Skip to content

Commit

Permalink
[SPARK-7788] Made KinesisReceiver.onStart() non-blocking
Browse files Browse the repository at this point in the history
KinesisReceiver calls worker.run() which is a blocking call (while loop) as per source code of kinesis-client library - https://github.com/awslabs/amazon-kinesis-client/blob/v1.2.1/src/main/java/com/amazonaws/services/kinesis/clientlibrary/lib/worker/Worker.java.
This results in infinite loop while calling sparkStreamingContext.stop(stopSparkContext = false, stopGracefully = true) perhaps because ReceiverTracker is never able to register the receiver (it's receiverInfo field is a empty map) causing it to be stuck in infinite loop while waiting for running flag to be set to false.

Author: Tathagata Das <[email protected]>

Closes apache#6348 from tdas/SPARK-7788 and squashes the following commits:

2584683 [Tathagata Das] Added receiver id in thread name
6cf1cd4 [Tathagata Das] Made KinesisReceiver.onStart non-blocking
  • Loading branch information
tdas committed May 23, 2015
1 parent 3d8760d commit 1c388a9
Showing 1 changed file with 25 additions and 5 deletions.
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,8 @@ package org.apache.spark.streaming.kinesis

import java.util.UUID

import scala.util.control.NonFatal

import com.amazonaws.auth.{AWSCredentials, AWSCredentialsProvider, BasicAWSCredentials, DefaultAWSCredentialsProviderChain}
import com.amazonaws.services.kinesis.clientlibrary.interfaces.{IRecordProcessor, IRecordProcessorFactory}
import com.amazonaws.services.kinesis.clientlibrary.lib.worker.{InitialPositionInStream, KinesisClientLibConfiguration, Worker}
Expand Down Expand Up @@ -98,6 +100,9 @@ private[kinesis] class KinesisReceiver(
*/
private var worker: Worker = null

/** Thread running the worker */
private var workerThread: Thread = null

/**
* This is called when the KinesisReceiver starts and must be non-blocking.
* The KCL creates and manages the receiving/processing thread pool through Worker.run().
Expand Down Expand Up @@ -126,8 +131,19 @@ private[kinesis] class KinesisReceiver(
}

worker = new Worker(recordProcessorFactory, kinesisClientLibConfiguration)
worker.run()

workerThread = new Thread() {
override def run(): Unit = {
try {
worker.run()
} catch {
case NonFatal(e) =>
restart("Error running the KCL worker in Receiver", e)
}
}
}
workerThread.setName(s"Kinesis Receiver ${streamId}")
workerThread.setDaemon(true)
workerThread.start()
logInfo(s"Started receiver with workerId $workerId")
}

Expand All @@ -137,10 +153,14 @@ private[kinesis] class KinesisReceiver(
* The KCL will do its best to drain and checkpoint any in-flight records upon shutdown.
*/
override def onStop() {
if (worker != null) {
worker.shutdown()
if (workerThread != null) {
if (worker != null) {
worker.shutdown()
worker = null
}
workerThread.join()
workerThread = null
logInfo(s"Stopped receiver for workerId $workerId")
worker = null
}
workerId = null
}
Expand Down

0 comments on commit 1c388a9

Please sign in to comment.