go-kit Log & Metrics & Tracing

微服务监控3大核心 Log & Metrics & Tracing

Log

Log 模块源码有待分析（分析完再补上）

Metrics

主要是封装 Metrics 接口，及各个 Metrics(Prometheus,InfluxDB,StatsD,expvar) 中间件的封装。

// Counter describes a metric that accumulates values monotonically.
// An example of a counter is the number of received HTTP requests.
type Counter interface {With(labelValues ...string) CounterAdd(delta float64)
}// Gauge describes a metric that takes specific values over time.
// An example of a gauge is the current depth of a job queue.
type Gauge interface {With(labelValues ...string) GaugeSet(value float64)Add(delta float64)
}// Histogram describes a metric that takes repeated observations of the same
// kind of thing, and produces a statistical summary of those observations,
// typically expressed as quantiles or buckets. An example of a histogram is
// HTTP request latencies.
type Histogram interface {With(labelValues ...string) HistogramObserve(value float64)
}

Tracing

Tracing 主要是对 Zipkin, OpenTracing, OpenCensus 的封装。

Zipkin

func TraceEndpoint(tracer *zipkin.Tracer, name string) endpoint.Middleware {return func(next endpoint.Endpoint) endpoint.Endpoint {return func(ctx context.Context, request interface{}) (interface{}, error) {var sc model.SpanContextif parentSpan := zipkin.SpanFromContext(ctx); parentSpan != nil {sc = parentSpan.Context()}sp := tracer.StartSpan(name, zipkin.Parent(sc))defer sp.Finish()ctx = zipkin.NewContext(ctx, sp)return next(ctx, request)}}
}

OpenTracing

// TraceServer returns a Middleware that wraps the `next` Endpoint in an
// OpenTracing Span called `operationName`.
//
// If `ctx` already has a Span, it is re-used and the operation name is
// overwritten. If `ctx` does not yet have a Span, one is created here.
func TraceServer(tracer opentracing.Tracer, operationName string) endpoint.Middleware {return func(next endpoint.Endpoint) endpoint.Endpoint {return func(ctx context.Context, request interface{}) (interface{}, error) {serverSpan := opentracing.SpanFromContext(ctx)if serverSpan == nil {// All we can do is create a new root span.serverSpan = tracer.StartSpan(operationName)} else {serverSpan.SetOperationName(operationName)}defer serverSpan.Finish()otext.SpanKindRPCServer.Set(serverSpan)ctx = opentracing.ContextWithSpan(ctx, serverSpan)return next(ctx, request)}}
}// TraceClient returns a Middleware that wraps the `next` Endpoint in an
// OpenTracing Span called `operationName`.
func TraceClient(tracer opentracing.Tracer, operationName string) endpoint.Middleware {return func(next endpoint.Endpoint) endpoint.Endpoint {return func(ctx context.Context, request interface{}) (interface{}, error) {var clientSpan opentracing.Spanif parentSpan := opentracing.SpanFromContext(ctx); parentSpan != nil {clientSpan = tracer.StartSpan(operationName,opentracing.ChildOf(parentSpan.Context()),)} else {clientSpan = tracer.StartSpan(operationName)}defer clientSpan.Finish()otext.SpanKindRPCClient.Set(clientSpan)ctx = opentracing.ContextWithSpan(ctx, clientSpan)return next(ctx, request)}}
}

OpenCensus

func TraceEndpoint(name string, options ...EndpointOption) endpoint.Middleware {if name == "" {name = TraceEndpointDefaultName}cfg := &EndpointOptions{}for _, o := range options {o(cfg)}return func(next endpoint.Endpoint) endpoint.Endpoint {return func(ctx context.Context, request interface{}) (response interface{}, err error) {ctx, span := trace.StartSpan(ctx, name)if len(cfg.Attributes) > 0 {span.AddAttributes(cfg.Attributes...)}defer span.End()defer func() {if err != nil {if lberr, ok := err.(lb.RetryError); ok {// handle errors originating from lb.Retryattrs := make([]trace.Attribute, 0, len(lberr.RawErrors))for idx, rawErr := range lberr.RawErrors {attrs = append(attrs, trace.StringAttribute("gokit.retry.error."+strconv.Itoa(idx+1), rawErr.Error(),))}span.AddAttributes(attrs...)span.SetStatus(trace.Status{Code:    trace.StatusCodeUnknown,Message: lberr.Final.Error(),})return}// generic errorspan.SetStatus(trace.Status{Code:    trace.StatusCodeUnknown,Message: err.Error(),})return}// test for business errorif res, ok := response.(endpoint.Failer); ok && res.Failed() != nil {span.AddAttributes(trace.StringAttribute("gokit.business.error", res.Failed().Error()),)if cfg.IgnoreBusinessError {span.SetStatus(trace.Status{Code: trace.StatusCodeOK})return}// treating business error as real error in span.span.SetStatus(trace.Status{Code:    trace.StatusCodeUnknown,Message: res.Failed().Error(),})return}// no errors identifiedspan.SetStatus(trace.Status{Code: trace.StatusCodeOK})}()response, err = next(ctx, request)return}}
}

使用

参考 examples/set.go

var concatEndpoint endpoint.EndpointconcatEndpoint = MakeConcatEndpoint(svc)
concatEndpoint = opentracing.TraceServer(otTracer, "Concat")(concatEndpoint)
concatEndpoint = zipkin.TraceEndpoint(zipkinTracer, "Concat")(concatEndpoint)
concatEndpoint = LoggingMiddleware(log.With(logger, "method", "Concat"))(concatEndpoint)
concatEndpoint = InstrumentingMiddleware(duration.With("method", "Concat"))(concatEndpoint)

小结

通过把第三方中间件封装成 endpoint.Middleware, 可以与其它 go-kit 中间件组合。