[IMPALA-5381] Add query option to control join strategy when tables have no stats

• Type: New Feature
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: Impala 2.9.0
• Component/s: Frontend
• Target Version: Impala 2.9.0
• Assignee: Alexander Behm
• Reporter: Greg Rahn
• Created: 26/May/17 22:16
• Updated: 05/Jun/17 18:43
• Resolved: 05/Jun/17 18:43

Description

In IMPALA-5120 the join strategy was changed from bcast to shuffle when tables have no stats. Adding a query option to specify the behavior lowers the risk that users may have come to rely on this behavior. This would allow them to revert back to the previous behavior.

Query option proposal:

default_join_distribution_mode = [ broadcast | shuffle ] 

Ideally, the default would be shuffle, but in the spirit of preserving existing behavior it will stay broadcast. We should re-evaluate this choice in a compatibility-breaking release.

Issue Links

• is related to: IMPALA-5120 Consider defaulting to partitioned join when no stats are available.

Sub-Tasks

• Document default_join_distribution_mode query option: Resolved