com.alipay.sofa.jraft

接口 ReplicatorGroup

所有超级接口:

Describer

所有已知实现类:

ReplicatorGroupImpl

public interface ReplicatorGroup
extends Describer

Replicators in a raft group.

作者:

boyan (boyan@alibaba-inc.com) 2018-Apr-08 5:35:26 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
default boolean	addReplicator(PeerId peer) Adds a replicator for follower(ReplicatorType.Follower).
default boolean	addReplicator(PeerId peer, ReplicatorType replicatorType) Add a replicator attached with |peer| will be a notification when the replicator catches up according to the arguments.
boolean	addReplicator(PeerId peer, ReplicatorType replicatorType, boolean sync) Try to add a replicator attached with |peer| will be a notification when the replicator catches up according to the arguments.
void	checkReplicator(PeerId peer, boolean lockNode) Check replicator state, if it's not started, start it; if it is blocked, unblock it.
void	clearFailureReplicators() Clear failure to start replicators
boolean	contains(PeerId peer) Returns true if the there's a replicator attached to the given |peer|
PeerId	findTheNextCandidate(ConfigurationEntry conf) Find the follower with the most log entries in this group, which is likely becomes the leader according to the election algorithm of raft.
long	getLastRpcSendTimestamp(PeerId peer) Get peer's last rpc send timestamp (monotonic time in milliseconds).
ThreadId	getReplicator(PeerId peer) Get replicator id by peer, null if not found.
boolean	init(NodeId nodeId, ReplicatorGroupOptions opts) Init the replicator group.
List<ThreadId>	listReplicators() Returns all replicators.
boolean	resetElectionTimeoutInterval(int newIntervalMs) Reset the interval of electionTimeout for replicator.
boolean	resetHeartbeatInterval(int newIntervalMs) Reset the interval of heartbeat, This method is supposed to be called when the very candidate becomes the leader, so we suppose that there are no running replicators.
boolean	resetTerm(long newTerm) Reset the term of all to-add replicators.
void	sendHeartbeat(PeerId peer, RpcResponseClosure<RpcRequests.AppendEntriesResponse> closure) Send heartbeat to a peer.
boolean	stopAll() Stop all replicators.
ThreadId	stopAllAndFindTheNextCandidate(ConfigurationEntry conf) Stop all the replicators except for the one that we think can be the candidate of the next leader, which has the largest `last_log_id' among peers in |current_conf|.
boolean	stopReplicator(PeerId peer) Stop replicator for the peer.
boolean	stopTransferLeadership(PeerId peer) Stop transferring leadership to the given |peer|
boolean	transferLeadershipTo(PeerId peer, long logIndex) Transfer leadership to the given |peer|
boolean	waitCaughtUp(PeerId peer, long maxMargin, long dueTime, CatchUpClosure done) Wait the peer catchup.

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

describe

方法详细资料

init

boolean init(NodeId nodeId,
             ReplicatorGroupOptions opts)

Init the replicator group.

参数:

nodeId - node id

opts - options of replicator grop

返回:

true if init success

addReplicator

default boolean addReplicator(PeerId peer)

Adds a replicator for follower(ReplicatorType.Follower).

参数:

peer - target peer

返回:

true on success

另请参阅:

addReplicator(PeerId, ReplicatorType)

addReplicator

default boolean addReplicator(PeerId peer,
                              ReplicatorType replicatorType)

Add a replicator attached with |peer| will be a notification when the replicator catches up according to the arguments. NOTE: when calling this function, the replicators starts to work immediately, and might call Node#stepDown which might have race with the caller, you should deal with this situation.

参数:

peer - target peer

replicatorType - replicator type

返回:

true on success

addReplicator

boolean addReplicator(PeerId peer,
                      ReplicatorType replicatorType,
                      boolean sync)

Try to add a replicator attached with |peer| will be a notification when the replicator catches up according to the arguments. NOTE: when calling this function, the replicators starts to work immediately, and might call Node#stepDown which might have race with the caller, you should deal with this situation.

参数:

peer - target peer

replicatorType - replicator type

sync - synchronous

返回:

true on success

sendHeartbeat

void sendHeartbeat(PeerId peer,
                   RpcResponseClosure<RpcRequests.AppendEntriesResponse> closure)

Send heartbeat to a peer.

参数:

peer - target peer

closure - callback

getReplicator

ThreadId getReplicator(PeerId peer)

Get replicator id by peer, null if not found.

参数:

peer - peer of replicator

返回:

the replicator id

checkReplicator

void checkReplicator(PeerId peer,
                     boolean lockNode)

Check replicator state, if it's not started, start it; if it is blocked, unblock it. It should be called by leader.

参数:

peer - peer of replicator

lockNode - if lock with node

clearFailureReplicators

void clearFailureReplicators()

Clear failure to start replicators

waitCaughtUp

boolean waitCaughtUp(PeerId peer,
                     long maxMargin,
                     long dueTime,
                     CatchUpClosure done)

Wait the peer catchup.

getLastRpcSendTimestamp

long getLastRpcSendTimestamp(PeerId peer)

Get peer's last rpc send timestamp (monotonic time in milliseconds).

参数:

peer - the peer of replicator

stopAll

boolean stopAll()

Stop all replicators.

stopReplicator

boolean stopReplicator(PeerId peer)

Stop replicator for the peer.

参数:

peer - the peer of replicator

返回:

true on success

resetTerm

boolean resetTerm(long newTerm)

Reset the term of all to-add replicators. This method is supposed to be called when the very candidate becomes the leader, so we suppose that there are no running replicators. Return true on success, false otherwise

参数:

newTerm - new term num

返回:

true on success

resetHeartbeatInterval

boolean resetHeartbeatInterval(int newIntervalMs)

Reset the interval of heartbeat, This method is supposed to be called when the very candidate becomes the leader, so we suppose that there are no running replicators. return true when success, false otherwise.

参数:

newIntervalMs - new heartbeat interval millis

返回:

true on success

resetElectionTimeoutInterval

boolean resetElectionTimeoutInterval(int newIntervalMs)

Reset the interval of electionTimeout for replicator.

参数:

newIntervalMs - new election timeout millis

返回:

true on success

contains

boolean contains(PeerId peer)

Returns true if the there's a replicator attached to the given |peer|

参数:

peer - target peer

返回:

true on contains

transferLeadershipTo

boolean transferLeadershipTo(PeerId peer,
                             long logIndex)

Transfer leadership to the given |peer|

参数:

peer - target peer

logIndex - log index

返回:

true on success

stopTransferLeadership

boolean stopTransferLeadership(PeerId peer)

Stop transferring leadership to the given |peer|

参数:

peer - target peer

返回:

true on success

stopAllAndFindTheNextCandidate

ThreadId stopAllAndFindTheNextCandidate(ConfigurationEntry conf)

Stop all the replicators except for the one that we think can be the candidate of the next leader, which has the largest `last_log_id' among peers in |current_conf|. |candidate| would be returned if we found one and the caller is responsible for stopping it, or an invalid value if we found none. Returns candidate replicator id on success and null otherwise.

参数:

conf - configuration of all replicators

返回:

candidate replicator id on success

findTheNextCandidate

PeerId findTheNextCandidate(ConfigurationEntry conf)

Find the follower with the most log entries in this group, which is likely becomes the leader according to the election algorithm of raft. Returns the follower peerId on success and null otherwise.

参数:

conf - configuration of all replicators

返回:

the follower peerId on success

listReplicators

List<ThreadId> listReplicators()

Returns all replicators.