com.alipay.sofa.jraft.storage

接口 SnapshotExecutor

所有超级接口:

Describer, Lifecycle<SnapshotExecutorOptions>

所有已知实现类:

SnapshotExecutorImpl

public interface SnapshotExecutor
extends Lifecycle<SnapshotExecutorOptions>, Describer

Executing Snapshot related stuff.

作者:

boyan (boyan@alibaba-inc.com) 2018-Mar-22 2:27:02 PM

嵌套类概要

从接口继承的嵌套类/接口 com.alipay.sofa.jraft.util.Describer

Describer.DefaultPrinter, Describer.Printer

方法概要

限定符和类型	方法和说明
void	doSnapshot(Closure done) Start to snapshot StateMachine, and |done| is called after the execution finishes or fails.
NodeImpl	getNode() Return the owner NodeImpl
SnapshotStorage	getSnapshotStorage() Returns the backing snapshot storage
void	installSnapshot(RpcRequests.InstallSnapshotRequest request, RpcRequests.InstallSnapshotResponse.Builder response, RpcRequestClosure done) Install snapshot according to the very RPC from leader After the installing succeeds (StateMachine is reset with the snapshot) or fails, done will be called to respond Errors: - Term mismatches: which happens interrupt_downloading_snapshot was called before install_snapshot, indicating that this RPC was issued by the old leader
void	interruptDownloadingSnapshots(long newTerm) Interrupt the downloading if possible.
boolean	isInstallingSnapshot() Returns true if this is currently installing a snapshot, either downloading or loading.
void	join() Block the current thread until all the running job finishes (including failure)

从接口继承的方法 com.alipay.sofa.jraft.Lifecycle

init, shutdown

从接口继承的方法 com.alipay.sofa.jraft.util.Describer

describe

方法详细资料

getNode

NodeImpl getNode()

Return the owner NodeImpl

doSnapshot

void doSnapshot(Closure done)

Start to snapshot StateMachine, and |done| is called after the execution finishes or fails.

参数:

done - snapshot callback

installSnapshot

void installSnapshot(RpcRequests.InstallSnapshotRequest request,
                     RpcRequests.InstallSnapshotResponse.Builder response,
                     RpcRequestClosure done)

Install snapshot according to the very RPC from leader After the installing succeeds (StateMachine is reset with the snapshot) or fails, done will be called to respond Errors: - Term mismatches: which happens interrupt_downloading_snapshot was called before install_snapshot, indicating that this RPC was issued by the old leader. - Interrupted: happens when interrupt_downloading_snapshot is called or a new RPC with the same or newer snapshot arrives - Busy: the state machine is saving or loading snapshot

interruptDownloadingSnapshots

void interruptDownloadingSnapshots(long newTerm)

Interrupt the downloading if possible. This is called when the term of node increased to |new_term|, which happens when receiving RPC from new peer. In this case, it's hard to determine whether to keep downloading snapshot as the new leader possibly contains the missing logs and is going to send AppendEntries. To make things simplicity and leader changing during snapshot installing is very rare. So we interrupt snapshot downloading when leader changes, and let the new leader decide whether to install a new snapshot or continue appending log entries. NOTE: we can't interrupt the snapshot installing which has finished downloading and is reseting the State Machine.

参数:

newTerm - new term num

isInstallingSnapshot

boolean isInstallingSnapshot()

Returns true if this is currently installing a snapshot, either downloading or loading.

getSnapshotStorage

SnapshotStorage getSnapshotStorage()

Returns the backing snapshot storage

join

void join()
   throws InterruptedException

Block the current thread until all the running job finishes (including failure)

抛出:

InterruptedException