Hadoop 中的 MultiOutputFormat

2022-09-03 07:56:21

我是Hadoop的新手。我正在尝试Wordcount程序。

现在要尝试多个输出文件,我使用.这个链接帮助我做到了。http://hadoop.apache.org/common/docs/r0.19.0/api/org/apache/hadoop/mapred/lib/MultipleOutputs.htmlMultipleOutputFormat

在我的司机班上,我有

    MultipleOutputs.addNamedOutput(conf, "even",
            org.apache.hadoop.mapred.TextOutputFormat.class, Text.class,
            IntWritable.class);

    MultipleOutputs.addNamedOutput(conf, "odd",
            org.apache.hadoop.mapred.TextOutputFormat.class, Text.class,
            IntWritable.class);`

我的简化类变成了这个

public static class Reduce extends MapReduceBase implements
        Reducer<Text, IntWritable, Text, IntWritable> {
    MultipleOutputs mos = null;

    public void configure(JobConf job) {
        mos = new MultipleOutputs(job);
    }

    public void reduce(Text key, Iterator<IntWritable> values,
            OutputCollector<Text, IntWritable> output, Reporter reporter)
            throws IOException {
        int sum = 0;
        while (values.hasNext()) {
            sum += values.next().get();
        }
        if (sum % 2 == 0) {
            mos.getCollector("even", reporter).collect(key, new IntWritable(sum));
        }else {
            mos.getCollector("odd", reporter).collect(key, new IntWritable(sum));
        }
        //output.collect(key, new IntWritable(sum));
    }
    @Override
    public void close() throws IOException {
        // TODO Auto-generated method stub
    mos.close();
    }
}

事情工作正常,但我得到很多文件,(一个奇数和一个偶数对于每个map-reduce)

问题是:我怎么能只有2个输出文件(奇数和偶数),以便每个map-reduce的每个奇数输出都被写入该奇数文件,偶数也是如此。


答案 1

每个化简器都使用输出格式将记录写入。这就是为什么你得到一组奇数和偶数文件每个化简器。这是设计使然,以便每个化简器可以并行执行写入操作。

如果只需要一个奇数和单个偶数文件,则需要将 mapred.reduce.tasks 设置为 1。但性能会受到影响,因为所有映射器都将馈送到单个化简器中。

另一种选择是更改读取这些文件的过程以接受多个输入文件,或者编写一个单独的过程将这些文件合并在一起。


答案 2

我为此写了一门课。只需使用它你的工作:

job.setOutputFormatClass(m_customOutputFormatClass);

这是我的类:

import java.io.IOException;
import java.util.HashMap;
import java.util.Map;
import java.util.Map.Entry;

import org.apache.hadoop.fs.Path;
import org.apache.hadoop.mapreduce.RecordWriter;
import org.apache.hadoop.mapreduce.TaskAttemptContext;
import org.apache.hadoop.mapreduce.lib.output.TextOutputFormat;

/**
 * TextOutputFormat extension which enables writing the mapper/reducer's output in multiple files.<br>
 * <p>
 * <b>WARNING</b>: The number of different folder shuoldn't be large for one mapper since we keep an
 * {@link RecordWriter} instance per folder name.
 * </p>
 * <p>
 * In this class the folder name is defined by the written entry's key.<br>
 * To change this behavior simply extend this class and override the
 * {@link HdMultipleFileOutputFormat#getFolderNameExtractor()} method and create your own
 * {@link FolderNameExtractor} implementation.
 * </p>
 * 
 * 
 * @author ykesten
 * 
 * @param <K> - Keys type
 * @param <V> - Values type
 */
public class HdMultipleFileOutputFormat<K, V> extends TextOutputFormat<K, V> {

    private String folderName;

    private class MultipleFilesRecordWriter extends RecordWriter<K, V> {

        private Map<String, RecordWriter<K, V>> fileNameToWriter;
        private FolderNameExtractor<K, V> fileNameExtractor;
        private TaskAttemptContext job;

        public MultipleFilesRecordWriter(FolderNameExtractor<K, V> fileNameExtractor, TaskAttemptContext job) {
            fileNameToWriter = new HashMap<String, RecordWriter<K, V>>();
            this.fileNameExtractor = fileNameExtractor;
            this.job = job;
        }

        @Override
        public void write(K key, V value) throws IOException, InterruptedException {
            String fileName = fileNameExtractor.extractFolderName(key, value);
            RecordWriter<K, V> writer = fileNameToWriter.get(fileName);
            if (writer == null) {
                writer = createNewWriter(fileName, fileNameToWriter, job);
                if (writer == null) {
                    throw new IOException("Unable to create writer for path: " + fileName);
                }
            }
            writer.write(key, value);
        }

        @Override
        public void close(TaskAttemptContext context) throws IOException, InterruptedException {
            for (Entry<String, RecordWriter<K, V>> entry : fileNameToWriter.entrySet()) {
                entry.getValue().close(context);
            }
        }

    }

    private synchronized RecordWriter<K, V> createNewWriter(String folderName,
            Map<String, RecordWriter<K, V>> fileNameToWriter, TaskAttemptContext job) {
        try {
            this.folderName = folderName;
            RecordWriter<K, V> writer = super.getRecordWriter(job);
            this.folderName = null;
            fileNameToWriter.put(folderName, writer);
            return writer;
        } catch (Exception e) {
            e.printStackTrace();
            return null;
        }
    }

    @Override
    public Path getDefaultWorkFile(TaskAttemptContext context, String extension) throws IOException {
        Path path = super.getDefaultWorkFile(context, extension);
        if (folderName != null) {
            String newPath = path.getParent().toString() + "/" + folderName + "/" + path.getName();
            path = new Path(newPath);
        }
        return path;
    }

    @Override
    public RecordWriter<K, V> getRecordWriter(TaskAttemptContext job) throws IOException, InterruptedException {
        return new MultipleFilesRecordWriter(getFolderNameExtractor(), job);
    }

    public FolderNameExtractor<K, V> getFolderNameExtractor() {
        return new KeyFolderNameExtractor<K, V>();
    }

    public interface FolderNameExtractor<K, V> {
        public String extractFolderName(K key, V value);
    }

    private static class KeyFolderNameExtractor<K, V> implements FolderNameExtractor<K, V> {
        public String extractFolderName(K key, V value) {
            return key.toString();
        }
    }

}

推荐