Well , the OOM issue about Pig Bag has been fixed by me.  I've push it into 
production and no OOM about Bag any more. I've contributed a patch here: 
https://issues.apache.org/jira/browse/PIG-2812 . Hope this can draw someone's 
attention and merge it into trunk.


Haitao Yao
[email protected]
weibo: @haitao_yao
Skype:  haitao.yao.final

在 2012-8-22,上午6:05, Jonathan Coveney 写道:

> Make sure that any Pig process has at least 1gb. I've seen bag-related
> OOM's when people don't give enough heap to Pig.
> 
> 2012/8/21 Haitao Yao <[email protected]>
> 
>> Thank you very much for this.
>> But I still can not find which snippet of the script caused this OOM. The
>> heap dump is generated in the midnight, nobody is standing by.
>> 
>> I 've added some scripts to capture more information about the job. If
>> succeed,  I 'll share with you tomorrow.
>> Thanks again.
>> 
>> 
>> 
>> Haitao Yao
>> [email protected]
>> weibo: @haitao_yao
>> Skype:  haitao.yao.final
>> 
>> 在 2012-8-21,下午5:22, Subir S 写道:
>> 
>>> I think it helps to get some context, if you can spin a small snippet
>> from your script which causes this issue. This may give a better
>> understanding of what you are trying to achieve, and if there is some work
>> around. However this is upto you!
>>> 
>>> Thanks, Subir
>>> 
>>> On Tue, Aug 21, 2012 at 2:12 PM, Haitao Yao <[email protected]> wrote:
>>> I 'v found the reason: org.apache.hadoop.io.DataInputBuffer , there's a
>> big byte array referenced by DataInputBuffer.
>>> 
>>> But there's no any way to close the buffer.
>>> 
>>> Is  there any other solution?
>>> 
>>> BTW, my script is more than 500 lines, sorry i can not show you.
>>> 
>>> here's the screenshot:
>>> 
>>> <aa.jpg>
>>> 
>>> 
>>> 
>>> 
>>> <bb.jpg>
>>> 
>>> 
>>> 
>>> Haitao Yao
>>> [email protected]
>>> weibo: @haitao_yao
>>> Skype:  haitao.yao.final
>>> 
>>> 在 2012-8-20,上午9:49, Haitao Yao 写道:
>>> 
>>>> Hi, all,
>>>>     I got an OOME , on org.apache.hadoop.mapreduce.Reducer$Context,
>> here's the snapshot of the heap dump:
>>>> <aa.jpg>
>>>> 
>>>> Well, does pig have to report so many data through  the Reducer$Context?
>>>> Can this be closed?
>>>> 
>>>> thanks.
>>>> 
>>>> 
>>>> 
>>>> Haitao Yao
>>>> [email protected]
>>>> weibo: @haitao_yao
>>>> Skype:  haitao.yao.final
>>>> 
>>> 
>>> 
>> 
>> 

Reply via email to