[HBASE-1963] Output to multiple tables from Hadoop MR without use of HTable

• Type: New Feature
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 0.20.1
• Fix Version/s: 0.90.0
• Component/s: mapreduce
• Hadoop Flags: Reviewed
• Release Note: MultiTableOutputFormat allows output from a map/reduce job to be written to multiple tables. An example illustrates use for creating secondary indexes from an existing table.
• Assignee: Kevin Peterson
• Reporter: Kevin Peterson
• Created: 09/Nov/09 10:12
• Updated: 20/Nov/15 13:01
• Resolved: 10/Nov/09 17:22

Description

o.a.h.h.mapreduce.TableOutputFormat allows writing to a single table as output from a map/reduce job in the natural way. It requires that the user specify the table name ahead of time and can only write to one table. I had a need to write to multiple tables from the same job (write my data to one table, and also write to index tables), and I wanted to have a consistent API whether writing to one or many tables.

Attached MultiTableOutputFormat takes the table name as the key and the Put or Delete as the value. Also included is an example demonstrating the usage.