There are a lot of new APIs with Spark 2 and more importantly the new entry point is SparkSession (available as spark variable).
I think Z built against Spark 2 should leverage spark variable by default instead of using SC. Please ignore this if the current Spark2 support already includes this feature. Thanks, Vinay On Wed, Aug 3, 2016 at 9:14 PM, Jeff Zhang <zjf...@gmail.com> wrote: > Thanks moon, if that works, it seem not necessary to have different > profiles for different spark version. We can always build it with the > latest spark version. > > On Thu, Aug 4, 2016 at 11:26 AM, moon soo Lee <m...@apache.org> wrote: > >> Hi, >> >> Could you try remove 'SPARK_HOME' from conf/zeppelin-env.sh and add >> 'SPARK_HOME' property (in different version of spark directory) in each >> individual spark interpreter setting on GUI? This should work. >> >> Best, >> moon >> >> On Wed, Aug 3, 2016 at 6:42 PM Jeff Zhang <zjf...@gmail.com> wrote: >> >>> I build zeppelin with spark-2.0 profile enabled, and it seems I can also >>> run spark 1.6 interpreter. But I am not sure whether it is officially >>> supported to run different versions of spark interpreter in one zeppelin >>> build ? My guess maybe it is not, otherwise we don't need profiles for >>> different spark version. If it is not, then I think support multiple >>> versions of spark interpreter in one zeppelin build might be useful, >>> otherwise user need to setup multiple zeppelin servers for different spark >>> versions. >>> >>> >>> >>> -- >>> Best Regards >>> >>> Jeff Zhang >>> >> > > > -- > Best Regards > > Jeff Zhang >
