Fix agent panic on prometheus_text metric parsing (#5166)

Set model.NameValidationScheme to LegacyValidation in an init() to
prevent the prometheus/common text parser from panicking when the
validation scheme is unset. Additionally, wrap extractMetrics in a
deferred recover so any future parsing panic cannot crash the agent.

Signed-off-by: Sarvodaya Kumar <sarvodaya.kumar.ctr@sumologic.com>

Author: sarvodaya-kumar-26

agent/check_handler.go

@@ -425,7 +425,18 @@ func (a *Agent) sendFailure(event *corev2.Event, err error) {
 	}
 }
 
-func extractMetrics(event *corev2.Event) []*corev2.MetricPoint {
+func extractMetrics(event *corev2.Event) (points []*corev2.MetricPoint) {
+	defer func() {
+		if r := recover(); r != nil {
+			logger.WithFields(logrus.Fields{
+				"namespace": event.Check.Namespace,
+				"check":     event.Check.Name,
+				"format":    event.Check.OutputMetricFormat,
+			}).Errorf("recovered from panic during metric extraction: %v", r)
+			points = nil
+		}
+	}()
+
 	var transformer Transformer
 	if !event.HasCheck() {
 		logger.WithError(transformers.ErrMetricExtraction).Error("event must contain a check to parse and extract metrics")

agent/check_handler_internal_test.go

@@ -588,6 +588,24 @@ func TestExtractMetrics(t *testing.T) {
 	}
 }
 
+func TestExtractMetricsPanicRecovery(t *testing.T) {
+	event := &corev2.Event{
+		Check: &corev2.Check{
+			ObjectMeta: corev2.ObjectMeta{
+				Name:      "panic-check",
+				Namespace: "default",
+			},
+			Output:             "bogus output",
+			OutputMetricFormat: corev2.PrometheusOutputMetricFormat,
+		},
+	}
+
+	assert.NotPanics(t, func() {
+		metrics := extractMetrics(event)
+		_ = metrics
+	})
+}
+
 func TestFailOnAssetCheckWithDisabledAssets(t *testing.T) {
 	config, cleanup := FixtureConfig()
 	defer cleanup()

agent/transformers/prometheus.go

@@ -12,6 +12,10 @@ import (
 	"github.com/sirupsen/logrus"
 )
 
+func init() {
+	model.NameValidationScheme = model.LegacyValidation
+}
+
 const (
 	PromTypeTagName = "prom_type"
 	PromHelpTagName = "prom_help"

agent/transformers/prometheus_test.go

@@ -214,6 +214,20 @@ func TestParsePromTags(t *testing.T) {
 	}
 }
 
+func TestParsePromValidationScheme(t *testing.T) {
+	assert := assert.New(t)
+
+	event := types.FixtureEvent("test", "test")
+	event.Check.Output = "# HELP node_cpu_seconds_total Seconds the CPUs spent in each mode.\n# TYPE node_cpu_seconds_total counter\nnode_cpu_seconds_total{cpu=\"0\",mode=\"idle\"} 123456.78\n"
+
+	assert.NotPanics(func() {
+		prom := ParseProm(event)
+		assert.NotEmpty(prom)
+		points := prom.Transform()
+		assert.Equal("node_cpu_seconds_total", points[0].Name)
+	})
+}
+
 func TestTransformProm(t *testing.T) {
 	assert := assert.New(t)
 	ts := time.Now().Unix()