[HIVE-13285] Orc concatenation may drop old files from moving to final path

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 0.14.0, 1.0.0, 1.1.0, 1.2.0, 1.3.0, 2.0.0, 2.1.0
• Fix Version/s: 1.3.0, 2.0.1, 2.1.0
• Component/s: ORC
• Target Version/s: 1.3.0, 2.0.1, 2.1.0
• Assignee: Prasanth Jayachandran
• Reporter: Prasanth Jayachandran
• Created: 15/Mar/16 00:29
• Updated: 21/Jun/16 15:44
• Resolved: 16/Mar/16 21:09

Description

ORC concatenation uses combine hive input format for merging files. Under specific case where all files within a combine split are incompatible for merge (old files without stripe statistics) then these files are added to incompatible file set. But this file set is not processed as closeOp() will not be called (no output file writer will exist which will skip super.closeOp()). As a result, these incompatible files are not moved to final path.

Issue Links

• is related to: HIVE-7509 Fast stripe level merging for ORC