MON-1157: Test metrics for prometheus best practice

test/extended/prometheus/prometheus.go

@@ -40,6 +40,7 @@ import (
 	"sigs.k8s.io/yaml"
 
 	configv1 "github.com/openshift/api/config/v1"
+	ote "github.com/openshift-eng/openshift-tests-extension/pkg/ginkgo"
 
 	testresult "github.com/openshift/origin/pkg/test/ginkgo/result"
 	exutil "github.com/openshift/origin/test/extended/util"
@@ -951,6 +952,48 @@ var _ = g.Describe("[sig-instrumentation] Prometheus [apigroup:image.openshift.i
 	})
 })
 
+var _ = g.Describe("[sig-instrumentation][Late] Prometheus metric naming", func() {
+	defer g.GinkgoRecover()
+	oc := exutil.NewCLIWithoutNamespace("prometheus-lint")
+
+	g.It("should follow Prometheus naming best practices", ote.Informing(), func(ctx g.SpecContext) {
+		promClient := oc.NewPrometheusClient(ctx)
+
+		g.By("fetching all metric metadata from Prometheus")
+		metadata, err := promClient.Metadata(ctx, "", "")
+		o.Expect(err).NotTo(o.HaveOccurred(), "failed to query Prometheus metadata")
+		o.Expect(metadata).NotTo(o.BeEmpty(), "Prometheus metadata should not be empty")
+		e2e.Logf("fetched metadata for %d metrics", len(metadata))
+
+		families := helper.MetadataToMetricFamilies(metadata)
+		e2e.Logf("built %d metric families (recording rules excluded)", len(families))
+
+		g.By("finding camelCase labels")
+		camelLabels, err := helper.FindCamelCaseLabels(ctx, promClient)
+		o.Expect(err).NotTo(o.HaveOccurred(), "failed to query camelCase labels")
+		if len(camelLabels) > 0 {
+			e2e.Logf("found camelCase labels on %d metrics", len(camelLabels))
+		}
+		helper.SetCamelCaseLabels(families, camelLabels)
+
+		g.By("linting metrics with promlint")
+		linter := helper.NewLinterWithMetricFamilies(families)
+		problems, err := linter.Lint()
+		o.Expect(err).NotTo(o.HaveOccurred(), "promlint returned an error")
+
+		if len(problems) > 0 {
+			e2e.Logf("found %d metric naming violations:", len(problems))
+			for _, p := range problems {
+				e2e.Logf("  %s", p)
+			}
+			e2e.Logf("to add these to the exception list, use the following entries:\n%s",
+				helper.FormatExceptionEntries(problems))
+		}
+		o.Expect(problems).To(o.BeEmpty(),
+			"metrics violating Prometheus naming best practices found; see test log for details and exception list entries")
+	})
+})
+
 func all(errs ...error) []error {
 	var result []error
 	for _, err := range errs {

test/extended/util/prometheus/metrics.go

@@ -0,0 +1,119 @@
+package prometheus
+
+import (
+	"context"
+	"fmt"
+	"regexp"
+	"strings"
+	"time"
+
+	prometheusv1 "github.com/prometheus/client_golang/api/prometheus/v1"
+	dto "github.com/prometheus/client_model/go"
+	"github.com/prometheus/common/model"
+)
+
+var camelCaseRe = regexp.MustCompile(`[a-z][A-Z]`)
+
+// MetadataToMetricFamilies converts Prometheus API metadata into a slice of
+// MetricFamily suitable for promlint validation.
+func MetadataToMetricFamilies(metadata map[string][]prometheusv1.Metadata) []*dto.MetricFamily {
+	var families []*dto.MetricFamily
+
+	for name, entries := range metadata {
+		if len(entries) == 0 {
+			continue
+		}
+		// Recording rules use colons in their names and are not subject to
+		// metric naming conventions.
+		if strings.Contains(name, ":") {
+			continue
+		}
+
+		entry := entries[0]
+		name := name
+		help := entry.Help
+		metricType := convertMetricType(entry.Type)
+		families = append(families, &dto.MetricFamily{
+			Name:   &name,
+			Help:   &help,
+			Type:   &metricType,
+			Metric: []*dto.Metric{{}},
+		})
+	}
+
+	return families
+}
+
+// SetCamelCaseLabels attaches camelCase label pairs to the corresponding
+// metric families so that promlint's LintCamelCase validation can detect them.
+func SetCamelCaseLabels(families []*dto.MetricFamily, labelsPerMetric map[string][]*dto.LabelPair) {
+	for i, family := range families {
+		labels, ok := labelsPerMetric[family.GetName()]
+		if ok {
+			families[i].Metric[0].Label = labels
+		}
+	}
+}
+
+// FindCamelCaseLabels queries Prometheus for all label names that contain
+// camelCase, then for each such label, determines which metrics use it.
+func FindCamelCaseLabels(ctx context.Context, promClient prometheusv1.API) (map[string][]*dto.LabelPair, error) {
+	labelNames, _, err := promClient.LabelNames(ctx, nil, time.Time{}, time.Time{})
+	if err != nil {
+		return nil, fmt.Errorf("failed to get label names: %w", err)
+	}
+
+	var camelCaseLabels []string
+	for _, label := range labelNames {
+		if camelCaseRe.MatchString(label) {
+			camelCaseLabels = append(camelCaseLabels, label)
+		}
+	}
+
+	if len(camelCaseLabels) == 0 {
+		return nil, nil
+	}
+
+	result := make(map[string][]*dto.LabelPair)
+	for _, label := range camelCaseLabels {
+		query := fmt.Sprintf(`count({__name__=~".+",%s=~".+"}) by (__name__)`, label)
+		val, _, err := promClient.Query(ctx, query, time.Time{})
+		if err != nil {
+			return nil, fmt.Errorf("failed to query metrics for label %q: %w", label, err)
+		}
+
+		vector, ok := val.(model.Vector)
+		if !ok {
+			continue
+		}
+		for _, sample := range vector {
+			metricName := string(sample.Metric[model.MetricNameLabel])
+			if metricName == "" {
+				continue
+			}
+			lbl := label
+			val := ""
+			result[metricName] = append(result[metricName], &dto.LabelPair{
+				Name:  &lbl,
+				Value: &val,
+			})
+		}
+	}
+
+	return result, nil
+}
+
+func convertMetricType(mt prometheusv1.MetricType) dto.MetricType {
+	switch mt {
+	case prometheusv1.MetricTypeCounter:
+		return dto.MetricType_COUNTER
+	case prometheusv1.MetricTypeGauge:
+		return dto.MetricType_GAUGE
+	case prometheusv1.MetricTypeHistogram:
+		return dto.MetricType_HISTOGRAM
+	case prometheusv1.MetricTypeSummary:
+		return dto.MetricType_SUMMARY
+	default:
+		return dto.MetricType_UNTYPED
+	}
+}