You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@hive.apache.org by Jan Stöcker <Ja...@q2web.de> on 2010/03/30 11:16:08 UTC

Hive/Hadoop Interface

Hi,

could someone help me with a question to better understand how Hive
interacts with Hadoop?
I understood that for each Map/Reduce-Job Hive wants to get done, it
creates one instance of "Task", which can either be a MapRedTask or an
ExecDriver. On what does it depend which of the two it is?

Thanks,
Jan

Re: Hive/Hadoop Interface

Posted by Ted Yu <yu...@gmail.com>.

When mapred.job.tracker is set to local (default value) or
hive.exec.submitviachild is set to true, MapRedTask instance is used.
Otherwise ExecDriver instance is used.

Please refer to
trunk\ql\src\java\org\apache\hadoop\hive\ql\exec\TaskFactory.java

On Tue, Mar 30, 2010 at 2:16 AM, Jan Stöcker <Ja...@q2web.de> wrote:

> Hi,
>
> could someone help me with a question to better understand how Hive
> interacts with Hadoop?
> I understood that for each Map/Reduce-Job Hive wants to get done, it
> creates one instance of "Task", which can either be a MapRedTask or an
> ExecDriver. On what does it depend which of the two it is?
>
> Thanks,
> Jan
>
>