Github user JoshRosen commented on the pull request:
https://github.com/apache/spark/pull/2432#issuecomment-56910689
Hmm, it looks like the first failure might be due to change in the
initialization order of the event logger and the task scheduler:
(in `ReplSuite.local-cluster-mode` test)
```
[info] java.lang.NullPointerException
[info] at
org.apache.spark.scheduler.cluster.SparkDeploySchedulerBackend.start(SparkDeploySchedulerBackend.scala:71)
[info] at
org.apache.spark.scheduler.TaskSchedulerImpl.start(TaskSchedulerImpl.scala:141)
[info] at org.apache.spark.SparkContext.<init>(SparkContext.scala:298)
[info] at
org.apache.spark.repl.SparkILoop.createSparkContext(SparkILoop.scala:986)
```
In this case, it looks like we pass the partially-constructed SparkContext
into the TaskScheduler constructor, which causes a problem because eventLogger
is null when SparkDeploySchedulerBackend tries to read it.
The trickiness here is that the current code assumes that we can send the
event log directory as part of the application registration message, but we
don't know that directory's location until _after_ we've registered and
obtained an application id.
---
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]
