hadoop-mapreduce-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Todd Lipcon <t...@cloudera.com>
Subject Re: How compile hadoop-0.20.1
Date Tue, 24 Nov 2009 01:53:06 GMT
Hi Pedro,

If you're trying to compile 0.20.1, that version is pre-project-split.
There's only one jar because it predates the point at which Hadoop was split
into common/hdfs/MR. I've answered the specific questions below:

On Mon, Nov 23, 2009 at 2:39 PM, psdc1978 <psdc1978@gmail.com> wrote:

>  Hi all,
>  1 - I'm trying to compile hadoop-0.20.1 but I'm facing several
> errors.  Hadoop includes several subprojects, but these subprojects
> are all compiled in one file called hadoop-0.20.1-core.jar. If I just
> want to update the code in the MapReduce component, do I have to
> compile all the project?

In 0.20.1, yes. But it compiles quickly - only a couple minutes.

>  2 - Why Hadoop only have one jar for all the project? For example,
> MapReduce subcomponent could be inside mapreduce.jar, Common
> subcomponent could be inside common.jar, HDFS subcomponent could be
> inside hdfs.jar, etc.
As noted above, 0.20.1 is pre-split. In 0.21 and above, it will be separate
jars as you expected.

>  3 - Is there any tutorial that helps how to compile Hadoop,
> especially only the MapReduce component?
> In 0.20.1, you can't do just one component, really. "ant bin-package" is a
good way to compile that won't require a lot of other junk on your system.

>  4 - Should I try to compile MapReduce in windows with Cygwin, or
> should I compile inside Ubuntu?
> Ubuntu should be easier, though I'm told it's possible to do so in Cygwin.


  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message