Metrics for discarded samples (soft errors)

Signed-off-by: Peter Štibraný <[email protected]>

Author: pstibrany

pkg/ingester/ingester_v2.go

@@ -314,6 +314,15 @@ func (i *Ingester) v2Push(ctx context.Context, req *client.WriteRequest) (*clien
 					firstPartialErr = errors.Wrapf(err, "series=%s, timestamp=%v", client.FromLabelAdaptersToLabels(ts.Labels).String(), model.Time(s.TimestampMs).Time().Format(time.RFC3339Nano))
 				}
 
+				switch cause {
+				case tsdb.ErrOutOfBounds:
+					validation.DiscardedSamples.WithLabelValues(sampleOutOfBounds, userID).Inc()
+				case tsdb.ErrOutOfOrderSample:
+					validation.DiscardedSamples.WithLabelValues(sampleOutOfOrder, userID).Inc()
+				case tsdb.ErrAmendSample:
+					validation.DiscardedSamples.WithLabelValues(newValueForTimestamp, userID).Inc()
+				}
+
 				continue
 			}
 

pkg/ingester/ingester_v2_test.go

@@ -45,6 +45,7 @@ func TestIngester_v2Push(t *testing.T) {
 		"cortex_ingester_memory_users",
 		"cortex_ingester_memory_series_created_total",
 		"cortex_ingester_memory_series_removed_total",
+		"cortex_discarded_samples_total",
 	}
 	userID := "test"
 
@@ -124,6 +125,9 @@ func TestIngester_v2Push(t *testing.T) {
 				# HELP cortex_ingester_memory_series_removed_total The total number of series that were removed per user.
 				# TYPE cortex_ingester_memory_series_removed_total counter
 				cortex_ingester_memory_series_removed_total{user="test"} 0
+				# HELP cortex_discarded_samples_total The total number of samples that were discarded.
+				# TYPE cortex_discarded_samples_total counter
+				cortex_discarded_samples_total{reason="sample-out-of-order",user="test"} 1
 			`,
 		},
 		"should soft fail on sample out of bound": {
@@ -160,6 +164,9 @@ func TestIngester_v2Push(t *testing.T) {
 				# HELP cortex_ingester_memory_series_removed_total The total number of series that were removed per user.
 				# TYPE cortex_ingester_memory_series_removed_total counter
 				cortex_ingester_memory_series_removed_total{user="test"} 0
+				# HELP cortex_discarded_samples_total The total number of samples that were discarded.
+				# TYPE cortex_discarded_samples_total counter
+				cortex_discarded_samples_total{reason="out-of-bounds",user="test"} 1
 			`,
 		},
 		"should soft fail on two different sample values at the same timestamp": {
@@ -196,6 +203,9 @@ func TestIngester_v2Push(t *testing.T) {
 				# HELP cortex_ingester_memory_series_removed_total The total number of series that were removed per user.
 				# TYPE cortex_ingester_memory_series_removed_total counter
 				cortex_ingester_memory_series_removed_total{user="test"} 0
+				# HELP cortex_discarded_samples_total The total number of samples that were discarded.
+				# TYPE cortex_discarded_samples_total counter
+				cortex_discarded_samples_total{reason="new-value-for-timestamp",user="test"} 1
 			`,
 		},
 	}
@@ -204,6 +214,9 @@ func TestIngester_v2Push(t *testing.T) {
 		t.Run(testName, func(t *testing.T) {
 			registry := prometheus.NewRegistry()
 
+			registry.MustRegister(validation.DiscardedSamples)
+			validation.DiscardedSamples.Reset()
+
 			// Create a mocked ingester
 			cfg := defaultIngesterTestConfig()
 			cfg.LifecyclerConfig.JoinAfter = 0

pkg/ingester/series.go

@@ -13,6 +13,14 @@ import (
 	"github.com/cortexproject/cortex/pkg/prom1/storage/metric"
 )
 
+const (
+	sampleOutOfOrder     = "sample-out-of-order"
+	newValueForTimestamp = "new-value-for-timestamp"
+	sampleOutOfBounds    = "out-of-bounds"
+	duplicateSample      = "duplicate-sample"
+	duplicateTimestamp   = "duplicate-timestamp"
+)
+
 type memorySeries struct {
 	metric labels.Labels
 
@@ -51,19 +59,19 @@ func (s *memorySeries) add(v model.SamplePair) error {
 		// If we don't know what the last sample value is, silently discard.
 		// This will mask some errors but better than complaining when we don't really know.
 		if !s.lastSampleValueSet {
-			return makeNoReportError("duplicate-timestamp")
+			return makeNoReportError(duplicateTimestamp)
 		}
 		// If both timestamp and sample value are the same as for the last append,
 		// ignore as they are a common occurrence when using client-side timestamps
 		// (e.g. Pushgateway or federation).
 		if v.Value.Equal(s.lastSampleValue) {
-			return makeNoReportError("duplicate-sample")
+			return makeNoReportError(duplicateSample)
 		}
-		return makeMetricValidationError("new-value-for-timestamp", s.metric,
+		return makeMetricValidationError(newValueForTimestamp, s.metric,
 			fmt.Errorf("sample with repeated timestamp but different value; last value: %v, incoming value: %v", s.lastSampleValue, v.Value))
 	}
 	if v.Timestamp < s.lastTime {
-		return makeMetricValidationError("sample-out-of-order", s.metric,
+		return makeMetricValidationError(sampleOutOfOrder, s.metric,
 			fmt.Errorf("sample timestamp out of order; last timestamp: %v, incoming timestamp: %v", s.lastTime, v.Timestamp))
 	}