[HIVE-15720] Bucketing: incorrect inserts with inserting from similarly bucketed table

• Type: Bug
• Status: Open
• Priority: Critical
• Resolution: Unresolved
• Affects Version/s: 2.2.0
• Component/s: Query Processor
• Assignee: Unassigned
• Reporter: Gopal Vijayaraghavan
• Created: 25/Jan/17 02:36
• Updated: 25/Jan/17 02:36

Description

create temporary table foo(x int, y int) clustered by (x) sorted by (y) into 11 buckets stored as orc;
create temporary table bar like foo;
insert into foo values (23,2); insert into foo values(23,2); insert into foo values(23,2);
insert overwrite table bar select * from foo;

At this point, the table bar is incorrectly loaded.

hive> select x, regexp_replace(INPUT__FILE__NAME, '.*/','') from bar ;
OK
23      000000_0
23      000001_0
23      000002_0
Time taken: 0.22 seconds, Fetched: 3 row(s)

hive> select x, regexp_replace(INPUT__FILE__NAME, '.*/','') from foo;
OK
23      000001_0
23      000001_0_copy_1
23      000001_0_copy_2
Time taken: 0.329 seconds, Fetched: 3 row(s)