You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mrql.apache.org by "Leonidas Fegaras (JIRA)" <ji...@apache.org> on 2014/02/02 21:46:08 UTC

[jira] [Created] (MRQL-27) Improve the build process and support hadoop 0.20.x

Leonidas Fegaras created MRQL-27:
------------------------------------

             Summary: Improve the build process and support hadoop 0.20.x
                 Key: MRQL-27
                 URL: https://issues.apache.org/jira/browse/MRQL-27
             Project: MRQL
          Issue Type: Improvement
            Reporter: Leonidas Fegaras
            Assignee: Leonidas Fegaras
            Priority: Critical


The attached patch improves the MRQL build process in many ways:
1) it now supports hadoop 0.20.x. It uses the new extra source directory src/main/java/MultipleInputs, which was taken from hadoop 1.x sources, because 0.20.x didn't support MultipleInputs.
2) it corrects the classpath for hadoop 2.x (yarn).
3) it changes the scripts bin/mrql and bin/mrql.bsp to build a single jar file that contains all classes. This was done because in some cases the hadoop/hama -libjar parameter does not work correctly in distributed mode.

Currently, the "mvn install" builds MRQL with Hadoop 1.2.1, Hama 0.6.3, and Spark 0.8.1-incubating.
To build MRQL on some other Hadoop 1.x, such as 1.0.3, use:
mvn -Dhadoop.version=1.0.3 install
To build MRQL on Hadoop 2.x (yarn) use:
mvn -Pyarn -Dyarn.version=2.2.0 -Dhadoop.version=1.2.1 install
To build MRQL on Hadoop 0.20.x use:
mvn -PMultipleInputs -Dhadoop.version=0.20.2 install




--
This message was sent by Atlassian JIRA
(v6.1.5#6160)