HADOOP_CONF_DIR would affect the whole zeppelin instance. and define it interpreter setting would affect that interpreter.
And all the capitalized property name would be taken as env variable. Serega Sheypak <serega.shey...@gmail.com>于2017年7月1日周六 上午3:20写道： > hi, thanks for your reply. How should I set this variable? > I'm looking at Spark interpreter config UI. It doesn't allow me to set env > variable. > > https://zeppelin.apache.org/docs/latest/interpreter/spark.html#1-export-spark_home > tells that HADOOP_CONF_DIR should be set once per whole Zeppelin instance. > > What do I miss? > Thanks! > > 2017-06-30 16:43 GMT+02:00 Jeff Zhang <zjf...@gmail.com>: > >> >> Right, create three spark interpreters for your 3 yarn cluster. >> >> >> >> Serega Sheypak <serega.shey...@gmail.com>于2017年6月30日周五 下午10:33写道： >> >>> Hi, thanks for your reply! >>> What do you mean by that? >>> I can have only one env variable HADOOP_CONF_DIR... >>> And how can user pick which env to run? >>> >>> Or you mean I have to create three Spark interpreters and each of them >>> would have it's own HADOOP_CONF_DIR pointed to single cluster config? >>> >>> 2017-06-30 16:21 GMT+02:00 Jeff Zhang <zjf...@gmail.com>: >>> >>>> >>>> Try set HADOOP_CONF_DIR for each yarn conf in interpreter setting. >>>> >>>> Serega Sheypak <serega.shey...@gmail.com>于2017年6月30日周五 下午10:11写道： >>>> >>>>> Hi I have several different hadoop clusters, each of them has it's own >>>>> YARN. >>>>> Is it possible to configure single Zeppelin instance to work with >>>>> different clusters? >>>>> I want to run spark on cluster A if data is there. Right now my >>>>> Zeppelin runs on single cluster and it sucks data from remote clusters >>>>> which is inefficient. Zeppelin can easily access any HDFS cluster, but >>>>> what >>>>> about YARN? >>>>> >>>>> What are the correct approaches to solve the problem? >>>>> >>>> >>> >
