Merge pull request #655 from smartcontractkit/feature/add-v2-txm-metrics

Feature/add v2 txm metrics

Author: cawthorne

relayer/CONFIG.md

@@ -21,6 +21,7 @@ OCR2CacheTTL = '1m' # Default
 RequestTimeout = '10s' # Default
 TxTimeout = '10s' # Default
 ConfirmationPoll = '5s' # Default
+FeeEstimationMaxAttempts = 5 # Default
 ```
 
 
@@ -72,6 +73,12 @@ ConfirmationPoll = '5s' # Default
 ```
 ConfirmationPoll is how often to confirmer checks for tx inclusion on chain.
 
+### FeeEstimationMaxAttempts
+```toml
+FeeEstimationMaxAttempts = 5 # Default
+```
+FeeEstimationMaxAttempts is the maximum number of retry attempts for fee estimation.
+
 ## Nodes
 ```toml
 [[Nodes]]

relayer/go.mod

@@ -8,10 +8,13 @@ require (
 	github.com/ethereum/go-ethereum v1.15.0
 	github.com/hashicorp/go-plugin v1.6.2
 	github.com/pelletier/go-toml/v2 v2.2.3
+	github.com/prometheus/client_golang v1.21.1
 	github.com/smartcontractkit/chainlink-common v0.7.1-0.20250602141924-3c15a42d8266
 	github.com/smartcontractkit/freeport v0.1.0
 	github.com/smartcontractkit/libocr v0.0.0-20250220133800-f3b940c4f298
 	github.com/stretchr/testify v1.10.0
+	go.opentelemetry.io/otel v1.35.0
+	go.opentelemetry.io/otel/metric v1.35.0
 	go.uber.org/zap v1.27.0
 	golang.org/x/exp v0.0.0-20250218142911-aa4b98e5adaa
 )
@@ -98,7 +101,6 @@ require (
 	github.com/oklog/run v1.1.0 // indirect
 	github.com/pkg/errors v0.9.1 // indirect
 	github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2 // indirect
-	github.com/prometheus/client_golang v1.21.1 // indirect
 	github.com/prometheus/client_model v0.6.1 // indirect
 	github.com/prometheus/common v0.63.0 // indirect
 	github.com/prometheus/procfs v0.16.0 // indirect
@@ -120,7 +122,6 @@ require (
 	github.com/yusufpapurcu/wmi v1.2.4 // indirect
 	go.opentelemetry.io/auto/sdk v1.1.0 // indirect
 	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.53.0 // indirect
-	go.opentelemetry.io/otel v1.35.0 // indirect
 	go.opentelemetry.io/otel/exporters/otlp/otlplog/otlploggrpc v0.0.0-20240823153156-2a54df7bffb9 // indirect
 	go.opentelemetry.io/otel/exporters/otlp/otlplog/otlploghttp v0.6.0 // indirect
 	go.opentelemetry.io/otel/exporters/otlp/otlpmetric/otlpmetricgrpc v1.35.0 // indirect
@@ -132,7 +133,6 @@ require (
 	go.opentelemetry.io/otel/exporters/stdout/stdoutmetric v1.28.0 // indirect
 	go.opentelemetry.io/otel/exporters/stdout/stdouttrace v1.28.0 // indirect
 	go.opentelemetry.io/otel/log v0.6.0 // indirect
-	go.opentelemetry.io/otel/metric v1.35.0 // indirect
 	go.opentelemetry.io/otel/sdk v1.35.0 // indirect
 	go.opentelemetry.io/otel/sdk/log v0.6.0 // indirect
 	go.opentelemetry.io/otel/sdk/metric v1.35.0 // indirect

relayer/pkg/chainlink/chain/chain.go

@@ -9,6 +9,7 @@ import (
 	"strconv"
 
 	"github.com/pelletier/go-toml/v2"
+	"go.opentelemetry.io/otel/metric"
 
 	"github.com/smartcontractkit/chainlink-common/pkg/chains"
 	"github.com/smartcontractkit/chainlink-common/pkg/logger"
@@ -33,9 +34,9 @@ type Chain interface {
 }
 
 type ChainOpts struct {
-	Logger logger.Logger
-	// the implementation used here needs to be co-ordinated with the starknet transaction manager keystore adapter
-	KeyStore loop.Keystore
+	Logger   logger.Logger
+	KeyStore loop.Keystore // the implementation used here needs to be co-ordinated with the starknet transaction manager keystore adapter
+	Meter    metric.Meter  // OpenTelemetry meter for beholder metrics
 }
 
 func (o *ChainOpts) Name() string {
@@ -69,14 +70,14 @@ func NewChain(cfg *config.TOMLConfig, opts ChainOpts) (Chain, error) {
 	if !cfg.IsEnabled() {
 		return nil, fmt.Errorf("cannot create new chain with ID %s: chain is disabled", *cfg.ChainID)
 	}
-	c, err := newChain(*cfg.ChainID, cfg, opts.KeyStore, opts.Logger)
+	c, err := newChain(*cfg.ChainID, cfg, opts.KeyStore, opts.Logger, opts.Meter)
 	if err != nil {
 		return nil, err
 	}
 	return c, nil
 }
 
-func newChain(id string, cfg *config.TOMLConfig, loopKs loop.Keystore, lggr logger.Logger) (*chain, error) {
+func newChain(id string, cfg *config.TOMLConfig, loopKs loop.Keystore, lggr logger.Logger, meter metric.Meter) (*chain, error) {
 	lggr = logger.Named(lggr, "Chain")
 	lggr = logger.With(lggr, "starknetChainID", id)
 	ch := &chain{
@@ -86,7 +87,7 @@ func newChain(id string, cfg *config.TOMLConfig, loopKs loop.Keystore, lggr logg
 	}
 
 	var err error
-	ch.txm, err = txm.New(lggr, loopKs, cfg, ch.getClient, ch.getFeederClient)
+	ch.txm, err = txm.New(lggr, loopKs, cfg, ch.ChainID(), meter, ch.getClient, ch.getFeederClient)
 	if err != nil {
 		return nil, err
 	}

relayer/pkg/chainlink/cmd/chainlink-starknet/main.go

@@ -8,6 +8,7 @@ import (
 	"github.com/hashicorp/go-plugin"
 	"github.com/pelletier/go-toml/v2"
 
+	"github.com/smartcontractkit/chainlink-common/pkg/beholder"
 	"github.com/smartcontractkit/chainlink-common/pkg/loop"
 	"github.com/smartcontractkit/chainlink-common/pkg/types/core"
 
@@ -75,9 +76,13 @@ func (c *pluginRelayer) NewRelayer(ctx context.Context, config string, loopKs lo
 	}
 	c.Logger.Infow("Creating relayer", "config", cfgStr)
 
+	// Get beholder meter for injection into chain/txm
+	meter := beholder.GetMeter()
+
 	chain, err := starkchain.NewChain(&cfg, starkchain.ChainOpts{
 		Logger:   c.Logger,
 		KeyStore: loopKs,
+		Meter:    meter,
 	})
 	if err != nil {
 		return nil, fmt.Errorf("failed to create chain: %w", err)

relayer/pkg/chainlink/config/config.go

@@ -40,8 +40,9 @@ type ConfigSet struct { //nolint:revive
 	RequestTimeout time.Duration
 
 	// txm config
-	TxTimeout        time.Duration
-	ConfirmationPoll time.Duration
+	TxTimeout                time.Duration
+	ConfirmationPoll         time.Duration
+	FeeEstimationMaxAttempts int
 }
 
 type Config interface {
@@ -55,11 +56,12 @@ type Config interface {
 }
 
 type Chain struct {
-	OCR2CachePollPeriod *config.Duration
-	OCR2CacheTTL        *config.Duration
-	RequestTimeout      *config.Duration
-	TxTimeout           *config.Duration
-	ConfirmationPoll    *config.Duration
+	OCR2CachePollPeriod      *config.Duration
+	OCR2CacheTTL             *config.Duration
+	RequestTimeout           *config.Duration
+	TxTimeout                *config.Duration
+	ConfirmationPoll         *config.Duration
+	FeeEstimationMaxAttempts *int
 }
 
 type Node struct {
@@ -174,6 +176,9 @@ func setFromChain(c, f *Chain) {
 	if f.ConfirmationPoll != nil {
 		c.ConfirmationPoll = f.ConfirmationPoll
 	}
+	if f.FeeEstimationMaxAttempts != nil {
+		c.FeeEstimationMaxAttempts = f.FeeEstimationMaxAttempts
+	}
 }
 
 func (c *TOMLConfig) ValidateConfig() (err error) {
@@ -248,6 +253,13 @@ func (c *TOMLConfig) ConfirmationPoll() time.Duration {
 	return c.Chain.ConfirmationPoll.Duration()
 }
 
+func (c *TOMLConfig) FeeEstimationMaxAttempts() int {
+	if c.Chain.FeeEstimationMaxAttempts == nil {
+		return 5 // Default value for fee estimation
+	}
+	return *c.Chain.FeeEstimationMaxAttempts
+}
+
 func (c *TOMLConfig) OCR2CachePollPeriod() time.Duration {
 	return c.Chain.OCR2CachePollPeriod.Duration()
 }

relayer/pkg/chainlink/config/config_test.go

@@ -30,11 +30,12 @@ func TestTOMLConfig_FullMarshal(t *testing.T) {
 		Enabled:   ptr(true),
 		FeederURL: config.MustParseURL("http://feeder.url"),
 		Chain: Chain{
-			OCR2CachePollPeriod: config.MustNewDuration(6 * time.Hour),
-			OCR2CacheTTL:        config.MustNewDuration(3 * time.Minute),
-			RequestTimeout:      config.MustNewDuration(1*time.Minute + 3*time.Second),
-			TxTimeout:           config.MustNewDuration(13 * time.Second),
-			ConfirmationPoll:    config.MustNewDuration(42 * time.Second),
+			OCR2CachePollPeriod:      config.MustNewDuration(6 * time.Hour),
+			OCR2CacheTTL:             config.MustNewDuration(3 * time.Minute),
+			RequestTimeout:           config.MustNewDuration(1*time.Minute + 3*time.Second),
+			TxTimeout:                config.MustNewDuration(13 * time.Second),
+			ConfirmationPoll:         config.MustNewDuration(42 * time.Second),
+			FeeEstimationMaxAttempts: ptr(8),
 		},
 		Nodes: []*Node{
 			{

relayer/pkg/chainlink/config/docs.toml

@@ -14,6 +14,8 @@ RequestTimeout = '10s' # Default
 TxTimeout = '10s' # Default
 # ConfirmationPoll is how often to confirmer checks for tx inclusion on chain.
 ConfirmationPoll = '5s' # Default
+# FeeEstimationMaxAttempts is the maximum number of retry attempts for fee estimation.
+FeeEstimationMaxAttempts = 5 # Default
 
 [[Nodes]]
 # Name is a unique (per-chain) identifier for this node.

relayer/pkg/chainlink/config/testdata/config-full.toml

@@ -6,6 +6,7 @@ OCR2CacheTTL = '3m0s'
 RequestTimeout = '1m3s'
 TxTimeout = '13s'
 ConfirmationPoll = '42s'
+FeeEstimationMaxAttempts = 8
 
 [[Nodes]]
 Name = 'primary'

relayer/pkg/chainlink/txm/config.go

@@ -8,4 +8,5 @@ import "time"
 type Config interface {
 	ConfirmationPoll() time.Duration
 	TxTimeout() time.Duration
+	FeeEstimationMaxAttempts() int
 }