[FLINK-10491] Deadlock during spilling data in SpillableSubpartition

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.5.4, 1.6.1
• Fix Version/s: 1.5.6, 1.6.3, 1.7.0
• Component/s: Runtime / Network
• Labels: pull-request-available
• Assignee: Zhijiang
• Reporter: Piotr Nowojski
• Created: 04/Oct/18 13:40
• Updated: 23/Oct/18 13:57
• Resolved: 23/Oct/18 13:57

Description

Originally reported here: https://lists.apache.org/thread.html/472c8f4a2711c5e217fadd9a88f8c73670218e7432bb81ba3f5076db@%3Cuser.flink.apache.org%3E

Thread dump (from 1.5.3 version) showing two deadlocked threads, because they are taking two locks in different order:

Thread-1
"DataSink (DATA#HadoopFileOutputFormat ) (1/2)@11002" prio=5 tid=0x3e2 nid=NA waiting for monitor entry
waiting for Map (Key Extractor) (1/10)@9967 to release lock on <0x2dfb> (a java.util.ArrayDeque)
at org.apache.flink.runtime.io.network.partition.SpillableSubpartition.releaseMemory(SpillableSubpartition.java:223)
at org.apache.flink.runtime.io.network.partition.ResultPartition.releaseMemory(ResultPartition.java:373)
at org.apache.flink.runtime.io.network.buffer.LocalBufferPool.setNumBuffers(LocalBufferPool.java:355)
- locked <0x2dfd> (a java.util.ArrayDeque)
at org.apache.flink.runtime.io.network.buffer.NetworkBufferPool.redistributeBuffers(NetworkBufferPool.java:402)
at org.apache.flink.runtime.io.network.buffer.NetworkBufferPool.recycleMemorySegments(NetworkBufferPool.java:203)
- locked <0x2da5> (a java.lang.Object)
at org.apache.flink.runtime.io.network.buffer.NetworkBufferPool.recycleMemorySegments(NetworkBufferPool.java:193)
at org.apache.flink.runtime.io.network.partition.consumer.SingleInputGate.returnExclusiveSegments(SingleInputGate.java:318)
at org.apache.flink.runtime.io.network.partition.consumer.RemoteInputChannel.releaseAllResources(RemoteInputChannel.java:259)
at org.apache.flink.runtime.io.network.partition.consumer.SingleInputGate.getNextBufferOrEvent(SingleInputGate.java:578)
at org.apache.flink.runtime.io.network.partition.consumer.SingleInputGate.pollNextBufferOrEvent(SingleInputGate.java:507)
at org.apache.flink.runtime.io.network.partition.consumer.UnionInputGate.waitAndGetNextInputGate(UnionInputGate.java:213)
at org.apache.flink.runtime.io.network.partition.consumer.UnionInputGate.getNextBufferOrEvent(UnionInputGate.java:163)
at org.apache.flink.runtime.io.network.api.reader.AbstractRecordReader.getNextRecord(AbstractRecordReader.java:86)
at org.apache.flink.runtime.io.network.api.reader.MutableRecordReader.next(MutableRecordReader.java:47)
at org.apache.flink.runtime.operators.util.ReaderIterator.next(ReaderIterator.java:73)
at org.apache.flink.runtime.operators.DataSinkTask.invoke(DataSinkTask.java:216)
at org.apache.flink.runtime.taskmanager.Task.run(Task.java:703)
at java.lang.Thread.run(Thread.java:745)


Thread-2
"Map (Key Extractor) (1/10)@9967" prio=5 tid=0xaab nid=NA waiting for monitor entry
java.lang.Thread.State: BLOCKED
blocks DataSink (DATA#HadoopFileOutputFormat ) (1/2)@11002
waiting for DataSink (DATA#HadoopFileOutputFormat ) (1/2)@11002 to release lock on <0x2dfd> (a java.util.ArrayDeque)
at org.apache.flink.runtime.io.network.buffer.LocalBufferPool.recycle(LocalBufferPool.java:261)
at org.apache.flink.runtime.io.network.buffer.NetworkBuffer.deallocate(NetworkBuffer.java:171)
at org.apache.flink.shaded.netty4.io.netty.buffer.AbstractReferenceCountedByteBuf.release(AbstractReferenceCountedByteBuf.java:106)
at org.apache.flink.runtime.io.network.buffer.NetworkBuffer.recycleBuffer(NetworkBuffer.java:146)
at org.apache.flink.runtime.io.network.buffer.BufferConsumer.close(BufferConsumer.java:110)
at org.apache.flink.runtime.io.network.partition.SpillableSubpartition.spillFinishedBufferConsumers(SpillableSubpartition.java:271)
at org.apache.flink.runtime.io.network.partition.SpillableSubpartition.add(SpillableSubpartition.java:117)
- locked <0x2dfb> (a java.util.ArrayDeque)
at org.apache.flink.runtime.io.network.partition.SpillableSubpartition.add(SpillableSubpartition.java:96)
- locked <0x2dfc> (a org.apache.flink.runtime.io.network.partition.SpillableSubpartition)
at org.apache.flink.runtime.io.network.partition.ResultPartition.addBufferConsumer(ResultPartition.java:255)
at org.apache.flink.runtime.io.network.api.writer.RecordWriter.requestNewBufferBuilder(RecordWriter.java:211)
at org.apache.flink.runtime.io.network.api.writer.RecordWriter.sendToTarget(RecordWriter.java:142)
at org.apache.flink.runtime.io.network.api.writer.RecordWriter.emit(RecordWriter.java:105)
at org.apache.flink.runtime.operators.shipping.OutputCollector.collect(OutputCollector.java:65)
at org.apache.flink.runtime.operators.util.metrics.CountingCollector.collect(CountingCollector.java:35)
at org.apache.flink.runtime.operators.MapDriver.run(MapDriver.java:103)
at org.apache.flink.runtime.operators.BatchTask.run(BatchTask.java:503)
at org.apache.flink.runtime.operators.BatchTask.invoke(BatchTask.java:368)
at org.apache.flink.runtime.taskmanager.Task.run(Task.java:703)
at java.lang.Thread.run(Thread.java:745)

The same pattern can occur on the master (but with slightly shifted line numbers)

CC NicoK

Issue Links

• links to: GitHub Pull Request #6809