Github user pwendell commented on a diff in the pull request:
https://github.com/apache/spark/pull/152#discussion_r10637519
--- Diff: core/src/main/scala/org/apache/spark/MapOutputTracker.scala ---
@@ -35,13 +35,21 @@ private[spark] case class
GetMapOutputStatuses(shuffleId: Int)
extends MapOutputTrackerMessage
private[spark] case object StopMapOutputTracker extends
MapOutputTrackerMessage
-private[spark] class MapOutputTrackerMasterActor(tracker:
MapOutputTrackerMaster)
+private[spark] class MapOutputTrackerMasterActor(tracker:
MapOutputTrackerMaster, conf: SparkConf)
extends Actor with Logging {
+ val maxAkkaFrameSize = AkkaUtils.maxFrameSizeBytes(conf)
+
def receive = {
case GetMapOutputStatuses(shuffleId: Int) =>
val hostPort = sender.path.address.hostPort
logInfo("Asked to send map output locations for shuffle " +
shuffleId + " to " + hostPort)
- sender ! tracker.getSerializedMapOutputStatuses(shuffleId)
+ val mapOutputStatuses =
tracker.getSerializedMapOutputStatuses(shuffleId)
+ val serializedSize = mapOutputStatuses.size
+ if (serializedSize > maxAkkaFrameSize) {
+ throw new SparkException(s"Map output statuses were
$serializedSize bytes which " +
+ s"exceeds spark.akka.frameSize ($maxAkkaFrameSize bytes).")
--- End diff --
Hm great point @markhamstra I forgot the actor will swallow it. I think our
only option here is to just log an error because we can't change the message
format between the master and worker (e.g. to allow the master to signal a
failure in it's response) in a maintenance release and this is something we are
trying to fix for 0.9.1.
---
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.
---
