[DRILL-4221] Not correct number of rows when rows are fetched from hbase storage

• Type: Bug
• Status: Open
• Priority: Critical
• Resolution: Unresolved
• Affects Version/s: 1.3.0
• Component/s: Storage - HBase

Environment

Linux 14.04, JAVA 1.7.0_67, hbase 1.1.1

• Flags: Important
• Assignee: Unassigned
• Reporter: Nikolay
• Created: 24/Dec/15 12:33
• Updated: 24/Dec/15 13:43

Description

Drill returns incorrect result when rows more than 832.
When I run query

SELECT count(row_key) FROM hbase.`ns:events`

query returns 833(correct result).

If I run query

SELECT count(*) FROM hbase.`ns:events`

query returns 832(incorrect result).

Also incorrect results can be for other type of queries. For example if I run query

SELECT count(`ns:events`.CF.column1) FROM hbase.`ns:events`

query returns 833(correct results) but if I run query with WHERE CLAUSE:

SELECT count(`ns:events`.CF.column1) FROM hbase.`ns:events` where row_key BETWEEN 'a' AND 'f'

In this case the query returns incorrect result(832, range from 'a' to 'f' is enough for getting all rows, all keys starts with 'b')

Also strange behavior is continued, when I added another 1000 rows, then the query that returns an incorrect number of rows (832), began to return the result 831