[HIVE-16799] Control the max number of task for a stage in a spark job

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.0.0
• Labels: TODOC3.0
• Release Note: Document the new configuration.
• Assignee: Xuefu Zhang
• Reporter: Xuefu Zhang
• Created: 01/Jun/17 03:16
• Updated: 22/May/18 23:57
• Resolved: 02/Jun/17 18:29

Description

HIVE-16552 gives admin an option to control the maximum number of tasks a Spark job may have. However, this may not be sufficient as this tends to penalize jobs that have many stages while favoring jobs that has fewer stages. Ideally, we should also limit the number of tasks in a stage, which is closer to the maximum number of mappers or reducers in a MR job.