[SPARK-21444] Fetch failure due to node reboot causes job failure

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.3.0
• Fix Version/s: 2.3.0
• Component/s: Scheduler, Spark Core
• Assignee: Josh Rosen
• Reporter: Sital Kedia
• Created: 17/Jul/17 23:38
• Updated: 17/May/20 17:48
• Resolved: 18/Jul/17 03:41

Description

We started seeing this issue after merging the PR - https://github.com/apache/spark/pull/17955.

This PR introduced a change to keep the map-output statuses only in the map output tracker and whenever there is a fetch failure, the scheduler tries to invalidate the map-output statuses by talking all the the block manager slave end point. However, in case the fetch failure is because node reboot, the block manager slave end point would not be reachable and so the driver fails the job as a result. See exception below -

2017-07-15 05:20:50,255 WARN  storage.BlockManagerMaster (Logging.scala:logWarning(87)) - Failed to remove broadcast 10 with removeFromMaster = true - Connection reset by peer
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 sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:380)
        at io.netty.buffer.PooledUnsafeDirectByteBuf.setBytes(PooledUnsafeDirectByteBuf.java:313)
        at io.netty.buffer.AbstractByteBuf.writeBytes(AbstractByteBuf.java:881)
        at io.netty.channel.socket.nio.NioSocketChannel.doReadBytes(NioSocketChannel.java:242)
        at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:119)
        at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:511)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:468)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:382)
        at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:354)
        at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:111)
        at java.lang.Thread.run(Thread.java:745)
2017-07-15 05:20:50,275 ERROR scheduler.DAGSchedulerEventProcessLoop (Logging.scala:logError(91)) - DAGSchedulerEventProcessLoop failed; shutting down SparkContext
org.apache.spark.SparkException: Exception thrown in awaitResult
        at org.apache.spark.rpc.RpcTimeout$$anonfun$1.applyOrElse(RpcTimeout.scala:77)
        at org.apache.spark.rpc.RpcTimeout$$anonfun$1.applyOrElse(RpcTimeout.scala:75)
        at scala.runtime.AbstractPartialFunction.apply(AbstractPartialFunction.scala:36)
        at org.apache.spark.rpc.RpcTimeout$$anonfun$addMessageIfTimeout$1.applyOrElse(RpcTimeout.scala:59)
        at org.apache.spark.rpc.RpcTimeout$$anonfun$addMessageIfTimeout$1.applyOrElse(RpcTimeout.scala:59)
        at scala.PartialFunction$OrElse.apply(PartialFunction.scala:167)
        at org.apache.spark.rpc.RpcTimeout.awaitResult(RpcTimeout.scala:83)
        at org.apache.spark.storage.BlockManagerMaster.removeBroadcast(BlockManagerMaster.scala:143)
        at org.apache.spark.broadcast.TorrentBroadcast$.unpersist(TorrentBroadcast.scala:271)
        at org.apache.spark.broadcast.TorrentBroadcast.doDestroy(TorrentBroadcast.scala:165)
        at org.apache.spark.broadcast.Broadcast.destroy(Broadcast.scala:111)
        at org.apache.spark.broadcast.Broadcast.destroy(Broadcast.scala:98)
        at org.apache.spark.ShuffleStatus.invalidateSerializedMapOutputStatusCache(MapOutputTracker.scala:197)
        at org.apache.spark.ShuffleStatus.removeMapOutput(MapOutputTracker.scala:105)
        at org.apache.spark.MapOutputTrackerMaster.unregisterMapOutput(MapOutputTracker.scala:420)
        at org.apache.spark.scheduler.DAGScheduler.handleTaskCompletion(DAGScheduler.scala:1324)
        at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.doOnReceive(DAGScheduler.scala:1715)
        at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:1673)
        at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:1662)
        at org.apache.spark.util.EventLoop$$anon$1.run(EventLoop.scala:48)
Caused by: java.io.IOException: Connection reset by peer
        at sun.nio.ch.FileDispatcherImpl.read0(Native Method)
        at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:39)
        at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:223)
        at sun.nio.ch.IOUtil.read(IOUtil.java:192)
        at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:380)
        at io.netty.buffer.PooledUnsafeDirectByteBuf.setBytes(PooledUnsafeDirectByteBuf.java:313)
        at io.netty.buffer.AbstractByteBuf.writeBytes(AbstractByteBuf.java:881)
        at io.netty.channel.socket.nio.NioSocketChannel.doReadBytes(NioSocketChannel.java:242)
        at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:119)
        at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:511)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:468)
        at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:382)
        at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:354)
        at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:111)
        at java.lang.Thread.run(Thread.java:745)

Issue Links

• relates to: SPARK-20715 MapStatuses shouldn't be redundantly stored in both ShuffleMapStage and MapOutputTracker
• links to: [Github] Pull Request #18662 (JoshRosen)