Github user HeartSaVioR commented on a diff in the pull request:
https://github.com/apache/spark/pull/21385#discussion_r190131693
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/continuous/shuffle/UnsafeRowReceiver.scala
---
@@ -56,20 +69,73 @@ private[shuffle] class UnsafeRowReceiver(
override def receiveAndReply(context: RpcCallContext):
PartialFunction[Any, Unit] = {
case r: UnsafeRowReceiverMessage =>
- queue.put(r)
+ queues(r.writerId).put(r)
context.reply(())
}
override def read(): Iterator[UnsafeRow] = {
new NextIterator[UnsafeRow] {
- override def getNext(): UnsafeRow = queue.take() match {
- case ReceiverRow(r) => r
- case ReceiverEpochMarker() =>
- finished = true
- null
+ // An array of flags for whether each writer ID has gotten an epoch
marker.
+ private val writerEpochMarkersReceived =
+ mutable.Map.empty[Int, Boolean].withDefaultValue(false)
+
+ private val executor =
Executors.newFixedThreadPool(numShuffleWriters)
--- End diff --
And I'm also now seeing this approach as alternative to deal with alignment
(not buffer rows explicitly but just don't read after epoch comes in). Nice
approach.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
