>> As you shall see, >> https://github.com/apache/incubator-zeppelin/blob/master/spark/src/main/java/org/apache/zeppelin/spark/SparkInterpreter.java#L354 >> >> <https://github.com/apache/incubator-zeppelin/blob/master/spark/src/main/java/org/apache/zeppelin/spark/SparkInterpreter.java#L354> >> >> Env var overrides property setting, so if env var is set in zeppelin-env.sh >> then the user setting the property on the Interpreter page would have no >> effect. So that behavior seems to be opposite to what you want?
To summarize what I document below, I do not observe the behavior that Felix documents above. It appears that setting env vars provides defaults and the user can indeed override them in the interpreter settings. So here’s what I did. I build zeppelin using the following: mvn clean package -Pspark-1.5 -Dhadoop.version=2.6.0 -Phadoop-2.6 -DskipTests -P build-distr Then I: tar -zxvf ~/Development/zeppelin/cching/incubator-zeppelin/zeppelin-distribution/target/zeppelin-0.6.0-incubating-SNAPSHOT.tar.gz cd zeppelin-0.6.0-incubating-SNAPSHOT ls -l conf <Here I note that conf/interpreters.json does not exist> echo 'export MASTER=spark://hyperdawn:7077' >conf/zeppelin-env.sh echo 'export SPARK_HOME=/Users/cching/apps/spark/spark-1.5.1-bin-hadoop2.6' >>conf/zeppelin-env.sh echo 'export ZEPPELIN_PORT=8082' >>conf/zeppelin-env.sh <I have a port conflict on 8080, hence I use port 8082 instead> ➜ cat conf/zeppelin-env.sh export MASTER=spark://hyperdawn:7077 export SPARK_HOME=/Users/cching/apps/spark/spark-1.5.1-bin-hadoop2.6 export ZEPPELIN_PORT=8082 bin/zeppelin.sh I then go to http://localhost:8082/#/interpreter <http://localhost:8082/#/interpreter> and note: master spark://hyperdawn:7077 spark.home /Users/cching/apps/spark/spark-1.5.1-bin-hadoop2.6 I then change “master” to “spark://localhost:7077 <spark://localhost:7077>” and spark.home to a blank value and save. I am prompted to confirm that I want to save my changes and restart, though I’m not sure if a restart actually occurred. Certainly on the console it doesn’t appear to have done anything. So I manually restart with a Ctrl+C and bin/zeppelin.sh. I refresh http://localhost:8082/#/interpreter <http://localhost:8082/#/interpreter> and check “master” and “spark.home”: master spark://localhost:7077 spark.home They appear to be what I, as a user, have set them to and not to what’s set in the environment variables. This appears to be exactly what I want. My immediate need is to simply be able to set spark.app.name, but ideally the above is the behavior I want. So, what is my recourse here? I note that IMO setting spark.app.name through environment variables is not the best way to do things, but with spark master and spark.home being set this way, I don’t see that what I want to do is inconsistent with that. Felix or any other developers, I really would appreciate why what I want to do is unacceptable. I am a reasonable person, but I need to solve my problem ;-) Cheers, Craig
