From 9e964e28cbf6d42b722e6970d20af0817d7bc522 Mon Sep 17 00:00:00 2001
From: Mikhail <misha_mst@mail.ru>
Date: Sat, 16 May 2026 11:59:27 +0300
Subject: [PATCH] Emit capacity endpoint observations

---
 .../rap-node-agent/cmd/rap-node-agent/main.go | 20 +++++++++++++++++++
 .../cmd/rap-node-agent/main_test.go           | 14 +++++++++++++
 .../DISTRIBUTED_FABRIC_NODE_PROTOCOL_PLAN.md  |  3 +++
 3 files changed, 37 insertions(+)

diff --git a/agents/rap-node-agent/cmd/rap-node-agent/main.go b/agents/rap-node-agent/cmd/rap-node-agent/main.go
index aa3b05c..1b2e662 100644
--- a/agents/rap-node-agent/cmd/rap-node-agent/main.go
+++ b/agents/rap-node-agent/cmd/rap-node-agent/main.go
@@ -576,6 +576,25 @@ func (s *vpnFabricEndpointObservationStore) ObserveFailure(endpointID string, re
 	s.pruneLocked(observation.ObservedAt, vpnFabricEndpointObservationMaxAge, maxVPNFabricEndpointObservationEntries)
 }
 
+func (s *vpnFabricEndpointObservationStore) ObserveCapacity(endpointID string) {
+	if s == nil || strings.TrimSpace(endpointID) == "" {
+		return
+	}
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	observation := s.observations[endpointID]
+	observation.EndpointID = endpointID
+	observation.Source = "local_vpn_fabric_session"
+	observation.ReporterNodeID = s.reporterNodeID
+	observation.LastFailureReason = "capacity_limited"
+	if observation.ReliabilityScore <= 0 {
+		observation.ReliabilityScore = 90
+	}
+	observation.ObservedAt = time.Now().UTC()
+	s.observations[endpointID] = observation
+	s.pruneLocked(observation.ObservedAt, vpnFabricEndpointObservationMaxAge, maxVPNFabricEndpointObservationEntries)
+}
+
 func fabricTransportLabelIsQUIC(label string) bool {
 	switch strings.ToLower(strings.TrimSpace(label)) {
 	case "quic", "direct_quic", "udp_quic", "quic_udp":
@@ -5063,6 +5082,7 @@ func fabricSessionGatewayTransportForAssignment(ctx context.Context, identity st
 			reason := fabricSessionOpenFailureReason(err)
 			if reason == "capacity_limited" {
 				meshState.VPNFabricSessionDialStats.ObserveCapacityLimited(selectedTarget)
+				meshState.VPNFabricEndpointObservations.ObserveCapacity(selectedTarget.EndpointID)
 			} else {
 				meshState.VPNFabricSessionDialStats.ObserveCandidateFailure(reason)
 				meshState.VPNFabricEndpointObservations.ObserveFailure(selectedTarget.EndpointID, reason)
diff --git a/agents/rap-node-agent/cmd/rap-node-agent/main_test.go b/agents/rap-node-agent/cmd/rap-node-agent/main_test.go
index 10418cb..9c92c73 100644
--- a/agents/rap-node-agent/cmd/rap-node-agent/main_test.go
+++ b/agents/rap-node-agent/cmd/rap-node-agent/main_test.go
@@ -851,6 +851,20 @@ func TestVPNFabricEndpointObservationStoreTagsLocalSource(t *testing.T) {
 	}
 }
 
+func TestVPNFabricEndpointObservationStoreRecordsCapacityWithoutFailure(t *testing.T) {
+	store := newVPNFabricEndpointObservationStore("node-a")
+	store.ObserveCapacity("endpoint-a")
+	snapshot := store.Snapshot()
+	observation := snapshot["endpoint-a"]
+	if observation.LastFailureReason != "capacity_limited" ||
+		observation.FailureCount != 0 ||
+		observation.ReliabilityScore != 90 ||
+		observation.Source != "local_vpn_fabric_session" ||
+		observation.ReporterNodeID != "node-a" {
+		t.Fatalf("unexpected capacity observation: %+v", observation)
+	}
+}
+
 func TestVPNFabricEndpointObservationStorePrunesOldAndExcessEntries(t *testing.T) {
 	store := newVPNFabricEndpointObservationStore()
 	now := time.Now().UTC()
diff --git a/docs/architecture/DISTRIBUTED_FABRIC_NODE_PROTOCOL_PLAN.md b/docs/architecture/DISTRIBUTED_FABRIC_NODE_PROTOCOL_PLAN.md
index 43ffc49..7ff927c 100644
--- a/docs/architecture/DISTRIBUTED_FABRIC_NODE_PROTOCOL_PLAN.md
+++ b/docs/architecture/DISTRIBUTED_FABRIC_NODE_PROTOCOL_PLAN.md
@@ -377,6 +377,9 @@ observations.
 Endpoint ranking treats `capacity_limited` observations as a soft pressure
 penalty instead of a hard recent failure, enabling load spreading without
 marking the carrier unhealthy.
+Local QUIC stream-limit pressure is now emitted as a capacity observation with
+no failure-count increment, allowing control plane to spread load without
+treating saturation as packet-path breakage.
 Cached QUIC carrier idle TTL is configurable through
 `RAP_VPN_FABRIC_QUIC_IDLE_TTL_SECONDS` / `-vpn-fabric-quic-idle-ttl` and
 propagated by host-agent install profiles.