apache-spark - 如何解决向 Hive 表发送大文件时的连接问题？

Question

我在 hdfs 上有文件，我想将其放入 Hive 表中。该操作由 Java 应用程序中的 Spark 批处理执行。 执行该任务的代码如下:

[...]

final Dataset<File> fileDs = rawDs.map(record -> {
        return FileService.map(record.getList(2));
      }, Encoders.bean(File.class));

final Dataset<Row> fileDsWithId = fileDs.withColumn("id", functions.lit(id));
fileDsWithId.repartition(fileDsWithId.col("id")); 

fileWithId.write().mode(SaveMode.Append)
  .format("orc")
  .partitionBy("id")
  .option("path", hdfs://..../mydatabase.db/mytable")
  .saveAsTable("mydatabase.mytable");

当我使用小文件(1 或 2 行数据)时，应用程序运行正常，作业在 30 秒内成功结束。该表是在 Hive 中创建的，我可以使用 Select * 查询显示数据。当表已经存在时它也可以工作。数据只需添加到现有数据中即可。 hive 中的生成表结构看起来不错。它与我的数据相符。

但是当我尝试处理更大的文件(3.7Mo，大约 1000 行数据)时，作业在 15 分钟后失败。相应的orc文件在hdfs中创建，但它是空的，Hive不知道它。

日志文件显示了几个如下错误:

2019-05-31 14:20:07,500 - [ERROR] [                           dispatcher-event-loop-3] pache.spark.scheduler.cluster.YarnClusterScheduler - [{}] - Lost executor 31 on XXXXXX: Container marked as failed: container_e71_1559121287708_0019_02_000032 on host: XXXXXXXXX. Exit status: 143. Diagnostics: Container killed on request. Exit code is 143
Container exited with a non-zero exit code 143.
Killed by external signal
[...]
java.lang.RuntimeException: java.io.IOException: Connection reset by peer
        at org.spark_project.guava.base.Throwables.propagate(Throwables.java:160)
        at org.apache.spark.network.client.TransportClient.sendRpcSync(TransportClient.java:273)
        at org.apache.spark.network.crypto.AuthClientBootstrap.doSparkAuth(AuthClientBootstrap.java:105)
        at org.apache.spark.network.crypto.AuthClientBootstrap.doBootstrap(AuthClientBootstrap.java:79)
...
        at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:338)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
        at java.lang.Thread.run(Thread.java:748)
Caused by: java.io.IOException: Connection reset by peer
        at sun.nio.ch.FileDispatcherImpl.read0(Native Method)
        at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:39)
        at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:223)
        at sun.nio.ch.IOUtil.read(IOUtil.java:192)
...
        at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:442)
        at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:131)
        at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:144)
        ... 1 more
[...]       
        2019-05-31 14:20:17,898 - [ERROR] [                                shuffle-client-4-1]    org.apache.spark.network.client.TransportClient - [{}] - Failed to send RPC 9035939448873337359 to XXXXXXXX: java.nio.channels.ClosedChannelExceptionsg
java.nio.channels.ClosedChannelException
        at io.netty.channel.AbstractChannel$AbstractUnsafe.write(...)(Unknown Source)
2019-05-31 14:20:17,899 - [ERROR] [          Executor task launch worker for task 244] apache.spark.network.client.TransportClientFactory - [{}] - Exception while bootstrapping client after 5999 mssg
java.lang.RuntimeException: java.io.IOException: Failed to send RPC 9035939448873337359 to XXXXXXXXX: java.nio.channels.ClosedChannelException
        at org.spark_project.guava.base.Throwables.propagate(Throwables.java:160)
        at org.apache.spark.network.client.TransportClient.sendRpcSync(TransportClient.java:273)
        at org.apache.spark.network.sasl.SaslClientBootstrap.doBootstrap(SaslClientBootstrap.java:70)
        at org.apache.spark.network.crypto.AuthClientBootstrap.doSaslAuth(AuthClientBootstrap.java:115)
     ...
        at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:338)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
        at java.lang.Thread.run(Thread.java:748)
Caused by: java.io.IOException: Failed to send RPC 9035939448873337359 to XXXXXXXXXXXX: java.nio.channels.ClosedChannelException
        at org.apache.spark.network.client.TransportClient.lambda$sendRpc$2(TransportClient.java:237)
        at io.netty.util.concurrent.DefaultPromise.notifyListener0(DefaultPromise.java:507)
        at io.netty.util.concurrent.DefaultPromise.notifyListenersNow(DefaultPromise.java:481)
        at io.netty.util.concurrent.DefaultPromise.notifyListeners(DefaultPromise.java:420)
        at io.netty.util.concurrent.DefaultPromise.tryFailure(DefaultPromise.java:122)
        at io.netty.channel.AbstractChannel$AbstractUnsafe.safeSetFailure(AbstractChannel.java:852)
at io.netty.channel.AbstractChannel$AbstractUnsafe.write(AbstractChannel.java:738)
        at io.netty.channel.DefaultChannelPipeline$HeadContext.write(DefaultChannelPipeline.java:1251)
        at io.netty.channel.AbstractChannelHandlerContext.invokeWrite0(AbstractChannelHandlerContext.java:733)
        at io.netty.channel.AbstractChannelHandlerContext.invokeWrite(AbstractChannelHandlerContext.java:725)
        at io.netty.channel.AbstractChannelHandlerContext.access$1900(AbstractChannelHandlerContext.java:35)
...
        at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:446)
        at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:131)
        at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:144)
        ... 1 more
[...]       
2019-05-31 14:20:22,907 - [INFO ] [                             Block Fetch Retry-6-1] .apache.spark.network.shuffle.RetryingBlockFetcher - [{}] - Retrying fetch (2/3) for 1 outstanding blocks after 5000 mssg
2019-05-31 14:20:27,909 - [ERROR] [                             Block Fetch Retry-6-2] .apache.spark.network.shuffle.RetryingBlockFetcher - [{}] - Exception while beginning fetch of 1 outstanding blocks (after 2 retries)sg
java.io.IOException: Failed to connect to XXXXXXXXX
        at org.apache.spark.network.client.TransportClientFactory.createClient(TransportClientFactory.java:232)
        at org.apache.spark.network.client.TransportClientFactory.createClient(TransportClientFactory.java:182)
...
        at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:144)
        at java.lang.Thread.run(Thread.java:748)
Caused by: io.netty.channel.AbstractChannel$AnnotatedConnectException: Connexion refused: XXXXXXXX
        at sun.nio.ch.SocketChannelImpl.checkConnect(Native Method)
        at sun.nio.ch.SocketChannelImpl.finishConnect(SocketChannelImpl.java:717)
        at io.netty.channel.socket.nio.NioSocketChannel.doFinishConnect(NioSocketChannel.java:257)
...
        at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:131)
        ... 2 more
[...]       
        2019-05-31 14:20:32,915 - [WARN ] [          Executor task launch worker for task 244]              org.apache.spark.storage.BlockManager - [{}] - Failed to fetch remote block broadcast_2_piece0 from BlockManagerId(1, XXXXXXX, 44787, None) (failed attempt 1)sg
org.apache.spark.SparkException: Exception thrown in awaitResult:
        at org.apache.spark.util.ThreadUtils$.awaitResult(ThreadUtils.scala:205)
        at org.apache.spark.network.BlockTransferService.fetchBlockSync(BlockTransferService.scala:105)
        at org.apache.spark.storage.BlockManager.getRemoteBytes(BlockManager.scala:642)
...
        at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:338)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
        at java.lang.Thread.run(Thread.java:748)
Caused by: java.io.IOException: Failed to connect to XXXXXXXXX
        at org.apache.spark.network.client.TransportClientFactory.createClient(TransportClientFactory.java:232)
        at org.apache.spark.network.client.TransportClientFactory.createClient(TransportClientFactory.java:182)
...
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
        at io.netty.util.concurrent.DefaultThreadFactory$DefaultRunnableDecorator.run(DefaultThreadFactory.java:144)
        ... 1 more
Caused by: io.netty.channel.AbstractChannel$AnnotatedConnectException: Connexion refused: XXXXXXXX
        at sun.nio.ch.SocketChannelImpl.checkConnect(Native Method)
        at sun.nio.ch.SocketChannelImpl.finishConnect(SocketChannelImpl.java:717)
...
        at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:131)
        ... 2 more

我不明白那里发生了什么。我已经检查过内存问题，但看起来没问题。这些机器用于处理更大的文件(通常是几十GB)。 为什么连接丢失/被拒绝/重置？ Spark 事先创建表方案是否存在任何问题可以解释这一点？

<小时/>

在 Ram Ghadiyaram 的回答后更新:
我尝试将 spark.network.timeout 设置为 6000 秒。环境中没有配置其他超时设置。 结果似乎是一样的。 10 分钟后作业失败，并在日志文件中显示相同的错误:“连接被对等方重置”、“无法发送 RPC”等

设置spark.core.connection.ack.wait.timeout、spark.storage.blockManagerSlaveTimeoutMs、spark.shuffle.io.connectionTimeout 、 spark.rpc.askTimeout 和 spark.rpc.lookupTimeout 设置为相同的值(6000s)似乎也不起作用。

我认为我的数据集太困惑，无论如何都无法正确处理。我将尝试更改数据模型，然后使用这些超时设置再次运行应用程序。

<小时/>

更新于 2019 年 1 月 7 日:
我简化了数据模型。该模型很复杂，导致数据集中出现一些空结构，因为系统无法通过继承链接某些字段。我已经扁平化了结构，以便每种可能的类型都作为泛型类的实际属性出现，因此我删除了继承。

总结一下，类似这样的:
文件.class
|-字段1
|-field2
|-field3
|- 通用类
|-Class1
|-2级
|-Class3

我没有使用带有一些子类的抽象类，而是使用其他类作为属性创建了一个通用类。 这很脏(我不建议这样做)，但这样数据集就干净多了。

执行此更改后，我不再遇到超时问题。我猜之前的模型太乱了，Spark 无法高效编写。

我尝试过以 ORC 和 Avro 格式编写，两者都很好。 在 Avro 中，我设法在一分钟内写入大约 300000 行，因此默认超时设置不再是问题。

Answer 1

Q:Why are the connection lost/denied/reset ?

org.apache.spark.SparkException: Exception thrown in awaitResult:
        at org.apache.spark.util.ThreadUtils$.awaitResult(ThreadUtils.scala:205)

似乎是大型或繁重工作负载的典型超时问题。

我不知道你用的是哪个版本的spark。但它在这里失败了。 基本上它的作用是等待一段时间(默认超时)然后失败。请参阅ThreadUtils

/**
   * Preferred alternative to `Await.result()`.
   *
   * This method wraps and re-throws any exceptions thrown by the underlying `Await` call, ensuring
   * that this thread's stack trace appears in logs.
   *
   * In addition, it calls `Awaitable.result` directly to avoid using `ForkJoinPool`'s
   * `BlockingContext`. Codes running in the user's thread may be in a thread of Scala ForkJoinPool.
   * As concurrent executions in ForkJoinPool may see some [[ThreadLocal]] value unexpectedly, this
   * method basically prevents ForkJoinPool from running other tasks in the current waiting thread.
   * In general, we should use this method because many places in Spark use [[ThreadLocal]] and it's
   * hard to debug when [[ThreadLocal]]s leak to other tasks.
   */
  @throws(classOf[SparkException])
  def awaitResult[T](awaitable: Awaitable[T], atMost: Duration): T = {
    try {
      // `awaitPermission` is not actually used anywhere so it's safe to pass in null here.
      // See SPARK-13747.
      val awaitPermission = null.asInstanceOf[scala.concurrent.CanAwait]
      awaitable.result(atMost)(awaitPermission)
    } catch {
      case e: SparkFatalException =>
        throw e.throwable
      // TimeoutException is thrown in the current thread, so not need to warp the exception.
      case NonFatal(t) if !t.isInstanceOf[TimeoutException] =>
        throw new SparkException("Exception thrown in awaitResult: ", t)
    }
  }

您必须增加超时，请参阅 networking docs

spark.network.timeout 120s Default timeout for all network interactions. This config will be used in place of spark.core.connection.ack.wait.timeout, spark.storage.blockManagerSlaveTimeoutMs, spark.shuffle.io.connectionTimeout, spark.rpc.askTimeout or spark.rpc.lookupTimeout if they are not configured.

<小时/>

To sumup : For small workloads timeout is enough for large work loads time out needs to be increasaed.