SWIM membership: failure detector, gossip dissemination, UDP transport

nodedb-cluster/Cargo.toml

@@ -22,6 +22,7 @@ thiserror = { workspace = true }
 tracing = { workspace = true }
 tokio = { workspace = true }
 rand = { workspace = true }
+async-trait = { workspace = true }
 
 # Serialization (compact binary for RPCs — matches nexar wire format)
 rkyv = { workspace = true }

nodedb-cluster/src/bootstrap/bootstrap_fn.rs

@@ -101,6 +101,7 @@ mod tests {
             data_dir: _dir.path().to_path_buf(),
             force_bootstrap: false,
             join_retry: Default::default(),
+            swim_udp_addr: None,
         };
 
         let state = bootstrap(&config, &catalog).unwrap();

nodedb-cluster/src/bootstrap/config.rs

@@ -84,6 +84,13 @@ pub struct ClusterConfig {
     /// (`8` attempts, `32 s` ceiling). Tests override this with a
     /// faster policy.
     pub join_retry: JoinRetryPolicy,
+    /// Optional UDP bind address for the SWIM failure detector. `None`
+    /// disables SWIM entirely — cluster startup then relies solely on
+    /// the existing raft transport for membership observations. When
+    /// `Some`, the operator is expected to spawn SWIM separately via
+    /// [`crate::spawn_swim`] after the cluster is up and feed the
+    /// seed list from `seed_nodes`.
+    pub swim_udp_addr: Option<SocketAddr>,
 }
 
 /// Result of cluster startup — everything needed to run the Raft loop.

nodedb-cluster/src/bootstrap/join.rs

@@ -449,6 +449,7 @@ mod tests {
             data_dir: _dir1.path().to_path_buf(),
             force_bootstrap: false,
             join_retry: Default::default(),
+            swim_udp_addr: None,
         };
         let state1 = bootstrap(&config1, &catalog1).unwrap();
 
@@ -497,6 +498,7 @@ mod tests {
             data_dir: _dir2.path().to_path_buf(),
             force_bootstrap: false,
             join_retry: Default::default(),
+            swim_udp_addr: None,
         };
 
         let lifecycle = ClusterLifecycleTracker::new();

nodedb-cluster/src/bootstrap/probe.rs

@@ -222,6 +222,7 @@ mod tests {
             data_dir: std::env::temp_dir(),
             force_bootstrap: false,
             join_retry: Default::default(),
+            swim_udp_addr: None,
         }
     }
 

nodedb-cluster/src/bootstrap/restart.rs

@@ -111,6 +111,7 @@ mod tests {
             data_dir: _dir.path().to_path_buf(),
             force_bootstrap: false,
             join_retry: Default::default(),
+            swim_udp_addr: None,
         };
 
         // Bootstrap first.

nodedb-cluster/src/lib.rs

@@ -78,4 +78,7 @@ pub use lifecycle::{
 pub use rdma_transport::{RdmaConfig, RdmaTransport};
 pub use rebalance_scheduler::{NodeMetrics, RebalanceScheduler, RebalanceTrigger, SchedulerConfig};
 pub use shard_split::{SplitPlan, SplitStrategy, plan_graph_split, plan_vector_split};
-pub use swim::{Incarnation, Member, MemberState, MembershipList, SwimConfig, SwimError};
+pub use swim::{
+    Incarnation, Member, MemberState, MembershipList, SwimConfig, SwimError, SwimHandle,
+    UdpTransport, spawn as spawn_swim,
+};