boyan created ZEPPELIN-6152:
-------------------------------
Summary: Unable so execute spark jobs after upgrade of Zeppelin
0.10.1 and Spark 3.3.3
Key: ZEPPELIN-6152
URL: https://issues.apache.org/jira/browse/ZEPPELIN-6152
Project: Zeppelin
Issue Type: Bug
Components: spark
Affects Versions: 0.10.1
Environment: Standalone Zeppelin Instance running in a container on
OpenShift cluster.
Used to trigger remote spark's jobs on remote yarn cluster.
Reporter: boyan
Hello Zeppelin Team,
I hope you can help us identify what is causing the issues. We are using
Zeppelin with Spark interpreter with Mapr (Hadoop). In our environment with
Zeppelin 0.9.0, Spark 2.4.4 and Mapr 6 we have no issues of executing simple
Spark commands (e.g. println("Hello World") ) using spark interpreter.
However, in our preparation for upgrade to Zeppelin 0.10.1, Spark 3.3.3 and
Mapr 7, we found an issue in our test environment. We could not trigger anymore
simple spark command. We are always getting this error:
INFO [2024-11-18 10:55:43,398] (\{RegisterThread}
RemoteInterpreterServer.java[run]:603) - Start registration INFO [2024-11-18
10:55:43,512] (\{RegisterThread} RemoteInterpreterServer.java[run]:617) -
Registering interpreter process ERROR [2024-11-18 10:55:43,519]
(\{RegisterThread} RemoteInterpreterServer.java[run]:621) - Error while
registering interpreter: RegisterInfo(host:x.x.x.x, port:29037,
interpreterGroupId:spark-shared_process), cause: {} java.lang.RuntimeException
at
org.apache.zeppelin.interpreter.remote.PooledRemoteClient.callRemoteFunction(PooledRemoteClient.java:119)
at
org.apache.zeppelin.interpreter.remote.RemoteInterpreterEventClient.callRemoteFunction(RemoteInterpreterEventClient.java:80)
at
org.apache.zeppelin.interpreter.remote.RemoteInterpreterEventClient.registerInterpreterProcess(RemoteInterpreterEventClient.java:88)
at
org.apache.zeppelin.interpreter.remote.RemoteInterpreterServer$RegisterRunnable.run(RemoteInterpreterServer.java:618)
at java.base/java.lang.Thread.run(Thread.java:829) INFO [2024-11-18
10:55:43,520] (\{RegisterThread} RemoteInterpreterServer.java[shutdown]:250) -
Unregister interpreter process ERROR [2024-11-18 10:55:43,526]
(\{RegisterThread} RemoteInterpreterServer.java[shutdown]:253) - Fail to
unregister remote interpreter process java.lang.RuntimeException at
org.apache.zeppelin.interpreter.remote.PooledRemoteClient.callRemoteFunction(PooledRemoteClient.java:119)
at
org.apache.zeppelin.interpreter.remote.RemoteInterpreterEventClient.callRemoteFunction(RemoteInterpreterEventClient.java:80)
at
org.apache.zeppelin.interpreter.remote.RemoteInterpreterEventClient.unRegisterInterpreterProcess(RemoteInterpreterEventClient.java:95)
at
org.apache.zeppelin.interpreter.remote.RemoteInterpreterServer.shutdown(RemoteInterpreterServer.java:251)
at
org.apache.zeppelin.interpreter.remote.RemoteInterpreterServer$RegisterRunnable.run(RemoteInterpreterServer.java:623)
at java.base/java.lang.Thread.run(Thread.java:829) INFO [2024-11-18
10:55:43,526] (\{RegisterThread} RemoteInterpreterServer.java[run]:640) -
Registration finished INFO [2024-11-18 10:55:43,526] (\{ShutdownThread}
RemoteInterpreterServer.java[run]:657) - Shutting down...
The way we trigger the spark jobs is with yarn cluster and we use the same
configurations as before. However, we found that Zeppelin's interpreter
triggering code has changed dramatically from the old version of Zeppelin.
We have no issues of triggering spark-submit jobs directly from the spark
folder or with Zeppelin's spark-submit interpreter.
Thank you for your support!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
