Go 全链路追踪OpenTelemetry 落地与最佳实践微服务时代离不开追踪。OpenTelemetry 是 CNCF 的下一代标准本文讲清它的核心 API 与 Go 实战。一、什么是 OTelOpenTelemetry OpenCensus OpenTracing合体的统一规范多语言 SDK标准化 exporterOTLP/Jaeger/Zipkin自动 手动埋点二、SDK 安装go get go.opentelemetry.io/otel go get go.opentelemetry.io/otel/sdk go get go.opentelemetry.io/otel/exporters/jaeger go get go.opentelemetry.io/contrib/instrumentation/...三、初始化import(go.opentelemetry.io/otelgo.opentelemetry.io/otel/sdk/tracego.opentelemetry.io/otel/exporters/jaeger)funcnewTraceProvider()*trace.TracerProvider{exp,_:jaeger.New(jaeger.WithCollectorEndpoint(http://jaeger:14268/api/traces))tp:trace.NewTracerProvider(trace.WithBatcher(exp),trace.WithSampler(trace.ParentBased(trace.TraceIDRatioBased(0.05))),)otel.SetTracerProvider(tp)returntp}funcmain(){tp:newTraceProvider()defertp.Shutdown(context.Background())// 业务}四、手动埋点ctx,span:otel.Tracer(my-service).Start(ctx,MyOperation)deferspan.End()// 添加事件span.AddEvent(processing start)span.SetAttributes(attribute.String(user_id,uid))五、自动埋点ginimportgo.opentelemetry.io/contrib/instrumentation/github.com/gin-gonic/gin/otelginr:gin.Default()r.Use(otelgin.Middleware(my-service))gin interceptor 自动注入 traceparent 头与 span。六、自动埋点gRPCimportgo.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc// servergrpc.UnaryInterceptor(otelgrpc.UnaryServerInterceptor())// clientgrpc.WithUnaryClient(otelgrpc.UnaryClientInterceptor())七、自动埋点databaseimportgithub.com/XSAM/otelsql// 注册 driver wrapperdb,err:otelsql.Open(mysql,dsn,otelsql.WithDBSystem(...))让 SQL 也带 trace 信息。八、Baggage业务参数跨服务流转bag:baggage.New(map[string]string{user_id:42})ctxbaggage.ContextWithBaggage(ctx,bag)下游可通过 ctx 取出 key。九、与 logs 关联slog.Info(error,slog.Any(error,err),slog.String(trace_id,trace.SpanFromContext(ctx).SpanContext().TraceID().String()),)日志自动带 trace id可定位到 Jaeger。十、采样策略AlwaysSample开发环境ParentBased(TraceIDRatioBased(0.05))5% 采样RatelimitedSampler每秒 N 条采样CompositeSample依据业务组合生产 5%-10% 即可避免全部 span 储存成本。十一、Metrics Logsimportgo.opentelemetry.io/otel/exporters/prometheusmetricExp,_:prometheus.New()meterProvider:sdkmetric.NewMeterProvider(sdkmetric.WithReader(sdkmetric.NewPeriodicReader(metricExp)))十二、生产实战自动 traceparent 注入otelPropagators:otel.GetTextMapPropagator().Fields()// traceparent跨服务调用需在 header 中携带carrier:propagation.HeaderCarrier(c.Request.Header)ctxotel.GetTextMapPropagator().Extract(ctx,carrier)十三、踩坑清单不是所有 trace 都需要业务指标够时不必连 trace采样要谨慎0% 就是完全关闭Exporter 高可用Jaeger/Prometheus 不可用时不要把服务挂掉log/slog 与 trace 关联在中间件里加字段十四、总结与展望OTel 是下一代可观测性标准。SDK 易用、自动埋点多、自带多种 exporter。未来eBPF-based trace、Trace AI 监控将成为下一个突破口。十五、参考文献OpenTelemetry 官方文档otelgin, otelgrpc 各库文档CNCF Observability Analysis TAG
Go 全链路追踪：OpenTelemetry 落地与最佳实践
Go 全链路追踪OpenTelemetry 落地与最佳实践微服务时代离不开追踪。OpenTelemetry 是 CNCF 的下一代标准本文讲清它的核心 API 与 Go 实战。一、什么是 OTelOpenTelemetry OpenCensus OpenTracing合体的统一规范多语言 SDK标准化 exporterOTLP/Jaeger/Zipkin自动 手动埋点二、SDK 安装go get go.opentelemetry.io/otel go get go.opentelemetry.io/otel/sdk go get go.opentelemetry.io/otel/exporters/jaeger go get go.opentelemetry.io/contrib/instrumentation/...三、初始化import(go.opentelemetry.io/otelgo.opentelemetry.io/otel/sdk/tracego.opentelemetry.io/otel/exporters/jaeger)funcnewTraceProvider()*trace.TracerProvider{exp,_:jaeger.New(jaeger.WithCollectorEndpoint(http://jaeger:14268/api/traces))tp:trace.NewTracerProvider(trace.WithBatcher(exp),trace.WithSampler(trace.ParentBased(trace.TraceIDRatioBased(0.05))),)otel.SetTracerProvider(tp)returntp}funcmain(){tp:newTraceProvider()defertp.Shutdown(context.Background())// 业务}四、手动埋点ctx,span:otel.Tracer(my-service).Start(ctx,MyOperation)deferspan.End()// 添加事件span.AddEvent(processing start)span.SetAttributes(attribute.String(user_id,uid))五、自动埋点ginimportgo.opentelemetry.io/contrib/instrumentation/github.com/gin-gonic/gin/otelginr:gin.Default()r.Use(otelgin.Middleware(my-service))gin interceptor 自动注入 traceparent 头与 span。六、自动埋点gRPCimportgo.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc// servergrpc.UnaryInterceptor(otelgrpc.UnaryServerInterceptor())// clientgrpc.WithUnaryClient(otelgrpc.UnaryClientInterceptor())七、自动埋点databaseimportgithub.com/XSAM/otelsql// 注册 driver wrapperdb,err:otelsql.Open(mysql,dsn,otelsql.WithDBSystem(...))让 SQL 也带 trace 信息。八、Baggage业务参数跨服务流转bag:baggage.New(map[string]string{user_id:42})ctxbaggage.ContextWithBaggage(ctx,bag)下游可通过 ctx 取出 key。九、与 logs 关联slog.Info(error,slog.Any(error,err),slog.String(trace_id,trace.SpanFromContext(ctx).SpanContext().TraceID().String()),)日志自动带 trace id可定位到 Jaeger。十、采样策略AlwaysSample开发环境ParentBased(TraceIDRatioBased(0.05))5% 采样RatelimitedSampler每秒 N 条采样CompositeSample依据业务组合生产 5%-10% 即可避免全部 span 储存成本。十一、Metrics Logsimportgo.opentelemetry.io/otel/exporters/prometheusmetricExp,_:prometheus.New()meterProvider:sdkmetric.NewMeterProvider(sdkmetric.WithReader(sdkmetric.NewPeriodicReader(metricExp)))十二、生产实战自动 traceparent 注入otelPropagators:otel.GetTextMapPropagator().Fields()// traceparent跨服务调用需在 header 中携带carrier:propagation.HeaderCarrier(c.Request.Header)ctxotel.GetTextMapPropagator().Extract(ctx,carrier)十三、踩坑清单不是所有 trace 都需要业务指标够时不必连 trace采样要谨慎0% 就是完全关闭Exporter 高可用Jaeger/Prometheus 不可用时不要把服务挂掉log/slog 与 trace 关联在中间件里加字段十四、总结与展望OTel 是下一代可观测性标准。SDK 易用、自动埋点多、自带多种 exporter。未来eBPF-based trace、Trace AI 监控将成为下一个突破口。十五、参考文献OpenTelemetry 官方文档otelgin, otelgrpc 各库文档CNCF Observability Analysis TAG
