com.alipay.sofa.jraft

接口 StateMachine

所有已知实现类:

StateMachineAdapter

public interface StateMachine

|StateMachine| is the sink of all the events of a very raft node. Implement a specific StateMachine for your own business logic. NOTE: All the interfaces are not guaranteed to be thread safe and they are called sequentially, saying that every single operation will block all the following ones.

作者:

boyan (boyan@alibaba-inc.com) 2018-Apr-08 5:43:21 PM

方法概要

限定符和类型	方法和说明
void	onApply(Iterator iter) Update the StateMachine with a batch a tasks that can be accessed through |iterator|.
void	onConfigurationCommitted(Configuration conf) Invoked when a configuration has been committed to the group.
void	onError(RaftException e) This method is called when a critical error was encountered, after this point, no any further modification is allowed to applied to this node until the error is fixed and this node restarts.
void	onLeaderStart(long term) Invoked when the belonging node becomes the leader of the group at |term| Default: Do nothing
void	onLeaderStop(Status status) Invoked when this node steps down from the leader of the replication group and |status| describes detailed information
void	onShutdown() Invoked once when the raft node was shut down.
boolean	onSnapshotLoad(SnapshotReader reader) User defined snapshot load function get and load snapshot Default: Load nothing and returns error.
void	onSnapshotSave(SnapshotWriter writer, Closure done) User defined snapshot generate function, this method will block StateMachine#onApply(Iterator).
void	onStartFollowing(LeaderChangeContext ctx) This method is called when a follower or candidate starts following a leader and its leaderId (should be NULL before the method is called) is set to the leader's id, situations including: 1. a candidate receives appendEntries request from a leader 2. a follower(without leader) receives appendEntries from a leader the parameter ctx gives the information(leaderId, term and status) about the very leader whom the follower starts to follow.
void	onStopFollowing(LeaderChangeContext ctx) This method is called when a follower stops following a leader and its leaderId becomes null, situations including: 1. handle election timeout and start preVote 2. receive requests with higher term such as VoteRequest from a candidate or appendEntries request from a new leader 3. receive timeoutNow request from current leader and start request vote.

方法详细资料

onApply

void onApply(Iterator iter)

Update the StateMachine with a batch a tasks that can be accessed through |iterator|. Invoked when one or more tasks that were passed to Node#apply(Task) have been committed to the raft group (quorum of the group peers have received those tasks and stored them on the backing storage). Once this function returns to the caller, we will regard all the iterated tasks through |iter| have been successfully applied. And if you didn't apply all the the given tasks, we would regard this as a critical error and report a error whose type is ERROR_TYPE_STATE_MACHINE.

参数:

iter - iterator of states

onShutdown

void onShutdown()

Invoked once when the raft node was shut down. Default do nothing

onSnapshotSave

void onSnapshotSave(SnapshotWriter writer,
                    Closure done)

User defined snapshot generate function, this method will block StateMachine#onApply(Iterator). user can make snapshot async when fsm can be cow(copy-on-write). call done.run(status) when snapshot finished. Default: Save nothing and returns error.

参数:

writer - snapshot writer

done - callback

onSnapshotLoad

boolean onSnapshotLoad(SnapshotReader reader)

User defined snapshot load function get and load snapshot Default: Load nothing and returns error.

参数:

reader - snapshot reader

返回:

true on success

onLeaderStart

void onLeaderStart(long term)

Invoked when the belonging node becomes the leader of the group at |term| Default: Do nothing

参数:

term - new term num

onLeaderStop

void onLeaderStop(Status status)

Invoked when this node steps down from the leader of the replication group and |status| describes detailed information

参数:

status - status info

onError

void onError(RaftException e)

This method is called when a critical error was encountered, after this point, no any further modification is allowed to applied to this node until the error is fixed and this node restarts.

参数:

e - raft error message

onConfigurationCommitted

void onConfigurationCommitted(Configuration conf)

Invoked when a configuration has been committed to the group.

参数:

conf - committed configuration

onStopFollowing

void onStopFollowing(LeaderChangeContext ctx)

This method is called when a follower stops following a leader and its leaderId becomes null, situations including: 1. handle election timeout and start preVote 2. receive requests with higher term such as VoteRequest from a candidate or appendEntries request from a new leader 3. receive timeoutNow request from current leader and start request vote. the parameter ctx gives the information(leaderId, term and status) about the very leader whom the follower followed before. User can reset the node's information as it stops following some leader.

参数:

ctx - context of leader change

onStartFollowing

void onStartFollowing(LeaderChangeContext ctx)

This method is called when a follower or candidate starts following a leader and its leaderId (should be NULL before the method is called) is set to the leader's id, situations including: 1. a candidate receives appendEntries request from a leader 2. a follower(without leader) receives appendEntries from a leader the parameter ctx gives the information(leaderId, term and status) about the very leader whom the follower starts to follow. User can reset the node's information as it starts to follow some leader.

参数:

ctx - context of leader change