feat: track store confirmation failures

cmd/server/main.go

@@ -65,6 +65,7 @@ func main() {
 	server.RegisterMetric(types.SentEnvelopeMetric, &metrics.SentEnvelope{})
 	server.RegisterMetric(types.PeerCountByShardMetric, &metrics.PeerCountByShard{})
 	server.RegisterMetric(types.PeerCountByOriginMetric, &metrics.PeerCountByOrigin{})
+	server.RegisterMetric(types.StoreConfirmationFailedMetric, &metrics.StoreConfirmationFailed{})
 
 	server.RegisterMetric(types.MessageCheckSuccessMetric, &metrics.MessageCheckSuccess{})
 	server.RegisterMetric(types.MessageCheckFailureMetric, &metrics.MessageCheckFailure{})

lib/common/test_utils.go

@@ -43,6 +43,7 @@ func DropTables(db *sql.DB) {
 		"messageCheckSuccess",
 		"messageCheckFailure",
 		"dialFailure",
+		"storeconfirmationfailed",
 		"schema_migrations",
 	}
 
@@ -94,6 +95,11 @@ func DropTables(db *sql.DB) {
 		log.Fatalf("an error '%s' was not expected when dropping the index", err)
 	}
 
+	_, err = tx.Exec("DROP INDEX IF EXISTS storeConfirmationFailed_unique")
+	if err != nil {
+		log.Fatalf("an error '%s' was not expected when dropping the index", err)
+	}
+
 	_, err = tx.Exec("DROP INDEX IF EXISTS peerCountByOrigin_unique")
 	if err != nil {
 		log.Fatalf("an error '%s' was not expected when dropping the index", err)

lib/database/sql/000017_store_confirmation_failure.up.sql

@@ -0,0 +1,15 @@
+CREATE TABLE IF NOT EXISTS storeConfirmationFailed (
+    id SERIAL PRIMARY KEY,
+    messageHash TEXT NOT NULL,
+    recordId INTEGER NOT NULL
+);
+
+ALTER TABLE storeConfirmationFailed ADD CONSTRAINT fk_storeConfirmationFailed_telemetryRecord
+            FOREIGN KEY (recordId) REFERENCES telemetryRecord(id);
+
+ALTER TABLE storeConfirmationFailed 
+ADD CONSTRAINT storeConfirmationFailed_unique 
+UNIQUE (
+    recordId,
+    messageHash
+);

lib/metrics/store_confirmation_failure.go

@@ -0,0 +1,59 @@
+package metrics
+
+import (
+	"context"
+	"database/sql"
+	"encoding/json"
+	"fmt"
+	"log"
+
+	"github.com/status-im/telemetry/lib/common"
+	"github.com/status-im/telemetry/pkg/types"
+)
+
+type StoreConfirmationFailed struct {
+	types.StoreConfirmationFailed
+}
+
+func (s *StoreConfirmationFailed) Process(ctx context.Context, db *sql.DB, errs *common.MetricErrors, data *types.TelemetryRequest) error {
+	if err := json.Unmarshal(*data.TelemetryData, &s); err != nil {
+		errs.Append(data.ID, fmt.Sprintf("Error decoding store confirmation failed: %v", err))
+		return err
+	}
+
+	if err := s.Put(ctx, db); err != nil {
+		errs.Append(data.ID, fmt.Sprintf("Error saving store confirmation failed: %v", err))
+		return err
+	}
+
+	log.Printf("AK: store confirmation metric saved: %v", s)
+	return nil
+}
+
+func (s *StoreConfirmationFailed) Put(ctx context.Context, db *sql.DB) error {
+	tx, err := db.BeginTx(ctx, nil)
+	if err != nil {
+		return err
+	}
+	defer tx.Rollback()
+
+	recordId, err := InsertTelemetryRecord(tx, &s.TelemetryRecord)
+	if err != nil {
+		return err
+	}
+	result := tx.QueryRow("INSERT INTO storeConfirmationFailed (recordId, messageHash) VALUES ($1, $2) RETURNING id;", recordId, s.MessageHash)
+	if result.Err() != nil {
+		return result.Err()
+	}
+
+	err = result.Scan(&s.ID)
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
+func (s *StoreConfirmationFailed) Clean(db *sql.DB, before int64) (int64, error) {
+	return common.Cleanup(db, "storeConfirmationFailed", before)
+}

pkg/types/types.go

@@ -5,19 +5,20 @@ import "encoding/json"
 type TelemetryType string
 
 const (
-	ProtocolStatsMetric        TelemetryType = "ProtocolStats"
-	ReceivedEnvelopeMetric     TelemetryType = "ReceivedEnvelope"
-	SentEnvelopeMetric         TelemetryType = "SentEnvelope"
-	UpdateEnvelopeMetric       TelemetryType = "UpdateEnvelope"
-	ReceivedMessagesMetric     TelemetryType = "ReceivedMessages"
-	ErrorSendingEnvelopeMetric TelemetryType = "ErrorSendingEnvelope"
-	PeerCountMetric            TelemetryType = "PeerCount"
-	PeerConnFailureMetric      TelemetryType = "PeerConnFailure"
-	PeerCountByShardMetric     TelemetryType = "PeerCountByShard"
-	PeerCountByOriginMetric    TelemetryType = "PeerCountByOrigin"
-	MessageCheckSuccessMetric  TelemetryType = "MessageCheckSuccess"
-	MessageCheckFailureMetric  TelemetryType = "MessageCheckFailure"
-	DialFailureMetric          TelemetryType = "DialFailure"
+	ProtocolStatsMetric           TelemetryType = "ProtocolStats"
+	ReceivedEnvelopeMetric        TelemetryType = "ReceivedEnvelope"
+	SentEnvelopeMetric            TelemetryType = "SentEnvelope"
+	UpdateEnvelopeMetric          TelemetryType = "UpdateEnvelope"
+	ReceivedMessagesMetric        TelemetryType = "ReceivedMessages"
+	ErrorSendingEnvelopeMetric    TelemetryType = "ErrorSendingEnvelope"
+	PeerCountMetric               TelemetryType = "PeerCount"
+	PeerConnFailureMetric         TelemetryType = "PeerConnFailure"
+	PeerCountByShardMetric        TelemetryType = "PeerCountByShard"
+	PeerCountByOriginMetric       TelemetryType = "PeerCountByOrigin"
+	MessageCheckSuccessMetric     TelemetryType = "MessageCheckSuccess"
+	MessageCheckFailureMetric     TelemetryType = "MessageCheckFailure"
+	DialFailureMetric             TelemetryType = "DialFailure"
+	StoreConfirmationFailedMetric TelemetryType = "StoreConfirmationFailed"
 )
 
 type Origin int64
@@ -168,3 +169,9 @@ type DialFailure struct {
 	Protocols string        `json:"protocols"`
 	Timestamp int64         `json:"timestamp"`
 }
+type StoreConfirmationFailed struct {
+	TelemetryRecord
+	ID          int    `json:"id"`
+	MessageHash string `json:"messageHash"`
+	Timestamp   int64  `json:"timestamp"`
+}