You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hbase.apache.org by Ioakim Perros <im...@gmail.com> on 2012/08/08 02:11:31 UTC
Presplitting table regions - #Reducers different from #Regions
By generalizing my question, in a M/R job, reading/writing from/to the
same Hbase table,
is there any way to work with any desirable number of reducers importing
data to a table,
which has its regions presplitted & specified?
Thanks in advance and excuse me for re-asking for help.
-------- Original Message --------
Subject: Re: Bulk load - #Reducers different from #Regions
Date: Tue, 07 Aug 2012 20:02:27 +0300
From: Ioakim Perros <im...@gmail.com>
To: user@hbase.apache.org
Excuse me for not well-defining.
I am bulk updating my hbase table through code, using
configureIncrementalLoad function of HFileOutputFormat. At the
respective documentation, I read that this function " Sets the number of
reduce tasks to match the current number of regions" ,but I was
wondering if I could explicitly avoid it, perhaps by another way of bulk
importing data.
PS: I try to insist on bulk importing, because I have understood (I hope
that this is correct), that it is much more efficient than going with
the traditional Hbase API. And as I require my job to be of iterative
nature, this way hopefully would end up giving a good boost-up, as
opposed to the Hbase API.
Thank you for responding.
On 08/07/2012 07:53 PM, Subir S wrote:
> Bulk load using
> ImportTsv with pre-splitted regions for target table?
>
> Do u mean to set number of reducers that ImportTsv must use?
>
> On 8/7/12, Ioakim Perros<im...@gmail.com> wrote:
>> HI,
>>
>> I am bulk importing (updating) data iteratively and I would like to be
>> able to set the number of reducers at a M/R task, to be different from
>> the number of regions of the table to which I am updating data.
>>
>> I tried it through job.setNumReduceTasks(#reducers), but the job ignored
>> it.
>>
>> Is there a way to avoid an intermediary job and to set the number of
>> reducers explicitly ?
>> I would be grateful if anyone could shed a light to this.
>>
>> Thanks and regards,
>> Ioakim
>>
>>