wbo4958 commented on code in PR #43494:
URL: https://github.com/apache/spark/pull/43494#discussion_r1376085306
##########
core/src/main/scala/org/apache/spark/resource/ResourceAllocator.scala:
##########
@@ -29,59 +65,45 @@ private[spark] trait ResourceAllocator {
protected def resourceName: String
protected def resourceAddresses: Seq[String]
- protected def slotsPerAddress: Int
/**
- * Map from an address to its availability, a value > 0 means the address is
available,
- * while value of 0 means the address is fully assigned.
- *
- * For task resources ([[org.apache.spark.scheduler.ExecutorResourceInfo]]),
this value
- * can be a multiple, such that each address can be allocated up to
[[slotsPerAddress]]
- * times.
+ * Map from an address to its availability default to RESOURCE_TOTAL_AMOUNT,
a value > 0 means
+ * the address is available, while value of 0 means the address is fully
assigned.
*/
- private lazy val addressAvailabilityMap = {
- mutable.HashMap(resourceAddresses.map(_ -> slotsPerAddress): _*)
+ protected lazy val addressAvailabilityMap = {
+ mutable.HashMap(resourceAddresses.map(address => address ->
RESOURCE_TOTAL_AMOUNT): _*)
}
/**
- * Sequence of currently available resource addresses.
- *
- * With [[slotsPerAddress]] greater than 1, [[availableAddrs]] can contain
duplicate addresses
- * e.g. with [[slotsPerAddress]] == 2, availableAddrs for addresses 0 and 1
can look like
- * Seq("0", "0", "1"), where address 0 has two assignments available, and 1
has one.
+ * Sequence of currently available resource addresses which is not fully
assigned.
Review Comment:
Done
##########
core/src/main/scala/org/apache/spark/deploy/master/WorkerInfo.scala:
##########
@@ -28,12 +29,24 @@ private[spark] case class WorkerResourceInfo(name: String,
addresses: Seq[String
override protected def resourceName = this.name
override protected def resourceAddresses = this.addresses
- override protected def slotsPerAddress: Int = 1
+ /**
+ * Acquire the resources.
+ * @param amount How many addresses are requesting.
+ * @return ResourceInformation
+ */
def acquire(amount: Int): ResourceInformation = {
- val allocated = availableAddrs.take(amount)
- acquire(allocated)
- new ResourceInformation(resourceName, allocated.toArray)
+
+ var count = amount
+ val allocated: mutable.HashMap[String, Double] = mutable.HashMap.empty
+ for (address <- availableAddrs if count > 0) {
+ if (addressAvailabilityMap(address) == RESOURCE_TOTAL_AMOUNT) {
Review Comment:
Good suggestion. Done
##########
core/src/main/scala/org/apache/spark/resource/ResourceAllocator.scala:
##########
@@ -29,59 +65,45 @@ private[spark] trait ResourceAllocator {
protected def resourceName: String
protected def resourceAddresses: Seq[String]
- protected def slotsPerAddress: Int
/**
- * Map from an address to its availability, a value > 0 means the address is
available,
- * while value of 0 means the address is fully assigned.
- *
- * For task resources ([[org.apache.spark.scheduler.ExecutorResourceInfo]]),
this value
- * can be a multiple, such that each address can be allocated up to
[[slotsPerAddress]]
- * times.
+ * Map from an address to its availability default to RESOURCE_TOTAL_AMOUNT,
a value > 0 means
+ * the address is available, while value of 0 means the address is fully
assigned.
*/
- private lazy val addressAvailabilityMap = {
- mutable.HashMap(resourceAddresses.map(_ -> slotsPerAddress): _*)
+ protected lazy val addressAvailabilityMap = {
+ mutable.HashMap(resourceAddresses.map(address => address ->
RESOURCE_TOTAL_AMOUNT): _*)
}
/**
- * Sequence of currently available resource addresses.
- *
- * With [[slotsPerAddress]] greater than 1, [[availableAddrs]] can contain
duplicate addresses
- * e.g. with [[slotsPerAddress]] == 2, availableAddrs for addresses 0 and 1
can look like
- * Seq("0", "0", "1"), where address 0 has two assignments available, and 1
has one.
+ * Sequence of currently available resource addresses which is not fully
assigned.
*/
def availableAddrs: Seq[String] = addressAvailabilityMap
- .flatMap { case (addr, available) =>
- (0 until available).map(_ => addr)
- }.toSeq.sorted
+ .filter(addresses => addresses._2 > 0).keys.toSeq.sorted
Review Comment:
Done.
##########
core/src/main/scala/org/apache/spark/scheduler/ExecutorResourcesAmounts.scala:
##########
@@ -0,0 +1,199 @@
+/*
+ * 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.HashMap
+
+import org.apache.spark.SparkException
+import org.apache.spark.resource.{ResourceInformation, ResourceProfile}
+import org.apache.spark.resource.ResourceAmountUtils.RESOURCE_TOTAL_AMOUNT
+
+/**
+ * Class to hold information about a series of resources belonging to an
executor.
+ * A resource could be a GPU, FPGA, etc. And it is used as a temporary
+ * class to calculate the resources amounts when offering resources to
+ * the tasks in the [[TaskSchedulerImpl]]
+ *
+ * One example is GPUs, where the addresses would be the indices of the GPUs
+ *
+ * @param resources The executor available resources and amount. eg,
+ * Map("gpu" -> mutable.Map("0" -> 0.2, "1" -> 1.0),
+ * "fpga" -> mutable.Map("a" -> 0.3, "b" -> 0.9)
+ * )
+ */
+private[spark] class ExecutorResourcesAmounts(
+ private val resources: Map[String, Map[String, Double]]) extends
Serializable {
+
+ resources.foreach { case (_, addressMount) =>
Review Comment:
Good catch. Thx, I will remove it.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
