I can confirm that this is a bug, and created https://issues.apache.org/jira/browse/ZEPPELIN-3531
Will fix it soon Jeff Zhang <zjf...@gmail.com>于2018年6月5日周二 下午9:01写道： > > hmm, it looks like a bug. I will check it tomorrow. > > > Thomas Bünger <thom.bu...@googlemail.com>于2018年6月5日周二 下午8:56写道： > >> $ ls /usr/lib/spark/python/lib >> py4j-0.10.6-src.zip PY4J_LICENSE.txt pyspark.zip >> >> So folder exists and contains both necessary zips. Please note, that in >> local or yarn-client mode the files are properly picked up from that very >> same location. >> >> How does yarn-cluster work under the hood? Could it be that environment >> variables (like SPARK_HOME) are lost, because they are only available in my >> local shell + zeppelin daemon process? Do I need to tell YARN somehow about >> SPARK_HOME? >> >> Am Di., 5. Juni 2018 um 14:48 Uhr schrieb Jeff Zhang <zjf...@gmail.com>: >> >>> >>> Could you check whether there's folder /usr/lib/spark/python/lib ? >>> >>> >>> Thomas Bünger <thom.bu...@googlemail.com>于2018年6月5日周二 下午8:45写道： >>> >>>> >>>> sys.env >>>> java.lang.NullPointerException at >>>> org.apache.zeppelin.spark.NewSparkInterpreter.setupConfForPySpark(NewSparkInterpreter.java:149) >>>> at >>>> org.apache.zeppelin.spark.NewSparkInterpreter.open(NewSparkInterpreter.java:90) >>>> at >>>> org.apache.zeppelin.spark.SparkInterpreter.open(SparkInterpreter.java:62) >>>> at >>>> org.apache.zeppelin.interpreter.LazyOpenInterpreter.open(LazyOpenInterpreter.java:69) >>>> at >>>> org.apache.zeppelin.interpreter.remote.RemoteInterpreterServer$InterpretJob.jobRun(RemoteInterpreterServer.java:617) >>>> at org.apache.zeppelin.scheduler.Job.run(Job.java:188) at >>>> org.apache.zeppelin.scheduler.FIFOScheduler$1.run(FIFOScheduler.java:140) >>>> at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) >>>> at java.util.concurrent.FutureTask.run(FutureTask.java:266) at >>>> java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$201(ScheduledThreadPoolExecutor.java:180) >>>> at >>>> java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:293) >>>> at >>>> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) >>>> at >>>> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) >>>> at java.lang.Thread.run(Thread.java:748) >>>> >>>> >>>> Am Di., 5. Juni 2018 um 14:41 Uhr schrieb Jeff Zhang <zjf...@gmail.com >>>> >: >>>> >>>>> Could you paste the full stracktrace ? >>>>> >>>>> >>>>> Thomas Bünger <thom.bu...@googlemail.com>于2018年6月5日周二 下午8:21写道： >>>>> >>>>>> I've tried the 0.8.0-rc4 on my EMR cluster using the preinstalled >>>>>> version of spark under /usr/lib/spark. >>>>>> >>>>>> This works fine in local or yarn-client mode, but in yarn-cluster >>>>>> mode i just get a >>>>>> >>>>>> java.lang.NullPointerException at >>>>>> org.apache.zeppelin.spark.NewSparkInterpreter.setupConfForPySpark(NewSparkInterpreter.java:149) >>>>>> >>>>>> Seems to be caused by an unsuccessful search for the py4j libraries. >>>>>> I've made sure that SPARK_HOME is actually set in .bash_rc, in >>>>>> zeppelin-env.sh and via the new %spark.conf, but somehow in the remote >>>>>> interpreter, something odd is going on. >>>>>> >>>>>> Best regards, >>>>>> Thomas >>>>>> >>>>>
