[
https://issues.apache.org/jira/browse/HBASE-13027?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sean Busbey updated HBASE-13027:
--------------------------------
Release Note: (was: The bindings for reading data from HBase using both
the Hadoop MapReduce mapred and mapreduce libraries now are initialized the
same way. Subclasses wishing to build on TableInputFormatBase that use the new
initialize method will have access to job configuration information.)
release note was on the wrong ticket. :/
removed since this issue just restores proper behavior for the 0.98 way of
configuring.
> mapreduce.TableInputFormatBase should create its own Connection if needed
> -------------------------------------------------------------------------
>
> Key: HBASE-13027
> URL: https://issues.apache.org/jira/browse/HBASE-13027
> Project: HBase
> Issue Type: Bug
> Components: mapreduce
> Affects Versions: 1.0.0, 1.1.0
> Reporter: Sean Busbey
> Assignee: Sean Busbey
> Priority: Blocker
> Labels: hive
> Fix For: 1.0.0, 2.0.0, 1.1.0
>
> Attachments: HBASE-13027-0.98.1.patch.txt, HBASE-13027.1.patch.txt,
> HBASE-13027.2.patch.txt, HBASE-13027.3.patch.txt
>
>
> Folks extending the mapreduce.TableInputFormatBase class and relying on the
> 0.98 way of doing table initialization (as given in the example in the
> javadoc) will currently always get an error.
> When taking in a HTable, the current implementation retrieves a Connection
> instance from the table and then an Admin instance from said Connection. The
> getAdmin call validates that it only is called on unmanaged connections, but
> when folks construct an HTable instance according to the examples, it ends up
> backed by a managed connection.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
