Add cluster replication features and membership management

- Introduced a new `ClusterManager` to handle cluster membership events and facilitate event replication across relay peers.
- Implemented HTTP endpoints for retrieving the latest serial and fetching events within a specified range.
- Enhanced event handling to process cluster membership events (Kind 39108) and update relay lists accordingly.
- Updated configuration to support cluster administrators and their management capabilities.
- Added comprehensive tests to validate the new cluster replication functionalities.
- Documented the cluster replication protocol in a new specification file.
- Bumped version to reflect these changes.

app/config/config.go

@@ -52,6 +52,7 @@ type C struct {
 	RelayAddresses      []string      `env:"ORLY_RELAY_ADDRESSES" usage:"comma-separated list of websocket addresses for this relay (e.g., wss://relay.example.com,wss://backup.example.com)"`
 	RelayPeers          []string      `env:"ORLY_RELAY_PEERS" usage:"comma-separated list of peer relay URLs for distributed synchronization (e.g., https://peer1.example.com,https://peer2.example.com)"`
 	RelayGroupAdmins    []string      `env:"ORLY_RELAY_GROUP_ADMINS" usage:"comma-separated list of npubs authorized to publish relay group configuration events"`
+	ClusterAdmins       []string      `env:"ORLY_CLUSTER_ADMINS" usage:"comma-separated list of npubs authorized to manage cluster membership"`
 	FollowListFrequency time.Duration `env:"ORLY_FOLLOW_LIST_FREQUENCY" usage:"how often to fetch admin follow lists (default: 1h)" default:"1h"`
 
 	// Blossom blob storage service level settings

app/handle-event.go

@@ -467,6 +467,13 @@ func (l *Listener) HandleEvent(msg []byte) (err error) {
 		}
 	}
 
+	// Handle cluster membership events (Kind 39108)
+	if env.E.Kind == 39108 && l.clusterManager != nil {
+		if err := l.clusterManager.HandleMembershipEvent(env.E); err != nil {
+			log.W.F("invalid cluster membership event %s: %v", hex.Enc(env.E.ID), err)
+		}
+	}
+
 	// Update serial for distributed synchronization
 	if l.syncManager != nil {
 		l.syncManager.UpdateSerial()

app/main.go

@@ -152,6 +152,23 @@ func Run(
 		}
 	}
 
+	// Initialize cluster manager for cluster replication
+	var clusterAdminNpubs []string
+	if len(cfg.ClusterAdmins) > 0 {
+		clusterAdminNpubs = cfg.ClusterAdmins
+	} else {
+		// Default to regular admins if no cluster admins specified
+		for _, admin := range cfg.Admins {
+			clusterAdminNpubs = append(clusterAdminNpubs, admin)
+		}
+	}
+
+	if len(clusterAdminNpubs) > 0 {
+		l.clusterManager = dsync.NewClusterManager(ctx, db, clusterAdminNpubs)
+		l.clusterManager.Start()
+		log.I.F("cluster replication manager initialized with %d admin npubs", len(clusterAdminNpubs))
+	}
+
 	// Initialize the user interface
 	l.UserInterface()
 

app/server.go

@@ -53,6 +53,7 @@ type Server struct {
 	spiderManager    *spider.Spider
 	syncManager      *dsync.Manager
 	relayGroupMgr    *dsync.RelayGroupManager
+	clusterManager   *dsync.ClusterManager
 	blossomServer    *blossom.Server
 }
 
@@ -259,6 +260,13 @@ func (s *Server) UserInterface() {
 		s.mux.HandleFunc("/blossom/", s.blossomHandler)
 		log.Printf("Blossom blob storage API enabled at /blossom")
 	}
+
+	// Cluster replication API endpoints
+	if s.clusterManager != nil {
+		s.mux.HandleFunc("/cluster/latest", s.clusterManager.HandleLatestSerial)
+		s.mux.HandleFunc("/cluster/events", s.clusterManager.HandleEventsRange)
+		log.Printf("Cluster replication API enabled at /cluster")
+	}
 }
 
 // handleFavicon serves orly-favicon.png as favicon.ico