tgravescs commented on a change in pull request #27636: [SPARK-30873][CORE][YARN]Handling Node Decommissioning for Yarn cluster manger in Spark URL: https://github.com/apache/spark/pull/27636#discussion_r393007435
########## File path: core/src/main/scala/org/apache/spark/scheduler/DecommissionTracker.scala ########## @@ -0,0 +1,405 @@ +/* + * 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 java.text.SimpleDateFormat +import java.util.concurrent.TimeUnit +import java.util.concurrent.atomic.{AtomicBoolean, AtomicInteger} + +import scala.collection.mutable.HashMap + +import org.apache.spark.{ExecutorAllocationClient, SparkConf, SparkContext} +import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.internal.Logging +import org.apache.spark.internal.config +import org.apache.spark.util.{Clock, SystemClock, ThreadUtils, Utils} + +/** + * DecommissionTracker tracks the list of decommissioned nodes. + * This decommission trackers maintains the decommissioned nodes state. + * Decommission tracker schedules the executor decommission and shuffle + * decommission for that node. + */ +private[scheduler] class DecommissionTracker ( + conf: SparkConf, + executorAllocClient: Option[ExecutorAllocationClient], + dagScheduler: Option[DAGScheduler], + clock: Clock = new SystemClock()) extends Logging { + + def this(sc: SparkContext, + client: Option[ExecutorAllocationClient], + dagScheduler: Option[DAGScheduler]) = { + this(sc.conf, client, dagScheduler) + } + + // Decommission thread of node decommissioning!! + private val decommissionThread = + ThreadUtils.newDaemonThreadPoolScheduledExecutor("node-decommissioning-thread", 20) Review comment: 20 threads to just handle node decommissions seems like a lot. How long does it normally take for a decommission? On Yarn I realize it has to send message to AM which could then be blocked on doing allocate call or something, so I'm assuming that is where possible delay is. But have you actually run this and decommissioned multiple nodes, how long does it take? You are also calling executedecomission separate from the shuffle decommission and then have delays to have them happen not at the same time. Couldn't they be done in sequence and then you know they aren't done at same time? what was the default yarn timeout on newer versions? I know you said 30 seconds on gap. ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
