>> getInputFormatClass() {
                return KeyValueTextInputFormat.class;
            }
            @Override
            public Class<? extends OutputFormat<?,?>> getOutputFormatClass() {
                return null;
            }
        };
        return setupJob(tranformation,conf);
    }  
 }

这样你就可以用不同的配置和类来指定多个作业。

解决方案2:

您可以创建本地配置并指定您提到的值

示例测试类:

public class ConfigurationTest extends TestCase {
  @Test
  public void test() throws IOException {
    Configuration conf = new Configuration();
    conf.addResource("hadoop-local.xml");
    assertThat(conf.get("mapred.reduce.tasks"), is("2"));
}
}