hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ted Yu <yuzhih...@gmail.com>
Subject Re: Region Server always Crash When M/R
Date Fri, 08 Nov 2013 04:15:19 GMT
It depends on the expected load on the region server and whether the same
machine is used by other processes (such as DataNode / TaskTracker).

As a first step, you can try 1.5GB

Cheers


On Thu, Nov 7, 2013 at 7:57 PM, GuoWei <wei.guo@wbkit.com> wrote:

> Can I increase to 1.5 G ?
>
> I have 4G total for region server.
>
> Thanks
> 在 2013年11月8日,上午11:57,Ted Yu <yuzhihong@gmail.com> 写道:
>
> Please increase heap size - 617MB is too small.
>
>
> On Thu, Nov 7, 2013 at 7:52 PM, GuoWei <wei.guo@wbkit.com> wrote:
>
>> My Region server configuration as below:
>>
>> <屏幕快照 2013-11-08 上午11.47.52.png>
>>
>>
>> Should I set Java Heap Size to more ?  The region server have 4 G memory
>> total.
>>
>>
>>
>>
>> 在 2013年11月8日,上午11:37,Ted Yu <yuzhihong@gmail.com> 写道:
>>
>> The attachment didn't go through.
>>
>> Can you pastebin the logs ?
>>
>> What was GC like before the crash ?
>>
>> How much heap did you give your region servers ?
>>
>> Cheers
>>
>>
>> On Thu, Nov 7, 2013 at 7:28 PM, GuoWei <wei.guo@wbkit.com> wrote:
>>
>>> Dear,
>>>
>>> Recently when I use CDH 4.0 to do M/R job. I always let Region Server
>>> Crashed. Attached are region server logs.
>>>
>>> And In my M/R program , I set job reduceTaskNum to 4. Because I have two
>>> data nodes each configured 2 reduce slots.
>>>
>>> Please help me find why ?
>>>
>>>
>>> Thanks a lot.
>>>
>>>
>>>
>>> BR
>>> Brad
>>
>>
>>
>>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message