Home >> All >> org >> jboss >> ha >> framework >> [ interfaces Javadoc ]

• org.jboss.ha.framework.interfaces: Javadoc index of package org.jboss.ha.framework.interfaces.


Package Samples:

• org.jboss.ha.framework.interfaces

Classes:

• DistributedState: DistributedState is a service on top of HAPartition that provides a cluster-wide distributed state. The DistributedState (DS) service provides a tuple map. Thus, any service, application, container, ... can request its own DS "private space" by working* in its own category (a string name). You work in a category like a Dictionary: you set values by key within a category. Each time a value is added/modified/removed, the modification is made cluster-wide, on all other nodes. Reading values is always made locally (no network access!) Objects can also subscribes to DS events to be notified when some ...
• HARMIProxyCallback: DistributedState is a service on top of HAPartition that provides a cluster-wide distributed state. The DistributedState (DS) service provides a tuple map. Thus, any service, application, container, ... can request its own DS "private space" by working* in its own category (a string name). You work in a category like a Dictionary: you set values by key within a category. Each time a value is added/modified/removed, the modification is made cluster-wide, on all other nodes. Reading values is always made locally (no network access!) Objects can also subscribes to DS events to be notified when some ...
• FirstAvailableIdenticalAllProxies: LoadBalancingPolicy implementation that always favor the first available target i.e. no load balancing occurs. Nevertheless, the first target is randomly selected. This does not mean that fail-over will not occur if the first member in the list dies. In this case, fail-over will occur, and a new target will become the first member and invocation will continously be invoked on the same new target until its death. Each proxy using this policy will *not* elect its own prefered target: the target *is* shared with all proxies that belong to the same family (for a different behaviour please take a look ...
• FirstAvailable: LoadBalancingPolicy implementation that always favor the first available target i.e. no load balancing occurs. Nevertheless, the first target is randomly selected. This does not mean that fail-over will not occur if the first member in the list dies. In this case, fail-over will occur, and a new target will become the first member and invocation will continously be invoked on the same new target until its death. Each proxy using this policy will elect its own prefered target: the target is not shared accross the proxy family (for this behaviour please take a look at FirstAvailableIdenticalAllP ...
• FamilyClusterInfo: Maintain information for a given proxy family. Proxies can statically reference objects implementing this interface: only the content will change as the cluster topology changes, not the FamilyClusterInfo object itself. Proxies or LoadBalancing policy implementations can use the cursor and object attribute to store arbitrary data that is then shared accross all proxies belonging to the same family. Initial access to this object is done through the ClusteringTargetsRepository singleton.
• GenericClusteringException: Generic clustering exception that can be used to replace other exceptions that occur on the server. Thus, only this class is needed on the client side and some specific server side exceptions class are not needed on the client side (such as JMX exceptions for example). Furhtermore, it has support for "COMPLETED" status flag a la IIOP.
• ClusterNode: Replacement for a JG IpAddress that doesn't base its representation on the JG address but on the computed node name added to the IPAddress instead. This is to avoid any problem in the cluster as some nodes may interpret a node name differently (IP resolution, name case, FQDN or host name, etc.)
• ClusteringTargetsRepository: JVM singleton that associates a list of targets (+ other info) contained in a FamilyClusterInfo to a proxy family. For example All remote proxies for a given EJB in a given cluster are part of the same proxy family. Note that home and remote for a same EJB form *2* proxy families.
• HARMIServer: When using HA-RMI, the RMI communication end-point on the server-side is an instance of this class. All invocations are sent through this servant that will route the call to the appropriate object and call the appropriate Java method.
• LoadBalancePolicy: Base interface for load-balancing policies. It is possible to implement many different load-balancing policies by implementing this simple interface and using it in the different clustered services (home interface of SLSB for example)
• HARMIResponse: When using HA-RMI, the result of an invocation is embedded in an instance of this class. It contains the response of the invocation and, if the list of targets has changed, a new view of the cluster.
• SubPartitionsInfo: Holder class that knows about a set of HA(sub)Partition currently building the overall cluster. Exchanged between HASessionState instances to share the same knowledge.
• RandomRobin: LoadBalancingPolicy implementation that always fully randomly select its target (without basing its decision on any historic).
• SubPartitionInfo: Holder class that knows about a particular HA(sub)Partition i.e. member nodes, partition name and some utility functions.
• RoundRobin: LoadBalancingPolicy implementation that always favor the next available target load balancing always occurs.
• FamilyClusterInfoImpl: Default implementation of FamilyClusterInfo
• DistributedReplicantManager
• HAPartition

Home | Contact Us | Privacy Policy | Terms of Service