Github user mridulm commented on a diff in the pull request:
https://github.com/apache/spark/pull/12113#discussion_r62129859
--- Diff: core/src/main/scala/org/apache/spark/MapOutputTracker.scala ---
@@ -296,10 +290,89 @@ private[spark] class MapOutputTrackerMaster(conf:
SparkConf)
protected val mapStatuses = new ConcurrentHashMap[Int,
Array[MapStatus]]().asScala
private val cachedSerializedStatuses = new ConcurrentHashMap[Int,
Array[Byte]]().asScala
+ private val maxRpcMessageSize = RpcUtils.maxMessageSizeBytes(conf)
+
+ // Kept in sync with cachedSerializedStatuses explicitly
+ // This is required so that the Broadcast variable remains in scope
until we remove
+ // the shuffleId explicitly or implicitly.
+ private val cachedSerializedBroadcast = new HashMap[Int,
Broadcast[Array[Byte]]]()
+
+ // This is to prevent multiple serializations of the same shuffle -
which happens when
+ // there is a request storm when shuffle start.
+ private val shuffleIdLocks = new ConcurrentHashMap[Int, AnyRef]()
+
+ // requests for map output statuses
+ private val mapOutputRequests = new
LinkedBlockingQueue[GetMapOutputMessage]
+
+ // Thread pool used for handling map output status requests. This is a
separate thread pool
+ // to ensure we don't block the normal dispatcher threads.
+ private val threadpool: ThreadPoolExecutor = {
+ val numThreads =
conf.getInt("spark.shuffle.mapOutput.dispatcher.numThreads", 8)
+ val pool = ThreadUtils.newDaemonFixedThreadPool(numThreads,
"map-output-dispatcher")
+ for (i <- 0 until numThreads) {
+ pool.execute(new MessageLoop)
+ }
+ pool
+ }
+
+ def post(message: GetMapOutputMessage): Unit = {
+ mapOutputRequests.offer(message)
+ }
+
+ /** Message loop used for dispatching messages. */
+ private class MessageLoop extends Runnable {
+ override def run(): Unit = {
+ try {
+ while (true) {
+ try {
+ val data = mapOutputRequests.take()
+ if (data == PoisonPill) {
+ // Put PoisonPill back so that other MessageLoops can see it.
+ mapOutputRequests.offer(PoisonPill)
+ return
+ }
+ val context = data.context
+ val shuffleId = data.shuffleId
+ val hostPort = context.senderAddress.hostPort
+ logDebug("Handling request to send map output locations for
shuffle " + shuffleId +
+ " to " + hostPort)
+ val mapOutputStatuses =
getSerializedMapOutputStatuses(shuffleId)
+ val serializedSize = mapOutputStatuses.length
+ if (serializedSize > maxRpcMessageSize) {
+ val msg = s"Map output statuses were $serializedSize bytes
which " +
+ s"exceeds spark.rpc.message.maxSize ($maxRpcMessageSize
bytes)."
+
+ // For SPARK-1244 we'll opt for just logging an error and
then sending it to
+ // the sender. A bigger refactoring (SPARK-1239) will
ultimately remove this
--- End diff --
@davies @tgravescs If we are to remove existing functionality, then we need
to add a check when initializing minSizeForBroadcast to ensure it is <=
maxRpcMessageSize.
In general, I dont like coupling unrelated functionalities - though here it
looks more logical to do so (alternative is failure at runtime !).
---
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]
