[HIVE-7429] Set replication for archive called before file exists

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 0.11.0, 0.12.0, 0.13.0, 0.13.1
• Fix Version/s: 0.14.0
• Assignee: Daniel Weeks
• Reporter: Daniel Weeks
• Created: 16/Jul/14 20:05
• Updated: 13/Nov/14 19:43
• Resolved: 17/Jul/14 01:02

Description

The call to set replication is called prior to uploading the archive file to hdfs, which does not throw an error, but the replication never gets set.

This has a significant impact on large jobs (especially hash joins) due to too many tasks hitting the data nodes.