MapReduce Usage

This page describes how to use the MongoDB Hadoop Connector with vanilla MapReduce.

Installation

1. Obtain the MongoDB Hadoop Connector. You can either build it or download the jars. The releases page also includes instructions for use with Maven and Gradle. For MapReduce, all you need is the "core" jar.

2. Get a JAR for the MongoDB Java Driver.

3. Each node in the cluster will need to have access to the MongoDB Hadoop Connector JARs as well as the JAR for the MongoDB Java Driver. You can provision each machine in the cluster with the necessary JARs in $HADOOP_HOME/share/hadoop/common, for example, or you may use the Hadoop DistributedCache to distribute the JARs to pre-existing nodes. This is easily done using the -libjars option in the hadoop jar command:

    hadoop jar \
      -libjars mongo-hadoop-core.jar,mongo-java-driver.jar \
      MyJob.jar com.mycompany.HadoopJob
   

See the instructions on the releases page for how to include the Mongo Hadoop Connector jars easily in your MapReduce project code.

Writing MapReduce Jobs

The MongoDB Hadoop Connector can do input/output with a live MongoDB cluster or BSON files (such as ones generated by mongodump:

Direction	MongoDB	BSON
Input	MongoInputFormat	BSONFileInputFormat
Output	MongoOutputFormat	BSONFileOutputFormat

These formats are found in the com.mongodb.hadoop package. For Mapreduce 1.x, these classes are in the com.mongodb.hadoop.mapred package.

Examples

There are a number of examples for writing MapReduce jobs using the MongoDB Hadoop Connector.