org.apache.flink.runtime.checkpoint

Class PendingCheckpoint

java.lang.Object

org.apache.flink.runtime.checkpoint.PendingCheckpoint

All Implemented Interfaces:

Checkpoint

@NotThreadSafe
public class PendingCheckpoint
extends Object
implements Checkpoint

A pending checkpoint is a checkpoint that has been started, but has not been acknowledged by all tasks that need to acknowledge it. Once all tasks have acknowledged it, it becomes a CompletedCheckpoint.

Note that the pending checkpoint, as well as the successful checkpoint keep the state handles always as serialized values, never as actual values.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
class	PendingCheckpoint.PendingCheckpointDiscardObject Implementation of Checkpoint.DiscardObject for PendingCheckpoint.
static class	PendingCheckpoint.TaskAcknowledgeResult Result of the acknowledgedTasks method.

Nested classes/interfaces inherited from interface org.apache.flink.runtime.checkpoint.Checkpoint

Checkpoint.DiscardObject

Field Summary

Fields inherited from interface org.apache.flink.runtime.checkpoint.Checkpoint

NOOP_DISCARD_OBJECT

Constructor Summary

Constructors

Constructor and Description
PendingCheckpoint(org.apache.flink.api.common.JobID jobId, long checkpointId, long checkpointTimestamp, CheckpointPlan checkpointPlan, Collection<OperatorID> operatorCoordinatorsToConfirm, Collection<String> masterStateIdentifiers, CheckpointProperties props, CompletableFuture<CompletedCheckpoint> onCompletionPromise, PendingCheckpointStats pendingCheckpointStats)

Method Summary

Modifier and Type	Method and Description
void	abort(CheckpointFailureReason reason, Throwable cause, CheckpointsCleaner checkpointsCleaner, Runnable postCleanup, Executor executor, CheckpointStatsTracker statsTracker) Aborts a checkpoint with reason and cause.
PendingCheckpoint.TaskAcknowledgeResult	acknowledgeCoordinatorState(OperatorInfo coordinatorInfo, ByteStreamStateHandle stateHandle)
void	acknowledgeMasterState(String identifier, MasterState state) Acknowledges a master state (state generated on the checkpoint coordinator) to the pending checkpoint.
PendingCheckpoint.TaskAcknowledgeResult	acknowledgeTask(ExecutionAttemptID executionAttemptId, TaskStateSnapshot operatorSubtaskStates, CheckpointMetrics metrics) Acknowledges the task with the given execution attempt id and the given subtask state.
boolean	canBeSubsumed() Checks whether this checkpoint can be subsumed or whether it should always continue, regardless of newer checkpoints in progress.
CompletedCheckpoint	finalizeCheckpoint(CheckpointsCleaner checkpointsCleaner, Runnable postCleanup, Executor executor)
long	getCheckpointId() Deprecated. use getCheckpointID()
long	getCheckpointID()
CheckpointPlan	getCheckpointPlan()
CheckpointStorageLocation	getCheckpointStorageLocation()
long	getCheckpointTimestamp()
CompletableFuture<CompletedCheckpoint>	getCompletionFuture() Returns the completion future.
CheckpointException	getFailureCause()
org.apache.flink.api.common.JobID	getJobId()
List<MasterState>	getMasterStates()
int	getNumberOfAcknowledgedTasks()
int	getNumberOfNonAcknowledgedOperatorCoordinators()
int	getNumberOfNonAcknowledgedTasks()
Map<OperatorID,OperatorState>	getOperatorStates()
boolean	isAcknowledgedBy(ExecutionAttemptID executionAttemptId)
boolean	isDisposed()
boolean	isFullyAcknowledged()
Checkpoint.DiscardObject	markAsDiscarded() This method precede the Checkpoint.DiscardObject.discard() method and should be called from the CheckpointCoordinator(under the lock) while Checkpoint.DiscardObject.discard() can be called from any thread/place.
boolean	setCancellerHandle(ScheduledFuture<?> cancellerHandle) Sets the handle for the canceller to this pending checkpoint.
void	setCheckpointTargetLocation(CheckpointStorageLocation targetLocation)
String	toString()

Methods inherited from class java.lang.Object

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

Constructor Detail

PendingCheckpoint

public PendingCheckpoint(org.apache.flink.api.common.JobID jobId,
                         long checkpointId,
                         long checkpointTimestamp,
                         CheckpointPlan checkpointPlan,
                         Collection<OperatorID> operatorCoordinatorsToConfirm,
                         Collection<String> masterStateIdentifiers,
                         CheckpointProperties props,
                         CompletableFuture<CompletedCheckpoint> onCompletionPromise,
                         @Nullable
                         PendingCheckpointStats pendingCheckpointStats)

Method Detail

getJobId

public org.apache.flink.api.common.JobID getJobId()

getCheckpointId

@Deprecated
public long getCheckpointId()

Deprecated. use getCheckpointID()

getCheckpointID

public long getCheckpointID()

Specified by:

getCheckpointID in interface Checkpoint

setCheckpointTargetLocation

public void setCheckpointTargetLocation(CheckpointStorageLocation targetLocation)

getCheckpointStorageLocation

public CheckpointStorageLocation getCheckpointStorageLocation()

getCheckpointTimestamp

public long getCheckpointTimestamp()

getNumberOfNonAcknowledgedTasks

public int getNumberOfNonAcknowledgedTasks()

getNumberOfNonAcknowledgedOperatorCoordinators

public int getNumberOfNonAcknowledgedOperatorCoordinators()

getCheckpointPlan

public CheckpointPlan getCheckpointPlan()

getNumberOfAcknowledgedTasks

public int getNumberOfAcknowledgedTasks()

getOperatorStates

public Map<OperatorID,OperatorState> getOperatorStates()

getMasterStates

public List<MasterState> getMasterStates()

isFullyAcknowledged

public boolean isFullyAcknowledged()

isAcknowledgedBy

public boolean isAcknowledgedBy(ExecutionAttemptID executionAttemptId)

isDisposed

public boolean isDisposed()

canBeSubsumed

public boolean canBeSubsumed()

Checks whether this checkpoint can be subsumed or whether it should always continue, regardless of newer checkpoints in progress.

Returns:

True if the checkpoint can be subsumed, false otherwise.

setCancellerHandle

public boolean setCancellerHandle(ScheduledFuture<?> cancellerHandle)

Sets the handle for the canceller to this pending checkpoint. This method fails with an exception if a handle has already been set.

Returns:

true, if the handle was set, false, if the checkpoint is already disposed;

getFailureCause

public CheckpointException getFailureCause()

getCompletionFuture

public CompletableFuture<CompletedCheckpoint> getCompletionFuture()

Returns the completion future.

Returns:

A future to the completed checkpoint

finalizeCheckpoint

public CompletedCheckpoint finalizeCheckpoint(CheckpointsCleaner checkpointsCleaner,
                                              Runnable postCleanup,
                                              Executor executor)
                                       throws IOException

Throws:

IOException

acknowledgeTask

public PendingCheckpoint.TaskAcknowledgeResult acknowledgeTask(ExecutionAttemptID executionAttemptId,
                                                               TaskStateSnapshot operatorSubtaskStates,
                                                               CheckpointMetrics metrics)

Acknowledges the task with the given execution attempt id and the given subtask state.

Parameters:

executionAttemptId - of the acknowledged task

operatorSubtaskStates - of the acknowledged task

metrics - Checkpoint metrics for the stats

Returns:

TaskAcknowledgeResult of the operation

acknowledgeCoordinatorState

public PendingCheckpoint.TaskAcknowledgeResult acknowledgeCoordinatorState(OperatorInfo coordinatorInfo,
                                                                           @Nullable
                                                                           ByteStreamStateHandle stateHandle)

acknowledgeMasterState

public void acknowledgeMasterState(String identifier,
                                   @Nullable
                                   MasterState state)

Acknowledges a master state (state generated on the checkpoint coordinator) to the pending checkpoint.

Parameters:

identifier - The identifier of the master state

state - The state to acknowledge

abort

public void abort(CheckpointFailureReason reason,
                  @Nullable
                  Throwable cause,
                  CheckpointsCleaner checkpointsCleaner,
                  Runnable postCleanup,
                  Executor executor,
                  CheckpointStatsTracker statsTracker)

Aborts a checkpoint with reason and cause.

markAsDiscarded

public Checkpoint.DiscardObject markAsDiscarded()

Description copied from interface: Checkpoint

This method precede the Checkpoint.DiscardObject.discard() method and should be called from the CheckpointCoordinator(under the lock) while Checkpoint.DiscardObject.discard() can be called from any thread/place.

Specified by:

markAsDiscarded in interface Checkpoint

toString

public String toString()

Overrides:

toString in class Object