start wrapping requests withRedFmetrics

cmd/zoekt-sourcegraph-indexserver/main.go

@@ -48,11 +48,10 @@ var (
 		Buckets: prometheus.ExponentialBuckets(1, 10, 6), // 1s -> 27min
 	})
 
-	metricResolveRevisionDuration = promauto.NewHistogramVec(prometheus.HistogramOpts{
-		Name:    "resolve_revision_seconds",
-		Help:    "A histogram of latencies for resolving a repository revision.",
-		Buckets: prometheus.ExponentialBuckets(.25, 2, 4), // 250ms -> 2s
-	}, []string{"success"}) // success=true|false
+	metricResolveRevisionObserver = NewRedfMetrics("resolve_revision",
+		WithLabels("success"), // success=true|false
+		WithDurationBuckets(prometheus.ExponentialBuckets(.25, 2, 4)), // 250ms -> 2s
+	)
 
 	metricGetIndexOptionsError = promauto.NewCounter(prometheus.CounterOpts{
 		Name: "get_index_options_error_total",

cmd/zoekt-sourcegraph-indexserver/metrics.go

@@ -0,0 +1,149 @@
+package main
+
+import (
+	"fmt"
+	"time"
+
+	"github.com/prometheus/client_golang/prometheus"
+)
+
+// Example Usage:
+//
+// observer := NewRedFMetrics("operation_name", WithLabels("factorA", "factorB"))
+//
+// start := time.now()
+// err := doOperation()
+//
+// operation.Observe(time.Since(start), err)
+//
+// m.Observe(duration duration, err error, "label0", "label1"...)
+
+// RedFMetrics contains four common metrics for an operation.
+// It's based on the RED method + some additional advice from
+// Google SRE's "Monitoring Distributed Systems".
+//
+// See:
+// - https://www.weave.works/blog/the-red-method-key-metrics-for-microservices-architecture/
+// - https://sre.google/sre-book/monitoring-distributed-systems/
+type RedFMetrics struct {
+	Count    *prometheus.CounterVec   // How often did this operation run successfully?
+	Duration *prometheus.HistogramVec // How long did this operation run for?
+
+	ErrorCount    *prometheus.CounterVec   // How often did this operation fail?
+	ErrorDuration *prometheus.HistogramVec // How long did the failures take?
+}
+
+func (m *RedFMetrics) Observe(d time.Duration, err error, lvals ...string) {
+	if err != nil {
+		m.ErrorCount.WithLabelValues(lvals...).Inc()
+		m.ErrorDuration.WithLabelValues(lvals...).Observe(d.Seconds())
+		return
+	}
+
+	m.Count.WithLabelValues(lvals...).Inc()
+	m.ErrorDuration.WithLabelValues(lvals...).Observe(d.Seconds())
+}
+
+type redfMetricOptions struct {
+	countHelp    string
+	durationHelp string
+
+	errorsCountHelp    string
+	errorsDurationHelp string
+
+	labels          []string
+	durationBuckets []float64
+}
+
+// RedfMetricsOption alter the default behavior of NewRedfMetrics.
+type RedfMetricsOption func(o *redfMetricOptions)
+
+// WithDurationHelp overrides the default help text for duration metrics.
+func WithDurationHelp(text string) RedfMetricsOption {
+	return func(o *redfMetricOptions) { o.durationHelp = text }
+}
+
+// WithCountHelp overrides the default help text for count metrics.
+func WithCountHelp(text string) RedfMetricsOption {
+	return func(o *redfMetricOptions) { o.countHelp = text }
+}
+
+// WithErrorsCountHelp overrides the default help text for error count metrics.
+func WithErrorsCountHelp(text string) RedfMetricsOption {
+	return func(o *redfMetricOptions) { o.errorsCountHelp = text }
+}
+
+// WithErrorsCountHelp overrides the default help text for error duration metrics.
+func WithErrorsDurationHelp(text string) RedfMetricsOption {
+	return func(o *redfMetricOptions) { o.errorsDurationHelp = text }
+}
+
+// WithLabels overrides the default labels for all metrics.
+func WithLabels(labels ...string) RedfMetricsOption {
+	return func(o *redfMetricOptions) { o.labels = labels }
+}
+
+// WithDurationBuckets overrides the default histogram bucket values for duration metrics.
+func WithDurationBuckets(buckets []float64) RedfMetricsOption {
+	return func(o *redfMetricOptions) {
+		if len(buckets) != 0 {
+			o.durationBuckets = buckets
+		}
+	}
+}
+
+func NewRedfMetrics(name string, overrides ...RedfMetricsOption) *RedFMetrics {
+	options := &redfMetricOptions{
+		countHelp:          fmt.Sprintf("Number of successful %s operations", name),
+		durationHelp:       fmt.Sprintf("Time in seconds spent performing %s operations", name),
+		errorsCountHelp:    fmt.Sprintf("Number of failed %s operations", name),
+		errorsDurationHelp: fmt.Sprintf("Time in seconds spent performing failed %s operations", name),
+
+		labels:          nil,
+		durationBuckets: prometheus.DefBuckets,
+	}
+
+	for _, override := range overrides {
+		override(options)
+	}
+
+	count := prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Name: fmt.Sprintf("%s_total", name),
+			Help: options.countHelp,
+		},
+		options.labels,
+	)
+
+	duration := prometheus.NewHistogramVec(
+		prometheus.HistogramOpts{
+			Name: fmt.Sprintf("%s_duration", name),
+			Help: options.countHelp,
+		},
+		options.labels,
+	)
+
+	errorsCount := prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Name: fmt.Sprintf("%s_errors_total", name),
+			Help: options.errorsCountHelp,
+		},
+		options.labels,
+	)
+
+	errorsDuration := prometheus.NewHistogramVec(
+		prometheus.HistogramOpts{
+			Name: fmt.Sprintf("%s_errors_duration", name),
+			Help: options.errorsDurationHelp,
+		},
+		options.labels,
+	)
+
+	return &RedFMetrics{
+		Count:    count,
+		Duration: duration,
+
+		ErrorCount:    errorsCount,
+		ErrorDuration: errorsDuration,
+	}
+}

cmd/zoekt-sourcegraph-indexserver/sg.go

@@ -113,7 +113,7 @@ func (s *sourcegraphClient) List(ctx context.Context, indexed []uint32) (*Source
 				first = false
 				s.configFingerprint.Store(lastFingerprint)
 
-				metricResolveRevisionDuration.WithLabelValues("false").Observe(time.Since(start).Seconds())
+				metricResolveRevisionObserver.Observe(time.Since(start), err, "false")
 				tr.LazyPrintf("failed fetching options batch: %v", err)
 				tr.SetError()
 				continue
@@ -125,10 +125,9 @@ func (s *sourcegraphClient) List(ctx context.Context, indexed []uint32) (*Source
 				s.configFingerprint.Store(fingerprint)
 			}
 
-			metricResolveRevisionDuration.WithLabelValues("true").Observe(time.Since(start).Seconds())
+			metricResolveRevisionObserver.Observe(time.Since(start), err, "false")
 			for _, opt := range opts {
 				if opt.Error != "" {
-					metricGetIndexOptionsError.Inc()
 					tr.LazyPrintf("failed fetching options for %v: %v", opt.Name, opt.Error)
 					tr.SetError()
 					continue