com.hp.util.dcord.peer

Class PubSubPeerMonitor<P>

java.lang.Object

com.hp.util.dcord.peer.PubSubPeerMonitor<P>

Type Parameters:

P - type of the peer's status

All Implemented Interfaces:

PeerMonitor<P>, Observable<PeerListener<P>>

public class PubSubPeerMonitor<P>
extends Object
implements PeerMonitor<P>

Implementation of PeerMonitor that uses a PublishSubscribeService to distribute the status across the cluster.

It is highly recommended to use a type unique to the application for the peer's class (Create a class that represent the peer). If two different applications created PubSubPeerMonitor's for a common peer type, they both would receive updates from both applications which will mislead the peer status.

Implementation overview:

• Every time updateLocalPeer(Object) is called a messaged is send with such status so all remote peers received the update.
• Every time an update is received from a remote peer, PeerListener.onUpdate(Node, Object) is notified.
• Every time ClusterListener.onJoin(Node) event is received, a recurrent task is scheduled to send a Peer Request message until the status for such node is received. This is where PublishSubscribeService does not behaves well because all PubSubPeerMonitor instances on all nodes will receive the request even though it is meant for a single PubSubPeerMonitor. PubSubPeerMonitor's which are not the target of the request will discard the message. The node that joined the cluster will receive multiple requests (one for each active node) and thus it will reply several times with the same status which will be received also several times for each active node.
• Every time ClusterListener.onLeave(Node) event is received, PeerListener.onBecomingUnknown(Node, Object) is notified.

Author:

Fabiel Zuniga

Method Summary

Methods

Modifier and Type	Method and Description
static <P> Restrictable<PeerMonitor<P>,Stoppable>	create(Class<P> peerClass, CoordinationService coordinationService, ThreadExecutor threadExecutor, org.slf4j.Logger logger) Crates a peer monitor service.
static <P> Restrictable<PeerMonitor<P>,Stoppable>	create(Class<P> peerClass, Serializer<P> peerSerializer, CoordinationService coordinationService, ThreadExecutor threadExecutor, org.slf4j.Logger logger) Crates a peer monitor service.
Collection<Pair<Node,P>>	getKnownPeers() Gets known peers (including the local peer if its initial state has been reported via PeerMonitor.updateLocalPeer(Object)) which are the peers for which the state is considered up to date.
Collection<Pair<Node,P>>	getUnknownPeers() Gets unknown peers (including the local peer if its initial state has not been reported via PeerMonitor.updateLocalPeer(Object)) which are the peers for which the state is considered unknown or out of date: These are peers that have never joined the cluster, peers that have not reported status after they left, or peers that previously reported status but the node they belong to are not part of ClusterService.getActiveNodes().
void	register(PeerListener<P> observer) Registers an observer (listener).
void	unregister(PeerListener<P> observer) Unregisters an observer (listener).
void	updateLocalPeer(P localPeer) Updates the state of the local peer.

Methods inherited from class java.lang.Object

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

Method Detail

create

public static <P> Restrictable<PeerMonitor<P>,Stoppable> create(Class<P> peerClass,
                                                CoordinationService coordinationService,
                                                ThreadExecutor threadExecutor,
                                                org.slf4j.Logger logger)
                                                     throws NullPointerException

Crates a peer monitor service.

Parameters:

peerClass - peer's class

coordinationService - coordination service

threadExecutor - thread executor

logger - logger

Returns:

a PeerMonitor

Throws:

NullPointerException - if either coordinationService or threadExecutor or logger is null

create

public static <P> Restrictable<PeerMonitor<P>,Stoppable> create(Class<P> peerClass,
                                                Serializer<P> peerSerializer,
                                                CoordinationService coordinationService,
                                                ThreadExecutor threadExecutor,
                                                org.slf4j.Logger logger)
                                                     throws NullPointerException

Crates a peer monitor service.

Parameters:

peerClass - peer's class

peerSerializer - peer serializer

coordinationService - coordination service

threadExecutor - thread executor

logger - logger

Returns:

a PeerMonitor

Throws:

NullPointerException - if either coordinationService or threadExecutor or logger is null

register

public void register(PeerListener<P> observer)

Description copied from interface: Observable

Registers an observer (listener).

Specified by:

register in interface Observable<PeerListener<P>>

Parameters:

observer - observer to register

unregister

public void unregister(PeerListener<P> observer)

Description copied from interface: Observable

Unregisters an observer (listener).

Specified by:

unregister in interface Observable<PeerListener<P>>

Parameters:

observer - observer to unregister

updateLocalPeer

public void updateLocalPeer(P localPeer)

Description copied from interface: PeerMonitor

Updates the state of the local peer.

This method should be called when the initial state of the local peer is available, and after that when such state changes.

Specified by:

updateLocalPeer in interface PeerMonitor<P>

Parameters:

localPeer - local peer

getKnownPeers

public Collection<Pair<Node,P>> getKnownPeers()

Description copied from interface: PeerMonitor

Gets known peers (including the local peer if its initial state has been reported via PeerMonitor.updateLocalPeer(Object)) which are the peers for which the state is considered up to date. These are peers that have reported status and for which the node they belong to are part of ClusterService.getActiveNodes(). Note that a peer is only known if it has reported status after joining the cluster; Thus, a node that is part of ClusterService.getActiveNodes() is not guaranteed to be known.

Specified by:

getKnownPeers in interface PeerMonitor<P>

Returns:

reachable peers.

getUnknownPeers

public Collection<Pair<Node,P>> getUnknownPeers()

Description copied from interface: PeerMonitor

Gets unknown peers (including the local peer if its initial state has not been reported via PeerMonitor.updateLocalPeer(Object)) which are the peers for which the state is considered unknown or out of date: These are peers that have never joined the cluster, peers that have not reported status after they left, or peers that previously reported status but the node they belong to are not part of ClusterService.getActiveNodes().

The peer's status included in the result is the last known if available, otherwise null.

Specified by:

getUnknownPeers in interface PeerMonitor<P>

Returns:

unreachable peers.