Add relayer version metric + dashboard. (#205)

Makefile

@@ -6,6 +6,9 @@ INTEGRATION_TESTS_DIR:=$(ROOT_DIR)/integration-tests
 RELAYER_DIR:=$(ROOT_DIR)/relayer
 BUILD_DIR?=$(ROOT_DIR)/build
 GIT_TAG:=$(shell git describe --tags --exact-match 2>/dev/null)
+ifeq ($(GIT_TAG),)
+GIT_TAG:=devel
+endif
 GIT_SHA:=$(shell git rev-parse HEAD)
 DOCKER_PUSH_TAG?=$(shell git describe --tags --exact-match 2>/dev/null || git rev-parse HEAD)
 LD_FLAGS:="-extldflags=-static \

infra/composer/grafana/dashboards/relayer.json

@@ -1609,9 +1609,107 @@
       ],
       "title": "Malicious behaviour",
       "type": "timeseries"
+    },
+    {
+      "datasource": {
+        "type": "prometheus",
+        "uid": "PBFA97CFB590B2093"
+      },
+      "fieldConfig": {
+        "defaults": {
+          "color": {
+            "mode": "palette-classic"
+          },
+          "custom": {
+            "axisBorderShow": false,
+            "axisCenteredZero": false,
+            "axisColorMode": "text",
+            "axisLabel": "",
+            "axisPlacement": "auto",
+            "barAlignment": 0,
+            "drawStyle": "line",
+            "fillOpacity": 0,
+            "gradientMode": "none",
+            "hideFrom": {
+              "legend": false,
+              "tooltip": false,
+              "viz": false
+            },
+            "insertNulls": false,
+            "lineInterpolation": "linear",
+            "lineWidth": 1,
+            "pointSize": 5,
+            "scaleDistribution": {
+              "type": "linear"
+            },
+            "showPoints": "auto",
+            "spanNulls": false,
+            "stacking": {
+              "group": "A",
+              "mode": "none"
+            },
+            "thresholdsStyle": {
+              "mode": "off"
+            }
+          },
+          "mappings": [],
+          "thresholds": {
+            "mode": "absolute",
+            "steps": [
+              {
+                "color": "green",
+                "value": null
+              },
+              {
+                "color": "red",
+                "value": 80
+              }
+            ]
+          },
+          "unitScale": true
+        },
+        "overrides": []
+      },
+      "gridPos": {
+        "h": 7,
+        "w": 8,
+        "x": 0,
+        "y": 40
+      },
+      "id": 19,
+      "options": {
+        "legend": {
+          "calcs": [],
+          "displayMode": "list",
+          "placement": "bottom",
+          "showLegend": true
+        },
+        "tooltip": {
+          "mode": "single",
+          "sort": "none"
+        }
+      },
+      "pluginVersion": "10.3.3",
+      "targets": [
+        {
+          "datasource": {
+            "type": "prometheus",
+            "uid": "PBFA97CFB590B2093"
+          },
+          "editorMode": "code",
+          "expr": "relayer_version{relayer_coreum_address=~\"$relayer_coreum_address\"}",
+          "format": "time_series",
+          "instant": false,
+          "legendFormat": "{{relayer_coreum_address}} | {{version}} ",
+          "range": true,
+          "refId": "A"
+        }
+      ],
+      "title": "Relayer version",
+      "type": "timeseries"
     }
   ],
-  "refresh": "auto",
+  "refresh": "5s",
   "schemaVersion": 39,
   "tags": [],
   "templating": {
@@ -1715,7 +1813,7 @@
     ]
   },
   "time": {
-    "from": "now-1h",
+    "from": "now-5m",
     "to": "now"
   },
   "timepicker": {},

integration-tests/processes/metrics_test.go

@@ -24,6 +24,7 @@ import (
 
 	coreumintegration "github.com/CoreumFoundation/coreum/v4/testutil/integration"
 	integrationtests "github.com/CoreumFoundation/coreumbridge-xrpl/integration-tests"
+	"github.com/CoreumFoundation/coreumbridge-xrpl/relayer/buildinfo"
 	"github.com/CoreumFoundation/coreumbridge-xrpl/relayer/coreum"
 	"github.com/CoreumFoundation/coreumbridge-xrpl/relayer/metrics"
 	"github.com/CoreumFoundation/coreumbridge-xrpl/relayer/runner"
@@ -761,7 +762,7 @@ func TestRelayerActivityMetrics(t *testing.T) {
 			runnerEnv.RunnerComponents[0].MetricsRegistry.RelayerActivityGaugeVec,
 			map[string]string{
 				metrics.RelayerCoremAddressLabel: relayer.CoreumAddress,
-				metrics.ContractActionLabel:      "save_evidence",
+				metrics.ActionLabel:              "save_evidence",
 			},
 			2,
 		)
@@ -772,7 +773,59 @@ func TestRelayerActivityMetrics(t *testing.T) {
 			runnerEnv.RunnerComponents[0].MetricsRegistry.RelayerActivityGaugeVec,
 			map[string]string{
 				metrics.RelayerCoremAddressLabel: relayer.CoreumAddress,
-				metrics.ContractActionLabel:      "save_signature",
+				metrics.ActionLabel:              "save_signature",
+			},
+			1,
+		)
+	}
+}
+
+func TestRelayerVersionMetrics(t *testing.T) {
+	t.Parallel()
+
+	ctx, chains := integrationtests.NewTestingContext(t)
+
+	envCfg := DefaultRunnerEnvConfig()
+	envCfg.RelayersCount = 2
+	envCfg.SigningThreshold = 2
+	runnerEnv := NewRunnerEnv(ctx, t, envCfg, chains)
+	runnerEnv.StartAllRunnerProcesses()
+	runnerEnv.StartAllRunnerPeriodicMetricCollectors()
+	runnerEnv.AllocateTickets(ctx, t, uint32(200))
+
+	coreumSenderAddress := chains.Coreum.GenAccount()
+	chains.Coreum.FundAccountWithOptions(ctx, t, coreumSenderAddress, coreumintegration.BalancesOptions{
+		Amount: sdkmath.NewIntWithDecimal(1, 6),
+	})
+
+	xrplSenderAddressAddress := chains.XRPL.GenAccount(ctx, t, 10)
+
+	valueToSendFromXRPLToCoreum, err := rippledata.NewValue("1.0", true)
+	require.NoError(t, err)
+
+	amountToSendFromXRPLtoCoreum := rippledata.Amount{
+		Value:    valueToSendFromXRPLToCoreum,
+		Currency: xrpl.XRPTokenCurrency,
+		Issuer:   xrpl.XRPTokenIssuer,
+	}
+	// save evidence to generate the tx with memo
+	runnerEnv.SendFromXRPLToCoreum(
+		ctx,
+		t,
+		xrplSenderAddressAddress.String(),
+		amountToSendFromXRPLtoCoreum,
+		coreumSenderAddress,
+	)
+
+	for _, relayer := range runnerEnv.BootstrappingConfig.Relayers {
+		awaitGaugeVecMetricState(
+			ctx,
+			t,
+			runnerEnv,
+			runnerEnv.RunnerComponents[0].MetricsRegistry.RelayerVersion,
+			map[string]string{
+				metrics.RelayerCoremAddressLabel: relayer.CoreumAddress,
+				metrics.VersionLabel:             buildinfo.VersionTag,
 			},
 			1,
 		)

relayer/coreum/contract.go

@@ -32,6 +32,8 @@ import (
 
 const (
 	contractLabel = "coreumbridge-xrpl"
+	// RelayerCoreumMemoPrefix is memo prefix for the relayer transaction.
+	RelayerCoreumMemoPrefix = "Coreum XRPL bridge relayer version:"
 
 	eventAttributeAction           = "action"
 	eventAttributeHash             = "hash"
@@ -1800,7 +1802,7 @@ func (c *ContractClient) getTxFactory() client.Factory {
 		WithKeybase(c.clientCtx.Keyring()).
 		WithChainID(c.clientCtx.ChainID()).
 		WithTxConfig(c.clientCtx.TxConfig()).
-		WithMemo(fmt.Sprintf("Coreum XRPL bridge relayer version: %s", buildinfo.VersionTag)).
+		WithMemo(fmt.Sprintf("%s %s", RelayerCoreumMemoPrefix, buildinfo.VersionTag)).
 		WithSimulateAndExecute(true)
 }
 

relayer/metrics/collector.go

@@ -82,6 +82,7 @@ type PeriodicCollector struct {
 	registry           *Registry
 	xrplRPCClient      XRPLRPCClient
 	contractClient     ContractClient
+	clientContext      client.Context
 	feemodelClient     feemodeltypes.QueryClient
 	coreumBankClient   banktypes.QueryClient
 	cometServiceClient sdktxtypes.ServiceClient
@@ -90,6 +91,7 @@ type PeriodicCollector struct {
 	transactionEvidencesCachedKeys map[string]struct{}
 	relayersBalancesCachedKeys     map[string]struct{}
 	relayerActivityCachedKeys      map[string]struct{}
+	relayerVersionCachedKeys       map[string]struct{}
 	cacheMu                        sync.Mutex
 }
 
@@ -113,6 +115,7 @@ func NewPeriodicCollector(
 		registry:           registry,
 		xrplRPCClient:      xrplRPCClient,
 		contractClient:     contractClient,
+		clientContext:      clientContext,
 		feemodelClient:     feemodeltypes.NewQueryClient(clientContext),
 		coreumBankClient:   banktypes.NewQueryClient(clientContext),
 		cometServiceClient: sdktxtypes.NewServiceClient(clientContext),
@@ -121,6 +124,7 @@ func NewPeriodicCollector(
 		transactionEvidencesCachedKeys: make(map[string]struct{}),
 		relayersBalancesCachedKeys:     make(map[string]struct{}),
 		relayerActivityCachedKeys:      make(map[string]struct{}),
+		relayerVersionCachedKeys:       make(map[string]struct{}),
 		cacheMu:                        sync.Mutex{},
 	}
 }
@@ -136,10 +140,10 @@ func (c *PeriodicCollector) Start(ctx context.Context) error {
 		transactionEvidencesMetricName:      c.collectTransactionEvidences,
 		relayerBalancesMetricName:           c.collectRelayerBalances,
 		fmt.Sprintf("%s/%s", freeContractTicketsMetricName, freeXRPLTicketsMetricName): c.collectFreeTickets,
-		bridgeStateMetricName:               c.collectBridgeState,
-		relayerActivityMetricName:           c.collectRelayerActivity,
-		xrplTokensCoreumSupplyMetricName:    c.collectXRPLTokensCoreumSupply,
-		xrplBridgeAccountReservesMetricName: c.collectXRPLBridgeAccountReserves,
+		bridgeStateMetricName: c.collectBridgeState,
+		fmt.Sprintf("%s/%s", relayerActivityMetricName, relayerVersionMetricName): c.collectRelayerActivityAndVersion,
+		xrplTokensCoreumSupplyMetricName:                                          c.collectXRPLTokensCoreumSupply,
+		xrplBridgeAccountReservesMetricName:                                       c.collectXRPLBridgeAccountReserves,
 	}
 	return parallel.Run(ctx, func(ctx context.Context, spawn parallel.SpawnFn) error {
 		for name, collector := range periodicCollectors {
@@ -411,12 +415,13 @@ func (c *PeriodicCollector) collectBridgeState(ctx context.Context) error {
 	return nil
 }
 
-func (c *PeriodicCollector) collectRelayerActivity(ctx context.Context) error {
+func (c *PeriodicCollector) collectRelayerActivityAndVersion(ctx context.Context) error {
 	contractCfg, err := c.contractClient.GetContractConfig(ctx)
 	if err != nil {
 		return errors.Wrap(err, "failed to get contract config")
 	}
-	currentValues := make(map[string]gaugeVecValue, 0)
+	currentRelayerActivityValues := make(map[string]gaugeVecValue, 0)
+	currentRelayerVersionValues := make(map[string]gaugeVecValue, 0)
 	// get sequentially to prevent rate limit
 	for _, relayer := range contractCfg.Relayers {
 		relayerAddress := relayer.CoreumAddress
@@ -438,14 +443,48 @@ func (c *PeriodicCollector) collectRelayerActivity(ctx context.Context) error {
 		}
 		// fill the metric with the data
 		for action, count := range actionCount {
-			currentValues[strings.Join([]string{relayerAddress.String(), action}, "-")] = gaugeVecValue{
+			currentRelayerActivityValues[strings.Join([]string{relayerAddress.String(), action}, "-")] = gaugeVecValue{
 				keys:  []string{relayerAddress.String(), action},
 				value: float64(count),
 			}
 		}
+
+		// fill relayer version
+		for _, txRes := range txResponses {
+			var sdkTx sdk.Tx
+			if err := c.clientContext.Codec().UnpackAny(txRes.Tx, &sdkTx); err != nil {
+				return errors.Errorf("failed to unpack sdk.Tx, tx:%v", sdkTx)
+			}
+			tx, ok := sdkTx.(*sdktxtypes.Tx)
+			if !ok {
+				continue
+			}
+			if !strings.HasPrefix(tx.Body.Memo, coreum.RelayerCoreumMemoPrefix) {
+				continue
+			}
+			version := strings.TrimSpace(strings.TrimPrefix(tx.Body.Memo, coreum.RelayerCoreumMemoPrefix))
+			if version == "" {
+				continue
+			}
+
+			currentRelayerVersionValues[strings.Join([]string{relayerAddress.String(), version}, "-")] = gaugeVecValue{
+				keys:  []string{relayerAddress.String(), version},
+				value: float64(1),
+			}
+			break
+		}
 	}
 
-	c.updateGaugeVecAndCachedValues(currentValues, c.relayerActivityCachedKeys, c.registry.RelayerActivityGaugeVec)
+	c.updateGaugeVecAndCachedValues(
+		currentRelayerActivityValues,
+		c.relayerActivityCachedKeys,
+		c.registry.RelayerActivityGaugeVec,
+	)
+	c.updateGaugeVecAndCachedValues(
+		currentRelayerVersionValues,
+		c.relayerVersionCachedKeys,
+		c.registry.RelayerVersion,
+	)
 
 	return nil
 }

relayer/metrics/registry.go

@@ -23,6 +23,7 @@ const (
 	relayerActivityMetricName                         = "relayer_activity"
 	xrplTokensCoreumSupplyMetricName                  = "xrpl_tokens_coreum_supply"
 	xrplBridgeAccountReservesMetricName               = "xrpl_bridge_account_reserves"
+	relayerVersionMetricName                          = "relayer_version"
 	xrplRPCDecodingErrorCounterMetricName             = "xrpl_rpc_decoding_errors_total"
 
 	// XRPLCurrencyIssuerLabel is XRPL currency issuer label.
@@ -37,8 +38,10 @@ const (
 	RelayerCoremAddressLabel = "relayer_coreum_address"
 	// MaliciousBehaviourKeyLabel malicious behaviour key label.
 	MaliciousBehaviourKeyLabel = "malicious_behaviour_key"
-	// ContractActionLabel is contract action label.
-	ContractActionLabel = "action"
+	// ActionLabel is action label.
+	ActionLabel = "action"
+	// VersionLabel is version label.
+	VersionLabel = "version"
 )
 
 // Registry contains metrics.
@@ -58,6 +61,7 @@ type Registry struct {
 	BridgeStateGauge                             prometheus.Gauge
 	MaliciousBehaviourGaugeVec                   *prometheus.GaugeVec
 	RelayerActivityGaugeVec                      *prometheus.GaugeVec
+	RelayerVersion                               *prometheus.GaugeVec
 	XRPLTokensCoreumSupplyGaugeVec               *prometheus.GaugeVec
 	XRPLBridgeAccountReservesGauge               prometheus.Gauge
 	XRPLRPCDecodingErrorCounter                  prometheus.Counter
@@ -153,7 +157,16 @@ func NewRegistry() *Registry {
 		},
 			[]string{
 				RelayerCoremAddressLabel,
-				ContractActionLabel,
+				ActionLabel,
+			},
+		),
+		RelayerVersion: prometheus.NewGaugeVec(prometheus.GaugeOpts{
+			Name: relayerVersionMetricName,
+			Help: "Relayer version",
+		},
+			[]string{
+				RelayerCoremAddressLabel,
+				VersionLabel,
 			},
 		),
 		XRPLTokensCoreumSupplyGaugeVec: prometheus.NewGaugeVec(prometheus.GaugeOpts{
@@ -194,6 +207,7 @@ func (m *Registry) Register(registry prometheus.Registerer) error {
 		m.BridgeStateGauge,
 		m.MaliciousBehaviourGaugeVec,
 		m.RelayerActivityGaugeVec,
+		m.RelayerVersion,
 		m.XRPLTokensCoreumSupplyGaugeVec,
 		m.XRPLBridgeAccountReservesGauge,
 		m.XRPLRPCDecodingErrorCounter,