I specifically mean visualisation via ZeppelinContext inside a Spark interpreter. (e.g. "z.show(...)") The visualisation of SparkSQL results inside a SparkSQLInterpreter work fine, also in yarn-cluster mode.
Am Do., 7. Juni 2018 um 14:30 Uhr schrieb Thomas Bünger < thom.bu...@googlemail.com>: > 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 >>>>>>>> >>>>>>>