[HIVE-21123] LLAP text cache : same sql and TPCDS data occurs different results using textfile fileformat

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.1.0
• Component/s: File Formats, llap, ORC
• Assignee: Unassigned
• Reporter: Butao Zhang
• Created: 15/Jan/19 06:06
• Updated: 01/Mar/19 04:14

Description

I found that llap execution occurs different result using textfile fileformat and same sql when i test llap with TPCDS textfile data. And the orc fileformat is ok. The attachment(llap_tpcds_text_500_query53.sql) are TPCDS query53.sql execution results with data tpcds_text_500(500GB textfile), and you can find the results is different every time. The attachment (tez_tpcds_text_500_query53.sql) is the tez container execution mode and it is the right result, which is same as llap with orc textfile fileformat.

Issue Links

• relates to: HIVE-15664 LLAP text cache: improve first query perf I