[
https://issues.apache.org/jira/browse/HADOOP-2508?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12556758#action_12556758
]
stack commented on HADOOP-2508:
-------------------------------
So, your mapper adds a special 'column' named hbase.mapred.specified.timestamp
and this is used timestamping records for a particular row? Its ugly but I do
not have a better idea.
Should we make a new data type for the value param rather than MapWritable?
Maybe we should pass a superclass of BatchUpdate here... That way we could
support deletes from MR jobs too -- not just puts. We'd need to subclass
BatchUpdate so we could squeeze a timestamp in there (Maybe BatchUpdate should
be modified so it carries a timestamp?).
How ever we do it, it should have documentation added to the class javadoc
comment.
FYI, there is a longToBytes and vice-versa in Writables. You might want to use
this instead of your own.
> [Hbase] TableOutputFormat could not assign a timestamp
> ------------------------------------------------------
>
> Key: HADOOP-2508
> URL: https://issues.apache.org/jira/browse/HADOOP-2508
> Project: Hadoop
> Issue Type: Improvement
> Components: contrib/hbase
> Affects Versions: 0.16.0
> Reporter: Edward Yoon
> Assignee: Edward Yoon
> Fix For: 0.16.0
>
> Attachments: 2508.patch
>
>
> When i use the TableOutputFormat, It is impossible to assign an specified
> timestamp to the row.
> We may want to assign an specified timestamp.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
