[SPARK-20801] Record accurate size of blocks in MapStatus when it's above threshold.

## What changes were proposed in this pull request?

Currently, when number of reduces is above 2000, HighlyCompressedMapStatus is used to store size of blocks. in HighlyCompressedMapStatus, only average size is stored for non empty blocks. Which is not good for memory control when we shuffle blocks. It makes sense to store the accurate size of block when it's above threshold.

## How was this patch tested?

Added test in MapStatusSuite.

Author: jinxing <[email protected]>

Closes apache#18031 from jinxing64/SPARK-20801.

Author: jinxing

core/src/main/scala/org/apache/spark/internal/config/package.scala

@@ -278,4 +278,13 @@ package object config {
         "spark.io.compression.codec.")
       .booleanConf
       .createWithDefault(false)
+
+  private[spark] val SHUFFLE_ACCURATE_BLOCK_THRESHOLD =
+    ConfigBuilder("spark.shuffle.accurateBlockThreshold")
+      .doc("When we compress the size of shuffle blocks in HighlyCompressedMapStatus, we will " +
+        "record the size accurately if it's above this config. This helps to prevent OOM by " +
+        "avoiding underestimating shuffle block size when fetch shuffle blocks.")
+      .bytesConf(ByteUnit.BYTE)
+      .createWithDefault(100 * 1024 * 1024)
+
 }

core/src/main/scala/org/apache/spark/scheduler/MapStatus.scala

@@ -19,8 +19,13 @@ package org.apache.spark.scheduler
 
 import java.io.{Externalizable, ObjectInput, ObjectOutput}
 
+import scala.collection.mutable
+import scala.collection.mutable.ArrayBuffer
+
 import org.roaringbitmap.RoaringBitmap
 
+import org.apache.spark.SparkEnv
+import org.apache.spark.internal.config
 import org.apache.spark.storage.BlockManagerId
 import org.apache.spark.util.Utils
 
@@ -121,48 +126,68 @@ private[spark] class CompressedMapStatus(
 }
 
 /**
- * A [[MapStatus]] implementation that only stores the average size of non-empty blocks,
+ * A [[MapStatus]] implementation that stores the accurate size of huge blocks, which are larger
+ * than spark.shuffle.accurateBlockThreshold. It stores the average size of other non-empty blocks,
  * plus a bitmap for tracking which blocks are empty.
  *
  * @param loc location where the task is being executed
  * @param numNonEmptyBlocks the number of non-empty blocks
  * @param emptyBlocks a bitmap tracking which blocks are empty
- * @param avgSize average size of the non-empty blocks
+ * @param avgSize average size of the non-empty and non-huge blocks
+ * @param hugeBlockSizes sizes of huge blocks by their reduceId.
  */
 private[spark] class HighlyCompressedMapStatus private (
     private[this] var loc: BlockManagerId,
     private[this] var numNonEmptyBlocks: Int,
     private[this] var emptyBlocks: RoaringBitmap,
-    private[this] var avgSize: Long)
+    private[this] var avgSize: Long,
+    @transient private var hugeBlockSizes: Map[Int, Byte])
   extends MapStatus with Externalizable {
 
   // loc could be null when the default constructor is called during deserialization
-  require(loc == null || avgSize > 0 || numNonEmptyBlocks == 0,
+  require(loc == null || avgSize > 0 || hugeBlockSizes.size > 0 || numNonEmptyBlocks == 0,
     "Average size can only be zero for map stages that produced no output")
 
-  protected def this() = this(null, -1, null, -1)  // For deserialization only
+  protected def this() = this(null, -1, null, -1, null)  // For deserialization only
 
   override def location: BlockManagerId = loc
 
   override def getSizeForBlock(reduceId: Int): Long = {
+    assert(hugeBlockSizes != null)
     if (emptyBlocks.contains(reduceId)) {
       0
     } else {
-      avgSize
+      hugeBlockSizes.get(reduceId) match {
+        case Some(size) => MapStatus.decompressSize(size)
+        case None => avgSize
+      }
     }
   }
 
   override def writeExternal(out: ObjectOutput): Unit = Utils.tryOrIOException {
     loc.writeExternal(out)
     emptyBlocks.writeExternal(out)
     out.writeLong(avgSize)
+    out.writeInt(hugeBlockSizes.size)
+    hugeBlockSizes.foreach { kv =>
+      out.writeInt(kv._1)
+      out.writeByte(kv._2)
+    }
   }
 
   override def readExternal(in: ObjectInput): Unit = Utils.tryOrIOException {
     loc = BlockManagerId(in)
     emptyBlocks = new RoaringBitmap()
     emptyBlocks.readExternal(in)
     avgSize = in.readLong()
+    val count = in.readInt()
+    val hugeBlockSizesArray = mutable.ArrayBuffer[Tuple2[Int, Byte]]()
+    (0 until count).foreach { _ =>
+      val block = in.readInt()
+      val size = in.readByte()
+      hugeBlockSizesArray += Tuple2(block, size)
+    }
+    hugeBlockSizes = hugeBlockSizesArray.toMap
   }
 }
 
@@ -178,11 +203,21 @@ private[spark] object HighlyCompressedMapStatus {
     // we expect that there will be far fewer of them, so we will perform fewer bitmap insertions.
     val emptyBlocks = new RoaringBitmap()
     val totalNumBlocks = uncompressedSizes.length
+    val threshold = Option(SparkEnv.get)
+      .map(_.conf.get(config.SHUFFLE_ACCURATE_BLOCK_THRESHOLD))
+      .getOrElse(config.SHUFFLE_ACCURATE_BLOCK_THRESHOLD.defaultValue.get)
+    val hugeBlockSizesArray = ArrayBuffer[Tuple2[Int, Byte]]()
     while (i < totalNumBlocks) {
-      var size = uncompressedSizes(i)
+      val size = uncompressedSizes(i)
       if (size > 0) {
         numNonEmptyBlocks += 1
-        totalSize += size
+        // Huge blocks are not included in the calculation for average size, thus size for smaller
+        // blocks is more accurate.
+        if (size < threshold) {
+          totalSize += size
+        } else {
+          hugeBlockSizesArray += Tuple2(i, MapStatus.compressSize(uncompressedSizes(i)))
+        }
       } else {
         emptyBlocks.add(i)
       }
@@ -195,6 +230,7 @@ private[spark] object HighlyCompressedMapStatus {
     }
     emptyBlocks.trim()
     emptyBlocks.runOptimize()
-    new HighlyCompressedMapStatus(loc, numNonEmptyBlocks, emptyBlocks, avgSize)
+    new HighlyCompressedMapStatus(loc, numNonEmptyBlocks, emptyBlocks, avgSize,
+      hugeBlockSizesArray.toMap)
   }
 }

core/src/test/scala/org/apache/spark/scheduler/MapStatusSuite.scala

@@ -17,11 +17,15 @@
 
 package org.apache.spark.scheduler
 
+import java.io.{ByteArrayInputStream, ByteArrayOutputStream, ObjectInputStream, ObjectOutputStream}
+
 import scala.util.Random
 
+import org.mockito.Mockito._
 import org.roaringbitmap.RoaringBitmap
 
-import org.apache.spark.{SparkConf, SparkFunSuite}
+import org.apache.spark.{SparkConf, SparkEnv, SparkFunSuite}
+import org.apache.spark.internal.config
 import org.apache.spark.serializer.JavaSerializer
 import org.apache.spark.storage.BlockManagerId
 
@@ -128,4 +132,26 @@ class MapStatusSuite extends SparkFunSuite {
     assert(size1 === size2)
     assert(!success)
   }
+
+  test("Blocks which are bigger than SHUFFLE_ACCURATE_BLOCK_THRESHOLD should not be " +
+    "underestimated.") {
+    val conf = new SparkConf().set(config.SHUFFLE_ACCURATE_BLOCK_THRESHOLD.key, "1000")
+    val env = mock(classOf[SparkEnv])
+    doReturn(conf).when(env).conf
+    SparkEnv.set(env)
+    // Value of element in sizes is equal to the corresponding index.
+    val sizes = (0L to 2000L).toArray
+    val status1 = MapStatus(BlockManagerId("exec-0", "host-0", 100), sizes)
+    val arrayStream = new ByteArrayOutputStream(102400)
+    val objectOutputStream = new ObjectOutputStream(arrayStream)
+    assert(status1.isInstanceOf[HighlyCompressedMapStatus])
+    objectOutputStream.writeObject(status1)
+    objectOutputStream.flush()
+    val array = arrayStream.toByteArray
+    val objectInput = new ObjectInputStream(new ByteArrayInputStream(array))
+    val status2 = objectInput.readObject().asInstanceOf[HighlyCompressedMapStatus]
+    (1001 to 2000).foreach {
+      case part => assert(status2.getSizeForBlock(part) >= sizes(part))
+    }
+  }
 }

docs/configuration.md

@@ -612,6 +612,15 @@ Apart from these, the following properties are also available, and may be useful
     <code>spark.io.compression.codec</code>.
   </td>
 </tr>
+<tr>
+  <td><code>spark.shuffle.accurateBlockThreshold</code></td>
+  <td>100 * 1024 * 1024</td>
+  <td>
+    When we compress the size of shuffle blocks in HighlyCompressedMapStatus, we will record the
+    size accurately if it's above this config. This helps to prevent OOM by avoiding
+    underestimating shuffle block size when fetch shuffle blocks.
+  </td>
+</tr>
 <tr>
   <td><code>spark.io.encryption.enabled</code></td>
   <td>false</td>