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

Switch to Threaded View
Flume >> mail # user >> Hbase Sink


Copy link to this message
-
Re: HBase sink
On Sun, Jul 29, 2012 at 9:32 AM, Brock Noland <[EMAIL PROTECTED]> wrote:

> Hi,
>
> You might want to write a class which implements HbaseEventSerializer. To
> test, SimpleHbaseEventSerializer might fit your needs. Using timestamps as
> keys will create hotspots. As such it's recommended you "bucket" so that
> each write is hitting say 1 of 32 buckets. Each agent would be configured
> as follows:
>
> ...rowPrefix = bucket
> ...suffix = timestamp
>
Yes my row keys are bucketed and regions would be split per bucket. I just
wanted to see how the row keys are generated. Looks like this needs to be
implemented, which probably makes sense. But it would be good to have some
support for timeseries data as well. Let me see if I can share my
EventSerializer after I get it working :) I'll take a look at
SimpleHbaseEventSerializer

>  Brock
>
> On Sun, Jul 29, 2012 at 11:15 AM, Mohit Anchlia <[EMAIL PROTECTED]>wrote:
>
>> I am looking at example in
>> http://flume.apache.org/FlumeUserGuide.html#hbasesinks but this doesn't
>> show me how to generate row keys. My row keys are timeseries data.
>
>
>
>
> --
> Apache MRUnit - Unit testing MapReduce -
> http://incubator.apache.org/mrunit/
>