wypoon commented on a change in pull request #23767: [SPARK-26329][CORE] Faster polling of executor memory metrics. URL: https://github.com/apache/spark/pull/23767#discussion_r270586163
########## File path: core/src/main/scala/org/apache/spark/executor/ExecutorMetricsPoller.scala ########## @@ -0,0 +1,202 @@ +/* + * 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.executor + +import java.lang.Long.{MAX_VALUE => LONG_MAX_VALUE} +import java.util.concurrent.{ConcurrentHashMap, TimeUnit} +import java.util.concurrent.atomic.{AtomicLong, AtomicLongArray} + +import scala.collection.mutable.HashMap + +import org.apache.spark.annotation.DeveloperApi +import org.apache.spark.internal.Logging +import org.apache.spark.memory.MemoryManager +import org.apache.spark.metrics.ExecutorMetricType +import org.apache.spark.util.{ThreadUtils, Utils} + +/** + * :: DeveloperApi :: + * A class that polls executor metrics, and tracks their peaks per task and per stage. + * Each executor keeps an instance of this class. + * The poll method polls the executor metrics, and is either run in its own thread or + * called by the executor's heartbeater thread, depending on configuration. + * The class keeps two ConcurrentHashMaps that are accessed (via its methods) by the + * executor's task runner threads concurrently with the polling thread. One thread may + * update one of these maps while another reads it, so the reading thread may not get + * the latest metrics, but this is ok. + * + * @param memoryManager the memory manager used by the executor. + * @param pollingInterval the polling interval in milliseconds. + */ +@DeveloperApi +private[spark] class ExecutorMetricsPoller( Review comment: I have done testing on an actual cluster. I ran a few applications with and without faster polling (100ms interval). In both cases, I used 1 second for the heartbeat interval. In both cases, I did not enable proc fs metrics. There was no noticeable difference to the task times in the executors or to the stage times. However, the workload is small and the tasks took ~15s to ~1 minute. ---------------------------------------------------------------- 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]
