Victsm commented on a change in pull request #30062: URL: https://github.com/apache/spark/pull/30062#discussion_r513608609
########## File path: common/network-shuffle/src/main/java/org/apache/spark/network/shuffle/RemoteBlockPushResolver.java ########## @@ -0,0 +1,893 @@ +/* + * 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.network.shuffle; + +import java.io.BufferedOutputStream; +import java.io.DataOutputStream; +import java.io.File; +import java.io.FileOutputStream; +import java.io.IOException; +import java.nio.ByteBuffer; +import java.nio.channels.FileChannel; +import java.nio.file.Files; +import java.nio.file.Path; +import java.nio.file.Paths; +import java.util.Arrays; +import java.util.Iterator; +import java.util.LinkedList; +import java.util.List; +import java.util.Map; +import java.util.concurrent.ConcurrentMap; +import java.util.concurrent.ExecutionException; +import java.util.concurrent.Executor; +import java.util.concurrent.Executors; + +import com.google.common.base.Objects; +import com.google.common.base.Preconditions; +import com.google.common.cache.CacheBuilder; +import com.google.common.cache.CacheLoader; +import com.google.common.cache.LoadingCache; +import com.google.common.cache.Weigher; +import com.google.common.collect.Maps; +import com.google.common.primitives.Ints; +import com.google.common.primitives.Longs; +import io.netty.buffer.ByteBuf; +import io.netty.buffer.Unpooled; +import org.roaringbitmap.RoaringBitmap; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import org.apache.spark.network.buffer.FileSegmentManagedBuffer; +import org.apache.spark.network.buffer.ManagedBuffer; +import org.apache.spark.network.client.StreamCallbackWithID; +import org.apache.spark.network.protocol.Encoders; +import org.apache.spark.network.shuffle.protocol.FinalizeShuffleMerge; +import org.apache.spark.network.shuffle.protocol.MergeStatuses; +import org.apache.spark.network.shuffle.protocol.PushBlockStream; +import org.apache.spark.network.util.JavaUtils; +import org.apache.spark.network.util.NettyUtils; +import org.apache.spark.network.util.TransportConf; + +/** + * An implementation of {@link MergedShuffleFileManager} that provides the most essential shuffle + * service processing logic to support push based shuffle. + */ +public class RemoteBlockPushResolver implements MergedShuffleFileManager { + + private static final Logger logger = LoggerFactory.getLogger(RemoteBlockPushResolver.class); + private static final String MERGE_MANAGER_DIR = "merge_manager"; + + private final ConcurrentMap<String, AppPathsInfo> appsPathInfo; + private final ConcurrentMap<AppShufflePartitionId, AppShufflePartitionInfo> partitions; + + private final Executor directoryCleaner; + private final TransportConf conf; + private final int minChunkSize; + private final String relativeMergeDirPathPattern; + private final ErrorHandler.BlockPushErrorHandler errorHandler; + + @SuppressWarnings("UnstableApiUsage") + private final LoadingCache<File, ShuffleIndexInformation> indexCache; + + @SuppressWarnings("UnstableApiUsage") + public RemoteBlockPushResolver(TransportConf conf, String relativeMergeDirPathPattern) { + this.conf = conf; + this.partitions = Maps.newConcurrentMap(); + this.appsPathInfo = Maps.newConcurrentMap(); + this.directoryCleaner = Executors.newSingleThreadExecutor( + // Add `spark` prefix because it will run in NM in Yarn mode. + NettyUtils.createThreadFactory("spark-shuffle-merged-shuffle-directory-cleaner")); + this.minChunkSize = conf.minChunkSizeInMergedShuffleFile(); + CacheLoader<File, ShuffleIndexInformation> indexCacheLoader = + new CacheLoader<File, ShuffleIndexInformation>() { + public ShuffleIndexInformation load(File file) throws IOException { + return new ShuffleIndexInformation(file); + } + }; + indexCache = CacheBuilder.newBuilder() + .maximumWeight(conf.mergedIndexCacheSize()) + .weigher((Weigher<File, ShuffleIndexInformation>) (file, indexInfo) -> indexInfo.getSize()) + .build(indexCacheLoader); + this.relativeMergeDirPathPattern = relativeMergeDirPathPattern; + this.errorHandler = new ErrorHandler.BlockPushErrorHandler(); + } + + /** + * Given an ID that uniquely identifies a given shuffle partition of an application, retrieves the + * associated metadata. If not present and the corresponding merged shuffle does not exist, + * initializes the metadata. + */ + private AppShufflePartitionInfo getOrCreateAppShufflePartitionInfo(AppShufflePartitionId id) { + return partitions.computeIfAbsent(id, key -> { + // It only gets here when the key is not present in the map. This could either + // be the first time the merge manager receives a pushed block for a given application + // shuffle partition, or after the merged shuffle file is finalized. We handle these + // two cases accordingly by checking if the file already exists. + File dataFile = getMergedShuffleDataFile(id); + File indexFile = getMergedShuffleIndexFile(id); + File metaFile = getMergedShuffleMetaFile(id); + try { + if (dataFile.exists()) { + return null; Review comment: The situations in which this method returns null is explained inside `receiveBlockDataAsStream`. There are 2 scenarios: the block is received after merge finalization, or the block is a duplicate of one that's already merged. Here, we are specifically checking for the first condition. Even if we are returning null here, we cannot delete the data file, since it might be reused later when stage retries. ---------------------------------------------------------------- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
