[SPARK-23027] optimizer a simple query using a non-existent data is too slow

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Invalid
• Affects Version/s: 2.0.1
• Component/s: Optimizer, SQL
• Assignee: Unassigned
• Reporter: wangminfeng
• Created: 10/Jan/18 11:56
• Updated: 17/May/20 17:58
• Resolved: 11/Jan/18 15:57

Description

When i use spark sql to do ad-hoc query, i have data partitioned by event_day, event_minute, event_hour, data is large enough，one day data is about 3T， we saved 3 month data.
But when query use a non-existent day， get optimizedPlan is too slow.
i use “sparkSession.sessionState.executePlan(logicalPlan).optimizedPlan” get optimized plan， for five minutes，i can not get it. Query is simple enough, like:
SELECT
event_day
FROM db.table t1
WHERE (t1.event_day='20170104' and t1.event_hour='23' and t1.event_minute='55')
LIMIT 1