使用MapReduce将Hadoop HDFS中的日志文件导入HBase中（二）

最近准备面试，所以上次写完《使用JAVA将Hadoop HDFS中的日志文件导入HBase中（一）》后去研究了下HBase原理及架构，所以一直没有进行下一步的记录，今天来记录一下由HDFS高效率导入HBase的两种导入方式。

1.使用Map+Reduce方式

public class MapReduceImport {

    /**
     * Mapper
     */
    static class HMapper extends Mapper<LongWritable, Text, LongWritable, Text> {
        Text v2 = new Text();

        protected void map(LongWritable key, Text value, Context context) throws java.io.IOException, InterruptedException {
            String[] splited = value.toString().split(" ");
            if (splited.length != 6)//清洗不符合标准的数据
                return;
            try {
                //GetRowKey.getRowKeyString方法是自己定义生成rowkey的方法
                //rowkey设计为IP_TimeStamp这种方式
                v2.set(GetRowKey.getRowKeyString(splited[2], splited[4]) + " " + value.toString());
                context.write(key, v2);
            } catch (NumberFormatException e) {
                System.out.println("出错了" + e.getMessage());
            }
        }
    }

    /**
     * Reducer
     */
    static class HReducer extends TableReducer<LongWritable, Text, NullWritable> {

        protected void reduce(LongWritable key, java.lang.Iterable values, Context context) throws java.io.IOException, InterruptedException {
            for (Text text : values) {
                String[] splited = text.toString().split(" ");

                Put put = new Put(Bytes.toBytes(splited[0]));

                for (int j = 1; j < splited.length; j++) {
                    put.addColumn(Bytes.toBytes(HConfiguration.colFamily), Bytes.toBytes("log" + j), Bytes.toBytes(splited[j]));
                }

                context.write(NullWritable.get(), put);
            }
        }
    }

    /**
     * Main
     *
     * @param args
     * @throws Exception
     */
    public static void main(String[] args) throws Exception {


        Configuration configuration = new Configuration();
        //设置zookeeper
        configuration.set("hbase.zookeeper.quorum", HConfiguration.hbase_zookeeper_quorum);
        configuration.set("hbase.zookeeper.property.clientPort", "2181");
        //设置hbase表名称
        configuration.set(TableOutputFormat.OUTPUT_TABLE, HConfiguration.tableName);
        //将该值改大，防止hbase超时退出
        configuration.set("dfs.socket.timeout", "180000");

        MRDriver myDriver = MRDriver.getInstance();

        try {
            myDriver.createTableIfExistDelete(HConfiguration.tableName, HConfiguration.colFamily);
        } catch (Exception e) {
            e.printStackTrace();
        }

        Job job = new Job(configuration, "Map+ReduceImport");

        job.setMapperClass(HMapper.class);
        job.setReducerClass(HReducer.class);

        job.setMapOutputKeyClass(LongWritable.class);
        job.setMapOutputValueClass(Text.class);

        job.setInputFormatClass(TextInputFormat.class);
        //不再设置输出路径，而是设置输出格式类型TableOutputFormat
        job.setOutputFormatClass(TableOutputFormat.class);

        FileInputFormat.setInputPaths(job, HConfiguration.mapreduce_inputPath);

        job.waitForCompletion(true);
    }
}

———————-分—-割—-线————————–

2.只使用Map的方式

public class OnlyMapImport {
    /**
     * Mapper
     */
    static class ImportMapper extends Mapper<LongWritable, Text, ImmutableBytesWritable, Put> {

        @Override
        public void map(LongWritable offset, Text value, Context context) {
            String[] splited = value.toString().split(" ");
            if (splited.length != 6)
                return;
            try {
                //GetRowKey.getRowKeyString方法是自己定义生成rowkey的方法
                //rowkey设计为IP_TimeStamp这种方式
                byte[] rowkey = Bytes.toBytes(GetRowKey.getRowKeyString(splited[2], splited[4]));
                Put put = new Put(rowkey);

                for (int j = 0; j < splited.length; j++) {
                    put.addColumn(Bytes.toBytes(HConfiguration.colFamily), Bytes.toBytes("log" + j), Bytes.toBytes(splited[j]));
                }

                context.write(new ImmutableBytesWritable(rowkey), put);

            } catch (NumberFormatException e) {
                System.out.println("出错了" + e.getMessage());
            } catch (IOException e) {
                e.printStackTrace();
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        }
    }

    /**
     * Main
     *
     * @param args
     * @throws Exception
     */
    public static void main(String[] args) throws Exception {


        Configuration configuration = new Configuration();
        //设置zookeeper
        configuration.set("hbase.zookeeper.quorum", HConfiguration.hbase_zookeeper_quorum);
        configuration.set("hbase.zookeeper.property.clientPort", "2181");
        //设置hbase表名称
        configuration.set(TableOutputFormat.OUTPUT_TABLE, HConfiguration.tableName);
        //将该值改大，防止hbase超时退出
        configuration.set("dfs.socket.timeout", "180000");

        MRDriver myDriver = MRDriver.getInstance();

        try {
            myDriver.createTableIfExistDelete(HConfiguration.tableName, HConfiguration.colFamily);
        } catch (Exception e) {
            e.printStackTrace();
        }

        Job job = new Job(configuration, "HBaseBatchImport");

        job.setJarByClass(OnlyMapImport.class);
        job.setMapperClass(ImportMapper.class);
        //设置map的输出，不设置reduce的输出类型
        job.setMapOutputKeyClass(ImmutableBytesWritable.class);
        job.setMapOutputValueClass(Writeable.class);
        job.setNumReduceTasks(0);

        job.setInputFormatClass(TextInputFormat.class);
        //不再设置输出路径，而是设置输出格式类型
        job.setOutputFormatClass(TableOutputFormat.class);

        FileInputFormat.setInputPaths(job, HConfiguration.mapreduce_inputPath);

        job.waitForCompletion(true);
    }
}

经过测试，导入时间明显减少。