Hadoop reducer 未被调用

Question

全部

我有简单的 map/reduce 实现。 Mapper 被调用并完成其工作，但 reducer 从未被调用。

这是映射器:

static public class InteractionMap extends Mapper<LongWritable, Text, Text, InteractionWritable> {

    @Override
    protected void map(LongWritable offset, Text text, Context context) throws IOException, InterruptedException {
        System.out.println("mapper");
        String[] tokens = text.toString().split(",");
        for (int idx = 0; idx < tokens.length; idx++) {
            String sourceUser = tokens[1];
            String targetUser = tokens[2];
            int points = Integer.parseInt(tokens[4]);
            context.write(new Text(sourceUser), new InteractionWritable(targetUser, points));
            }
        }
    }
}

这是我的 reducer :

static public class InteractionReduce extends Reducer<Text, InteractionWritable, Text, Text> {

    @Override
    protected void reduce(Text token, Iterable<InteractionWritable> counts, Context context) throws IOException, InterruptedException {
        System.out.println("REDUCER");
        Iterator<InteractionWritable> i = counts.iterator();
        while (i.hasNext()) {
            InteractionWritable interaction = i.next();
            context.write(token, new Text(token.toString() + " " + interaction.getTargetUser().toString() + " " + interaction.getPoints().get()));
        }
    }

}

还有，这里是配置部分:

@Override
public int run(String[] args) throws Exception {
    Configuration configuration = getConf();
    Job job = new Job(configuration, "Interaction Count");
    job.setJarByClass(InteractionMapReduce.class);
    job.setMapperClass(InteractionMap.class);
    job.setCombinerClass(InteractionReduce.class);
    job.setReducerClass(InteractionReduce.class);
    job.setInputFormatClass(TextInputFormat.class);
    job.setOutputFormatClass(TextOutputFormat.class);
    job.setOutputKeyClass(Text.class);
    job.setOutputValueClass(Text.class);
    FileInputFormat.addInputPath(job, new Path(args[0]));
    FileOutputFormat.setOutputPath(job, new Path(args[1]));
    return job.waitForCompletion(true) ? 0 : -1;
}

有谁知道为什么没有调用 reducer？

Answer 1

好吧，果然是我的错。工作配置不好。它应该是这样的:

Configuration configuration = getConf();

Job job = new Job(configuration, "Interaction Count");
job.setJarByClass(InteractionMapReduce.class);
job.setMapperClass(InteractionMap.class);
job.setReducerClass(InteractionReduce.class);
job.setMapOutputKeyClass(Text.class);
job.setMapOutputValueClass(InteractionWritable.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(Text.class);

FileInputFormat.addInputPath(job, new Path(args[0]));
FileOutputFormat.setOutputPath(job, new Path(args[1]));

return job.waitForCompletion(true) ? 0 : -1;

问题的发生是因为 map 和 reduce 阶段的输出类型不同。作业在调用 context.write 方法后静默失败。所以，我必须添加的是这些行:

job.setMapOutputKeyClass(Text.class);
job.setMapOutputValueClass(InteractionWritable.class);
job.setOutputKeyClass(Text.class);
job.setOutputValueClass(Text.class);