Hadoop 3.1.0 ErasureCodingPolicy导致spark streaming的任务失败问题分析

一、ErasureCodePolicy问题，导致streaming任务退出

1、任务失败的原因，当执行block时有uncaught 异常时，stop sparkcontext，如下

具体源码错误路径，感兴趣的，可以根据错误日志跟踪一下，这里就不具体跟踪了，只显示比较重要的源码信息
AsyncEventQueue

  private val dispatchThread = new Thread(s"spark-listener-group-$name") {
    setDaemon(true)
    override def run(): Unit = Utils.tryOrStopSparkContext(sc) {
      dispatch()
    }
  }

  private def dispatch(): Unit = LiveListenerBus.withinListenerThread.withValue(true) {
    try {
      var next: SparkListenerEvent = eventQueue.take()
      while (next != POISON_PILL) {
        val ctx = processingTime.time()
        try {
          super.postToAll(next)
        } finally {
          ctx.stop()
        }
        eventCount.decrementAndGet()
        next = eventQueue.take()
      }
      eventCount.decrementAndGet()
    } catch {
      case ie: InterruptedException =>
        logInfo(s"Stopping listener queue $name.", ie)
    }
  }

org.apache.spark.util.Utils

  /**
   * Execute a block of code that evaluates to Unit, stop SparkContext if there is any uncaught
   * exception
   *
   * NOTE: This method is to be called by the driver-side components to avoid stopping the
   * user-started JVM process completely; in contrast, tryOrExit is to be called in the
   * spark-started JVM process .
   */
  def tryOrStopSparkContext(sc: SparkContext)(block: => Unit) {
    try {
      block
    } catch {
      case e: ControlThrowable => throw e
      case t: Throwable =>
        val currentThreadName = Thread.currentThread().getName
        if (sc != null) {
          logError(s"uncaught error in thread $currentThreadName, stopping SparkContext", t)
          sc.stopInNewThread()
        }
        if (!NonFatal(t)) {
          logError(s"throw uncaught fatal error in thread $currentThreadName", t)
          throw t
        }
    }
  }

2、首先是个告警信息，如下

2018-11-30 16:35:53 WARN  Client:66 - Neither spark.yarn.jars nor spark.yarn.archive is set, falling back to uploading libraries under SPARK_HOME.
2018-11-30 16:35:54 WARN  ErasureCodeNative:55 - ISA-L support is not available in your platform... using builtin-java codec where applicable

3、当spark streaming任务跑完一批后，就开始报错，一直循环同样的报错，跑不了多久任务就挂掉了，报错信息如下

2018-11-30 16:37:21 ERROR AsyncEventQueue:91 - Listener EventLoggingListener threw an exception
java.lang.IllegalStateException
	at com.google.common.base.Preconditions.checkState(Preconditions.java:133)
	at org.apache.hadoop.hdfs.DFSStripedOutputStream$CellBuffers.addTo(DFSStripedOutputStream.java:238)
	at org.apache.hadoop.hdfs.DFSStripedOutputStream$CellBuffers.access$700(DFSStripedOutputStream.java:203)
	at org.apache.hadoop.hdfs.DFSStripedOutputStream.writeChunk(DFSStripedOutputStream.java:520)
	at org.apache.hadoop.fs.FSOutputSummer.writeChecksumChunks(FSOutputSummer.java:217)
	at org.apache.hadoop.fs.FSOutputSummer.flushBuffer(FSOutputSummer.java:164)
	at org.apache.hadoop.fs.FSOutputSummer.flushBuffer(FSOutputSummer.java:145)
	at org.apache.hadoop.fs.FSOutputSummer.write1(FSOutputSummer.java:136)
	at org.apache.hadoop.fs.FSOutputSummer.write(FSOutputSummer.java:111)
	at org.apache.hadoop.fs.FSDataOutputStream$PositionCache.write(FSDataOutputStream.java:57)
	at java.io.DataOutputStream.write(DataOutputStream.java:107)
	at java.io.BufferedOutputStream.write(BufferedOutputStream.java:122)
	at sun.nio.cs.StreamEncoder.writeBytes(StreamEncoder.java:221)
	at sun.nio.cs.StreamEncoder.implWrite(StreamEncoder.java:282)
	at sun.nio.cs.StreamEncoder.write(StreamEncoder.java:125)
	at java.io.OutputStreamWriter.write(OutputStreamWriter.java:207)
	at java.io.BufferedWriter.flushBuffer(BufferedWriter.java:129)
	at java.io.BufferedWriter.write(BufferedWriter.java:230)
	at java.io.PrintWriter.write(PrintWriter.java:456)
	at java.io.PrintWriter.write(PrintWriter.java:473)
	at java.io.PrintWriter.print(PrintWriter.java:603)
	at java.io.PrintWriter.println(PrintWriter.java:739)
	at org.apache.spark.scheduler.EventLoggingListener$$anonfun$logEvent$1.apply(EventLoggingListener.scala:143)
	at org.apache.spark.scheduler.EventLoggingListener$$anonfun$logEvent$1.apply(EventLoggingListener.scala:143)
	at scala.Option.foreach(Option.scala:257)
	at org.apache.spark.scheduler.EventLoggingListener.logEvent(EventLoggingListener.scala:143)
	at org.apache.spark.scheduler.EventLoggingListener.onTaskEnd(EventLoggingListener.scala:164)
	at org.apache.spark.scheduler.SparkListenerBus$class.doPostEvent(SparkListenerBus.scala:45)
	at org.apache.spark.scheduler.AsyncEventQueue.doPostEvent(AsyncEventQueue.scala:37)
	at org.apache.spark.scheduler.AsyncEventQueue.doPostEvent(AsyncEventQueue.scala:37)
	at org.apache.spark.util.ListenerBus$class.postToAll(ListenerBus.scala:82)
	at org.apache.spark.scheduler.AsyncEventQueue.org$apache$spark$scheduler$AsyncEventQueue$$super$postToAll(AsyncEventQueue.scala:89)
	at org.apache.spark.scheduler.AsyncEventQueue$$anonfun$org$apache$spark$scheduler$AsyncEventQueue$$dispatch$1.apply(AsyncEventQueue.scala:89)
	at scala.util.DynamicVariable.withValue(DynamicVariable.scala:58)
	at org.apache.spark.scheduler.AsyncEventQueue.org$apache$spark$scheduler$AsyncEventQueue$$dispatch(AsyncEventQueue.scala:83)
	at org.apache.spark.scheduler.AsyncEventQueue$$anon$1$$anonfun$run$1.apply$mcV$sp(AsyncEventQueue.scala:79)
	at org.apache.spark.util.Utils$.tryOrStopSparkContext(Utils.scala:1319)
	at org.apache.spark.scheduler.AsyncEventQueue$$anon$1.run(AsyncEventQueue.scala:78)

二、问题分析

报错的源码大致意思是

1、在 onTaskEnd 的时候会调用 PrintWriter 输出 even log 信息

2、而这个 PrintWriter 的write 最终由 DFSStripedOutputStream 的 writeChunk去写

3、写之前有个 checkState 就是在这里报的错

4、相关参数有个 ErasureCodingPolicy 的一个 cellSize

5、当输出的 position 大于 cellsize 时就抛这个异常了

6、而且 spark.util.Utils 的 tryOrStopSparkContext 发现执行 block 有 uncaught 异常，就stop sparkContext

7、从这来看任务在新集群没跑一会儿就挂了很大原因就是这个引起的

8、EC分析：日志输出时，会将其分块，然后写入EC的缓冲池中，当缓冲池满后，flush将其清空，导致异常当两个很重要当原因：A、缓冲块分的大 B、缓冲池比较小，由于这两个原因导致异常发生

三、问题解决

将日志输出改为本地，不写HDFS，或关闭日志输出的EC

四、代码分析

报错代码

类：DFSStripedOutputStream

private int addTo(int i, byte[] b, int off, int len) {
    ByteBuffer buf = this.buffers[i];
    int pos = buf.position() + len;
    Preconditions.checkState(pos <= DFSStripedOutputStream.this.cellSize);
    buf.put(b, off, len);
    return pos;
}

public ErasureCodingPolicy(String name, ECSchema schema, int cellSize, byte id) {
    Preconditions.checkNotNull(name);
    Preconditions.checkNotNull(schema);
    Preconditions.checkArgument(cellSize > 0, "cellSize must be positive");
    Preconditions.checkArgument(cellSize % 1024 == 0, "cellSize must be 1024 aligned");
    this.name = name;
    this.schema = schema;
    this.cellSize = cellSize;
    this.id = id;
}

有关EC导致的直接问题，目前还没真正解决

Hadoop 3.1.0 ErasureCodingPolicy导致spark streaming的任务失败问题分析

猜你喜欢