spark-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From aarondav <...@git.apache.org>
Subject [GitHub] spark pull request: [SPARK-3796] Create external service which can...
Date Fri, 31 Oct 2014 01:48:58 GMT
Github user aarondav commented on a diff in the pull request:

    https://github.com/apache/spark/pull/3001#discussion_r19647880
  
    --- Diff: network/shuffle/src/main/java/org/apache/spark/network/shuffle/StandaloneShuffleBlockManager.java
---
    @@ -0,0 +1,151 @@
    +/*
    + * 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.DataInputStream;
    +import java.io.File;
    +import java.io.FileInputStream;
    +import java.io.IOException;
    +import java.util.concurrent.ConcurrentHashMap;
    +
    +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.util.JavaUtils;
    +
    +/**
    + * Manages converting shuffle BlockIds into physical segments of local files. Each Executor
must
    + * register its own configuration about where it stores its files (local dirs) and how
(shuffle
    + * manager). The logic for retrieval of individual files is replicated from Spark's
    + * FileShuffleBlockManager and IndexShuffleBlockManager.
    + *
    + * Executors with shuffle file consolidation are not currently supported, as the index
is stored in
    + * the Executor's memory, unlike the IndexShuffleBlockManager.
    + */
    +public class StandaloneShuffleBlockManager {
    +  private final Logger logger = LoggerFactory.getLogger(StandaloneShuffleBlockManager.class);
    +
    +  // Map from "appId-execId" to the executor's configuration.
    +  private final ConcurrentHashMap<String, ExecutorShuffleConfig> executors =
    +    new ConcurrentHashMap<String, ExecutorShuffleConfig>();
    +
    +  // Returns an id suitable for a single executor within a single application.
    +  private String getAppExecId(String appId, String execId) {
    +    return appId + "-" + execId;
    +  }
    +
    +  /** Registers a new Executor with all the configuration we need to find its shuffle
files. */
    +  public void registerExecutor(
    +      String appId,
    +      String execId,
    +      ExecutorShuffleConfig executorConfig) {
    +    String fullId = getAppExecId(appId, execId);
    +    executors.put(fullId, executorConfig);
    +  }
    +
    +  /**
    +   * Obtains a FileSegmentManagedBuffer from a shuffle block id. We expect the blockId
has the
    +   * format "shuffle_ShuffleId_MapId_ReduceId", and additionally make assumptions about
how the
    +   * hash and sort based shuffles store their data.
    +   */
    +  public ManagedBuffer getBlockData(String appId, String execId, String blockId) {
    +    String[] blockIdParts = blockId.split("_");
    +    if (blockIdParts.length < 4) {
    +      throw new IllegalArgumentException("Unexpected block id format: " + blockId);
    +    } else if (!blockIdParts[0].equals("shuffle")) {
    +      throw new IllegalArgumentException("Expected shuffle block id, got: " + blockId);
    +    }
    +    int shuffleId = Integer.parseInt(blockIdParts[1]);
    +    int mapId = Integer.parseInt(blockIdParts[2]);
    +    int reduceId = Integer.parseInt(blockIdParts[3]);
    +
    +    ExecutorShuffleConfig executor = executors.get(getAppExecId(appId, execId));
    +    if (executor == null) {
    +      throw new RuntimeException(
    +        String.format("Executor is not registered (appId=%s, execId=%s)", appId, execId));
    +    }
    +
    +    if ("org.apache.spark.shuffle.hash.HashShuffleManager".equals(executor.shuffleManager))
{
    +      return getHashBasedShuffleBlockData(executor, blockId);
    +    } else if ("org.apache.spark.shuffle.sort.SortShuffleManager".equals(executor.shuffleManager))
{
    +      return getSortBasedShuffleBlockData(executor, shuffleId, mapId, reduceId);
    +    } else {
    +      throw new UnsupportedOperationException(
    +        "Unsupported shuffle manager: " + executor.shuffleManager);
    +    }
    +  }
    +
    +  /**
    +   * Hash-based shuffle data is simply stored as one file per block.
    +   * This logic is from FileShuffleBlockManager.
    +   */
    +  // TODO: Support consolidated hash shuffle files
    --- End diff --
    
    The exception wouldn't be great, it would say it couldn't find the shuffle block. I will
add a check in BlockManager that disallows the co-occurrence of externalShuffleService and
consolidateFiles (and not Sort-based shuffle).


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org


Mime
View raw message