Hey Jeff, I tried your changes and now it works nicely. Thank you very much!
But I still can't use any of the forms and visualizations in yarn-cluster? I was hoping that this got resolved with the new SparkInterpreter so that I can switch from yarn-client to yarn-cluster mode in 0.8, but I'm still getting errors like "error: not found: value z" Was this not in scope of that change? Is this a bug? Or is it known limitation and also not supported in 0.8? Best regards, Thomas Am Mi., 6. Juni 2018 um 03:28 Uhr schrieb Jeff Zhang <zjf...@gmail.com>: > > 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 >>>>>>> >>>>>>