add pubsub raw tracer for metrics

network/p2p/pubsub.go

@@ -98,6 +98,7 @@ func makePubSub(ctx context.Context, cfg config.Local, host host.Host) (*pubsub.
 		pubsub.WithMessageSignaturePolicy(pubsub.StrictNoSign),
 		// pubsub.WithValidateThrottle(cfg.TxBacklogSize),
 		pubsub.WithValidateWorkers(incomingThreads),
+		pubsub.WithRawTracer(pubsubTracer{}),
 	}
 
 	return pubsub.NewGossipSub(ctx, host, options...)

network/p2p/pubsubTracer.go

@@ -0,0 +1,86 @@
+// Copyright (C) 2019-2024 Algorand, Inc.
+// This file is part of go-algorand
+//
+// go-algorand is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as
+// published by the Free Software Foundation, either version 3 of the
+// License, or (at your option) any later version.
+//
+// go-algorand is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with go-algorand.  If not, see <https://www.gnu.org/licenses/>.
+
+package p2p
+
+import (
+	pubsub "github.com/libp2p/go-libp2p-pubsub"
+	"github.com/libp2p/go-libp2p/core/peer"
+	"github.com/libp2p/go-libp2p/core/protocol"
+
+	"github.com/algorand/go-algorand/util/metrics"
+)
+
+var _ = pubsub.RawTracer(pubsubTracer{})
+
+var transactionMessagesDroppedFromBacklog = metrics.MakeCounter(metrics.TransactionMessagesDroppedFromBacklog)
+var transactionMessagesDupRawMsg = metrics.MakeCounter(metrics.TransactionMessagesDupRawMsg)
+
+// pubsubTracer is a tracer for pubsub events used to track metrics.
+type pubsubTracer struct{}
+
+// AddPeer is invoked when a new peer is added.
+func (t pubsubTracer) AddPeer(p peer.ID, proto protocol.ID) {}
+
+// RemovePeer is invoked when a peer is removed.
+func (t pubsubTracer) RemovePeer(p peer.ID) {}
+
+// Join is invoked when a new topic is joined
+func (t pubsubTracer) Join(topic string) {}
+
+// Leave is invoked when a topic is abandoned
+func (t pubsubTracer) Leave(topic string) {}
+
+// Graft is invoked when a new peer is grafted on the mesh (gossipsub)
+func (t pubsubTracer) Graft(p peer.ID, topic string) {}
+
+// Prune is invoked when a peer is pruned from the message (gossipsub)
+func (t pubsubTracer) Prune(p peer.ID, topic string) {}
+
+// ValidateMessage is invoked when a message first enters the validation pipeline.
+func (t pubsubTracer) ValidateMessage(msg *pubsub.Message) {}
+
+// DeliverMessage is invoked when a message is delivered
+func (t pubsubTracer) DeliverMessage(msg *pubsub.Message) {}
+
+// RejectMessage is invoked when a message is Rejected or Ignored.
+// The reason argument can be one of the named strings Reject*.
+func (t pubsubTracer) RejectMessage(msg *pubsub.Message, reason string) {
+	if reason == pubsub.RejectValidationThrottled || reason == pubsub.RejectValidationQueueFull {
+		transactionMessagesDroppedFromBacklog.Inc(nil)
+	}
+}
+
+// DuplicateMessage is invoked when a duplicate message is dropped.
+func (t pubsubTracer) DuplicateMessage(msg *pubsub.Message) {
+	transactionMessagesDupRawMsg.Inc(nil)
+}
+
+// ThrottlePeer is invoked when a peer is throttled by the peer gater.
+func (t pubsubTracer) ThrottlePeer(p peer.ID) {}
+
+// RecvRPC is invoked when an incoming RPC is received.
+func (t pubsubTracer) RecvRPC(rpc *pubsub.RPC) {}
+
+// SendRPC is invoked when a RPC is sent.
+func (t pubsubTracer) SendRPC(rpc *pubsub.RPC, p peer.ID) {}
+
+// DropRPC is invoked when an outbound RPC is dropped, typically because of a queue full.
+func (t pubsubTracer) DropRPC(rpc *pubsub.RPC, p peer.ID) {}
+
+// UndeliverableMessage is invoked when the consumer of Subscribe is not reading messages fast enough and
+// the pressure release mechanism trigger, dropping messages.
+func (t pubsubTracer) UndeliverableMessage(msg *pubsub.Message) {}

network/wsNetwork.go

@@ -124,6 +124,7 @@ var networkIncomingBufferMicros = metrics.MakeCounter(metrics.MetricName{Name: "
 var networkHandleMicros = metrics.MakeCounter(metrics.MetricName{Name: "algod_network_rx_handle_micros_total", Description: "microseconds spent by protocol handlers in the receive thread"})
 
 var networkBroadcasts = metrics.MakeCounter(metrics.MetricName{Name: "algod_network_broadcasts_total", Description: "number of broadcast operations"})
+var networkBroadcastQueueFull = metrics.MakeCounter(metrics.MetricName{Name: "algod_network_broadcast_queue_full_total", Description: "number of messages that were drops due to full broadcast queue"})
 var networkBroadcastQueueMicros = metrics.MakeCounter(metrics.MetricName{Name: "algod_network_broadcast_queue_micros_total", Description: "microseconds broadcast requests sit on queue"})
 var networkBroadcastSendMicros = metrics.MakeCounter(metrics.MetricName{Name: "algod_network_broadcast_send_micros_total", Description: "microseconds spent broadcasting"})
 var networkBroadcastsDropped = metrics.MakeCounter(metrics.MetricName{Name: "algod_broadcasts_dropped_total", Description: "number of broadcast messages not sent to any peer"})
@@ -135,7 +136,6 @@ var networkPeerAlreadyClosed = metrics.MakeCounter(metrics.MetricName{Name: "alg
 
 var networkSlowPeerDrops = metrics.MakeCounter(metrics.MetricName{Name: "algod_network_slow_drops_total", Description: "number of peers dropped for being slow to send to"})
 var networkIdlePeerDrops = metrics.MakeCounter(metrics.MetricName{Name: "algod_network_idle_drops_total", Description: "number of peers dropped due to idle connection"})
-var networkBroadcastQueueFull = metrics.MakeCounter(metrics.MetricName{Name: "algod_network_broadcast_queue_full_total", Description: "number of messages that were drops due to full broadcast queue"})
 
 var minPing = metrics.MakeGauge(metrics.MetricName{Name: "algod_network_peer_min_ping_seconds", Description: "Network round trip time to fastest peer in seconds."})
 var meanPing = metrics.MakeGauge(metrics.MetricName{Name: "algod_network_peer_mean_ping_seconds", Description: "Network round trip time to average peer in seconds."})