I have just installed 0.9.0 preview2 on a system intended for instructional use. As such, all interpreters are configured to be per-user isolated. There are a number of issues that you might consider bugs, but others using a similar set up might want to know about. I am using Java 8, Python 3.8, Spark 3.0.0, Scala 2.12, R 3.5.3
I believe some of this is specific to the isolated per-user setup. /usr/bin/python must link to the right python. It's not enough to configure it. ...interpreter/spark has to have a copy of spark-scala-2.12-0.9.0-preview2.jar. Copy it from the right subdirectory. Starting an interpreter must cd to user's home dir or things fail trying to create files in /usr/zeppelin or / I replace bash in the impersonate command with a script that does cd then calls bash The current version of R is 4.0.2. This won't work with SparkR. I had to install 3.5.3 In the spark R examples, the sql function doesn't need sparkcontext as an arg, and fails bkzep doesn't work. It complains about bundle being undefined, and none of the workarounds I could find worked. I'm omitting bkzep. The JS display mostly works, but I sometimes have to refresh the page to get graphics to show. Importing notes doesn't work in chrome. It displays the file rather than uploading it. It works in firefox and Safari, though Safari is otherwise unusable. Safari Mac display is odd and editing is unusable. -- Sent from: http://apache-zeppelin-users-incubating-mailing-list.75479.x6.nabble.com/
