[DRILL-4155] Query with two-way join and flatten fails with "IllegalArgumentException: maxCapacity"

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: Execution - Flow, Storage - JSON
• Assignee: Unassigned
• Reporter: Abhishek Girish
• Created: 03/Dec/15 02:21
• Updated: 03/Dec/15 02:29

Description

The following query on the Yelp Academic dataset fails to execute:

select u.name, b.name , flatten(b.categories) from maprfs.yelp_tutorial.`yelp_academic_dataset_user.json` u, maprfs.yelp_tutorial.`yelp_academic_dataset_business.json` b where u.average_stars = b.stars limit 10

Query Failed: An Error Occurred
org.apache.drill.common.exceptions.UserRemoteException: SYSTEM ERROR: IllegalArgumentException: maxCapacity: -104845 (expected: >= 0) Fragment 1:0 [Error Id: b0d99a6c-3434-49ce-8aa6-181993cdd853 on atsqa6c62.qa.lab:31010]

Tried on multiple setups in distributed mode - consistently fails.

Dataset can be accessed from : https://s3.amazonaws.com/apache-drill/files/yelp.tgz (uncompressed tar archive)
Log attached.