[KYLIN-1792] behaviours for non-aggregated queries

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: v1.5.2
• Fix Version/s: v1.5.3
• Assignee: Wang Cheng
• Reporter: Hongbin Ma
• Created: 16/Jun/16 05:31
• Updated: 28/Jul/16 06:48
• Resolved: 05/Jul/16 03:02

Description

Even though kylin is not designed for non-aggregated queries, people do frequently use queries like

select * from fact limit 

to verify their cubes. Recently many new users reported problems when running such queries, we'll use this JIRA as an umbrella to track all related issues.

for such queries on fact table:

• select * from fact : should return results from base cuboid (thus not raw fact table data) and MIGHT crash query server or region server
• select * from fact limit x : should return results from base cuboid (thus not raw fact table data) and SHOULD NOT crash query server or region server. Current there's a known issue KYLIN-1787 fails to push down the limit value
• select a,b from fact: should return results from base cuboid (thus not raw fact table data) and MIGHT crash query server or region server (when a or b contains is a measure, the query only works when you defined a sum() measure )
• select a,b from fact limit x: should return results from base cuboid (thus not raw fact table data) and SHOULD NOT crash query server or region server (when a or b contains is a measure, the query only works when you defined a sum() measure)

such four kinds of queries on look up tables should return correct results from the lookup table snapshot directly, and even without the limit clause, they have a low possibility to crash query server or region server because lookup table tends to be small. However, it's still better to make sure limit is taking effect.

Issue Links

• relates to: KYLIN-1787 Properly deal with limit clause in CubeHBaseEndpointRPC (SELECT * problem)