org.apache.flink.runtime.state

Interface CheckpointStreamFactory

All Known Subinterfaces:

CheckpointStorageLocation

All Known Implementing Classes:

FsCheckpointStorageLocation, FsCheckpointStreamFactory, FsMergingCheckpointStorageLocation, MemCheckpointStreamFactory, NonPersistentMetadataCheckpointStorageLocation, PersistentMetadataCheckpointStorageLocation

public interface CheckpointStreamFactory

A factory for checkpoint output streams, which are used to persist data for checkpoints.

Stream factories can be created from the CheckpointStorageAccess through CheckpointStorageWorkerView.resolveCheckpointStorageLocation(long, CheckpointStorageLocationReference).

Method Summary

Modifier and Type	Method and Description
boolean	canFastDuplicate(StreamStateHandle stateHandle, CheckpointedStateScope scope) Tells if we can duplicate the given StreamStateHandle into the path corresponding to the given CheckpointedStateScope.
default boolean	couldReuseStateHandle(StreamStateHandle stateHandle) A pre-check hook before the checkpoint writer want to reuse a state handle, if this returns false, it is not recommended for the writer to rewrite the state file considering the space amplification.
CheckpointStateOutputStream	createCheckpointStateOutputStream(CheckpointedStateScope scope) Creates an new CheckpointStateOutputStream.
List<StreamStateHandle>	duplicate(List<StreamStateHandle> stateHandles, CheckpointedStateScope scope) Duplicates StreamStateHandle into the path corresponding to * the given CheckpointedStateScope.
default void	reusePreviousStateHandle(Collection<? extends StreamStateHandle> previousHandle) A callback method when some previous handle is reused.

Method Detail

createCheckpointStateOutputStream

CheckpointStateOutputStream createCheckpointStateOutputStream(CheckpointedStateScope scope)
                                                       throws IOException

Creates an new CheckpointStateOutputStream. When the stream is closed, it returns a state handle that can retrieve the state back.

Parameters:

scope - The state's scope, whether it is exclusive or shared.

Returns:

An output stream that writes state for the given checkpoint.

Throws:

IOException - Exceptions may occur while creating the stream and should be forwarded.

canFastDuplicate

boolean canFastDuplicate(StreamStateHandle stateHandle,
                         CheckpointedStateScope scope)
                  throws IOException

Tells if we can duplicate the given StreamStateHandle into the path corresponding to the given CheckpointedStateScope.

This should be a rather cheap operation, preferably not involving any remote accesses.

Parameters:

stateHandle - The handle to duplicate

scope - Scope determining the location to duplicate into

Returns:

true, if we can perform the duplication

Throws:

IOException

duplicate

List<StreamStateHandle> duplicate(List<StreamStateHandle> stateHandles,
                                  CheckpointedStateScope scope)
                           throws IOException

Duplicates StreamStateHandle into the path corresponding to * the given CheckpointedStateScope.

You should first check if you can duplicate with canFastDuplicate(StreamStateHandle, CheckpointedStateScope).

Parameters:

stateHandles - The handles to duplicate

scope - Scope determining the location to duplicate into

Returns:

The duplicated handle

Throws:

IOException

reusePreviousStateHandle

default void reusePreviousStateHandle(Collection<? extends StreamStateHandle> previousHandle)

A callback method when some previous handle is reused. It is needed by the file merging mechanism (FLIP-306) which will manage the life cycle of underlying files by file-reusing information.

Parameters:

previousHandle - the previous handles that will be reused.

couldReuseStateHandle

default boolean couldReuseStateHandle(StreamStateHandle stateHandle)

A pre-check hook before the checkpoint writer want to reuse a state handle, if this returns false, it is not recommended for the writer to rewrite the state file considering the space amplification.

Parameters:

stateHandle - the handle to be reused.

Returns:

true if it can be reused.