I'm pretty consistently seeing a few reduce tasks fail with OutOfMemoryError
(below). It doesn't kill the job, but it slows it down.

 

In my current case the reducer is pretty darn simple, the algorithm
basically does:

1.       Do you have 2 values for this key?

2.       If so, build a json string and emit a NullWritable and Text value.

 

The string buffer I use to build the json is re-used, and I can't see
anywhere in my code that would be taking more than ~50k of memory at any
point in time.

 

But I want to verify, is there a way to get the heap dump and all after this
error? I'm running on AWS MapReduce v1.0.3 of Hadoop.

 

Error: java.lang.OutOfMemoryError: Java heap space

        at
org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.shuffleInMe
mory(ReduceTask.java:1711)

        at
org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.getMapOutpu
t(ReduceTask.java:1571)

        at
org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.copyOutput(
ReduceTask.java:1412)

        at
org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.run(ReduceT
ask.java:1344)

 

 

David,

I faced the same issue due to too much of logging that fills the task
tracker log folder.

Cheers!
Manoj.
On Sat, Dec 22, 2012 at 9:10 PM, Stephen Fritz <[EMAIL PROTECTED]>wrote:

> Troubleshooting OOMs in the map/reduce tasks can be tricky, see page 118
> of Hadoop Operations<http://books.google.com/books?id=W5VWrrCOuQ8C&pg=PA123&lpg=PA123&dq=mapred+child+address+space+size&source=bl&ots=PCdqGFbU-Z&sig=ArgpJroU7UEmMqMB_hwXoCq7whk&hl=en&sa=X&ei=TNPVUMjjHsS60AGHtoHQDA&ved=0CEUQ6AEwAw#v=onepage&q=mapred%20child%20address%20space%20size&f=false>for a couple of settings which could affect the frequency of OOMs which
> aren't necessarily intuitive.
>
> To answer your question about getting the heap dump, you should be able to
> add "-XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/some/path" to your
> mapred.child.java.opts, then look for the heap dump in that path next time
> you see the OOM.
>
>
> On Fri, Dec 21, 2012 at 11:33 PM, David Parks <[EMAIL PROTECTED]>wrote:
>
>> I’m pretty consistently seeing a few reduce tasks fail with
>> OutOfMemoryError (below). It doesn’t kill the job, but it slows it down.*
>> ***
>>
>> ** **
>>
>> In my current case the reducer is pretty darn simple, the algorithm
>> basically does:****
>>
>> **1.       **Do you have 2 values for this key?****
>>
>> **2.       **If so, build a json string and emit a NullWritable and Text
>> value.****
>>
>> ** **
>>
>> The string buffer I use to build the json is re-used, and I can’t see
>> anywhere in my code that would be taking more than ~50k of memory at any
>> point in time.****
>>
>> ** **
>>
>> But I want to verify, is there a way to get the heap dump and all after
>> this error? I’m running on AWS MapReduce v1.0.3 of Hadoop.****
>>
>> ** **
>>
>> Error: java.lang.OutOfMemoryError: Java heap space****
>>
>>         at
>> org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.shuffleInMemory(ReduceTask.java:1711)
>> ****
>>
>>         at
>> org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.getMapOutput(ReduceTask.java:1571)
>> ****
>>
>>         at
>> org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.copyOutput(ReduceTask.java:1412)
>> ****
>>
>>         at
>> org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.run(ReduceTask.java:1344)
>> ****
>>
>> ** **
>>
>> ** **
>>
>
>