There is a Phoenix / mapreduce integration. If you bypass Hbase you will need to take care to not miss edits that are only in memory and WAL.
If you bypass both Phoenix and Hbase you will have to write code that can interpret both...Possible, yes, but not a good use of your time. Is there some machine learning algorithm you want to use that isn't included in Spark, or that you wouldn't be able to integrate with either Spark or a MapReduce job? > On May 23, 2017, at 1:39 PM, Ash N <[email protected]> wrote: > > Thanks Jonathan. .. > > But am looking to access data directly from HDFS. not go through > phoenix/hbase fir access. > > Is this possible? > > > Best regards > > On May 23, 2017 3:35 PM, "Jonathan Leech" <[email protected]> wrote: > I think you would use Spark for that, via the Phoenix spark plugin. > > > On May 23, 2017, at 12:24 PM, Ash N <[email protected]> wrote: > > > > Hi All, > > > > This may be a silly question. we are storing data through Apache Phoenix. > > Is there anything special we have to do so that machine learning and other > > analytics workloads can access this data from HDFS layer? > > > > Considering HBase stores data in HDFS. > > > > > > thanks, > > -ash >
