[Hadoop-common-dev] Hadoop-Git-Eclipse

Hi Prajakta,

I have Eclipse setup with M2E plugins. And once thats done, I merely
clone a repo and import projects in via M2E's "Import existing maven
projects" feature. This seems to work just fine for apache/hadoop's
trunk.

Hi Harsh

Appreciate the response. I was able to configure and implement basic JUnits
within eclipse and get some code running. Still getting familiar with the
new YARN and federation architecture.

I was, however, not able to check the MR jobs submitted within eclipse for
a sample WordCount program on the
http://localhost:8088/<http://localhost:8080/>page. I am starting my
namenode/datanode/resourcemanager/nodemanager/historyserver as instructed
on the wiki page. And then executing JUnit tests from eclipse.

I believe a single MR job will be submitted as a single application in the
new framework, right? The eclipse console shows a successful execution (the
details are pretty neat). However, the webpage shows 'No applications
submitted'. Do I have to tweak with any config properties to get this done?

Please let me know.

Regards,
Prajakta

Good to know your progress Prajakta!

Did your submission surely go via the RM/NM or did it execute via the
LocalJobRunner (logs show this classname)?

You would ideally want to set the config "mapreduce.framework.name" to
value "yarn" (in either config object before you use it, or in local
mapred-site.xml), for it to use the YARN framework. A set of general
configs for YARN deployment may be found at http://bit.ly/M2Eobz or at
http://bit.ly/LW3Var.

Does this help?

Hi Harsh

You were right - it was executed via the LocalJobRunner. What settings do I
need to change to ensure it goes through RM/NM? I did specify to use the
yarn framework in the mapred-site.xml. The eclipse console even shows to
check http://localhost:8080 to track job progress.

I will again cross-check all the parameters today with the ones specified
in the urls you gave me. By the way, I found this link <
http://hadoop.apache.org/common/docs/r2.0.0-alpha/hadoop-project-dist/hadoop-common/DeprecatedProperties.html
>
which gives a list of deprecated properties. I have, in fact, specified a
lot of old properties in my core-site.xml, hdfs-site.xml, mapred-site.xml
files. Will that cause a problem?

Regards,
Prajakta

Hey,

The presence of deprecated properties should not really hamper your
submissions. The config for YARN (pseudo or fully distributed) I
usually follow are all listed at the links I provided earlier. When
running programs from within the IDE, I place the directory they are
contained in onto my launcher's (Called 'Run Configuration' in
eclipse, for instance) classpath and they are picked up by my program.

Hi Harsh

I can actually run the jobs (and browse them on localhost:8088) when
executed from command line. The same program gives me problems from
eclipse. :(

Moreover, I cannot even browse DFS locations from eclipse (indigo). I am
using hadoop-0.20.3-dev-eclipse-plugin.jar and following steps from
shaswat's video to setup mapreduce environment in eclipse. It gives me
"Error: IPC server version 7 trying to communicate with client version 3".

I believe it is because of using the old (0.20.3) plugin and I could not
find a newer plugin for latest Hadoop versions. Is there a workaround?

I think (not sure) this is the reason that my jobs launched from eclipse
are getting executed by the LocalJobRunner (inspite of specifying details
of classpath in run configurations in eclipse as you suggested).

One more thing, does maven automatically assume jre path even if i specify
JAVA_HOME to my jdk? I had to comment out the following lines and create
the artifact for jdk.tools in ~/.m2/repository manually as eclipse kept
converting the jdk path to jre.
<!-- scope>system</scope>
<systemPath>${java.home}/../lib/tools.jar</systemPath-->

The 'effective pom' view for the hadoop-project/pom.xml showed the jre path
instead of jdk.

I saw your old post for such problems at <
http://www.harshj.com/2010/07/18/making-the-eclipse-plugin-work-for-hadoop/>
but I believe the patch is already committed in the trunk now. Any idea why
eclipse is troubling me like this? :(

Regards,
Prajakta

Hi Prajakta,

can u please tell me the steps to do MapReduce programming using Eclipse .
I really appreciate.. ur help. i m new to hadoop . want to learn

thanks
shant

Follow this video...:

http://www.youtube.com/watch?v=TavehEdfNDk



On Thu, Jun 14, 2012 at 10:14 AM, shant.....
wrote:

Hi Shant

Sorry for the delay in reply (I was not well). Did you manage to install
using Shaswat's video? The video gives good demo for setting up basic
Mapreduce programming environment.

If you need steps for cloning and setting up Hadoop in Eclipse, please
follow this link. <http://wiki.apache.org/hadoop/EclipseEnvironment>
For setting up properties, follow this link <
http://hadoop.apache.org/common/docs/r2.0.0-alpha/hadoop-yarn/hadoop-yarn-site/SingleCluster.html>
Please ensure to specify appropriate values for host:port combinations.

Please let me know if this helps.

Regards,
Prajakta