Github user vanzin commented on a diff in the pull request: https://github.com/apache/spark/pull/19041#discussion_r143801968 --- Diff: core/src/main/scala/org/apache/spark/CacheRecoveryManager.scala --- @@ -0,0 +1,250 @@ +/* + * 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 + +import java.util.concurrent.{ScheduledFuture, TimeUnit} + +import scala.collection.mutable +import scala.concurrent.{ExecutionContext, Future} +import scala.util.Failure + +import org.apache.spark.internal.Logging +import org.apache.spark.internal.config.DYN_ALLOCATION_RECOVER_CACHE_TIMEOUT +import org.apache.spark.rpc.RpcEndpointRef +import org.apache.spark.storage.{BlockId, BlockManagerId, RDDBlockId} +import org.apache.spark.storage.BlockManagerMessages.{GetCachedBlocks, GetMemoryStatus, GetSizeOfBlocks, ReplicateOneBlock} +import org.apache.spark.util.ThreadUtils + +/** + * Responsible for asynchronously replicating all of an executors cached blocks, and then shutting + * it down. + */ +final private class CacheRecoveryManager( + state: CacheRecoveryManagerState, + conf: SparkConf) + extends Logging { + + private val threadPool = ThreadUtils.newDaemonCachedThreadPool("recover-cache-shutdown-pool") + private implicit val asyncExecutionContext: ExecutionContext = + ExecutionContext.fromExecutorService(threadPool) + + /** + * Start the recover cache shutdown process for these executors + * + * @param execIds the executors to start shutting down + */ + def startExecutorKill(execIds: Seq[String]): Unit = { + logDebug(s"Recover cached data before shutting down executors ${execIds.mkString(", ")}.") + checkForReplicableBlocks(execIds) + } + + /** + * Stops all thread pools + * + * @return + */ + def stop(): java.util.List[Runnable] = { + threadPool.shutdownNow() + state.stop() + } + + /** + * Get list of cached blocks from BlockManagerMaster. If there are cached blocks, replicate them, + * otherwise kill the executors + * + * @param execIds the executors to check + */ + private def checkForReplicableBlocks(execIds: Seq[String]) = state.getBlocks(execIds).foreach { + case (executorId, HasCachedBlocks) => replicateBlocks(executorId) + case (executorId, NoMoreBlocks | NotEnoughMemory) => state.killExecutor(executorId) + } + + /** + * Replicate one cached block on an executor. If there are more, repeat. If there are none, check + * with the block manager master again. If there is an error, go ahead and kill executor. + * + * @param execId the executor to save a block one + */ + private def replicateBlocks(execId: String): Unit = { + import scala.util.Success + val (response, blockId) = state.replicateFirstBlock(execId) + response.onComplete { + case Success(true) => + logTrace(s"Finished replicating block ${blockId.getOrElse("unknown")} on exec $execId.") --- End diff -- s/exec/executor. That is if you even want to keep this log around...
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org