Skip recursive discovery query if no useful ENRs

beacon_node/lighthouse_network/src/metrics.rs

@@ -77,6 +77,12 @@ pub static DISCOVERY_SESSIONS: LazyLock<Result<IntGauge>> = LazyLock::new(|| {
         "The number of active discovery sessions with peers",
     )
 });
+pub static DISCOVERY_NO_USEFUL_ENRS: LazyLock<Result<IntCounter>> = LazyLock::new(|| {
+    try_create_int_counter(
+        "discovery_no_useful_enrs_found",
+        "Total number of counts a query returned no useful ENRs to dial",
+    )
+});
 
 pub static PEERS_PER_CLIENT: LazyLock<Result<IntGaugeVec>> = LazyLock::new(|| {
     try_create_int_gauge_vec(

beacon_node/lighthouse_network/src/peer_manager/mod.rs

@@ -323,6 +323,7 @@ impl<E: EthSpec> PeerManager<E> {
     #[allow(clippy::mutable_key_type)]
     pub fn peers_discovered(&mut self, results: HashMap<Enr, Option<Instant>>) {
         let mut to_dial_peers = 0;
+        let results_count = results.len();
         let connected_or_dialing = self.network_globals.connected_or_dialing_peers();
         for (enr, min_ttl) in results {
             // There are two conditions in deciding whether to dial this peer.
@@ -354,8 +355,19 @@ impl<E: EthSpec> PeerManager<E> {
             }
         }
 
-        // Queue another discovery if we need to
-        self.maintain_peer_count(to_dial_peers);
+        // The heartbeat will attempt new discovery queries every N seconds if the node needs more
+        // peers. As an optimization, this function can recursively trigger new discovery queries
+        // immediatelly if we don't fulfill our peers needs after completing a query. This
+        // recursiveness results in an infinite loop in networks where there not enough peers to
+        // reach out target. To prevent the infinite loop, if a query returns no useful peers, we
+        // will cancel the recursiveness and wait for the heartbeat to trigger another query latter.
+        if results_count > 0 && to_dial_peers == 0 {
+            debug!(self.log, "Skipping recursive discovery query after finding no useful results"; "results" => results_count);
+            metrics::inc_counter(&metrics::DISCOVERY_NO_USEFUL_ENRS);
+        } else {
+            // Queue another discovery if we need to
+            self.maintain_peer_count(to_dial_peers);
+        }
     }
 
     /// A STATUS message has been received from a peer. This resets the status timer.