[jira] [Created] (ZEPPELIN-4136) Class Cast Exception with Spark Implementations that Backported SparkR Security Fix
Russell Spitzer created ZEPPELIN-4136: - Summary: Class Cast Exception with Spark Implementations that Backported SparkR Security Fix Key: ZEPPELIN-4136 URL: https://issues.apache.org/jira/browse/ZEPPELIN-4136 Project: Zeppelin Issue Type: Bug Components: security, spark Affects Versions: 0.8.1 Reporter: Russell Spitzer Zeppelin uses a version check to determine the return type of the SparkR channel https://github.com/apache/zeppelin/blob/8e6974fdc33e834bc01a5ee594e2cfca4ff3045f/spark/interpreter/src/main/java/org/apache/zeppelin/spark/SparkVersion.java#L92-L97 and https://github.com/apache/zeppelin/blob/735064fdc57ae958fabae85b399bb5af3cb79144/spark/interpreter/src/main/scala/org/apache/spark/SparkRBackend.scala#L34-L44 Datastax Enterprise build of Spark includes this security fix in 2.2.2.X, but since Zeppelin doesn't have knowledge of this (for obvious reasons) it attempts to connect without the secret. While I know this isn't an issue for everyone I think we could fix this issue by attempting to match on return type and then we could remove the version check portion of the code. This may end up looking a bit cleaner too although that may just be my opinion -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Created] (ZEPPELIN-2416) Make Spark Properties Optional in Spark Interperter
Russell Spitzer created ZEPPELIN-2416: - Summary: Make Spark Properties Optional in Spark Interperter Key: ZEPPELIN-2416 URL: https://issues.apache.org/jira/browse/ZEPPELIN-2416 Project: Zeppelin Issue Type: Improvement Components: Interpreters Affects Versions: 0.7.1 Reporter: Russell Spitzer Priority: Trivial I was doing some experimentation connecting Zepplin to DataStax Enterprise and I noticed that the Spark Interperter requires the "master" interpreter parameter to be present in order to work. There are several locations which do null-unsafe operations against the return value of getProperty("master") See https://github.com/apache/zeppelin/blob/v0.7.1/spark/src/main/java/org/apache/zeppelin/spark/SparkInterpreter.java#L310 Then the value of this property is set in the SparkConf even if the value is already set there. https://github.com/apache/zeppelin/blob/v0.7.1/spark/src/main/java/org/apache/zeppelin/spark/SparkInterpreter.java#L335 and https://github.com/apache/zeppelin/blob/v0.7.1/spark/src/main/java/org/apache/zeppelin/spark/SparkInterpreter.java#L449 For this and other parameters it might make sense to allow values that are already set to be used and not overwritten. This would allow Zepplin to work better with "spark-defaults" and those vendors which preset these parameters in the environment. I would propose that all interpreter properties be optional. Operationally this means that the "master" is allowed to be unset if it can be automatically inherited from the environment in the SparkConf. Other parameters would simply need a null check, where they would not be set if they are already null (unset). -- This message was sent by Atlassian JIRA (v6.3.15#6346)