Github user pwendell commented on a diff in the pull request:
https://github.com/apache/spark/pull/3026#discussion_r19652761
--- Diff:
streaming/src/main/scala/org/apache/spark/streaming/scheduler/ReceivedBlockTracker.scala
---
@@ -0,0 +1,207 @@
+/*
+ * 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.streaming.scheduler
+
+import java.nio.ByteBuffer
+
+import scala.collection.mutable
+import scala.language.implicitConversions
+
+import org.apache.hadoop.conf.Configuration
+import org.apache.hadoop.fs.Path
+
+import org.apache.spark.{Logging, SparkConf}
+import org.apache.spark.storage.StreamBlockId
+import org.apache.spark.streaming.Time
+import org.apache.spark.streaming.util.{Clock, WriteAheadLogManager}
+import org.apache.spark.util.Utils
+
+/** Trait representing any action done in the ReceivedBlockTracker */
+private[streaming] sealed trait ReceivedBlockTrackerAction
+
+private[streaming] case class BlockAddition(receivedBlockInfo:
ReceivedBlockInfo)
+ extends ReceivedBlockTrackerAction
+private[streaming] case class BatchAllocations(time: Time,
allocatedBlocks: AllocatedBlocks)
+ extends ReceivedBlockTrackerAction
+private[streaming] case class BatchCleanup(times: Seq[Time])
+ extends ReceivedBlockTrackerAction
+
+
+/** Class representing the blocks of all the streams allocated to a batch
*/
+case class AllocatedBlocks(streamIdToAllocatedBlocks: Map[Int,
Seq[ReceivedBlockInfo]]) {
+ def apply(streamId: Int) = streamIdToAllocatedBlocks(streamId)
+}
+
+/**
+ * Class that keep track of all the received blocks, and allocate them to
batches
+ * when required. All actions taken by this class can be saved to a write
ahead log,
+ * so that the state of the tracker (received blocks and block-to-batch
allocations)
+ * can be recovered after driver failure.
+ */
+private[streaming]
+class ReceivedBlockTracker(
+ conf: SparkConf, hadoopConf: Configuration, streamIds: Seq[Int],
clock: Clock,
+ checkpointDirOption: Option[String]) extends Logging {
+
+ private type ReceivedBlockQueue = mutable.Queue[ReceivedBlockInfo]
+
+ private val streamIdToUnallocatedBlockInfo = new mutable.HashMap[Int,
ReceivedBlockQueue]
+ private val timeToAllocatedBlockInfo = new mutable.HashMap[Time,
AllocatedBlocks]
+
+ private val logManagerRollingIntervalSecs = conf.getInt(
+
"spark.streaming.receivedBlockTracker.writeAheadLog.rotationIntervalSecs", 60)
+ private val logManagerOption = checkpointDirOption.map { checkpointDir =>
+ new WriteAheadLogManager(
+ ReceivedBlockTracker.checkpointDirToLogDir(checkpointDir),
+ hadoopConf,
+ rollingIntervalSecs = logManagerRollingIntervalSecs,
+ callerName = "ReceivedBlockHandlerMaster",
+ clock = clock
+ )
+ }
+
+ // Recover block information from write ahead logs
+ recoverFromWriteAheadLogs()
+
+ /** Add received block */
+ def addBlock(receivedBlockInfo: ReceivedBlockInfo): Boolean =
synchronized {
+ try {
+ writeToLog(BlockAddition(receivedBlockInfo))
+ getReceivedBlockQueue(receivedBlockInfo.streamId) +=
receivedBlockInfo
+ logDebug(s"Stream ${receivedBlockInfo.streamId} received " +
+ s"block ${receivedBlockInfo.blockStoreResult.blockId}")
+ true
+ } catch {
+ case e: Exception =>
+ logError("Error adding block " + receivedBlockInfo, e)
+ false
+ }
+ }
+
+ /** Get blocks that have been added but not yet allocated to any batch */
+ def getUnallocatedBlocks(streamId: Int): Seq[ReceivedBlockInfo] =
synchronized {
+ getReceivedBlockQueue(streamId).toSeq
+ }
+
+ /** Get the blocks allocated to a batch, or allocate blocks to the batch
and then get them */
+ def getOrAllocateBlocksToBatch(batchTime: Time, streamId: Int):
Seq[ReceivedBlockInfo] = {
--- End diff --
Would it be possible to write this in a different way? It's hard to reason
about interactions here because the behavior is defined entirely by the order
and timing of invocations. This is used by both the ReceiverInputDStream and
the JobGenerator... do those both expect to generate "new" allocations, or
could one of them instead use a simpler API such as "getBlocksForBatch", where
you assume that the batch already has blocks.
The semantics also get hidden up the call chain. `ReceiverTracker` has a
function `getReceivedBlocks` that doesn't say anything about the fact that it
might create new assignments.
If there is any way to change the code to make it more explicit when new
allocations are happening that would be helpful.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
