Github user vanzin commented on a diff in the pull request:
https://github.com/apache/spark/pull/3233#discussion_r24363190
--- Diff: yarn/src/main/scala/org/apache/spark/deploy/yarn/Client.scala ---
@@ -318,6 +314,13 @@ private[spark] class Client(
env("SPARK_YARN_STAGING_DIR") = stagingDir
env("SPARK_USER") =
UserGroupInformation.getCurrentUser().getShortUserName()
+ // Propagate SPARK_HOME to the containers. This is needed for pyspark
to
+ // work, since the executor's PYTHONPATH is built based on the location
+ // of SPARK_HOME.
+
sparkConf.getOption("spark.home").orElse(sys.env.get("SPARK_HOME")).foreach {
path =>
+ env("SPARK_HOME") = path
+ }
--- End diff --
I think I figured this one out.
The assembly generated by maven includes the python files needed to run
pyspark. The assembly generated by sbt doesn't. So if you deploy the sbt
assembly, since `SPARK_HOME` is not propagated, it won't include the py4j and
pyspark modules in `PYTHONPATH`, and pyspark jobs will fail.
In any case, unrelated to this issue and this is probably the wrong fix
anyway, so I'll revert this part.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
