@Internal @NotThreadSafe public class CheckpointBarrierUnaligner extends CheckpointBarrierHandler
CheckpointBarrierUnaligner is used for triggering checkpoint while reading the first
barrier and keeping track of the number of received barriers and consumed barriers.| 限定符和类型 | 方法和说明 |
|---|---|
void |
abortPendingCheckpoint(long checkpointId,
org.apache.flink.runtime.checkpoint.CheckpointException exception) |
void |
close() |
CompletableFuture<Void> |
getAllBarriersReceivedFuture(long checkpointId) |
Optional<org.apache.flink.runtime.io.network.buffer.BufferReceivedListener> |
getBufferReceivedListener() |
long |
getLatestCheckpointId() |
boolean |
hasInflightData(long checkpointId,
org.apache.flink.runtime.checkpoint.channel.InputChannelInfo channelInfo)
Returns true if there is in-flight data in the buffers for the given channel and checkpoint.
|
protected boolean |
isCheckpointPending() |
void |
processBarrier(org.apache.flink.runtime.io.network.api.CheckpointBarrier receivedBarrier,
org.apache.flink.runtime.checkpoint.channel.InputChannelInfo channelInfo)
We still need to trigger checkpoint via
ThreadSafeUnaligner#notifyBarrierReceived(CheckpointBarrier, InputChannelInfo) while reading
the first barrier from one channel, because this might happen earlier than the previous async
trigger via mailbox by netty thread. |
void |
processCancellationBarrier(org.apache.flink.runtime.io.network.api.CancelCheckpointMarker cancelBarrier) |
void |
processEndOfPartition() |
String |
toString() |
executeInTaskThread, getAlignmentDurationNanos, getCheckpointStartDelayNanos, isBlocked, markCheckpointStart, notifyAbort, notifyAbortOnCancellationBarrier, notifyCheckpoint, releaseBlocksAndResetBarrierspublic void processBarrier(org.apache.flink.runtime.io.network.api.CheckpointBarrier receivedBarrier,
org.apache.flink.runtime.checkpoint.channel.InputChannelInfo channelInfo)
throws Exception
ThreadSafeUnaligner#notifyBarrierReceived(CheckpointBarrier, InputChannelInfo) while reading
the first barrier from one channel, because this might happen earlier than the previous async
trigger via mailbox by netty thread.
Note this is also suitable for the trigger case of local input channel.
processBarrier 在类中 CheckpointBarrierHandlerExceptionpublic void abortPendingCheckpoint(long checkpointId,
org.apache.flink.runtime.checkpoint.CheckpointException exception)
throws IOException
public void processCancellationBarrier(org.apache.flink.runtime.io.network.api.CancelCheckpointMarker cancelBarrier)
throws Exception
processCancellationBarrier 在类中 CheckpointBarrierHandlerExceptionpublic void processEndOfPartition()
throws Exception
processEndOfPartition 在类中 CheckpointBarrierHandlerExceptionpublic long getLatestCheckpointId()
public void close()
throws IOException
close 在接口中 Closeableclose 在接口中 AutoCloseableclose 在类中 CheckpointBarrierHandlerIOExceptionpublic boolean hasInflightData(long checkpointId,
org.apache.flink.runtime.checkpoint.channel.InputChannelInfo channelInfo)
CheckpointBarrierHandlerpublic CompletableFuture<Void> getAllBarriersReceivedFuture(long checkpointId)
public Optional<org.apache.flink.runtime.io.network.buffer.BufferReceivedListener> getBufferReceivedListener()
protected boolean isCheckpointPending()
Copyright © 2014–2021 The Apache Software Foundation. All rights reserved.