[FALCON-1787] Ooozie pig-action.xml requires hive sharelib for HCatalog use

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.6.1
• Fix Version/s: 0.10
• Component/s: oozie

Environment

HDP-2.3.2.0-2950
Pig 0.15.0.2.3
Hive 1.2.1.2.3
Oozie 4.2.0.2.3
Falcon 0.6.1.2.3

• Assignee: Sowmya Ramesh
• Reporter: Mark Greene
• Created: 27/Jan/16 22:28
• Updated: 04/Feb/16 14:13
• Resolved: 04/Feb/16 14:13

Description

I have a Pig script that I am using as the workflow for my Falcon process. The pig script uses HCatalogStorer to write to a HCatalog URI that is the output feed defined in my Falcon Process Entity. The Pig action in the resulting Ooozie Workflow generated by Falcon fails with the attached stack trace. The root is that it is missing a class definitions of org/apache/hadoop/hive/shims/ShimLoader.

Running the script manually using pig -x tex -useHCatalog <all the -params passed by Oozie> <path to pig script> results in a successful execution. It's only once this is called as a Pig activity in the Falcon-generated Oozie workflow that the missing class definitions manifests.

After some investigation I found that the Oozie workflow.xml is missing a required sharelib decleration.

From the workflow.xml generated by Falcon:
<property>
<name>oozie.action.sharelib.for.pig</name>
<value>pig,hcatalog</value>
</property>

If I modify the value to include hive sharelib then the Pig action succeeds and does not throw a missing class definition error.

Modified workflow.xml property (works):
<property>
<name>oozie.action.sharelib.for.pig</name>
<value>hive,pig,hcatalog</value>
</property>