Hi,

I'm trying to debug map and reduce tasks for a quite long time, and it
seems that it's impossible. MR are launched in new process and there's no
way to debug them. Even with IsolationRunner class it's impossible. This
isn't good because I really need to debug the class, to understand some
changes that I made to the code.

I wonder how MapReduce programmers could debug the code that they
implemented?

--
Best regards,

Can I suggest unit testing with MRUnit:

https://cwiki.apache.org/confluence/display/MRUNIT/Index
https://cwiki.apache.org/confluence/display/MRUNIT/MRUnit+Tutorial

Brock

On Thu, Mar 29, 2012 at 10:33 AM, Pedro Costa <[EMAIL PROTECTED]> wrote:
> Hi,
>
> I'm trying to debug map and reduce tasks for a quite long time, and it seems
> that it's impossible. MR are launched in new process and there's no way to
> debug them. Even with IsolationRunner class it's impossible. This isn't good
> because I really need to debug the class, to understand some changes that I
> made to the code.
>
> I wonder how MapReduce programmers could debug the code that they
> implemented?
>
> --
> Best regards,
>

--
Apache MRUnit - Unit testing MapReduce - http://incubator.apache.org/mrunit/

You can also decompose your problem so that you have code that is testable outside the map-reduce process.  This way, you can ensure all your "helper" classes/methods are behaving as you expect, and you can debug this locally.  Once that is done, you are left with only debugging things related to MR plumbing.  MRUnit is good for this as Brock pointed out.  But even if you can't use that for some reason, knowing the helper logic is correct is a big help with complex jobs.

-paul

-----Original Message-----
From: Brock Noland [mailto:[EMAIL PROTECTED]]
Sent: Thursday, March 29, 2012 12:08 PM
To: [EMAIL PROTECTED]
Subject: Re: Debug MR tasks impossible.

Can I suggest unit testing with MRUnit:

https://cwiki.apache.org/confluence/display/MRUNIT/Index
https://cwiki.apache.org/confluence/display/MRUNIT/MRUnit+Tutorial

Brock

On Thu, Mar 29, 2012 at 10:33 AM, Pedro Costa <[EMAIL PROTECTED]> wrote:
> Hi,
>
> I'm trying to debug map and reduce tasks for a quite long time, and it seems
> that it's impossible. MR are launched in new process and there's no way to
> debug them. Even with IsolationRunner class it's impossible. This isn't good
> because I really need to debug the class, to understand some changes that I
> made to the code.
>
> I wonder how MapReduce programmers could debug the code that they
> implemented?
>
> --
> Best regards,
>

--
Apache MRUnit - Unit testing MapReduce - http://incubator.apache.org/mrunit/

You can run a job in local mode without a cluster and run in a debugger
enabling breakpoints single stepping - examination of variables ...
If you do this
1)  Choose a SMALL problem
2) Beware of the fact that there is only one VM - that means be cautions
with static variables - they cause leakage from mappers to reducers.
3) Keep testing on the cluster
My rule is never try a task on a cluster until it works well in local mode

On Thu, Mar 29, 2012 at 9:13 AM, Wydra, Paul <[EMAIL PROTECTED]> wrote:

> You can also decompose your problem so that you have code that is testable
> outside the map-reduce process.  This way, you can ensure all your "helper"
> classes/methods are behaving as you expect, and you can debug this locally.
>  Once that is done, you are left with only debugging things related to MR
> plumbing.  MRUnit is good for this as Brock pointed out.  But even if you
> can't use that for some reason, knowing the helper logic is correct is a
> big help with complex jobs.
>
> -paul
>
> -----Original Message-----
> From: Brock Noland [mailto:[EMAIL PROTECTED]]
> Sent: Thursday, March 29, 2012 12:08 PM
> To: [EMAIL PROTECTED]
> Subject: Re: Debug MR tasks impossible.
>
> Can I suggest unit testing with MRUnit:
>
> https://cwiki.apache.org/confluence/display/MRUNIT/Index
> https://cwiki.apache.org/confluence/display/MRUNIT/MRUnit+Tutorial
>
> Brock
>
> On Thu, Mar 29, 2012 at 10:33 AM, Pedro Costa <[EMAIL PROTECTED]> wrote:
> > Hi,
> >
> > I'm trying to debug map and reduce tasks for a quite long time, and it
> seems
> > that it's impossible. MR are launched in new process and there's no way
> to
> > debug them. Even with IsolationRunner class it's impossible. This isn't
> good
> > because I really need to debug the class, to understand some changes
> that I
> > made to the code.
> >
> > I wonder how MapReduce programmers could debug the code that they
> > implemented?
> >
> > --
> > Best regards,
> >
>
>
>
> --
> Apache MRUnit - Unit testing MapReduce -
> http://incubator.apache.org/mrunit/
>

--
Steven M. Lewis PhD
4221 105th Ave NE
Kirkland, WA 98033
206-384-1340 (cell)
Skype lordjoe_com

Hi Pedro,

I know its sub-optimal but you should be able to put in as many
System.out.println / log messages as you want and you should be able to see
them in stdout, and syslog files. Which version of hadoop are you using?

On Thu, Mar 29, 2012 at 10:33 AM, Pedro Costa <[EMAIL PROTECTED]> wrote:

> Hi,
>
> I'm trying to debug map and reduce tasks for a quite long time, and it
> seems that it's impossible. MR are launched in new process and there's no
> way to debug them. Even with IsolationRunner class it's impossible. This
> isn't good because I really need to debug the class, to understand some
> changes that I made to the code.
>
> I wonder how MapReduce programmers could debug the code that they
> implemented?
>
> --
> Best regards,
>
>

Hi Pedro,

    You can use the below property to pass the debug params to the child jvm. And also you should make sure that have only one task running at a time by giving the input appropriately.

<property>
  <name>mapred.child.java.opts</name>
  <value>-Xmx200m</value>
  <description>Java opts for the task tracker child processes.
  The following symbol, if present, will be interpolated: @taskid@ is replaced
  by current TaskID. Any other occurrences of '@' will go unchanged.
  For example, to enable verbose gc logging to a file named for the taskid in
  /tmp and to set the heap maximum to be a gigabyte, pass a 'value' of:
        -Xmx1024m -verbose:gc -Xloggc:/tmp/@taskid@.gc

  The configuration variable mapred.child.ulimit can be used to control the
  maximum virtual memory of the child processes.
  </description>
</property>

Thanks
Devaraj
________________________________
From: Ravi Prakash [[EMAIL PROTECTED]]
Sent: Thursday, March 29, 2012 9:36 PM
To: [EMAIL PROTECTED]
Subject: Re: Debug MR tasks impossible.

Hi Pedro,

I know its sub-optimal but you should be able to put in as many System.out.println / log messages as you want and you should be able to see them in stdout, and syslog files. Which version of hadoop are you using?

On Thu, Mar 29, 2012 at 10:33 AM, Pedro Costa <[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>> wrote:
Hi,

I'm trying to debug map and reduce tasks for a quite long time, and it seems that it's impossible. MR are launched in new process and there's no way to debug them. Even with IsolationRunner class it's impossible. This isn't good because I really need to debug the class, to understand some changes that I made to the code.

I wonder how MapReduce programmers could debug the code that they implemented?

--
Best regards,

On 0.23 I usually breakpoint on DefaultContainerExecutor.java : 139. Modify
the script mentioned in nmPrivateContainerScriptPath to include JAVA remote
debugging options. Connect to the nodemanager and then breakpoint again in
MRAppMaster.java. Tedious. I know. :(

On Thu, Mar 29, 2012 at 11:06 AM, Ravi Prakash <[EMAIL PROTECTED]> wrote:

> Hi Pedro,
>
> I know its sub-optimal but you should be able to put in as many
> System.out.println / log messages as you want and you should be able to see
> them in stdout, and syslog files. Which version of hadoop are you using?
>
>
>
>
> On Thu, Mar 29, 2012 at 10:33 AM, Pedro Costa <[EMAIL PROTECTED]> wrote:
>
>> Hi,
>>
>> I'm trying to debug map and reduce tasks for a quite long time, and it
>> seems that it's impossible. MR are launched in new process and there's no
>> way to debug them. Even with IsolationRunner class it's impossible. This
>> isn't good because I really need to debug the class, to understand some
>> changes that I made to the code.
>>
>> I wonder how MapReduce programmers could debug the code that they
>> implemented?
>>
>> --
>> Best regards,
>>
>>
>