Comments on: Debugging Hadoop applications using your Eclipse

By: Pedro

Pedro — Wed, 28 Mar 2012 14:57:35 +0000

This might help you for standalone. You probably can step through the main method, but you can't debug the mapreduce classes. Try doing the following: Add this line to your code:
jobConf.set("mapred.job.tracker","local");
That should force Hadoop to run the mapreduce in the local process and therefore you should be able to step through. There are ways to do it externally (via XML) but this is the quickest way I know.

By: Nitin

Nitin — Fri, 09 Mar 2012 05:36:24 +0000

it works only in standalone mode, would not work in pseudo and real cluster mode.

By: Jack Parsons

Jack Parsons — Mon, 05 Jul 2010 02:30:49 +0000

I have the same problem running a map-reduce job locally inside Eclipse. Is there some weird thing it does to the classpath or something that nukes breakpoints? Does it generate new code from the existing class bytecodes?

Thanks – newbie at Hadoop in Eclipse.

By: elton tian

elton tian — Mon, 12 Apr 2010 03:38:08 +0000

Hello,

I tried the way you described above. I got the connection established when dubugging, which is great!
The problem is the breakpoint I set is not actually "break" the program from running. It's still keep running to the end.

Do you have any idea on this issue?

Regards,
Elton