[HIVE-21209] [Improvement] Exchange partitition to be metadata only change?

• Type: Improvement
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 2.1.1
• Component/s: Hive
• Assignee: Unassigned
• Reporter: Naveen Gangam
• Created: 04/Feb/19 15:22
• Updated: 04/Feb/19 15:25

Description

https://issues.apache.org/jira/browse/HIVE-14560
Current implementation of the above jira is a metadata and a "copy" of the partition data on the DFS. Could possibly take a long time to copy the data for large partition data especially different storage clusters. When exchanging a partition from a HDFS to S3a or vice versa the data is copied and this is client copy operation and it can be very slow if the partition is very large.

The customer would like the "exchange partition" operation to purely metadata. I would like to start a discussion on whether this improvement is to be made. Obviously, the current behavior will be supported but and option for it to be a metadata operation only needs to be evaluated.

Issue Links

• is related to: HIVE-14560 Support exchange partition between s3 and hdfs tables