org.apache.flink.runtime.io.network.partition

Class PipelinedApproximateSubpartition

java.lang.Object

org.apache.flink.runtime.io.network.partition.ResultSubpartition

org.apache.flink.runtime.io.network.partition.PipelinedSubpartition

org.apache.flink.runtime.io.network.partition.PipelinedApproximateSubpartition

All Implemented Interfaces:

ChannelStateHolder

public class PipelinedApproximateSubpartition
extends PipelinedSubpartition

A pipelined in-memory only subpartition, which allows to reconnect after failure. Only one view is allowed at a time to read teh subpartition.

Nested Class Summary

Nested classes/interfaces inherited from class org.apache.flink.runtime.io.network.partition.ResultSubpartition

ResultSubpartition.BufferAndBacklog

Field Summary

Fields inherited from class org.apache.flink.runtime.io.network.partition.ResultSubpartition

ADD_BUFFER_ERROR_CODE, parent, subpartitionInfo

Method Summary

Modifier and Type	Method and Description
PipelinedSubpartitionView	createReadView(BufferAvailabilityListener availabilityListener) To simply the view releasing threading model, releaseView() is called only before creating a new view.
boolean	isSupportChannelStateRecover()

Methods inherited from class org.apache.flink.runtime.io.network.partition.PipelinedSubpartition

abortCheckpoint, acknowledgeAllDataProcessed, add, alignedBarrierTimeout, bufferSize, finish, flush, getAvailabilityAndBacklog, getBuffersInBacklogUnsafe, getChannelStateCheckpointId, getNumberOfQueuedBuffers, getTotalNumberOfBuffersUnsafe, getTotalNumberOfBytesUnsafe, isReleased, release, setChannelStateWriter, toString, unsynchronizedGetNumberOfQueuedBuffers

Methods inherited from class org.apache.flink.runtime.io.network.partition.ResultSubpartition

add, getSubPartitionIndex, getSubpartitionInfo, onConsumedSubpartition

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Method Detail

createReadView

public PipelinedSubpartitionView createReadView(BufferAvailabilityListener availabilityListener)

To simply the view releasing threading model, releaseView() is called only before creating a new view.

There is still one corner case when a downstream task fails continuously in a short period of time then multiple netty worker threads can createReadView at the same time. TODO: This problem will be solved in FLINK-19774

Overrides:

createReadView in class PipelinedSubpartition

isSupportChannelStateRecover

public boolean isSupportChannelStateRecover()

Overrides:

isSupportChannelStateRecover in class PipelinedSubpartition