Github user nishkamravi2 commented on a diff in the pull request:
https://github.com/apache/spark/pull/6607#discussion_r32879522
--- Diff:
streaming/src/main/scala/org/apache/spark/streaming/scheduler/ReceiverTracker.scala
---
@@ -308,12 +336,25 @@ class ReceiverTracker(ssc: StreamingContext,
skipReceiverLaunch: Boolean = false
supervisor.start()
supervisor.awaitTermination()
}
+
// Run the dummy Spark job to ensure that all slaves have registered.
// This avoids all the receivers to be scheduled on the same node.
if (!ssc.sparkContext.isLocal) {
ssc.sparkContext.makeRDD(1 to 50, 50).map(x => (x,
1)).reduceByKey(_ + _, 20).collect()
}
+ // Get the list of executors and schedule receivers
+ val executors = getExecutors(ssc)
+ val locations = scheduleReceivers(receivers, executors)
+ val tempRDD =
+ if (locations(0) != null) {
--- End diff --
That's fine. Following, I think, is slightly cleaner:
if(!executors.isEmpty){
scheduleReceivers
make RDD with returned result
}else{
make RDD
}
Avoids the redundant invocation to scheduleReceivers and subsequent memory
allocations. Optimizes away the extra logic (check on local) and assumptions
about how locations is formatted. If this sounds good, I think we can get the
final iteration of this PR going.
---
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]
