Details
-
Bug
-
Resolution: Fixed
-
Major
-
6.5.1
-
Security Level: Public
-
Untriaged
-
Centos 64-bit
-
1
-
Unknown
Description
Untar the attachments, use following command to load data into two buckets KS1 and KS2:
cbdocloader -c couchbase://127.0.0.1 -u Administrator -p password -m 100 -b KS1 -d ./KS1json |
cbdocloader -c couchbase://127.0.0.1 -u Administrator -p password -m 100 -b KS2 -d ./KS2json |
Then create the following indexes:
create index rand_idx_KS1 on KS1(rand_i);
|
create index uniq_idx_KS2 on KS2(uniq_i);
|
create index zipf_idx_KS2 on KS2(zipf_i);
|
Then run the following query in a loop:
select count(y.zipf_i) from KS1 x USE INDEX(rand_idx_KS1) join KS2 y on x.rand_i <= 16384 and x.uniq_i = y.uniq_i; |
The query should return a count of 1000, however when run repeatedly, it sometimes produces a count that's less than 1000 (early termination). The query uses intersect scan on KS2 (or y) under nested-loop join.
Attachments
Issue Links
- Clones
-
MB-39684 Intersect Scan under inner of nested-loop join may produce wrong results
- Closed