We use Geomesa on Accumulo with Spark and Zeppelin on a Kerberized cluster (hdp3). We've had a number of issues, but that one doesn't look familiar. >From memory, we had to:
Build geomesa spark with Accumulo version to match our cluster, and libthrift to match Accumulo, and another version change I forget right now. Build zeppelin 0.8.1 with libthrift to match hive (different to Accumulo) (or use hwx zeppelin) Sometimes specify geomesa jar in the config file SPARK_SUBMIT_OPTIONS rather than in the interpreter config. This was definitely an issue <0.8 On Thu, 23 May 2019, 21:53 Krentz, <cpkre...@gmail.com> wrote: > Hello - I am looking for insight into an issue I have been having with our > Zeppelin cluster for a while. We are adding a Geomesa-Accumulo-Spark jar to > the Spark interpreter. The notebook paragraphs run fine until we try to > access the data, at which point we get an "Unread Block Data" error from > the Spark process. However, this error only occurs when the interpreter > setting "zeppelin.spark.useNew" is set to true. If this parameter is set to > false, the paragraph works just fine. Here is a paragraph that fails: > > %sql > select linktype,count(linktype) from linkageview group by linktype > > The error we get as a result is this: > java.lang.IllegalStateException: unread block data > at > java.io.ObjectInputStream$BlockDataInputStream.setBlockDataMode(ObjectInputStream.java:2783) > at java.io.ObjectInputStream.readObject0(ObjectInputStream.java:1605) > at java.io.ObjectInputStream.defaultReadFields(ObjectInputStream.java:2287) > at java.io.ObjectInputStream.readSerialData(ObjectInputStream.java:2211) > at > java.io.ObjectInputStream.readOrdinaryObject(ObjectInputStream.java:2069) > at java.io.ObjectInputStream.readObject0(ObjectInputStream.java:1573) > at java.io.ObjectInputStream.readObject(ObjectInputStream.java:431) > at > org.apache.spark.serializer.JavaDeserializationStream.readObject(JavaSerializer.scala:75) > at > org.apache.spark.serializer.JavaSerializerInstance.deserialize(JavaSerializer.scala:114) > at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:258) > 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) > > > If I drill down and inspect the Spark job itself, I get an error saying > "readObject can't find class > org.apache.accumulo.core.client.mapreduce.impl.BatchInputSplit." The full > stack trace is attached. We dug into and opened up the __spark_conf and > __spark_libs files associated with the Spark job (under > /user/root/.sparkStaging/application_<pid>/ but they did not contain the > jar file containing this method. However, it was not present in both the > spark.useNew true version false version. > > Basically I am just trying to figure out why the spark.useNew option would > cause the error to happen when it works fine turned off. We can move > forward with it turned off for now, but I would like to get to the bottom > of this issue in case there is something deeper going wrong. > > Thanks so much, > Chris Krentz > > >
