Details
-
Bug
-
Status: Resolved
-
Not specified
-
Resolution: Duplicate
-
None
-
None
-
None
Description
I did a HASH join, the results were vastly different to expected (W20160708-130935), filtered the inputs to the join to just a single record (which looked like it had the wrong result) on each side and re-ran the exact same join (W20160708-132255) getting a different (and now correct looking) result for the pair.
I then went back to the full files and instead of doing a HASH join explicitly distributed the two inputs and used a LOCAL join instead. This then gave the correct results (W20160708-181510).
Looking at the graph it would appear the original HASH join was somehow run as a LOCAL join, not sure why.
Attachments
Issue Links
- duplicates
-
HPCC-15204 Distribution should be modified for globals executed on a different cluster
-
- Resolved
-