[HIVE-2313] RPM/DEB package integration with Hadoop

• Type: New Feature
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 0.8.0
• Component/s: Build Infrastructure

Environment

Java 6, Linux

• Assignee: Unassigned
• Reporter: Eric Yang
• Created: 27/Jul/11 01:07
• Updated: 13/Sep/11 21:09

Description

In the past Hadoop related projects are self organized to have file layout look like this:

$PROJECT_HOME/bin
             /conf/$PROJECT-env.sh
             /lib/*.jar
             /docs
             /src

Software are deployed in isolated directory, and dependent projects are using environment variables to reference each other. This approach works fine for tarball and small software stack. In order to scale the Hadoop stack of software, and there is a proposal in HADOOP-6255 to improve the file system layout as a merged system. Hadoop related projects are recommended to use this new layout:

$PROJECT_PREFIX/bin
               /etc/$PROJECT
               /lib (native C libraries)
               /share/$PROJECT (jar files)
               /share/$PROJECT/lib (third party dependent jar files)
               /share/doc/$PROJECT

Where PROJECT_PREFIX is by default /usr. Hive can locate hadoop jar files by referencing to $PROJECT_PREFIX/share/hadoop/common/hadoop-*-core.jar. The system can operate by untar the tarball or install RPM packages to the same PREFIX location.