Interface DiscoveryStrategy
- All Known Implementing Classes:
AbstractDiscoveryStrategy
,MulticastDiscoveryStrategy
DiscoveryStrategy
itself is the actual implementation to discover
nodes based on whatever environment is used to run the Hazelcast cloud. The
internal implementation is completely vendor specific and might use external Java
libraries to query APIs or send requests to a external REST service.
DiscoveryStrategies
are also free to define any kind of properties that might
be necessary to discover eligible nodes based on configured attributes, tags or any
other kind of metadata.
Using the simple lifecycle management strategies like multicast discovery is able to register and destroy sockets based on Hazelcast’s lifecycle. Deactivated services will also never be started.
Attention: Instead of implementing this interface directly, implementors should consider
using the abstract class AbstractDiscoveryStrategy
instead for easier upgradeability
while this interface evolves.
- Since:
- 3.6
-
Method Summary
Modifier and TypeMethodDescriptionvoid
destroy()
Thestop
method is used to stop internal services, sockets or to destroy any kind of internal state.Returns a map with discovered metadata provided by the runtime environment.Returns a set of all discovered nodes based on the defined properties that were used to create theDiscoveryStrategy
instance.default PartitionGroupStrategy
Deprecated.default PartitionGroupStrategy
getPartitionGroupStrategy
(Collection<? extends Member> allMembers) Returns a custom implementation of aPartitionGroupStrategy
to override default behavior of zone aware backup strategies or to provide a specific behavior in case the discovery environment does not provide information about the infrastructure to be used for automatic configuration.Fetches a set ofAddress
marked as unhealthy by the underlying implementation.default void
markEndpointAsUnhealthy
(Address address) Marks the passedAddress
as unhealthy, which prevents it from being offered as a viable endpoint in someDiscoveryStrategy
implementations, usually prompting this endpoint to be periodically probed for liveliness.void
start()
Thestart
method is used to initialize internal state and perform any kind of startup procedure like multicast socket creation.
-
Method Details
-
start
void start()Thestart
method is used to initialize internal state and perform any kind of startup procedure like multicast socket creation. The behavior of this method might change based on theDiscoveryNode
instance passed to theDiscoveryStrategyFactory
. -
discoverNodes
Iterable<DiscoveryNode> discoverNodes()Returns a set of all discovered nodes based on the defined properties that were used to create theDiscoveryStrategy
instance.- Returns:
- a set of all discovered nodes
-
destroy
void destroy()Thestop
method is used to stop internal services, sockets or to destroy any kind of internal state. -
getPartitionGroupStrategy
Returns a custom implementation of aPartitionGroupStrategy
to override default behavior of zone aware backup strategies or to provide a specific behavior in case the discovery environment does not provide information about the infrastructure to be used for automatic configuration.- Parameters:
allMembers
- Current state of Cluster data members, excluding lite members- Returns:
- a custom implementation of a
PartitionGroupStrategy
otherwisenull
in case of the default implementation is to be used - Since:
- 4.2.1
-
getPartitionGroupStrategy
Deprecated.- use the above method that takes allMember arguments Returns a custom implementation of aPartitionGroupStrategy
to override default behavior of zone aware backup strategies or to provide a specific behavior in case the discovery environment does not provide information about the infrastructure to be used for automatic configuration.- Returns:
- a custom implementation of a
PartitionGroupStrategy
otherwisenull
in case of the default implementation is to be used - Since:
- 3.7
-
discoverLocalMetadata
Returns a map with discovered metadata provided by the runtime environment. Those information may include, but are not limited, to location information like datacenter, node name or additional tags to be used for custom purpose.Information discovered from this method are shaded into the
Member
s attributes. Existing attributes will not be overridden, that way local attribute configuration overrides provided metadata.The default implementation provides an empty map with no further metadata configured. Returning
null
is not permitted and will most probably result in anNullPointerException
inside the cluster system.- Returns:
- a map of discovered metadata as provided by the runtime environment
- Since:
- 3.7
-
markEndpointAsUnhealthy
Marks the passedAddress
as unhealthy, which prevents it from being offered as a viable endpoint in someDiscoveryStrategy
implementations, usually prompting this endpoint to be periodically probed for liveliness. If not supported by the underlying implementation, then this call does nothing.- Parameters:
address
- the address to mark as unhealthy- Since:
- 5.4
-
getUnhealthyEndpoints
Fetches a set ofAddress
marked as unhealthy by the underlying implementation. If not supported by the underlying implementation, then this call returns an empty set.- Returns:
- set of
Address
which are currently marked as unhealthy if supported by the underlying implementation, otherwise an empty set. - Since:
- 5.4
-
PartitionGroupStrategy
to override default behavior of zone aware backup strategies or to provide a specific behavior in case the discovery environment does not provide information about the infrastructure to be used for automatic configuration.