Read and write data to/from ElasticSearch within Hadoop/MapReduce libraries. Automatically converts data to/from JSON. Supports MapReduce, Cascading, Hive and Pig.

ElasticSearch cluster accessible through REST. That's it! Significant effort has been invested to create a small, dependency-free, self-contained jar that can be downloaded and put to use without any dependencies. Simply make it available to your job classpath and you're set.

This project is released under version 2.0 of the Apache License

We're working towards a first release and we plan to have nightly builds soon available. In the meantime please build the project yourself.

We're interested in your feedback! You can find us on the User mailing list - please append [Hadoop] to the post subject to filter it out. For more details, see the community page.

These properties are read from Hadoop Configuration if the user hasn't specified them.

es.host=<ES host address> # optional, defaults to localhost
es.port=<ES REST port>    # optional, defaults to 9200
es.location=<ES resource location, relative to the host/port specified above. Can be an index or a query> # required

For basic, low-level or performance-sensitive environments, ES-Hadoop provides dedicated InputFormat and OutputFormat that read and write data to ElasticSearch. To use them, add the es-hadoop jar to your job classpath (either by bundling the library along - it's less then 40kB and there are no-dependencies), using the DistributedCache or by provisioning the cluster manually.

Note that es-hadoop supports both the so-called 'old' and the 'new' API through its ESInputFormat and ESOutputFormat classes.

To read data from ES, configure the ESInputFormat on your job configuration along with the relevant properties:

JobConf conf = new JobConf();
conf.setInputFormat(ESInputFormat.class);
conf.set("es.location", "radio/artists/_search?q=me*"); // replace this with the relevant query
...
JobClient.runJob(conf);

Same configuration template can be used for writing but using ESOuputFormat:

JobConf conf = new JobConf();
conf.setOutputFormat(ESOutputFormat.class);
conf.set("es.location", "radio/artists"); // index or indices used for storing data
...
JobClient.runJob(conf);

Configuration conf = new Configuration();
conf.set("es.location", "radio/artists/_search?q=me*"); // replace this with the relevant query
Job job = new Job(conf)
job.setInputFormat(ESInputFormat.class);
...
job.waitForCompletion(true);

Configuration conf = new Configuration();
conf.set("es.location", "radio/artists"); // index or indices used for storing data
Job job = new Job(conf)
job.setOutputFormat(ESOutputFormat.class);
...
job.waitForCompletion(true);

ES-Hadoop provides a Hive storage handler for ElasticSearch, meaning one can define an external table on top of ES.

Add es-hadoop-.jar to hive.aux.jars.path or register it manually in your Hive script (recommended):

ADD JAR /path_to_jar/es-hadoop-<version>.jar;

To read data from ES, define a table backed by the desired index:

CREATE EXTERNAL TABLE artists (
    id      BIGINT,
    name    STRING,
    links   STRUCT<url:STRING, picture:STRING>)
STORED BY 'org.elasticsearch.hadoop.hive.ESStorageHandler'
TBLPROPERTIES('es.location' = 'radio/artists/_search?q=me*');

The fields defined in the table are mapped to the JSON when communicating with ElasticSearch. Notice the use of TBLPROPERTIES to define the location, that is the query used for reading from this table:

SELECT * FROM artists;

To write data, a similar definition is used but with a different es.location:

CREATE EXTERNAL TABLE artists (
    id      BIGINT,
    name    STRING,
    links   STRUCT<url:STRING, picture:STRING>)
STORED BY 'org.elasticsearch.hadoop.hive.ESStorageHandler'
TBLPROPERTIES('es.location' = 'radio/artists/');

Any data passed to the table is then passed down to ElasticSearch; for example considering a table s, mapped to a TSV/CSV file, one can index it to ElasticSearch like this:

INSERT OVERWRITE TABLE artists 
    SELECT NULL, s.name, named_struct('url', s.url, 'picture', s.picture) FROM source s;

As one can note, currently the reading and writing are treated separately but we're working on unifying the two and automatically translating HiveQL to ElasticSearch queries.

ES-Hadoop provides both read and write functions for Pig so you can access ElasticSearch from Pig scripts.

Register ES-Hadoop jar into your script or add it to your Pig classpath:

REGISTER /path_to_jar/es-hadoop-<version>.jar;

Additionally one can define an alias to save some chars:

%define ESSTORAGE org.elasticsearch.hadoop.pig.ESStorage()

and use $ESSTORAGE for storage definition.

To read data from ES, use ESStorage and specify the query through the LOAD function:

A = LOAD 'radio/artists/_search?q=me*' USING org.elasticsearch.hadoop.pig.ESStorage();
DUMP A;

Use the same Storage to write data to ElasticSearch:

A = LOAD 'src/test/resources/artists.dat' USING PigStorage() AS (id:long, name, url:chararray, picture: chararray);
B = FOREACH A GENERATE name, TOTUPLE(url, picture) AS links;
STORE B INTO 'radio/artists' USING org.elasticsearch.hadoop.pig.ESStorage();

ES-Hadoop offers a dedicate ElasticSearch Tap, ESTap that can be used both as a sink or a source. Note that ESTap can be used in both local (LocalFlowConnector) and Hadoop (HadoopFlowConnector) flows:

Tap in = new ESTap("radio/artists/_search?q=me*");
Tap out = new StdOut(new TextLine());
new LocalFlowConnector().connect(in, out, new Pipe("read-from-ES")).complete();

Tap in = Lfs(new TextDelimited(new Fields("id", "name", "url", "picture")), "src/test/resources/artists.dat");
Tap out = new ESTap("radio/artists", new Fields("name", "url", "picture"));
new HadoopFlowConnector().connect(in, out, new Pipe("write-to-ES")).complete();

ElasticSearch Hadoop uses Gradle for its build system and it is not required to have it installed on your machine.

To create a distributable jar, run gradlew -x test build from the command line; once completed you will find the jar in build\libs.