[FLINK-18832] BoundedBlockingSubpartition does not work with StreamTask

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.10.1, 1.11.1, 1.12.0
• Fix Version/s: 1.10.3, 1.11.2, 1.12.0
• Component/s: Runtime / Network, Runtime / Task
• Labels: pull-request-available
• Assignee: Zhijiang
• Reporter: Till Rohrmann
• Created: 05/Aug/20 13:56
• Updated: 08/Sep/20 08:10
• Resolved: 07/Sep/20 15:02

Description

The BoundedBlockingSubpartition does not work with a StreamTask because the StreamTask instantiates an OutputFlusher which concurrently accesses the BoundedBlockingSubpartition. This concurrency can lead to a double closing of the underlying BufferConsumer which manifests in this stack trace:

[9:15 PM] Caused by: org.apache.flink.shaded.netty4.io.netty.util.IllegalReferenceCountException: refCnt: 0, increment: 1
	at org.apache.flink.shaded.netty4.io.netty.util.internal.ReferenceCountUpdater.retain0(ReferenceCountUpdater.java:123)
	at org.apache.flink.shaded.netty4.io.netty.util.internal.ReferenceCountUpdater.retain(ReferenceCountUpdater.java:110)
	at org.apache.flink.shaded.netty4.io.netty.buffer.AbstractReferenceCountedByteBuf.retain(AbstractReferenceCountedByteBuf.java:80)
	at org.apache.flink.runtime.io.network.buffer.NetworkBuffer.retainBuffer(NetworkBuffer.java:174)
	at org.apache.flink.runtime.io.network.buffer.NetworkBuffer.retainBuffer(NetworkBuffer.java:47)
	at org.apache.flink.runtime.io.network.buffer.ReadOnlySlicedNetworkBuffer.retainBuffer(ReadOnlySlicedNetworkBuffer.java:127)
	at org.apache.flink.runtime.io.network.buffer.ReadOnlySlicedNetworkBuffer.retainBuffer(ReadOnlySlicedNetworkBuffer.java:41)
	at org.apache.flink.runtime.io.network.buffer.BufferConsumer.build(BufferConsumer.java:108)
	at org.apache.flink.runtime.io.network.partition.BoundedBlockingSubpartition.writeAndCloseBufferConsumer(BoundedBlockingSubpartition.java:156)
	at org.apache.flink.runtime.io.network.partition.BoundedBlockingSubpartition.flushCurrentBuffer(BoundedBlockingSubpartition.java:144)
	at org.apache.flink.runtime.io.network.partition.BoundedBlockingSubpartition.flush(BoundedBlockingSubpartition.java:135)
	at org.apache.flink.runtime.io.network.partition.ResultPartition.flushAll(ResultPartition.java:245)
	at org.apache.flink.runtime.io.network.api.writer.RecordWriter.flushAll(RecordWriter.java:183)
	at org.apache.flink.streaming.runtime.io.RecordWriterOutput.flush(RecordWriterOutput.java:156)
	at org.apache.flink.streaming.runtime.tasks.OperatorChain.flushOutputs(OperatorChain.java:344)
	at org.apache.flink.streaming.runtime.tasks.StreamTask.afterInvoke(StreamTask.java:602)
	at org.apache.flink.streaming.runtime.tasks.StreamTask.invoke(StreamTask.java:544)
	at org.apache.flink.runtime.taskmanager.Task.doRun(Task.java:721)
	at org.apache.flink.runtime.taskmanager.Task.run(Task.java:546)
	at java.lang.Thread.run(Thread.java:748)

Issue Links

• is related to: FLINK-15750 Execute network flushing actions in the mailbox instead of OutputFlusher
• links to: GitHub Pull Request #13209
• links to: GitHub Pull Request #13345
• links to: GitHub Pull Request #13346