feat: handle metric for peer connection failure

telemetry/peer_count.go

@@ -33,3 +33,32 @@ func (r *PeerCount) put(db *sql.DB) error {
 
 	return nil
 }
+
+type PeerConnFailure struct {
+	ID            int    `json:"id"`
+	CreatedAt     int64  `json:"createdAt"`
+	Timestamp     int64  `json:"timestamp"`
+	NodeName      string `json:"nodeName"`
+	NodeKeyUid    string `json:"nodeKeyUid"`
+	PeerId        string `json:"peerId"`
+	StatusVersion string `json:"statusVersion"`
+}
+
+func (r *PeerConnFailure) put(db *sql.DB) error {
+	stmt, err := db.Prepare("INSERT INTO peerConnFailure (timestamp, nodeName, nodeKeyUid, peerId, statusVersion, createdAt) VALUES ($1, $2, $3, $4, $5, $6) RETURNING id;")
+	if err != nil {
+		return err
+	}
+
+	defer stmt.Close()
+
+	r.CreatedAt = time.Now().Unix()
+	lastInsertId := 0
+	err = stmt.QueryRow(r.Timestamp, r.NodeName, r.NodeKeyUid, r.PeerId, r.StatusVersion, r.CreatedAt).Scan(&lastInsertId)
+	if err != nil {
+		return err
+	}
+	r.ID = lastInsertId
+
+	return nil
+}

telemetry/server.go

@@ -67,6 +67,7 @@ const (
 	ReceivedMessagesMetric     TelemetryType = "ReceivedMessages"
 	ErrorSendingEnvelopeMetric TelemetryType = "ErrorSendingEnvelope"
 	PeerCountMetric            TelemetryType = "PeerCount"
+	PeerConnFailureMetric      TelemetryType = "PeerConnFailure"
 )
 
 type TelemetryRequest struct {
@@ -139,6 +140,16 @@ func (s *Server) createTelemetryData(w http.ResponseWriter, r *http.Request) {
 				errorDetails = append(errorDetails, map[string]interface{}{"Id": data.Id, "Error": fmt.Sprintf("Error saving peer count: %v", err)})
 				continue
 			}
+		case PeerConnFailureMetric:
+			var peerConnFailure PeerConnFailure
+			if err := json.Unmarshal(*data.TelemetryData, &peerConnFailure); err != nil {
+				errorDetails = append(errorDetails, map[string]interface{}{"Id": data.Id, "Error": fmt.Sprintf("Error decoding peer connection failure: %v", err)})
+				continue
+			}
+			if err := peerConnFailure.put(s.DB); err != nil {
+				errorDetails = append(errorDetails, map[string]interface{}{"Id": data.Id, "Error": fmt.Sprintf("Error saving peer connection failure: %v", err)})
+				continue
+			}
 		default:
 			errorDetails = append(errorDetails, map[string]interface{}{"Id": data.Id, "Error": fmt.Sprintf("Unknown telemetry type: %s", data.TelemetryType)})
 		}

telemetry/sql/000010_peer_conn_failure.up.sql

@@ -0,0 +1,10 @@
+CREATE TABLE IF NOT EXISTS peerConnFailure (
+    id SERIAL PRIMARY KEY,
+    createdAt INTEGER NOT NULL,
+    peerId VARCHAR(255) NOT NULL,
+    nodeName VARCHAR(255) NOT NULL,
+    nodeKeyUid VARCHAR(255) NOT NULL,
+    timestamp INTEGER NOT NULL,
+    statusVersion VARCHAR(31),
+    CONSTRAINT peerConnFailure_unique unique(timestamp, nodeName, nodeKeyUid, peerId, statusVersion)
+);