将HDFS数据移至MongoDB

Question

I am trying to move HDFS data into MongoDB. I know how to export data into mysql by using sqoop. I dont think I can use sqoop for MongoDb. I need help understanding how to do that.

Answer 1

This recipe will use the MongoOutputFormat class to load data from an HDFS instance into a MongoDB collection.

Getting ready

The easiest way to get started with the Mongo Hadoop Adaptor is to clone the Mongo-Hadoop project from GitHub and build the project configured for a specific version of Hadoop. A Git client must be installed to clone this project. This recipe assumes that you are using the CDH3 distribution of Hadoop. The official Git Client can be found at http://git-scm.com/downloads .

The Mongo Hadoop Adaptor can be found on GitHub at https://github.com/mongodb/ mongo-hadoop . This project needs to be built for a specific version of Hadoop. The resulting JAR file must be installed on each node in the $HADOOP_HOME/lib folder. The Mongo Java Driver is required to be installed on each node in the $HADOOP_HOME/ lib folder. It can be found at https://github.com/mongodb/mongo-java-driver/ downloads .

How to do it...

 Complete the following steps to copy data form HDFS into MongoDB:
    1.   Clone the mongo-hadoop repository with the following command line:
    git clone https://github.com/mongodb/mongo-hadoop.git


    2.   Switch to the stable release 1.0 branch:
    git checkout release-1.0


    3.   Set the Hadoop version which mongo-hadoop should target. In the folder
    that mongo-hadoop was cloned to, open the build.sbt file with a text editor.
    Change the following line:
    hadoopRelease in ThisBuild := "default"
    to
    hadoopRelease in ThisBuild := "cdh3"


    4.   Build mongo-hadoop :
    ./sbt package
    This will create a file named mongo-hadoop-core_cdh3u3-1.0.0.jar in the
    core/target folder.


    5.   Download the MongoDB Java Driver Version 2.8.0 from https://github.com/
    mongodb/mongo-java-driver/downloads .


    6.   Copy mongo-hadoop and the MongoDB Java Driver to $HADOOP_HOME/lib on
    each node:


    cp mongo-hadoop-core_cdh3u3-1.0.0.jar mongo-2.8.0.jar $HADOOP_
    HOME/lib


    7.   Create a Java MapReduce program that will read the weblog_entries.txt file
    from HDFS and write them to MongoDB using the MongoOutputFormat class:


import java.io.*;
import org.apache.commons.logging.*;
import org.apache.hadoop.conf.*;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.*;
import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
import org.apache.hadoop.mapreduce.lib.input.TextInputFormat;
import org.apache.hadoop.mapreduce.*;
import org.bson.*;
import org.bson.types.ObjectId;
import com.mongodb.hadoop.*;
import com.mongodb.hadoop.util.*;
public class ExportToMongoDBFromHDFS {
private static final Log log = LogFactory.getLog(ExportToMongoDBFromHDFS.class);
public static class ReadWeblogs extends Mapper<LongWritable, Text, ObjectId, BSONObject>{
public void map(Text key, Text value, Context context)
throws IOException, InterruptedException{
           System.out.println("Key: " + key);
           System.out.println("Value: " + value);
           String[] fields = value.toString().split("\t");
           String md5 = fields[0];
           String url = fields[1];
           String date = fields[2];
           String time = fields[3];
           String ip = fields[4];
           BSONObject b = new BasicBSONObject();
            b.put("md5", md5);
            b.put("url", url);
            b.put("date", date);
            b.put("time", time);
              b.put("ip", ip);
          context.write( new ObjectId(), b);
       }
}
public static void main(String[] args) throws Exception{
       final Configuration conf = new Configuration();
          MongoConfigUtil.setOutputURI(conf,"mongodb://<HOST>:<PORT>/test.        weblogs");
System.out.println("Configuration: " + conf);
        final Job job = new Job(conf, "Export to Mongo");
         Path in = new Path("/data/weblogs/weblog_entries.txt");
            FileInputFormat.setInputPaths(job, in);
            job.setJarByClass(ExportToMongoDBFromHDFS.class);
            job.setMapperClass(ReadWeblogs.class);
            job.setOutputKeyClass(ObjectId.class);
            job.setOutputValueClass(BSONObject.class);
            job.setInputFormatClass(TextInputFormat.class);
            job.setOutputFormatClass(MongoOutputFormat.class);
            job.setNumReduceTasks(0);
           System.exit(job.waitForCompletion(true) ? 0 : 1 );
       }
}

8.   Export as a runnable JAR file and run the job:
hadoop jar ExportToMongoDBFromHDFS.jar
9.   Verify that the weblogs MongoDB collection was populated from the Mongo shell:
db.weblogs.find();

Answer 2

The basic problem is that mongo stores its data in BSON format (binary JSON), while you hdfs data may have different formats (txt, sequence, avro). The easiest thing to do would be to use pig to load your results using this driver:

https://github.com/mongodb/mongo-hadoop/tree/master/pig

into mongo db. You'll have to map your values to your collection - there's a good example on the git hub page.