fix(raft): deduplicate votes by peer ID and make election timeout configurable

Use a HashSet keyed by peer ID in handle_request_vote_response so that
duplicate grants from the same peer cannot inflate the vote count and
cause premature leader promotion.

Add election_timeout_min/max fields to ClusterConfig and propagate them
through bootstrap, join, and restart paths. Expose matching fields in
ClusterTransportTuning (default 2–5 s) and wire them into MultiRaft so
every startup path honours the configured window rather than using a
hardcoded constant. Adjust network tuning defaults from 60/120 s to 2/5 s
to match real-world cluster behaviour.

Author: farhan-syah

nodedb-cluster/src/bootstrap/bootstrap_fn.rs

@@ -35,7 +35,8 @@ pub(super) fn bootstrap(config: &ClusterConfig, catalog: &ClusterCatalog) -> Res
     );
 
     // Create MultiRaft with all groups (single-node, no peers).
-    let mut multi_raft = MultiRaft::new(config.node_id, routing.clone(), config.data_dir.clone());
+    let mut multi_raft = MultiRaft::new(config.node_id, routing.clone(), config.data_dir.clone())
+        .with_election_timeout(config.election_timeout_min, config.election_timeout_max);
     for group_id in routing.group_ids() {
         multi_raft.add_group(group_id, vec![])?;
     }
@@ -81,6 +82,7 @@ fn generate_cluster_id() -> u64 {
 mod tests {
     use super::*;
     use crate::catalog::ClusterCatalog;
+    use std::time::Duration;
 
     fn temp_catalog() -> (tempfile::TempDir, ClusterCatalog) {
         let dir = tempfile::tempdir().unwrap();
@@ -102,6 +104,8 @@ mod tests {
             force_bootstrap: false,
             join_retry: Default::default(),
             swim_udp_addr: None,
+            election_timeout_min: Duration::from_millis(150),
+            election_timeout_max: Duration::from_millis(300),
         };
 
         let state = bootstrap(&config, &catalog).unwrap();

nodedb-cluster/src/bootstrap/config.rs

@@ -91,6 +91,10 @@ pub struct ClusterConfig {
     /// [`crate::spawn_swim`] after the cluster is up and feed the
     /// seed list from `seed_nodes`.
     pub swim_udp_addr: Option<SocketAddr>,
+    /// Raft election timeout range. Controls how long a follower waits
+    /// before starting an election after losing contact with the leader.
+    pub election_timeout_min: Duration,
+    pub election_timeout_max: Duration,
 }
 
 /// Result of cluster startup — everything needed to run the Raft loop.

nodedb-cluster/src/bootstrap/join.rs

@@ -288,7 +288,8 @@ fn apply_join_response(
     //    learners). A learner-started group boots in the `Learner`
     //    role and will not run an election until a subsequent
     //    `PromoteLearner` conf change is applied.
-    let mut multi_raft = MultiRaft::new(config.node_id, routing.clone(), config.data_dir.clone());
+    let mut multi_raft = MultiRaft::new(config.node_id, routing.clone(), config.data_dir.clone())
+        .with_election_timeout(config.election_timeout_min, config.election_timeout_max);
     for g in &resp.groups {
         let is_voter = g.members.contains(&config.node_id);
         let is_learner = g.learners.contains(&config.node_id);
@@ -450,6 +451,8 @@ mod tests {
             force_bootstrap: false,
             join_retry: Default::default(),
             swim_udp_addr: None,
+            election_timeout_min: Duration::from_millis(150),
+            election_timeout_max: Duration::from_millis(300),
         };
         let state1 = bootstrap(&config1, &catalog1).unwrap();
 
@@ -499,6 +502,8 @@ mod tests {
             force_bootstrap: false,
             join_retry: Default::default(),
             swim_udp_addr: None,
+            election_timeout_min: Duration::from_millis(150),
+            election_timeout_max: Duration::from_millis(300),
         };
 
         let lifecycle = ClusterLifecycleTracker::new();

nodedb-cluster/src/bootstrap/probe.rs

@@ -223,6 +223,8 @@ mod tests {
             force_bootstrap: false,
             join_retry: Default::default(),
             swim_udp_addr: None,
+            election_timeout_min: Duration::from_millis(150),
+            election_timeout_max: Duration::from_millis(300),
         }
     }
 

nodedb-cluster/src/bootstrap/restart.rs

@@ -35,7 +35,8 @@ pub(super) fn restart(
     // as a learner on restart; dropping the group entirely would
     // leave the node permanently without any copy of it and
     // silently broken.
-    let mut multi_raft = MultiRaft::new(config.node_id, routing.clone(), config.data_dir.clone());
+    let mut multi_raft = MultiRaft::new(config.node_id, routing.clone(), config.data_dir.clone())
+        .with_election_timeout(config.election_timeout_min, config.election_timeout_max);
     for (group_id, info) in routing.group_members() {
         let is_voter = info.members.contains(&config.node_id);
         let is_learner = info.learners.contains(&config.node_id);
@@ -91,6 +92,7 @@ mod tests {
     use super::super::bootstrap_fn::bootstrap;
     use super::*;
     use crate::catalog::ClusterCatalog;
+    use std::time::Duration;
 
     fn temp_catalog() -> (tempfile::TempDir, ClusterCatalog) {
         let dir = tempfile::tempdir().unwrap();
@@ -112,6 +114,8 @@ mod tests {
             force_bootstrap: false,
             join_retry: Default::default(),
             swim_udp_addr: None,
+            election_timeout_min: Duration::from_millis(150),
+            election_timeout_max: Duration::from_millis(300),
         };
 
         // Bootstrap first.

nodedb-cluster/src/multi_raft/core.rs

@@ -77,8 +77,8 @@ impl MultiRaft {
             node_id,
             groups: HashMap::new(),
             routing,
-            election_timeout_min: Duration::from_millis(150),
-            election_timeout_max: Duration::from_millis(300),
+            election_timeout_min: Duration::from_secs(2),
+            election_timeout_max: Duration::from_secs(5),
             heartbeat_interval: Duration::from_millis(50),
             data_dir,
         }

nodedb-raft/src/node/core.rs

@@ -5,6 +5,7 @@
 //! replication) live in [`super::internal`]. RPC handlers live in
 //! [`super::rpc`].
 
+use std::collections::HashSet;
 use std::time::Instant;
 
 use crate::error::{RaftError, Result};
@@ -61,7 +62,7 @@ pub struct RaftNode<S: LogStorage> {
     /// When the next heartbeat should be sent (leader only).
     pub(super) heartbeat_deadline: Instant,
     /// Votes received in current election.
-    pub(super) votes_received: Vec<u64>,
+    pub(super) votes_received: HashSet<u64>,
     /// Pending ready output.
     pub(super) ready: Ready,
     /// Known leader ID (0 = unknown).
@@ -89,7 +90,7 @@ impl<S: LogStorage> RaftNode<S> {
             leader_state: None,
             election_deadline: now + config.election_timeout_max,
             heartbeat_deadline: now,
-            votes_received: Vec::new(),
+            votes_received: HashSet::new(),
             ready: Ready::default(),
             leader_id: 0,
             config,

nodedb-raft/src/node/rpc.rs

@@ -166,7 +166,7 @@ impl<S: LogStorage> RaftNode<S> {
     }
 
     /// Handle RequestVote response (candidate only).
-    pub fn handle_request_vote_response(&mut self, _peer: u64, resp: &RequestVoteResponse) {
+    pub fn handle_request_vote_response(&mut self, peer: u64, resp: &RequestVoteResponse) {
         if resp.term > self.hard_state.current_term {
             self.become_follower(resp.term);
             return;
@@ -177,7 +177,7 @@ impl<S: LogStorage> RaftNode<S> {
         }
 
         if resp.vote_granted {
-            self.votes_received.push(resp.term);
+            self.votes_received.insert(peer);
             let vote_count = self.votes_received.len() + 1; // +1 for self-vote
 
             if vote_count >= self.config.quorum() {

nodedb-types/src/config/tuning/network.rs

@@ -223,10 +223,10 @@ fn default_raft_tick_interval_ms() -> u64 {
     10
 }
 fn default_election_timeout_min_secs() -> u64 {
-    60
+    2
 }
 fn default_election_timeout_max_secs() -> u64 {
-    120
+    5
 }
 fn default_rpc_timeout_secs() -> u64 {
     5