New metric for ingester errors

CHANGELOG.md

@@ -56,6 +56,7 @@
 * [ENHANCEMENT] Distributor: Add native histograms max sample size bytes limit validation. #6834
 * [ENHANCEMENT] Querier: Support caching parquet labels file in parquet queryable. #6835
 * [ENHANCEMENT] Querier: Support query limits in parquet queryable. #6870
+* [ENHANCEMENT] Ingester: Add new metric `cortex_ingester_push_errors_total` to track reasons for ingester request failures. #6901
 * [BUGFIX] Ingester: Avoid error or early throttling when READONLY ingesters are present in the ring #6517
 * [BUGFIX] Ingester: Fix labelset data race condition. #6573
 * [BUGFIX] Compactor: Cleaner should not put deletion marker for blocks with no-compact marker. #6576

pkg/ingester/ingester.go

@@ -1167,6 +1167,11 @@ func (i *Ingester) Push(ctx context.Context, req *cortexpb.WriteRequest) (*corte
 	span, ctx := opentracing.StartSpanFromContext(ctx, "Ingester.Push")
 	defer span.Finish()
 
+	userID, err := tenant.TenantID(ctx)
+	if err != nil {
+		return nil, err
+	}
+
 	// We will report *this* request in the error too.
 	inflight := i.inflightPushRequests.Inc()
 	i.maxInflightPushRequests.Track(inflight)
@@ -1175,6 +1180,7 @@ func (i *Ingester) Push(ctx context.Context, req *cortexpb.WriteRequest) (*corte
 	gl := i.getInstanceLimits()
 	if gl != nil && gl.MaxInflightPushRequests > 0 {
 		if inflight > gl.MaxInflightPushRequests {
+			i.metrics.pushErrorsTotal.WithLabelValues(userID, pushErrTooManyInflightRequests).Inc()
 			return nil, errTooManyInflightPushRequests
 		}
 	}
@@ -1186,11 +1192,6 @@ func (i *Ingester) Push(ctx context.Context, req *cortexpb.WriteRequest) (*corte
 	defer req.Free()
 	defer cortexpb.ReuseSlice(req.Timeseries)
 
-	userID, err := tenant.TenantID(ctx)
-	if err != nil {
-		return nil, err
-	}
-
 	il := i.getInstanceLimits()
 	if il != nil && il.MaxIngestionRate > 0 {
 		if rate := i.ingestionRate.Rate(); rate >= il.MaxIngestionRate {

pkg/ingester/ingester_test.go

@@ -6515,7 +6515,8 @@ func TestIngester_inflightPushRequests(t *testing.T) {
 	cfg.InstanceLimitsFn = func() *InstanceLimits { return &limits }
 	cfg.LifecyclerConfig.JoinAfter = 0
 
-	i, err := prepareIngesterWithBlocksStorage(t, cfg, prometheus.NewRegistry())
+	reg := prometheus.NewRegistry()
+	i, err := prepareIngesterWithBlocksStorage(t, cfg, reg)
 	require.NoError(t, err)
 	require.NoError(t, services.StartAndAwaitRunning(context.Background(), i))
 	defer services.StopAndAwaitTerminated(context.Background(), i) //nolint:errcheck
@@ -6553,6 +6554,11 @@ func TestIngester_inflightPushRequests(t *testing.T) {
 
 		_, err := i.Push(ctx, req)
 		require.Equal(t, errTooManyInflightPushRequests, err)
+		require.NoError(t, testutil.GatherAndCompare(reg, bytes.NewBufferString(`
+		# HELP cortex_ingester_push_errors_total The total number of push errors per user.
+		# TYPE cortex_ingester_push_errors_total counter
+		cortex_ingester_push_errors_total{reason="tooManyInflightRequests",user="test"} 1
+	`), "cortex_ingester_push_errors_total"))
 		return nil
 	})
 

pkg/ingester/instance_limits.go

@@ -13,6 +13,8 @@ var (
 	errMaxSeriesLimitReached          = errors.New("cannot add series: ingesters's max series limit reached")
 	errTooManyInflightPushRequests    = errors.New("cannot push: too many inflight push requests in ingester")
 	errTooManyInflightQueryRequests   = errors.New("cannot push: too many inflight query requests in ingester")
+
+	pushErrTooManyInflightRequests = "tooManyInflightRequests"
 )
 
 // InstanceLimits describes limits used by ingester. Reaching any of these will result in error response to the call.

pkg/ingester/metrics.go

@@ -44,6 +44,7 @@ type ingesterMetrics struct {
 	memMetadataCreatedTotal *prometheus.CounterVec
 	memSeriesRemovedTotal   *prometheus.CounterVec
 	memMetadataRemovedTotal *prometheus.CounterVec
+	pushErrorsTotal         *prometheus.CounterVec
 
 	activeSeriesPerUser   *prometheus.GaugeVec
 	activeNHSeriesPerUser *prometheus.GaugeVec
@@ -165,6 +166,10 @@ func newIngesterMetrics(r prometheus.Registerer,
 			Name: "cortex_ingester_memory_metadata_removed_total",
 			Help: "The total number of metadata that were removed per user.",
 		}, []string{"user"}),
+		pushErrorsTotal: promauto.With(r).NewCounterVec(prometheus.CounterOpts{
+			Name: "cortex_ingester_push_errors_total",
+			Help: "The total number of push errors per user.",
+		}, []string{"user", "reason"}),
 
 		maxUsersGauge: promauto.With(r).NewGaugeFunc(prometheus.GaugeOpts{
 			Name:        instanceLimits,
@@ -295,6 +300,7 @@ func (m *ingesterMetrics) deletePerUserMetrics(userID string) {
 	m.activeNHSeriesPerUser.DeleteLabelValues(userID)
 	m.usagePerLabelSet.DeletePartialMatch(prometheus.Labels{"user": userID})
 	m.limitsPerLabelSet.DeletePartialMatch(prometheus.Labels{"user": userID})
+	m.pushErrorsTotal.DeletePartialMatch(prometheus.Labels{"user": userID})
 
 	if m.memSeriesCreatedTotal != nil {
 		m.memSeriesCreatedTotal.DeleteLabelValues(userID)