[jboss-jira] [JBoss JIRA] (WFLY-9521) Marshalling of org.wildfly.clustering.group.Node implementations is suboptimal

[ https://issues.jboss.org/browse/WFLY-9521?page=com.atlassian.jira.plugin.... ] Paul Ferraro updated WFLY-9521: ------------------------------- Description: The Node implementations currently serialize: * org.jgroups.Address * The logical name (typically a host name) * The socket binding address/port of its transport This is overkill. The Node should only need to serialize enough information to reconstruct itself using a NodeFactory<Address>, which should only include: * its Address * something to minimally identify the NodeFactory<Address>, with which to use to reconstruct the Node, e.g. a group name Unfortunately, getting a reference to the NodeFactory during deserialization isn't trivial, as these are on-demand services. Another approach, is to avoid referencing the Node in cache keys/values altogether. Currently, org.wildfly.clustering.registry.Registry uses a Cache<Node, Map.Entry<K, V>> and org.wildfly.clustering.provider.ServiceProviderRegistry uses a Cache<K, Set<Node>>. In both cases, we can replace Node with Address, and thus avoid unnecessary marshalling of the logical name and physical address of a member. was: The Node implementations currently serialize: * org.jgroups.Address * The logical name (typically a host name) * The socket binding address/port of its transport This is overkill. The Node should only need to serialize enough information to reconstruct itself using a NodeFactory<Address>, which should only include: * its Address * something to minimally identify the NodeFactory<Address>, with which to use to reconstruct the Node, e.g. a group name -- This message was sent by Atlassian JIRA (v7.5.0#75005)