Github user viirya commented on a diff in the pull request:
https://github.com/apache/spark/pull/15541#discussion_r84618189
--- Diff: core/src/main/scala/org/apache/spark/scheduler/TaskAssigner.scala
---
@@ -0,0 +1,229 @@
+/*
+ * 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.scheduler
+
+import scala.collection.mutable.ArrayBuffer
+import scala.collection.mutable.PriorityQueue
+import scala.util.Random
+
+import org.apache.spark.internal.{config, Logging}
+import org.apache.spark.SparkConf
+import org.apache.spark.util.Utils
+
+/** Tracks the current state of the workers with available cores and
assigned task list. */
+private[scheduler] class OfferState(val workOffer: WorkerOffer) {
+ /** The current remaining cores that can be allocated to tasks. */
+ var coresAvailable: Int = workOffer.cores
+ /** The list of tasks that are assigned to this WorkerOffer. */
+ val tasks = new ArrayBuffer[TaskDescription](coresAvailable)
+
+ def assignTask(task: TaskDescription, cpu: Int): Unit = {
+ tasks += task
+ coresAvailable -= cpu
+ assert(coresAvailable >= 0)
+ }
+}
+
+/**
+ * TaskAssigner is the base class for all task assigner implementations,
and can be
+ * extended to implement different task scheduling algorithms.
+ * Together with [[org.apache.spark.scheduler.TaskScheduler
TaskScheduler]], TaskAssigner
+ * is used to assign tasks to workers with available cores. Internally,
when TaskScheduler
+ * performs task assignment given available workers, it first sorts the
candidate tasksets,
+ * and then for each taskset, it takes multiple rounds to request
TaskAssigner for task
+ * assignment with different locality restrictions until there is either
no qualified
+ * workers or no valid tasks to be assigned.
+ *
+ * TaskAssigner is responsible to maintain the worker availability state
and task assignment
+ * information. The contract between
[[org.apache.spark.scheduler.TaskScheduler TaskScheduler]]
+ * and TaskAssigner is as follows.
+ *
+ * First, TaskScheduler invokes construct() of TaskAssigner to initialize
the its internal
+ * worker states at the beginning of resource offering.
+ *
+ * Second, before each round of task assignment for a taskset,
TaskScheduler invokes the init()
+ * of TaskAssigner to initialize the data structure for the round.
+ *
+ * Third, when performing real task assignment, hasNext/next() is used by
TaskScheduler
+ * to check the worker availability and retrieve current offering from
TaskAssigner.
+ *
+ * Fourth, TaskScheduler calls offerAccepted() to notify the TaskAssigner
so that
+ * TaskAssigner can decide whether the current offer is valid or not for
the next request.
+ *
+ * Fifth, after task assignment is done, TaskScheduler invokes the
function tasks to
+ * retrieve all the task assignment information.
+ */
+
+private[scheduler] sealed abstract class TaskAssigner {
+ protected var offer: Seq[OfferState] = _
+ protected var cpuPerTask = 1
+
+ protected def withCpuPerTask(cpuPerTask: Int): TaskAssigner = {
+ this.cpuPerTask = cpuPerTask
--- End diff --
Do we need to add sanity check for `cpuPerTask`?
---
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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
