Github user srowen commented on a diff in the pull request: https://github.com/apache/spark/pull/15618#discussion_r85663898 --- Diff: streaming/src/main/scala/org/apache/spark/streaming/scheduler/ReceiverTracker.scala --- @@ -154,6 +147,17 @@ class ReceiverTracker(ssc: StreamingContext, skipReceiverLaunch: Boolean = false throw new SparkException("ReceiverTracker already started") } + if (receivedBlockTracker == null) { --- End diff -- It looks like this is meant to be restartable, in that it will allow a stopped tracker to call `start()` again. If that's the case, then `receivedBlockTracker` needs to be nulled in `stop()`, and then this check isn't needed I think because it should really be null. In fact if you like you can require that it's null, and that endpoint is null, in `start()`? But then ... I think this method needs to throw an exception also if `isTrackerStopping` at the start. It won't observe the intermediate state because of syncrhonization, but it could be left in this state if stopping failed. I am wary of going too far down a path of fixing or changing the corner cases here. I do agree that there's a legitimate mismatch between where the tracker is initialized and stopped, and I think the changes you made and that possibly the ones I mention here together improve the situation without complicating it further.
--- 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 infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org