Github user JoshRosen commented on a diff in the pull request:
https://github.com/apache/spark/pull/7276#discussion_r34519959
--- Diff:
streaming/src/main/scala/org/apache/spark/streaming/scheduler/ReceiverScheduler.scala
---
@@ -0,0 +1,87 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.spark.streaming.scheduler
+
+import scala.collection.mutable
+import scala.util.Random
+
+import org.apache.spark.streaming.scheduler.ReceiverState._
+
+private[streaming] case class ReceiverTrackingInfo(
+ receiverId: Int,
+ state: ReceiverState,
+ scheduledLocations: Option[Seq[String]],
+ runningLocation: Option[String])
+
+private[streaming] trait ReceiverScheduler {
+
+ /**
+ * Return a candidate executor list to run the receiver. If the list is
empty, the caller can run
+ * this receiver in arbitrary executor.
+ */
+ def scheduleReceiver(
+ receiverId: Int,
+ preferredLocation: Option[String],
+ receiverTrackingInfoMap: Map[Int, ReceiverTrackingInfo],
+ executors: Seq[String]): Seq[String]
+}
+
+/**
+ * A ReceiverScheduler trying to balance executors' load.
+ */
+private[streaming] class LoadBalanceReceiverSchedulerImpl extends
ReceiverScheduler {
+
+ def scheduleReceiver(
+ receiverId: Int,
+ preferredLocation: Option[String],
+ receiverTrackingInfoMap: Map[Int, ReceiverTrackingInfo],
+ executors: Seq[String]): Seq[String] = {
+ if (executors.isEmpty) {
+ return Seq.empty
+ }
+
+ // Always try to schedule to the preferred locations
+ val locations = mutable.Set[String]()
+ locations ++= preferredLocation
+
+ val executorWeights = receiverTrackingInfoMap.filter { case (id, _) =>
+ // Ignore the receiver to be scheduled. It may be still running.
+ id != receiverId
+ }.values.flatMap { receiverTrackingInfo =>
+ receiverTrackingInfo.state match {
+ case ReceiverState.INACTIVE => Nil
+ case ReceiverState.SCHEDULED =>
+ val scheduledLocations =
receiverTrackingInfo.scheduledLocations.get
+ // The probability that a scheduled receiver will run in an
executor is
+ // 1.0 / scheduledLocations.size
+ scheduledLocations.map(location => (location, 1.0 /
scheduledLocations.size))
+ case ReceiverState.ACTIVE =>
Seq(receiverTrackingInfo.runningLocation.get -> 1.0)
+ }
+ }.groupBy(_._1).mapValues(_.map(_._2).sum) // Sum weights for each
executor
--- End diff --
Ignore this comment if this code is not performance-critical, but I wonder
whether writing imperative code that updates a mutable hashmap might be more
efficient here than materializing all of the values for each group. If we don't
go the imperative route, maybe there's still a more efficient functional way to
handle this by doing an aggregate-by-key instead of group-by followed by
aggregate. Disregard this comment if perf. isn't a factor here, though.
---
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]
