hive小文件造成map多_map数量太多

public InputSplit[] getSplits(JobConf job, int numSplits) throws IOException {

    //加载CombineFileInputFormatShim，这个类继承了org.apache.hadoop.mapred.lib.CombineFileInputFormat

    CombineFileInputFormatShim combine = ShimLoader.getHadoopShims()

        .getCombineFileInputFormat();

if (combine == null) {

//若为空则采用HiveInputFormat的方式，下同

      return super.getSplits(job, numSplits);

    }

    Path[] paths = combine.getInputPathsShim(job);

for (Path path : paths) {

//若是外部表，则按照HiveInputFormat方式分片

      if ((tableDesc != null) && tableDesc.isNonNative()) {

        return super.getSplits(job, numSplits);

      }

      Class inputFormatClass = part.getInputFileFormatClass();

      String inputFormatClassName = inputFormatClass.getName();

      InputFormat inputFormat = getInputFormatFromCache(inputFormatClass, job);

      if (this.mrwork != null && !this.mrwork.getHadoopSupportsSplittable()) {

        if (inputFormat instanceof TextInputFormat) {

         if ((new CompressionCodecFactory(job)).getCodec(path) != null)

//在未开启hive.hadoop.supports.splittable.combineinputformat(MAPREDUCE-1597)参数情况下，对于TextInputFormat并且为压缩则采用HiveInputFormat分片算法

                    return super.getSplits(job, numSplits);

        }

      }

    //对于连接式同上

      if (inputFormat instanceof SymlinkTextInputFormat) {

        return super.getSplits(job, numSplits);

      }

      CombineFilter f = null;

      boolean done = false;

Path filterPath = path;

//由参数hive.mapper.cannot.span.multiple.partitions控制，默认false;如果没true，则对每一个partition创建一个pool，以下省略为true的处理；对于同一个表的同一个文件格式的split创建一个pool为combine做准备；

      if (!mrwork.isMapperCannotSpanPartns()) {

        opList = HiveFileFormatUtils.doGetWorksFromPath(

                   pathToAliases, aliasToWork, filterPath);

        f = poolMap.get(new CombinePathInputFormat(opList, inputFormatClassName));

      }

      if (!done) {

        if (f == null) {

          f = new CombineFilter(filterPath);

          combine.createPool(job, f);

        } else {

          f.addPath(filterPath);

        }

      }

    }

if (!mrwork.isMapperCannotSpanPartns()) {

//到这里才调用combine的分片算法，继承了org.apache.hadoop.mapred.lib.CombineFileInputFormat extends 新版本CombineFileInputformat

      iss = Arrays.asList(combine.getSplits(job, 1));

}

//对于sample查询特殊处理

    if (mrwork.getNameToSplitSample() != null && !mrwork.getNameToSplitSample().isEmpty()) {

      iss = sampleSplits(iss);

}

//封装结果返回

    for (InputSplitShim is : iss) {

      CombineHiveInputSplit csplit = new CombineHiveInputSplit(job, is);

      result.add(csplit);

    }

    return result.toArray(new CombineHiveInputSplit[result.size()]);

  }

为TextInputFormat，并非实际存储的压缩格式，因为缺少相关参数设定造成了配置失效，开始走hiveinputformat计算 。

public InputSplit[] getSplits(JobConf job, int numSplits) throws IOException {

    //扫描每一个分区

    for (Path dir : dirs) {

      PartitionDesc part = getPartitionDescFromPath(pathToPartitionInfo, dir);

    //获取分区的输入格式

      Class inputFormatClass = part.getInputFileFormatClass();

      InputFormat inputFormat = getInputFormatFromCache(inputFormatClass, job);

    //按照相应格式的分片算法获取分片

    //注意：这里的Inputformat只是old version API：org.apache.hadoop.mapred而不是org.apache.hadoop.mapreduce，因此不能采用新的API，否则在查询时会报异常：Input format must implement InputFormat.区别就是新的API的计算inputsplit size（Math.max(minSize, Math.min(maxSize, blockSize)）和老的（Math.max(minSize, Math.min(goalSize, blockSize))）不一样；

      InputSplit[] iss = inputFormat.getSplits(newjob, numSplits / dirs.length);

      for (InputSplit is : iss) {

    //封装结果，返回

        result.add(new HiveInputSplit(is, inputFormatClass.getName()));

      }

    }

    return result.toArray(new HiveInputSplit[result.size()]);

}

解决方案：

设定hive.hadoop.supports.splittable.combineinputformat参数

相关全部参数：

set hive.merge.mapfiles=true;

set hive.hadoop.supports.splittable.combineinputformat=true;

set hive.merge.size.per.task=2147483648;

set hive.merge.smallfiles.avgsize=2147483648;

set hive.merge.size.smallfiles.avgsize=2147483648;

set mapreduce.input.fileinputformat.split.maxsize=2147483648;

set mapred.max.split.size=2147483648;

set mapred.min.split.size.per.node=2147483648;

set mapred.min.split.size.per.rack=2147483648;

set hive.exec.reducers.bytes.per.reducer=2147483648;

set hive.input.format=org.apache.hadoop.hive.ql.io.CombineHiveInputFormat;

验证结果：

map数由11236降低至2363个map，证明参数修改生效。