Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Plain View
HDFS >> mail # user >> Re: Memory Leak while using LocalJobRunner


Copy link to this message
-
Re: Memory Leak while using LocalJobRunner
Hi Vinod,

Thanks Vinod…..
I have raised the issue.
https://issues.apache.org/jira/browse/MAPREDUCE-5258

Cheers,
Subroto Sanyal
On May 17, 2013, at 7:33 PM, Vinod Kumar Vavilapalli wrote:

>
> It's inconvenient, but you can try DefaultMetricsSystem.shutDown(). Please see if that works and file a ticket to make LocalJobRunner automatically do this on finish.
>
> Thanks,
> +Vinod Kumar Vavilapalli
> Hortonworks Inc.
> http://hortonworks.com/
>
> On May 17, 2013, at 7:39 AM, Subroto wrote:
>
>> Hi,
>>
>> I have an application which uses LocalJobRunner to process MR Jobs. From the heap dump of my application I can see multiple object with Name $Proxy90;  happens to be created by org.apache.hadoop.metrics2.impl.MetricsSystemImpl.register(Callback).
>> Every-time a LocalJobRunner is launched it creates JobTrackerInstrumentation and QueueMetrics.
>>
>> While creating this MetricsSystem ; it registers and adds a Callback to ArrayList which keeps on growing as the DefaultMetricsSystem is Singleton. Is it a known bug? Is there any workaround to empty this list?
>>
>> Cheers,
>> Subroto Sanyal
>