[OOZIE-3643] Oozie should pass mapreduce.job.tags to Beeline in Hive2 action

• Type: Task
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 5.2.1
• Assignee: Unassigned
• Reporter: Dénes Bodó
• Created: 02/Dec/21 09:54
• Updated: 02/Dec/21 09:56

Description

Currently if we start a Hive2 action which starts a Tez child job and that runs for long and then we try to kill the Oozie workflow which started that Hive2 action we experience that the started Tez child job remains running.

To solve this problem Hive on Tez configurations should be set: hive.server2.tez.initialize.default.sessions=false and Oozie should pass the launcher job's tag to beeline to be able to find the Tez child job. In Oozie we could do something like:

arguments.add("--hiveconf");
arguments.add("hive.query.tag=" + actionConf.get(MAPREDUCE_JOB_TAGS)); 

in Hive2Main class.

As I experienced in my test environment when I did killed such a workflow it killed the Tez child job too but unfortunately a new instance of my workflow wasn't able to start the Hive query, the launcher job stuck.

Issue Links

• relates to: HIVE-20549 Allow user set query tag, and kill query with tag