Merge branch 'main' into dependency-docs

go.mod

@@ -9,14 +9,14 @@ require (
 	github.com/ava-labs/teleporter v1.0.5
 	github.com/aws/aws-sdk-go-v2 v1.30.4
 	github.com/aws/aws-sdk-go-v2/config v1.27.9
-	github.com/aws/aws-sdk-go-v2/service/kms v1.35.4
+	github.com/aws/aws-sdk-go-v2/service/kms v1.35.5
 	github.com/ethereum/go-ethereum v1.13.8
 	github.com/hashicorp/golang-lru/v2 v2.0.7
 	github.com/onsi/ginkgo/v2 v2.20.1
 	github.com/onsi/gomega v1.34.1
 	github.com/pingcap/errors v0.11.4
 	github.com/pkg/errors v0.9.1
-	github.com/prometheus/client_golang v1.20.1
+	github.com/prometheus/client_golang v1.20.2
 	github.com/redis/go-redis/v9 v9.6.1
 	github.com/spf13/pflag v1.0.5
 	github.com/spf13/viper v1.16.0

go.sum

@@ -84,8 +84,8 @@ github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.11.1 h1:EyBZibR
 github.com/aws/aws-sdk-go-v2/service/internal/accept-encoding v1.11.1/go.mod h1:JKpmtYhhPs7D97NL/ltqz7yCkERFW5dOlHyVl66ZYF8=
 github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.11.6 h1:b+E7zIUHMmcB4Dckjpkapoy47W6C9QBv/zoUP+Hn8Kc=
 github.com/aws/aws-sdk-go-v2/service/internal/presigned-url v1.11.6/go.mod h1:S2fNV0rxrP78NhPbCZeQgY8H9jdDMeGtwcfZIRxzBqU=
-github.com/aws/aws-sdk-go-v2/service/kms v1.35.4 h1:mG1MH6yPwT5gNEeBrhig3FHc4mK0QaZOXsmQUbphP6Y=
-github.com/aws/aws-sdk-go-v2/service/kms v1.35.4/go.mod h1:A5CS0VRmxxj2YKYLCY08l/Zzbd01m6JZn0WzxgT1OCA=
+github.com/aws/aws-sdk-go-v2/service/kms v1.35.5 h1:XUomV7SiclZl1QuXORdGcfFqHxEHET7rmNGtxTfNB+M=
+github.com/aws/aws-sdk-go-v2/service/kms v1.35.5/go.mod h1:A5CS0VRmxxj2YKYLCY08l/Zzbd01m6JZn0WzxgT1OCA=
 github.com/aws/aws-sdk-go-v2/service/sso v1.20.3 h1:mnbuWHOcM70/OFUlZZ5rcdfA8PflGXXiefU/O+1S3+8=
 github.com/aws/aws-sdk-go-v2/service/sso v1.20.3/go.mod h1:5HFu51Elk+4oRBZVxmHrSds5jFXmFj8C3w7DVF2gnrs=
 github.com/aws/aws-sdk-go-v2/service/ssooidc v1.23.3 h1:uLq0BKatTmDzWa/Nu4WO0M1AaQDaPpwTKAeByEc6WFM=
@@ -518,8 +518,8 @@ github.com/pkg/errors v0.9.1/go.mod h1:bwawxfHBFNV+L2hUp1rHADufV3IMtnDRdf1r5NINE
 github.com/pkg/sftp v1.13.1/go.mod h1:3HaPG6Dq1ILlpPZRO0HVMrsydcdLt6HRDccSgb87qRg=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/prometheus/client_golang v1.20.1 h1:IMJXHOD6eARkQpxo8KkhgEVFlBNm+nkrFUyGlIu7Na8=
-github.com/prometheus/client_golang v1.20.1/go.mod h1:PIEt8X02hGcP8JWbeHyeZ53Y/jReSnHgO035n//V5WE=
+github.com/prometheus/client_golang v1.20.2 h1:5ctymQzZlyOON1666svgwn3s6IKWgfbjsejTMiXIyjg=
+github.com/prometheus/client_golang v1.20.2/go.mod h1:PIEt8X02hGcP8JWbeHyeZ53Y/jReSnHgO035n//V5WE=
 github.com/prometheus/client_model v0.0.0-20190812154241-14fe0d1b01d4/go.mod h1:xMI15A0UPsDsEKsMN9yxemIoYk6Tm2C1GtYGdfGttqA=
 github.com/prometheus/client_model v0.6.1 h1:ZKSh/rekM+n3CeS952MLRAdFwIKqeY8b62p8ais2e9E=
 github.com/prometheus/client_model v0.6.1/go.mod h1:OrxVMOVHjw3lKMa8+x6HeMGkHMQyHDk9E3jmP2AmGiY=

peers/app_request_network.go

@@ -61,7 +61,7 @@ func NewNetwork(
 	}
 
 	// Create the handler for handling inbound app responses
-	handler, err := NewRelayerExternalHandler(logger, prometheus.DefaultRegisterer)
+	handler, err := NewRelayerExternalHandler(logger, metrics)
 	if err != nil {
 		logger.Error(
 			"Failed to create p2p network handler",

peers/app_request_netrwork_metrics.go → peers/app_request_network_metrics.go

@@ -13,6 +13,8 @@ var (
 type AppRequestNetworkMetrics struct {
 	infoAPICallLatencyMS   prometheus.Histogram
 	pChainAPICallLatencyMS prometheus.Histogram
+	connects               prometheus.Counter
+	disconnects            prometheus.Counter
 }
 
 func newAppRequestNetworkMetrics(registerer prometheus.Registerer) (*AppRequestNetworkMetrics, error) {
@@ -40,8 +42,32 @@ func newAppRequestNetworkMetrics(registerer prometheus.Registerer) (*AppRequestN
 	}
 	registerer.MustRegister(pChainAPICallLatencyMS)
 
+	connects := prometheus.NewCounter(
+		prometheus.CounterOpts{
+			Name: "connects",
+			Help: "Number of connected events",
+		},
+	)
+	if connects == nil {
+		return nil, ErrFailedToCreateAppRequestNetworkMetrics
+	}
+	registerer.MustRegister(connects)
+
+	disconnects := prometheus.NewCounter(
+		prometheus.CounterOpts{
+			Name: "disconnects",
+			Help: "Number of disconnected events",
+		},
+	)
+	if disconnects == nil {
+		return nil, ErrFailedToCreateAppRequestNetworkMetrics
+	}
+	registerer.MustRegister(disconnects)
+
 	return &AppRequestNetworkMetrics{
 		infoAPICallLatencyMS:   infoAPICallLatencyMS,
 		pChainAPICallLatencyMS: pChainAPICallLatencyMS,
+		connects:               connects,
+		disconnects:            disconnects,
 	}, nil
 }

peers/external_handler.go

@@ -30,6 +30,7 @@ type RelayerExternalHandler struct {
 	responseChans  map[uint32]chan message.InboundMessage
 	responsesCount map[uint32]expectedResponses
 	timeoutManager timer.AdaptiveTimeoutManager
+	metrics        *AppRequestNetworkMetrics
 }
 
 // expectedResponses counts the number of responses and compares against the expected number of responses
@@ -41,7 +42,7 @@ type expectedResponses struct {
 // Teleporter application relayer, as well as handle timeouts.
 func NewRelayerExternalHandler(
 	logger logging.Logger,
-	registerer prometheus.Registerer,
+	metrics *AppRequestNetworkMetrics,
 ) (*RelayerExternalHandler, error) {
 	// TODO: Leaving this static for now, but we may want to have this as a config option
 	cfg := timer.AdaptiveTimeoutConfig{
@@ -52,7 +53,7 @@ func NewRelayerExternalHandler(
 		TimeoutHalflife:    constants.DefaultNetworkTimeoutHalflife,
 	}
 
-	timeoutManager, err := timer.NewAdaptiveTimeoutManager(&cfg, registerer)
+	timeoutManager, err := timer.NewAdaptiveTimeoutManager(&cfg, prometheus.DefaultRegisterer)
 	if err != nil {
 		logger.Error(
 			"Failed to create timeout manager",
@@ -69,6 +70,7 @@ func NewRelayerExternalHandler(
 		responseChans:  make(map[uint32]chan message.InboundMessage),
 		responsesCount: make(map[uint32]expectedResponses),
 		timeoutManager: timeoutManager,
+		metrics:        metrics,
 	}, nil
 }
 
@@ -102,13 +104,15 @@ func (h *RelayerExternalHandler) Connected(nodeID ids.NodeID, version *version.A
 		zap.Stringer("version", version),
 		zap.Stringer("subnetID", subnetID),
 	)
+	h.metrics.connects.Inc()
 }
 
 func (h *RelayerExternalHandler) Disconnected(nodeID ids.NodeID) {
 	h.log.Debug(
 		"Disconnected",
 		zap.Stringer("nodeID", nodeID),
 	)
+	h.metrics.disconnects.Inc()
 }
 
 // RegisterRequestID registers an AppRequest by requestID, and marks the number of

relayer/main/main.go

@@ -143,6 +143,9 @@ func main() {
 	// TODO: remove this from here once trackedSubnets are no longer referenced
 	// by ping messages in avalanchego
 	for _, sourceBlockchain := range cfg.SourceBlockchains {
+		if sourceBlockchain.GetSubnetID() == constants.PrimaryNetworkID {
+			continue
+		}
 		trackedSubnets.Add(sourceBlockchain.GetSubnetID())
 	}
 

signature-aggregator/aggregator/aggregator.go

@@ -121,6 +121,12 @@ func (s *SignatureAggregator) CreateSignedMessage(
 		s.metrics.FailuresToGetValidatorSet.Inc()
 		return nil, fmt.Errorf("%s: %w", msg, err)
 	}
+	s.metrics.ConnectedStakeWeightPercentage.WithLabelValues(
+		signingSubnet.String(),
+	).Set(
+		float64(connectedValidators.ConnectedWeight) /
+			float64(connectedValidators.TotalValidatorWeight) * 100,
+	)
 
 	if !utils.CheckStakeWeightPercentageExceedsThreshold(
 		big.NewInt(0).SetUint64(connectedValidators.ConnectedWeight),

signature-aggregator/metrics/metrics.go

@@ -31,6 +31,7 @@ var Opts = struct {
 	InvalidSignatureResponses          prometheus.CounterOpts
 	SignatureCacheHits                 prometheus.CounterOpts
 	SignatureCacheMisses               prometheus.CounterOpts
+	ConnectedStakeWeightPercentage     prometheus.GaugeOpts
 }{
 	AggregateSignaturesLatencyMS: prometheus.GaugeOpts{
 		Name: "agg_sigs_latency_ms",
@@ -72,6 +73,10 @@ var Opts = struct {
 		Name: "signature_cache_misses",
 		Help: "Number of signatures that were not found in the cache",
 	},
+	ConnectedStakeWeightPercentage: prometheus.GaugeOpts{
+		Name: "connected_stake_weight_percentage",
+		Help: "The percentage of connected stake weight for a specific subnet",
+	},
 }
 
 type SignatureAggregatorMetrics struct {
@@ -85,6 +90,7 @@ type SignatureAggregatorMetrics struct {
 	InvalidSignatureResponses          prometheus.Counter
 	SignatureCacheHits                 prometheus.Counter
 	SignatureCacheMisses               prometheus.Counter
+	ConnectedStakeWeightPercentage     *prometheus.GaugeVec
 
 	// TODO: consider other failures to monitor. Issue #384 requires
 	// "network failures", but we probably don't handle those directly.
@@ -129,6 +135,10 @@ func NewSignatureAggregatorMetrics(
 		SignatureCacheMisses: prometheus.NewCounter(
 			Opts.SignatureCacheMisses,
 		),
+		ConnectedStakeWeightPercentage: prometheus.NewGaugeVec(
+			Opts.ConnectedStakeWeightPercentage,
+			[]string{"subnetID"},
+		),
 	}
 
 	registerer.MustRegister(m.AggregateSignaturesLatencyMS)
@@ -141,6 +151,7 @@ func NewSignatureAggregatorMetrics(
 	registerer.MustRegister(m.InvalidSignatureResponses)
 	registerer.MustRegister(m.SignatureCacheHits)
 	registerer.MustRegister(m.SignatureCacheMisses)
+	registerer.MustRegister(m.ConnectedStakeWeightPercentage)
 
 	return &m
 }

tests/signature_aggregator_api.go

@@ -137,6 +137,15 @@ func SignatureAggregatorAPI(network interfaces.LocalNetwork) {
 		{metrics.Opts.InvalidSignatureResponses.Name, "==", 0},
 		{metrics.Opts.SignatureCacheHits.Name, "==", 0},
 		{metrics.Opts.SignatureCacheMisses.Name, "==", 0},
+		{
+			fmt.Sprintf(
+				"%s{subnetID=\"%s\"}",
+				metrics.Opts.ConnectedStakeWeightPercentage.Name,
+				subnetAInfo.SubnetID.String(),
+			),
+			"==",
+			100,
+		},
 	} {
 		Expect(metricsSample[m.name]).Should(
 			BeNumerically(m.op, m.value),
@@ -191,6 +200,7 @@ func sampleMetrics(port uint16) map[string]uint64 {
 			metrics.Opts.InvalidSignatureResponses.Name,
 			metrics.Opts.SignatureCacheHits.Name,
 			metrics.Opts.SignatureCacheMisses.Name,
+			metrics.Opts.ConnectedStakeWeightPercentage.Name,
 		} {
 			if strings.HasPrefix(
 				line,
@@ -199,7 +209,9 @@ func sampleMetrics(port uint16) map[string]uint64 {
 				log.Debug("Found metric line", "line", line)
 				parts := strings.Fields(line)
 
-				// Fetch the metric count from the last field of the line
+				metricName = strings.Replace(parts[0], "U__signature_2d_aggregator_", "", 1)
+
+				// Parse the metric count from the last field of the line
 				value, err := strconv.ParseUint(parts[len(parts)-1], 10, 64)
 				if err != nil {
 					log.Warn("failed to parse value from metric line")