updates

When converting `rawRows` to in-memory parts, it is possible to do it
without a mutex.

Signed-off-by: Artem Fetishev <rtm@victoriametrics.com>

app/vmagent/opentelemetry/request_handler.go

@@ -63,6 +63,7 @@ func insertRows(at *auth.Token, tss []prompb.TimeSeries, mms []prompb.MetricMeta
 
 	rowsTotal := 0
 	tssDst := ctx.WriteRequest.Timeseries[:0]
+	mmsDst := ctx.WriteRequest.Metadata[:0]
 	labels := ctx.Labels[:0]
 	samples := ctx.Samples[:0]
 	for i := range tss {
@@ -82,7 +83,19 @@ func insertRows(at *auth.Token, tss []prompb.TimeSeries, mms []prompb.MetricMeta
 
 	var metadataTotal int
 	if prommetadata.IsEnabled() {
-		ctx.WriteRequest.Metadata = mms
+		for i := range mms {
+			mm := &mms[i]
+			mmsDst = append(mmsDst, prompb.MetricMetadata{
+				MetricFamilyName: mm.MetricFamilyName,
+				Help:             mm.Help,
+				Type:             mm.Type,
+				Unit:             mm.Unit,
+
+				AccountID: mm.AccountID,
+				ProjectID: mm.ProjectID,
+			})
+		}
+		ctx.WriteRequest.Metadata = mmsDst
 		metadataTotal = len(mms)
 	}
 

app/vmselect/promql/binary_op.go

@@ -172,7 +172,13 @@ func newBinaryOpFunc(bf func(left, right float64, isBool bool) float64) binaryOp
 			left = removeEmptySeries(left)
 			right = removeEmptySeries(right)
 		}
-		if len(left) == 0 || len(right) == 0 {
+		if len(left) == 0 && len(right) == 0 {
+			return nil, nil
+		}
+		if len(left) == 0 && bfa.be.FillLeft == nil {
+			return nil, nil
+		}
+		if len(right) == 0 && bfa.be.FillRight == nil {
 			return nil, nil
 		}
 		left, right, dst, err := adjustBinaryOpTags(bfa.be, left, right)
@@ -183,6 +189,8 @@ func newBinaryOpFunc(bf func(left, right float64, isBool bool) float64) binaryOp
 			logger.Panicf("BUG: len(left) must match len(right) and len(dst); got %d vs %d vs %d", len(left), len(right), len(dst))
 		}
 		isBool := bfa.be.Bool
+		fillLeft := bfa.be.FillLeft
+		fillRight := bfa.be.FillRight
 		for i, tsLeft := range left {
 			leftValues := tsLeft.Values
 			rightValues := right[i].Values
@@ -193,6 +201,19 @@ func newBinaryOpFunc(bf func(left, right float64, isBool bool) float64) binaryOp
 			}
 			for j, a := range leftValues {
 				b := rightValues[j]
+				leftIsNaN := math.IsNaN(a)
+				rightIsNaN := math.IsNaN(b)
+				// apply the fill value when either the left or right side is NaN, but not both.
+				if leftIsNaN && rightIsNaN {
+					dstValues[j] = bf(a, b, isBool)
+					continue
+				}
+				if leftIsNaN && fillLeft != nil {
+					a = fillLeft.N
+				}
+				if rightIsNaN && fillRight != nil {
+					b = fillRight.N
+				}
 				dstValues[j] = bf(a, b, isBool)
 			}
 		}
@@ -226,7 +247,7 @@ func adjustBinaryOpTags(be *metricsql.BinaryOpExpr, left, right []*timeseries) (
 		}
 	}
 
-	// Slow path: `vector op vector` or `a op {on|ignoring} {group_left|group_right} b`
+	// Slow path: `vector op vector` or `a op {on|ignoring} {group_left|group_right} {fill|fill_left|fill_right} b`
 	var rvsLeft, rvsRight []*timeseries
 	mLeft, mRight := createTimeseriesMapByTagSet(be, left, right)
 	joinOp := strings.ToLower(be.JoinModifier.Op)
@@ -239,10 +260,27 @@ func adjustBinaryOpTags(be *metricsql.BinaryOpExpr, left, right []*timeseries) (
 		// Add __name__ to groupTags if metric name must be preserved.
 		groupTags = append(groupTags[:len(groupTags):len(groupTags)], "__name__")
 	}
+	// Add missing keys from mRight to mLeft when fill_left()/fill() modifier is used
+	if be.FillLeft != nil {
+		for k := range mRight {
+			if _, ok := mLeft[k]; !ok {
+				mLeft[k] = nil
+			}
+		}
+	}
 	for k, tssLeft := range mLeft {
 		tssRight := mRight[k]
+		if len(tssLeft) == 0 {
+			if be.FillLeft == nil {
+				logger.Panicf("BUG: unexpected empty tssLeft for key %q when FillLeft is nil", k)
+			}
+			tssLeft = []*timeseries{newFillTimeseries(be, tssRight[0])}
+		}
 		if len(tssRight) == 0 {
-			continue
+			if be.FillRight == nil {
+				continue
+			}
+			tssRight = []*timeseries{newFillTimeseries(be, tssLeft[0])}
 		}
 		switch joinOp {
 		case "group_left":
@@ -287,6 +325,28 @@ func adjustBinaryOpTags(be *metricsql.BinaryOpExpr, left, right []*timeseries) (
 	return rvsLeft, rvsRight, dst, nil
 }
 
+// newFillTimeseries returns a time series filled with NaN values for the fill_left()/fill_right()/fill() modifiers.
+// These NaN values will be replaced later with the fill value if needed.
+func newFillTimeseries(be *metricsql.BinaryOpExpr, src *timeseries) *timeseries {
+	var ts timeseries
+	ts.CopyFromShallowTimestamps(src)
+	if !be.KeepMetricNames {
+		ts.MetricName.ResetMetricGroup()
+	}
+	groupTags := be.GroupModifier.Args
+	switch strings.ToLower(be.GroupModifier.Op) {
+	case "on":
+		ts.MetricName.RemoveTagsOn(groupTags)
+	default:
+		ts.MetricName.RemoveTagsIgnoring(groupTags)
+	}
+	values := ts.Values
+	for i := range values {
+		values[i] = math.NaN()
+	}
+	return &ts
+}
+
 func ensureSingleTimeseries(side string, be *metricsql.BinaryOpExpr, tss []*timeseries) error {
 	if len(tss) == 0 {
 		logger.Panicf("BUG: tss must contain at least one value")

app/vmselect/promql/eval.go

@@ -424,18 +424,7 @@ func evalBinaryOp(qt *querytracer.Tracer, ec *EvalConfig, be *metricsql.BinaryOp
 	if bf == nil {
 		return nil, fmt.Errorf(`unknown binary op %q`, be.Op)
 	}
-	var err error
-	var tssLeft, tssRight []*timeseries
-	switch strings.ToLower(be.Op) {
-	case "and", "if":
-		// Fetch right-side series at first, since it usually contains
-		// lower number of time series for `and` and `if` operator.
-		// This should produce more specific label filters for the left side of the query.
-		// This, in turn, should reduce the time to select series for the left side of the query.
-		tssRight, tssLeft, err = execBinaryOpArgs(qt, ec, be.Right, be.Left, be)
-	default:
-		tssLeft, tssRight, err = execBinaryOpArgs(qt, ec, be.Left, be.Right, be)
-	}
+	tssLeft, tssRight, err := execBinaryOpArgs(qt, ec, be)
 	if err != nil {
 		return nil, fmt.Errorf("cannot execute %q: %w", be.AppendString(nil), err)
 	}
@@ -451,6 +440,29 @@ func evalBinaryOp(qt *querytracer.Tracer, ec *EvalConfig, be *metricsql.BinaryOp
 	return rv, nil
 }
 
+// binaryOpEvalOrder might change the order of evaluation of the left and right sides of a binary operation,
+// when there is chance to push down common label filters from exprFirst to exprSecond in the following executions.
+func binaryOpEvalOrder(be *metricsql.BinaryOpExpr) (exprFirst, exprSecond metricsql.Expr) {
+	exprFirst, exprSecond = be.Left, be.Right
+	switch strings.ToLower(be.Op) {
+	case "and", "if":
+		// For `and` and `if`, fetch the right-side series first, since it usually contains
+		// fewer time series and yields more specific filters for the left side.
+		exprFirst, exprSecond = be.Right, be.Left
+	}
+	if be.FillLeft != nil && be.FillRight == nil {
+		// For `fill_left(<value>)`, the unmatched series can only come from the right side, so evaluate it first.
+		exprFirst, exprSecond = be.Right, be.Left
+	}
+	return exprFirst, exprSecond
+}
+
+// canPushdownCommonFilters decides if common label filters can be pushed down from one side of a binary operation to the other.
+//
+// Common filters cannot be pushed down when:
+// - the operator is `or` or `default`;
+// - either side is an aggregation function without explicit grouping;
+// - fill(<value>) modifier is used.
 func canPushdownCommonFilters(be *metricsql.BinaryOpExpr) bool {
 	switch strings.ToLower(be.Op) {
 	case "or", "default":
@@ -459,6 +471,10 @@ func canPushdownCommonFilters(be *metricsql.BinaryOpExpr) bool {
 	if isAggrFuncWithoutGrouping(be.Left) || isAggrFuncWithoutGrouping(be.Right) {
 		return false
 	}
+	// Filters cannot be propagated when fill(<value>) modifier is used.
+	if be.FillLeft != nil && be.FillRight != nil {
+		return false
+	}
 	return true
 }
 
@@ -470,7 +486,15 @@ func isAggrFuncWithoutGrouping(e metricsql.Expr) bool {
 	return len(afe.Modifier.Args) == 0
 }
 
-func execBinaryOpArgs(qt *querytracer.Tracer, ec *EvalConfig, exprFirst, exprSecond metricsql.Expr, be *metricsql.BinaryOpExpr) ([]*timeseries, []*timeseries, error) {
+func execBinaryOpArgs(qt *querytracer.Tracer, ec *EvalConfig, be *metricsql.BinaryOpExpr) ([]*timeseries, []*timeseries, error) {
+	exprFirst, exprSecond := binaryOpEvalOrder(be)
+	firstIsLeft := exprFirst == be.Left
+	sortResult := func(tssFirst, tssSecond []*timeseries) ([]*timeseries, []*timeseries, error) {
+		if firstIsLeft {
+			return tssFirst, tssSecond, nil
+		}
+		return tssSecond, tssFirst, nil
+	}
 	if canPushdownCommonFilters(be) {
 		// Execute binary operation in the following way:
 		//
@@ -512,7 +536,7 @@ func execBinaryOpArgs(qt *querytracer.Tracer, ec *EvalConfig, exprFirst, exprSec
 		if err != nil {
 			return nil, nil, err
 		}
-		return tssFirst, tssSecond, nil
+		return sortResult(tssFirst, tssSecond)
 	}
 
 	// Execute exprFirst and exprSecond sequentially if there are cacheable repeated subexpressions
@@ -568,7 +592,7 @@ func execBinaryOpArgs(qt *querytracer.Tracer, ec *EvalConfig, exprFirst, exprSec
 	if errSecond != nil {
 		return nil, nil, errSecond
 	}
-	return tssFirst, tssSecond, nil
+	return sortResult(tssFirst, tssSecond)
 }
 
 func shouldOptimizeRepeatedBinaryOpSubexprs(ec *EvalConfig, exprFirst, exprSecond metricsql.Expr) bool {

app/vmselect/promql/exec_test.go

@@ -4006,6 +4006,275 @@ func TestExecSuccess(t *testing.T) {
 		resultExpected := []netstorage.Result{r1, r2}
 		f(q, resultExpected)
 	})
+	t.Run(`vector + vector fill()`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(1, "foo", "common")
+			or label_set(2, "foo", "left_only")
+		) + fill(0) (
+			label_set(3, "foo", "common")
+			or label_set(4, "foo", "right_only")
+		), "foo")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("common"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{2, 2, 2, 2, 2, 2},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("left_only"),
+		}}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("right_only"),
+		}}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
+	t.Run(`vector + vector fill() both sides NaN case`, func(t *testing.T) {
+		t.Parallel()
+		q := `(
+			label_set(time() <= 1200, "foo", "common")
+		) + fill(10) (
+			label_set(time() >= 1600, "foo", "common")
+		)`
+		r := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{1010, 1210, nan, 1610, 1810, 2010},
+			Timestamps: timestampsExpected,
+		}
+		r.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("common"),
+		}}
+		resultExpected := []netstorage.Result{r}
+		f(q, resultExpected)
+	})
+	t.Run(`vector + vector fill_left() fill_right()`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(1, "foo", "common")
+			or label_set(2, "foo", "left_only")
+		) + fill_left(10) fill_right(20) (
+			label_set(3, "foo", "common")
+			or label_set(4, "foo", "right_only")
+		), "foo")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("common"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{22, 22, 22, 22, 22, 22},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("left_only"),
+		}}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{14, 14, 14, 14, 14, 14},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("right_only"),
+		}}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
+	t.Run(`vector + vector fill_right() only`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(1, "foo", "common")
+			or label_set(2, "foo", "left_only")
+		) + fill_right(20) (
+			label_set(3, "foo", "common")
+			or label_set(4, "foo", "right_only")
+		), "foo")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("common"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{22, 22, 22, 22, 22, 22},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("left_only"),
+		}}
+		resultExpected := []netstorage.Result{r1, r2}
+		f(q, resultExpected)
+	})
+	t.Run(`vector + vector on() fill()`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(1, "foo", "common", "extra", "l")
+			or label_set(2, "foo", "left_only", "extra", "l")
+		) + on(foo) fill(0) (
+			label_set(3, "foo", "common", "extra", "r")
+			or label_set(4, "foo", "right_only", "extra", "r")
+		), "foo")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("common"),
+		}}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{2, 2, 2, 2, 2, 2},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("left_only"),
+		}}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{4, 4, 4, 4, 4, 4},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.Tags = []storage.Tag{{
+			Key:   []byte("foo"),
+			Value: []byte("right_only"),
+		}}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
+	t.Run(`vector + vector on() group_left() fill_right()`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(1, "method", "get", "code", "500")
+			or label_set(2, "method", "get", "code", "404")
+			or label_set(3, "method", "put", "code", "501")
+		) + on(method) group_left() fill_right(0) (
+			label_set(10, "method", "get")
+		), "method", "code")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{12, 12, 12, 12, 12, 12},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("code"),
+				Value: []byte("404"),
+			},
+			{
+				Key:   []byte("method"),
+				Value: []byte("get"),
+			},
+		}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{11, 11, 11, 11, 11, 11},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("code"),
+				Value: []byte("500"),
+			},
+			{
+				Key:   []byte("method"),
+				Value: []byte("get"),
+			},
+		}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{3, 3, 3, 3, 3, 3},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("code"),
+				Value: []byte("501"),
+			},
+			{
+				Key:   []byte("method"),
+				Value: []byte("put"),
+			},
+		}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
+	t.Run(`vector / vector ignoring() fill()`, func(t *testing.T) {
+		t.Parallel()
+		q := `sort_by_label((
+			label_set(6, "method", "get", "code", "500")
+			or label_set(1, "method", "put", "code", "500")
+		) / ignoring(code) fill(0) (
+			label_set(12, "method", "get")
+			or label_set(5, "method", "post")
+			or label_set(10, "method", "put")
+		), "method")`
+		r1 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{0.5, 0.5, 0.5, 0.5, 0.5, 0.5},
+			Timestamps: timestampsExpected,
+		}
+		r1.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("method"),
+				Value: []byte("get"),
+			},
+		}
+		r2 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{0, 0, 0, 0, 0, 0},
+			Timestamps: timestampsExpected,
+		}
+		r2.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("method"),
+				Value: []byte("post"),
+			},
+		}
+		r3 := netstorage.Result{
+			MetricName: metricNameExpected,
+			Values:     []float64{0.1, 0.1, 0.1, 0.1, 0.1, 0.1},
+			Timestamps: timestampsExpected,
+		}
+		r3.MetricName.Tags = []storage.Tag{
+			{
+				Key:   []byte("method"),
+				Value: []byte("put"),
+			},
+		}
+		resultExpected := []netstorage.Result{r1, r2, r3}
+		f(q, resultExpected)
+	})
 	t.Run(`histogram_quantile(scalar)`, func(t *testing.T) {
 		t.Parallel()
 		q := `histogram_quantile(0.6, time())`

app/vmstorage/main.go

@@ -30,6 +30,11 @@ var (
 		"See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#retention. See also -retentionFilter")
 	futureRetention = flagutil.NewRetentionDuration("futureRetention", "2d", "Data with timestamps bigger than now+futureRetention is automatically deleted. "+
 		"The minimum futureRetention is 2 days. See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#retention")
+	maxBackfillAge = flagutil.NewRetentionDuration("maxBackfillAge", "0", "The maximum allowed age for the ingested samples with historical timestamps. "+
+		"Samples with timestamps older than now-maxBackfillAge are rejected during data ingestion. "+
+		"By default, or when set to 0, -maxBackfillAge equals to -retentionPeriod, e.g. it is unlimited within the configured retention. "+
+		"This can be useful for limiting ingestion of historical samples, for example, when older data has been moved to another storage tier. "+
+		"See https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#retention")
 	vmselectAddr                  = flag.String("vmselectAddr", "", "TCP address to accept connections from vmselect services")
 	vmselectDisableRPCCompression = flag.Bool("rpc.disableCompression", false, "Whether to disable compression of the data sent from vmstorage to vmselect. "+
 		"This reduces CPU usage at the cost of higher network bandwidth usage")
@@ -146,6 +151,7 @@ func Init(vmselectMaxConcurrentRequests int, vmselectMaxQueueDuration time.Durat
 	opts := storage.OpenOptions{
 		Retention:                   retentionPeriod.Duration(),
 		FutureRetention:             futureRetention.Duration(),
+		MaxBackfillAge:              maxBackfillAge.Duration(),
 		DenyQueriesOutsideRetention: *denyQueriesOutsideRetention,
 		MaxHourlySeries:             getMaxHourlySeries(),
 		MaxDailySeries:              getMaxDailySeries(),
@@ -467,8 +473,10 @@ func (vms *VMStorage) writeStorageMetrics(w io.Writer) {
 	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="storage/inmemory"}`, tm.InmemorySizeBytes)
 	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="storage/small"}`, tm.SmallSizeBytes)
 	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="storage/big"}`, tm.BigSizeBytes)
+	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="storage/metaindex"}`, tm.MetaindexSizeBytes)
 	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="indexdb/inmemory"}`, idbm.InmemorySizeBytes)
 	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="indexdb/file"}`, idbm.FileSizeBytes)
+	metrics.WriteGaugeUint64(w, `vm_data_size_bytes{type="indexdb/metaindex"}`, idbm.MetaindexSizeBytes)
 
 	metrics.WriteCounterUint64(w, `vm_rows_received_by_storage_total`, m.RowsReceivedTotal)
 	metrics.WriteCounterUint64(w, `vm_rows_added_to_storage_total`, m.RowsAddedTotal)

app/vmui/packages/vmui/src/layouts/Header/HeaderNav/HeaderNav.tsx

@@ -17,10 +17,12 @@ interface HeaderNavProps {
 const HeaderNav: FC<HeaderNavProps> = ({ color, background, direction }) => {
   const { pathname } = useLocation();
   const [activeMenu, setActiveMenu] = useState(pathname);
+  const [openMenu, setOpenMenu] = useState<string | null>(null);
   const menu = useNavigationMenu();
 
   useEffect(() => {
     setActiveMenu(pathname);
+    setOpenMenu(null);
   }, [pathname]);
 
   return (
@@ -41,6 +43,8 @@ const HeaderNav: FC<HeaderNavProps> = ({ color, background, direction }) => {
               color={color}
               background={background}
               direction={direction}
+              openMenu={openMenu}
+              setOpenMenu={setOpenMenu}
             />
           )
           : (

app/vmui/packages/vmui/src/layouts/Header/HeaderNav/NavSubItem.tsx

@@ -1,11 +1,8 @@
-import { FC, useRef, useState } from "preact/compat";
-import { useLocation } from "react-router-dom";
+import { FC, useRef, useState, Dispatch, SetStateAction } from "preact/compat";
 import classNames from "classnames";
 import { ArrowDropDownIcon } from "../../../components/Main/Icons";
 import Popper from "../../../components/Main/Popper/Popper";
 import NavItem from "./NavItem";
-import { useEffect } from "react";
-import useBoolean from "../../../hooks/useBoolean";
 import { NavigationItem, NavigationItemType } from "../../../router/navigation";
 
 interface NavItemProps {
@@ -15,6 +12,8 @@ interface NavItemProps {
   color?: string
   background?: string
   direction?: "row" | "column"
+  openMenu: string | null,
+  setOpenMenu: Dispatch<SetStateAction<string | null>>,
 }
 
 const NavSubItem: FC<NavItemProps> = ({
@@ -23,21 +22,18 @@ const NavSubItem: FC<NavItemProps> = ({
   color,
   background,
   submenu,
-  direction = "row"
+  direction = "row",
+  openMenu,
+  setOpenMenu,
 }) => {
-  const { pathname } = useLocation();
-
   const [menuTimeout, setMenuTimeout] = useState<NodeJS.Timeout | null>(null);
   const buttonRef = useRef<HTMLDivElement>(null);
 
-  const {
-    value: openSubmenu,
-    setFalse: handleCloseSubmenu,
-    setTrue: setOpenSubmenu,
-  } = useBoolean(false);
+  const openSubmenu = openMenu === label;
+  const handleCloseSubmenu = () => setOpenMenu(prev => (prev === label ? null : prev));
 
   const handleOpenSubmenu = () => {
-    if (direction === "row" || !openSubmenu) setOpenSubmenu();
+    if (direction === "row" || !openSubmenu) setOpenMenu(label);
     if (direction === "column" && openSubmenu) handleCloseSubmenu();
     if (direction === "row" && menuTimeout) clearTimeout(menuTimeout);
   };
@@ -52,10 +48,6 @@ const NavSubItem: FC<NavItemProps> = ({
     if (menuTimeout) clearTimeout(menuTimeout);
   };
 
-  useEffect(() => {
-    handleCloseSubmenu();
-  }, [pathname]);
-
   return (
     <div
       className={classNames({

apptest/tests/max_backfill_age_test.go

@@ -0,0 +1,222 @@
+package tests
+
+import (
+	"fmt"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/apptest"
+)
+
+func TestSingleMaxBackfillAge(t *testing.T) {
+	tc := apptest.NewTestCase(t)
+	defer tc.Stop()
+
+	opts := maxBackfillAgeOpts{
+		start: func(retentionPeriod, maxBackfillAge string) apptest.PrometheusWriteQuerier {
+			return tc.MustStartVmsingle("vmsingle", []string{
+				"-storageDataPath=" + filepath.Join(tc.Dir(), "vmsingle"),
+				"-retentionPeriod=" + retentionPeriod,
+				"-maxBackfillAge=" + maxBackfillAge,
+			})
+		},
+		stop: func() {
+			tc.StopApp("vmsingle")
+		},
+	}
+
+	testMaxBackfillAge(tc, opts)
+}
+
+func TestClusterMaxBackfillAge(t *testing.T) {
+	tc := apptest.NewTestCase(t)
+	defer tc.Stop()
+
+	opts := maxBackfillAgeOpts{
+		start: func(retentionPeriod, maxBackfillAge string) apptest.PrometheusWriteQuerier {
+			return tc.MustStartCluster(&apptest.ClusterOptions{
+				Vmstorage1Instance: "vmstorage1",
+				Vmstorage1Flags: []string{
+					"-storageDataPath=" + filepath.Join(tc.Dir(), "vmstorage1"),
+					"-retentionPeriod=" + retentionPeriod,
+					"-maxBackfillAge=" + maxBackfillAge,
+				},
+				Vmstorage2Instance: "vmstorage2",
+				Vmstorage2Flags: []string{
+					"-storageDataPath=" + filepath.Join(tc.Dir(), "vmstorage2"),
+					"-retentionPeriod=" + retentionPeriod,
+					"-maxBackfillAge=" + maxBackfillAge,
+				},
+				VminsertInstance: "vminsert",
+				VminsertFlags:    []string{},
+				VmselectInstance: "vmselect",
+				VmselectFlags:    []string{},
+			})
+		},
+		stop: func() {
+			tc.StopApp("vminsert")
+			tc.StopApp("vmselect")
+			tc.StopApp("vmstorage1")
+			tc.StopApp("vmstorage2")
+		},
+	}
+
+	testMaxBackfillAge(tc, opts)
+}
+
+type maxBackfillAgeOpts struct {
+	start func(retentionPeriod, maxBackfillAge string) apptest.PrometheusWriteQuerier
+	stop  func()
+}
+
+func testMaxBackfillAge(tc *apptest.TestCase, opts maxBackfillAgeOpts) {
+	t := tc.T()
+
+	assertSeries := func(app apptest.PrometheusQuerier, prefix string, start, end int64, want []map[string]string) {
+		t.Helper()
+
+		query := fmt.Sprintf(`{__name__=~"metric_%s.*"}`, prefix)
+		tc.Assert(&apptest.AssertOptions{
+			Msg: "unexpected /api/v1/series response",
+			Got: func() any {
+				return app.PrometheusAPIV1Series(t, query, apptest.QueryOpts{
+					Start: fmt.Sprintf("%d", start),
+					End:   fmt.Sprintf("%d", end),
+				}).Sort()
+			},
+			Want: &apptest.PrometheusAPIV1SeriesResponse{
+				Status: "success",
+				Data:   want,
+			},
+			FailNow: true,
+		})
+	}
+
+	assertQueryResults := func(app apptest.PrometheusQuerier, prefix string, start, end, step int64, want []*apptest.QueryResult) {
+		t.Helper()
+
+		query := fmt.Sprintf(`{__name__=~"metric_%s.*"}`, prefix)
+		tc.Assert(&apptest.AssertOptions{
+			Msg: "unexpected /api/v1/query_range response",
+			Got: func() any {
+				return app.PrometheusAPIV1QueryRange(t, query, apptest.QueryOpts{
+					Start:       fmt.Sprintf("%d", start),
+					End:         fmt.Sprintf("%d", end),
+					Step:        fmt.Sprintf("%dms", step),
+					MaxLookback: fmt.Sprintf("%dms", step-1),
+					NoCache:     "1",
+				})
+			},
+			Want: &apptest.PrometheusAPIV1QueryResponse{
+				Status: "success",
+				Data: &apptest.QueryData{
+					ResultType: "matrix",
+					Result:     want,
+				},
+			},
+			FailNow: true,
+		})
+	}
+
+	const numMetrics = 1000
+	now := time.Now().UTC()
+	var start, end, step int64
+	emptySeries := []map[string]string{}
+	emptyQueryResults := []*apptest.QueryResult{}
+
+	// Start sut with the same -retentionPeriod and -maxBackfillAge.
+	sut := opts.start("1y", "1y")
+
+	// Verify that samples older than the retention period are rejected.
+	start = now.Add(-365 * 24 * time.Hour).Add(-time.Hour).UnixMilli()
+	end = now.Add(-365 * 24 * time.Hour).UnixMilli()
+	step = (end - start) / numMetrics
+	outsideRetention := genMaxBackfillAgeData("outside_retention", numMetrics, start, step)
+	sut.PrometheusAPIV1ImportPrometheus(t, outsideRetention.samples, apptest.QueryOpts{})
+	sut.ForceFlush(t)
+	assertSeries(sut, "outside_retention", start, end, emptySeries)
+	assertQueryResults(sut, "outside_retention", start, end, step, emptyQueryResults)
+
+	// Verify that samples within the retention period are accepted and
+	// searcheable.
+	start = now.Add(-365 * 24 * time.Hour).Add(time.Hour).UnixMilli()
+	end = now.Add(-365 * 24 * time.Hour).Add(2 * time.Hour).UnixMilli()
+	step = (end - start) / numMetrics
+	insideRetention := genMaxBackfillAgeData("inside_retention", numMetrics, start, step)
+	sut.PrometheusAPIV1ImportPrometheus(t, insideRetention.samples, apptest.QueryOpts{})
+	sut.ForceFlush(t)
+	assertSeries(sut, "inside_retention", start, end, insideRetention.wantSeries)
+	assertQueryResults(sut, "inside_retention", start, end, step, insideRetention.wantQueryResults)
+
+	// Restart sut with -maxBackfillAge shorter than the -retentionPeriod.
+	opts.stop()
+	sut = opts.start("1y", "6M")
+
+	// Verify that new samples older than max backfill age but still within the
+	// retention period are rejected but existing samples are still searcheable.
+	start = now.Add(-365 * 24 * time.Hour).Add(time.Hour).UnixMilli()
+	end = now.Add(-365 * 24 * time.Hour).Add(2 * time.Hour).UnixMilli()
+	step = (end - start) / numMetrics
+	insideRetention2 := genMaxBackfillAgeData("inside_retention2", numMetrics, start, step)
+	sut.PrometheusAPIV1ImportPrometheus(t, insideRetention2.samples, apptest.QueryOpts{})
+	sut.ForceFlush(t)
+	assertSeries(sut, "inside_retention2", start, end, emptySeries)
+	assertQueryResults(sut, "inside_retention2", start, end, step, emptyQueryResults)
+	assertSeries(sut, "inside_retention", start, end, insideRetention.wantSeries)
+	assertQueryResults(sut, "inside_retention", start, end, step, insideRetention.wantQueryResults)
+
+	// Verify that the metrics that are outside the backfill window can still
+	// be deleted.
+	sut.PrometheusAPIV1AdminTSDBDeleteSeries(t, `{__name__=~".*inside_retention.*"}`, apptest.QueryOpts{})
+	sut.ForceFlush(t)
+	assertSeries(sut, "inside_retention", start, end, emptySeries)
+	assertQueryResults(sut, "inside_retention", start, end, step, emptyQueryResults)
+
+	// Verify that the samples that are within the backfill window are accepted
+	// and searchable.
+	start = now.Add(-180 * 24 * time.Hour).UnixMilli()
+	end = now.Add(-180 * 24 * time.Hour).Add(1 * time.Hour).UnixMilli()
+	step = (end - start) / numMetrics
+	insideMaxBackfillAge := genMaxBackfillAgeData("inside_max_backfill_age", numMetrics, start, step)
+	sut.PrometheusAPIV1ImportPrometheus(t, insideMaxBackfillAge.samples, apptest.QueryOpts{})
+	sut.ForceFlush(t)
+	assertSeries(sut, "inside_max_backfill_age", start, end, insideMaxBackfillAge.wantSeries)
+	assertQueryResults(sut, "inside_max_backfill_age", start, end, step, insideMaxBackfillAge.wantQueryResults)
+
+	opts.stop()
+}
+
+type maxBackfillAgeData struct {
+	samples          []string
+	wantSeries       []map[string]string
+	wantQueryResults []*apptest.QueryResult
+}
+
+func genMaxBackfillAgeData(prefix string, numMetrics, start, step int64) maxBackfillAgeData {
+	samples := make([]string, numMetrics)
+	wantSeries := make([]map[string]string, numMetrics)
+	wantQueryResults := make([]*apptest.QueryResult, numMetrics)
+	for i := range numMetrics {
+		metricName := fmt.Sprintf("metric_%s_%04d", prefix, i)
+		labelName := fmt.Sprintf("label_%s_%04d", prefix, i)
+		labelValue := fmt.Sprintf("value_%s_%04d", prefix, i)
+		value := i
+		timestamp := start + i*step
+		samples[i] = fmt.Sprintf(`%s{%s="value", label="%s"} %d %d`, metricName, labelName, labelValue, value, timestamp)
+		wantSeries[i] = map[string]string{
+			"__name__": metricName,
+			labelName:  "value",
+			"label":    labelValue,
+		}
+		wantQueryResults[i] = &apptest.QueryResult{
+			Metric: map[string]string{
+				"__name__": metricName,
+				labelName:  "value",
+				"label":    labelValue,
+			},
+			Samples: []*apptest.Sample{{Timestamp: timestamp, Value: float64(value)}},
+		}
+	}
+	return maxBackfillAgeData{samples, wantSeries, wantQueryResults}
+}

docs/victoriametrics/Articles.md

@@ -154,6 +154,13 @@ See [our blog](https://victoriametrics.com/blog) for the latest articles written
 * [Why irate from Prometheus doesn't capture spikes](https://valyala.medium.com/why-irate-from-prometheus-doesnt-capture-spikes-45f9896d7832)
 * [VictoriaMetrics: PromQL compliance](https://medium.com/@romanhavronenko/victoriametrics-promql-compliance-d4318203f51e)
 * [How do open source solutions for logs work: Elasticsearch, Loki and VictoriaLogs](https://itnext.io/how-do-open-source-solutions-for-logs-work-elasticsearch-loki-and-victorialogs-9f7097ecbc2f)
+* [How vmagent Collects and Ships Metrics Fast with Aggregation, Deduplication, and More](https://victoriametrics.com/blog/vmagent-how-it-works/)
+* [When Metrics Meet vminsert: A Data-Delivery Story](https://victoriametrics.com/blog/vminsert-how-it-works/)
+* [How vmstorage Handles Data Ingestion From vminsert](https://victoriametrics.com/blog/vmstorage-how-it-handles-data-ingestion/)
+* [How vmstorage Processes Data: Retention, Merging, Deduplication,...](https://victoriametrics.com/blog/vmstorage-retention-merging-deduplication/)
+* [How vmstorage's IndexDB Works](https://victoriametrics.com/blog/vmstorage-how-indexdb-works/)
+* [How vmstorage Handles Query Requests From vmselect](https://victoriametrics.com/blog/vmstorage-how-it-handles-query-requests/)
+* [Inside vmselect: The Query Processing Engine of VictoriaMetrics](https://victoriametrics.com/blog/vmselect-how-it-works/)
 
 ### Tutorials, guides and how-to articles
 
@@ -171,6 +178,12 @@ See [our guides](https://docs.victoriametrics.com/guides/) for the up-to-date gu
 * [Prometheus storage: tech terms for humans](https://valyala.medium.com/prometheus-storage-technical-terms-for-humans-4ab4de6c3d48)
 * [Cardinality explorer](https://victoriametrics.com/blog/cardinality-explorer/)
 * [Rules backfilling via vmalert](https://victoriametrics.com/blog/rules-replay/)
+* [vmagent: Key Features Explained in Under 15 Minutes](https://victoriametrics.com/blog/vmagent-key-features-explained/)
+* [Prometheus Metrics Explained: Counters, Gauges, Histograms & Summaries](https://victoriametrics.com/blog/prometheus-monitoring-metrics-counters-gauges-histogram-summaries/)
+* [Prometheus Monitoring: Instant Queries and Range Queries Explained](https://victoriametrics.com/blog/prometheus-monitoring-instant-range-query/)
+* [Prometheus Monitoring: Functions, Subqueries, Operators, and Modifiers](https://victoriametrics.com/blog/prometheus-monitoring-function-operator-modifier/)
+* [Prometheus Alerting 101: Rules, Recording Rules, and Alertmanager](https://victoriametrics.com/blog/alerting-recording-rules-alertmanager/)
+* [Alerting Best Practices](https://victoriametrics.com/blog/alerting-best-practices/)
 
 ### Other articles
 

docs/victoriametrics/Cluster-VictoriaMetrics.md

@@ -57,6 +57,10 @@ Each service may scale independently and may run on the most suitable hardware.
 This is a [shared nothing architecture](https://en.wikipedia.org/wiki/Shared-nothing_architecture).
 It increases cluster availability, and simplifies cluster maintenance as well as cluster scaling.
 
+> Further reading, deep dives into how each service works internally:
+> - `vmstorage`: [How vmstorage Handles Data Ingestion From vminsert](https://victoriametrics.com/blog/vmstorage-how-it-handles-data-ingestion/), [How vmstorage's IndexDB Works](https://victoriametrics.com/blog/vmstorage-how-indexdb-works/), [How vmstorage Handles Query Requests From vmselect](https://victoriametrics.com/blog/vmstorage-how-it-handles-query-requests/).
+> - `vmselect`: [Inside vmselect: The Query Processing Engine of VictoriaMetrics](https://victoriametrics.com/blog/vmselect-how-it-works/).
+
 ![Cluster Scheme](Cluster-VictoriaMetrics-components.webp)
 
 ## vmui
@@ -834,7 +838,7 @@ This ensures that incoming metrics are evenly distributed across all `vmstorage`
 The downside is that a single slow vmstorage node can throttle the entire cluster.
 
 When `-disableRerouting=false` is enabled on `vminsert`, 
-the cluster will automatically re-route writes away from the slowest vmstorage node to preserve maximum ingestion throughput.
+the cluster will automatically [re-route writes](https://victoriametrics.com/blog/vminsert-how-it-works/#31-rerouting) away from the slowest vmstorage node to preserve maximum ingestion throughput.
 
 Re-routing occurs only when all of the following conditions hold:
 - the storage send buffer is full.
@@ -877,7 +881,7 @@ See also [resource usage limits docs](#resource-usage-limits).
 
 ## Rebalancing
 
-Every `vminsert` node evenly spreads (shards) incoming data among `vmstorage` nodes specified in the `-storageNode` command-line flag.
+Every `vminsert` node [evenly spreads (shards) incoming data](https://victoriametrics.com/blog/vminsert-how-it-works/#3-sharding-and-buffering) among `vmstorage` nodes specified in the `-storageNode` command-line flag.
 This guarantees even distribution of the ingested data among `vmstorage` nodes. When new `vmstorage` nodes are added to the `-storageNode`
 command-line flag at `vminsert`, then only newly ingested data is distributed evenly among old and new `vmstorage` nodes, while
 historical data remains on the old `vmstorage` nodes. This speeds up data ingestion and querying for the majority of production workloads,
@@ -1025,7 +1029,7 @@ By default, VictoriaMetrics offloads replication to the underlying storage point
 which guarantees data durability. VictoriaMetrics supports application-level replication if replicated durable persistent disks cannot be used for some reason.
 
 The replication can be enabled by passing `-replicationFactor=N` command-line flag to `vminsert`. This instructs `vminsert` to store `N` copies for every ingested sample
-on `N` distinct `vmstorage` nodes. This guarantees that all the stored data remains available for querying if up to `N-1` `vmstorage` nodes are unavailable.
+on `N` distinct `vmstorage` nodes. This guarantees that all the stored data remains available for querying if up to `N-1` `vmstorage` nodes are unavailable. See [how `vminsert` replicates each sample to `N` `vmstorage` nodes](https://victoriametrics.com/blog/vminsert-how-it-works/#4-replication-and-sending-data-to-vmstorage) for details.
 
 Passing `-replicationFactor=N` command-line flag to `vmselect` instructs it to not mark responses as `partial` if less than `-replicationFactor` vmstorage nodes are unavailable during the query.
 See [cluster availability docs](#cluster-availability) for details.
@@ -1060,7 +1064,7 @@ deduplication can't be guaranteed when samples and sample duplicates for the sam
 - when `vmstorage` node has no enough capacity for processing incoming data stream. Then `vminsert` re-routes new samples to other `vmstorage` nodes.
 
 It is recommended to set **the same** `-dedup.minScrapeInterval` command-line flag value to both `vmselect` and `vmstorage` nodes
-to ensure query results consistency, even if storage layer didn't complete deduplication yet.
+to ensure query results consistency, even if [storage layer didn't complete deduplication](https://victoriametrics.com/blog/vmstorage-retention-merging-deduplication/#deduplication) yet.
 
 ## Metrics Metadata
 

docs/victoriametrics/MetricsQL.md

@@ -59,6 +59,7 @@ Other PromQL functionality should work the same in MetricsQL.
 
 MetricsQL implements [PromQL](https://medium.com/@valyala/promql-tutorial-for-beginners-9ab455142085)
 and provides additional functionality mentioned below, which is aimed towards solving practical cases.
+See [operators and modifiers in MetricsQL](https://victoriametrics.com/blog/prometheus-monitoring-function-operator-modifier/#operators--modifiers) for details.
 Feel free [filing a feature request](https://github.com/VictoriaMetrics/VictoriaMetrics/issues) if you think MetricsQL misses certain useful functionality.
 
 This functionality can be evaluated at [VictoriaMetrics demo playground](https://play.victoriametrics.com/select/accounting/1/6a716b0f-38bc-4856-90ce-448fd713e3fe/prometheus/graph/)
@@ -156,7 +157,7 @@ MetricsQL provides the following functions:
 
 ### Rollup functions
 
-**Rollup functions** (aka range functions or window functions) calculate rollups over [raw samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples)
+**[Rollup functions](https://victoriametrics.com/blog/prometheus-monitoring-function-operator-modifier/#rollup-functions)** (aka range functions or window functions) calculate rollups over [raw samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples)
 on the given lookbehind window for the [selected time series](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#filtering).
 For example, `avg_over_time(temperature[24h])` calculates the average temperature over [raw samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) for the last 24 hours.
 
@@ -348,6 +349,7 @@ If the lookbehind window is skipped in square brackets, then it is automatically
 passed to [/api/v1/query_range](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#range-query) or [/api/v1/query](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#instant-query),
 while `scrape_interval` is the interval between [raw samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) for the selected time series.
 This allows avoiding unexpected gaps on the graph when `step` is smaller than the `scrape_interval`.
+See [range vector selectors](https://victoriametrics.com/blog/prometheus-monitoring-instant-range-query/#range-vector-selector) for details.
 
 #### delta
 
@@ -1110,7 +1112,7 @@ See also [zscore](#zscore), [range_trim_zscore](#range_trim_zscore) and [outlier
 
 ### Transform functions
 
-**Transform functions** calculate transformations over [rollup results](#rollup-functions).
+**[Transform functions](https://victoriametrics.com/blog/prometheus-monitoring-function-operator-modifier/#transformation-functions)** calculate transformations over [rollup results](#rollup-functions).
 For example, `abs(delta(temperature[24h]))` calculates the absolute value for every point of every time series
 returned from the rollup `delta(temperature[24h])`.
 
@@ -1846,7 +1848,7 @@ This function is supported by PromQL.
 
 ### Label manipulation functions
 
-**Label manipulation functions** perform manipulations with labels on the selected [rollup results](#rollup-functions).
+**[Label manipulation functions](https://victoriametrics.com/blog/prometheus-monitoring-function-operator-modifier/#label-manipulation-functions)** perform manipulations with labels on the selected [rollup results](#rollup-functions).
 
 Additional details:
 
@@ -2016,7 +2018,7 @@ See also [sort_by_label_numeric](#sort_by_label_numeric) and [sort_by_label_desc
 
 ### Aggregate functions
 
-**Aggregate functions** calculate aggregates over groups of [rollup results](#rollup-functions).
+**[Aggregate functions](https://victoriametrics.com/blog/prometheus-monitoring-function-operator-modifier/#aggregation-functions)** calculate aggregates over groups of [rollup results](#rollup-functions).
 
 Additional details:
 
@@ -2339,7 +2341,7 @@ See also [zscore_over_time](#zscore_over_time), [range_trim_zscore](#range_trim_
 
 ## Subqueries
 
-MetricsQL supports and extends PromQL subqueries. See [this article](https://valyala.medium.com/prometheus-subqueries-in-victoriametrics-9b1492b720b3) for details.
+MetricsQL supports and extends PromQL [subqueries](https://victoriametrics.com/blog/prometheus-monitoring-function-operator-modifier/#subqueries). See [this article](https://valyala.medium.com/prometheus-subqueries-in-victoriametrics-9b1492b720b3) for details.
 Any [rollup function](#rollup-functions) for something other than [series selector](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#filtering) form a subquery.
 Nested rollup functions can be implicit thanks to the [implicit query conversions](#implicit-query-conversions).
 For example, `delta(sum(m))` is implicitly converted to `delta(sum(default_rollup(m))[1i:1i])`, so it becomes a subquery,

docs/victoriametrics/README.md

@@ -403,6 +403,9 @@ Resources:
 * [Cardinality explorer blog post](https://victoriametrics.com/blog/cardinality-explorer/).
 * [skills/victoriametrics-cardinality-analysis](https://github.com/VictoriaMetrics/skills/blob/main/plugins/diagnostics/skills/victoriametrics-cardinality-analysis/SKILL.md) for [agent-assisted](https://docs.victoriametrics.com/ai-tools/#agent-skills) analysis.
 
+For monitoring or alerting on cardinality, use [vmestimator](https://docs.victoriametrics.com/victoriametrics/vmestimator/).
+vmestimator measures metrics cardinality across [arbitrary label dimensions](https://docs.victoriametrics.com/victoriametrics/vmestimator/#basic) in real time and exposes the [results as metrics](https://docs.victoriametrics.com/victoriametrics/vmestimator/#cardinality-metrics).
+
 ### Cardinality explorer statistic inaccuracy
 
 In [cluster version of VictoriaMetrics](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/) each vmstorage tracks the stored time series individually.
@@ -1311,6 +1314,7 @@ per each `-dedup.minScrapeInterval` discrete interval if `-dedup.minScrapeInterv
 For example, `-dedup.minScrapeInterval=60s` would leave a single raw sample with the biggest timestamp per each discrete
 `60s` interval.
 This aligns with the [staleness rules in Prometheus](https://prometheus.io/docs/prometheus/latest/querying/basics/#staleness).
+See [how deduplication works](https://victoriametrics.com/blog/vmstorage-retention-merging-deduplication/#deduplication) for details.
 
 If multiple raw samples have **the same timestamp** on the given `-dedup.minScrapeInterval` discrete interval,
 then the sample with **the biggest value** is kept.
@@ -1394,7 +1398,7 @@ in separate files under `part` directory - `timestamps.bin` and `values.bin`.
 The `part` directory also contains `index.bin` and `metaindex.bin` files - these files contain index
 for fast block lookups, which belong to the given `TSID` and cover the given time range.
 
-`Parts` are periodically merged into bigger parts in background. The background merge provides the following benefits:
+`Parts` are periodically merged into bigger parts in background. The [background merge](https://victoriametrics.com/blog/vmstorage-retention-merging-deduplication/#merge-process) provides the following benefits:
 
 * keeping the number of data files under control, so they don't exceed limits on open files
 * improved data compression, since bigger parts are usually compressed better than smaller parts
@@ -1529,6 +1533,7 @@ are **eventually deleted** during [background merge](https://medium.com/@valyala
 The time range covered by data part is **not limited by retention period unit**. One data part can cover hours or days of
 data. Hence, a data part can be deleted only **when fully outside the configured retention**.
 See more about partitions and parts in the [Storage section](#storage).
+See [how retention frees disk space](https://victoriametrics.com/blog/vmstorage-retention-merging-deduplication/#retention-free-disk-space-guard-and-downsampling) for details.
 
 The maximum disk space usage for a given `-retentionPeriod` is going to be (`-retentionPeriod` + 1) months.
 For example, if `-retentionPeriod` is set to 1, data for January is deleted on March 1st.
@@ -1548,6 +1553,18 @@ For example, the following command starts VictoriaMetrics, which accepts samples
 /path/to/victoria-metrics -futureRetention=1y
 ```
 
+By default, VictoriaMetrics accepts samples with timestamps as old as the configured `-retentionPeriod` allows, e.g. it accepts backfilled
+historical data as long as it fits into the retention. If you need rejecting samples with historical timestamps older than the specified
+duration, then specify the desired duration via the `-maxBackfillAge` command-line flag. This can be useful for limiting ingestion of
+historical samples, for example, when older data has been moved to another storage tier (nvme/hdd, hot/cold).
+`-maxBackfillAge` cannot exceed the configured `-retentionPeriod` - bigger values are automatically clamped to `-retentionPeriod`.
+
+For example, the following command starts VictoriaMetrics, which rejects ingested samples with timestamps older than 2 days:
+
+```sh
+/path/to/victoria-metrics -maxBackfillAge=2d
+```
+
 ### Multiple retentions
 
 Distinct retentions for distinct time series can be configured via [retention filters](#retention-filters)
@@ -1619,6 +1636,7 @@ See how to request a [free trial license](https://victoriametrics.com/products/e
 This command-line flag instructs leaving the last sample per each `interval` for [time series](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#time-series)
 [samples](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#raw-samples) older than the `offset`. The `offset` must be a multiple of `interval`. For example, `-downsampling.period=30d:5m` instructs leaving the last sample
 per each 5-minute interval for samples older than 30 days, while the rest of samples are dropped.
+See [Enterprise downsampling internals](https://victoriametrics.com/blog/vmstorage-retention-merging-deduplication/#retention-filters-and-downsampling-enterprise-plan) for details.
 
 The `-downsampling.period` command-line flag can be specified multiple times in order to apply different downsampling levels for different time ranges (aka multi-level downsampling).
 For example, `-downsampling.period=30d:5m,180d:1h` instructs leaving the last sample per each 5-minute interval for samples older than 30 days,

docs/victoriametrics/changelog/CHANGELOG.md

@@ -26,7 +26,17 @@ See also [LTS releases](https://docs.victoriametrics.com/victoriametrics/lts-rel
 
 ## tip
 
+* FEATURE: [MetricsQL](https://docs.victoriametrics.com/victoriametrics/metricsql/): support `fill` modifiers to allow missing series on either side of a binary operation to be filled with a provided default value. See [#10598](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10598).
+* FEATURE: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/) and [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/): Improve background discovery performance for [http_sd](https://docs.victoriametrics.com/victoriametrics/sd_configs/#http_sd_configs) discovery. See [#8838](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/8838).
+* FEATURE: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/) and [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/): allow overriding `max_scrape_size` on a per-target basis via the `__max_scrape_size__` label during target relabeling. See [#11188](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11188).
+* FEATURE: [vmstorage](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/) and [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/): add `-maxBackfillAge` command-line flag for limiting ingestion of samples with historical timestamps, for example, when older data has been moved between storage tiers (nvme/hdd, hot/cold). See [#11199](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11199). Thanks to @AshwinRamaniPsg for contribution.
+
+* BUGFIX: `vminsert` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): Now drops metadata blocks when communicating with vmstorage nodes over the legacy RPC protocol. To avoid this limitation, upgrade `vmstorage` to a version that supports the new RPC protocol (>= [v1.137.0](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/docs/victoriametrics/changelog/CHANGELOG.md#v11370)). See [#11146](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11146).
+* BUGFIX: [vmbackup](https://docs.victoriametrics.com/victoriametrics/vmbackup/) and [vmbackupmanager](https://docs.victoriametrics.com/victoriametrics/vmbackupmanager/): retry S3 requests failing with `HTTP 429` status code or `TooManyRequests` error code. Previously such requests were not retried, so a short burst of rate limiting would fail the whole backup. See [#11218](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11218). Thanks to @gautamrizwani for contribution.
 * BUGFIX: `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): properly apply limit to metrics metadata response. See [#11139](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11139).
+* BUGFIX: [vmui](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#vmui): keep only one header navigation dropdown (`Explore`, `Tools`) open at a time. Previously, hovering across two dropdowns could briefly leave both open due to the close delay. See [#11224](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11224). Thanks to @antedotee for contribution.
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): fix a possible data race when processing OpenTelemetry metadata. See [#11238](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11238). Thanks to @nevgeny for contribution.
+* BUGFIX: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): flush pending persistent queue data to chunk file before updating the metadata. This prevents the metadata writer offset from getting ahead of the chunk file size and avoids losing the persistent queue after an unclean shutdown. See [#11192](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11192).
 
 ## [v1.147.0](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.147.0)
 
@@ -38,6 +48,7 @@ Released at 2026-07-06
 * FEATURE: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/): reduces CPU usage by 10% at [sharding among remote storages](https://docs.victoriametrics.com/victoriametrics/vmagent/#sharding-among-remote-storages). See [#11113](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11113). Thanks to @bennf for contribution.
 * FEATURE: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/), `vminsert` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/) and [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/): introduce `64KiB` size limit for `metric metadata` fields - `Unit`, `Help` and `MetricFamilyName`. See [#11128](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11128).
 * FEATURE: [vmagent](https://docs.victoriametrics.com/victoriametrics/vmagent/) and [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/): reduce CPU usage for storing scrape target labels. See [#10919](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10919).
+* FEATURE: [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vmstorage` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): expose `vm_data_size_bytes{type="storage/metaindex"}` and `vm_data_size_bytes{type="indexdb/metaindex"}` metrics for tracking memory occupied by metaindex data. See [#11204](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/11204). Thanks to @SamarthBagga for contribution.
 * FEATURE: [vmsingle](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/) and `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/): add `optimize_repeated_binary_op_subexprs=1` query arg to [/api/v1/query_range](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#range-query) for executing binary operator sides sequentially when they share the same optimized aggregate rollup result expression. This allows the second side to reuse rollup result cache populated by the first side. See [#10575](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/10575). Thanks to @xhebox for the contribution.
 * FEATURE: [vmauth](https://docs.victoriametrics.com/victoriametrics/vmauth/): prevent possible password brute-force attacks with an artificial 2-3 second delay as recommended by [OWASP](https://owasp.org/Top10/2025/A07_2025-Authentication_Failures). See [#11180](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11180).
 * FEATURE: [alerts](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/deployment/docker/rules): add `InvalidAuthTokenRequestErrors` alerting rule to [vmauth alerts](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/deployment/docker/rules/alerts-vmauth.yml). The new rule notifies when vmauth receives requests with invalid or missing auth tokens, which may indicate a client misconfiguration, expired token use, or brute-force attack. See [#11180](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11180).

docs/victoriametrics/keyConcepts/keyConcepts.md

@@ -40,7 +40,7 @@ requests_total{path="/", code="403"}
 The meta-information - a set of `labels` in curly braces - gives us a context for which `path` and with what `code`
 the `request` was served. Label-value pairs are always of a `string` type. VictoriaMetrics data model is schemaless,
 which means there is no need to define metric names or their labels in advance. The user is free to add or change ingested
-metrics anytime.
+metrics anytime. See [metric names and labels explained](https://victoriametrics.com/blog/prometheus-monitoring-metrics-counters-gauges-histogram-summaries/#metric-name-and-labels) for details.
 
 Actually, the metric name is also a label with a special name `__name__`. 
 The `__name__` key could be omitted {{% available_from "v1.111.0" %}} for simplicity. So the following series are identical:
@@ -75,7 +75,7 @@ See [these docs](https://docs.victoriametrics.com/victoriametrics/faq/#what-is-h
 
 #### Raw samples
 
-Every unique time series may consist of an arbitrary number of `(value, timestamp)` data points (aka `raw samples`) sorted by `timestamp`.
+Every unique time series may consist of an arbitrary number of `(value, timestamp)` data points (aka [`raw samples`](https://victoriametrics.com/blog/prometheus-monitoring-metrics-counters-gauges-histogram-summaries/#sample)) sorted by `timestamp`.
 VictoriaMetrics stores all the `values` as [float64](https://en.wikipedia.org/wiki/Double-precision_floating-point_format)
 with [extra compression](https://faun.pub/victoriametrics-achieving-better-compression-for-time-series-data-than-gorilla-317bc1f95932) applied.
 This allows storing precise integer values with up to 12 decimal digits and any floating-point values with up to 12 significant decimal digits.
@@ -126,7 +126,7 @@ type exists specifically to help users to understand how the metric was measured
 
 #### Counter
 
-Counter is a metric, which counts some events. Its value increases or stays the same over time.
+[Counter](https://victoriametrics.com/blog/prometheus-monitoring-metrics-counters-gauges-histogram-summaries/#counter) is a metric, which counts some events. Its value increases or stays the same over time.
 It cannot decrease in general case. The only exception is e.g. `counter reset`,
 when the metric resets to zero. The `counter reset` can occur when the service, which exposes the counter, restarts.
 So, the `counter` metric shows the number of observed events since the service start.
@@ -157,7 +157,7 @@ by humans from other metric types.
 
 #### Gauge
 
-Gauge is used for measuring a value that can go up and down:
+[Gauge](https://victoriametrics.com/blog/prometheus-monitoring-metrics-counters-gauges-histogram-summaries/#gauge) is used for measuring a value that can go up and down:
 
 ![gauge](gauge.webp)
 
@@ -178,7 +178,7 @@ and [rollup functions](https://docs.victoriametrics.com/victoriametrics/metricsq
 
 #### Histogram
 
-Histogram is a set of [counter](#counter) metrics with different `vmrange` or `le` labels.
+[Histogram](https://victoriametrics.com/blog/prometheus-monitoring-metrics-counters-gauges-histogram-summaries/#histogram) is a set of [counter](#counter) metrics with different `vmrange` or `le` labels.
 The `vmrange` or `le` labels define measurement boundaries of a particular bucket.
 When the observed measurement hits a particular bucket, then the corresponding counter is incremented.
 
@@ -282,7 +282,7 @@ We recommend reading the following articles before you start using histograms:
 
 #### Summary
 
-Summary metric type is quite similar to [histogram](#histogram) and is used for
+[Summary](https://victoriametrics.com/blog/prometheus-monitoring-metrics-counters-gauges-histogram-summaries/#summary) metric type is quite similar to [histogram](#histogram) and is used for
 [quantiles](https://prometheus.io/docs/practices/histograms/#quantiles) calculations. The main difference
 is that calculations are made on the client-side, so metrics exposition format already contains pre-defined
 quantiles:
@@ -504,6 +504,7 @@ Params:
 The result of Instant query is a list of [time series](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#time-series)
 matching the filter in `query` expression. Each returned series contains exactly one `(timestamp, value)` entry,
 where `timestamp` equals to the `time` query arg, while the `value` contains `query` result at the requested `time`.
+See [instant vectors explained](https://victoriametrics.com/blog/prometheus-monitoring-instant-range-query/#instant-vector) for details.
 
 To understand how instant queries work, let's begin with a data sample:
 
@@ -606,6 +607,7 @@ at `start`, `start+step`, `start+2*step`, ..., `start+N*step` timestamps. In oth
 executed independently at `start`, `start+step`, ..., `start+N*step` timestamps with the only difference that an instant query
 does not return `ephemeral` samples (see below). Instead, if the database does not contain any samples for the requested time and step,
 it simply returns an empty result.
+See [range vectors explained](https://victoriametrics.com/blog/prometheus-monitoring-instant-range-query/#range-vector) for details.
 
 
 For example, to get the values of `foo_bar` during the time range from `2022-05-10T07:59:00Z` to `2022-05-10T08:17:00Z`,

docs/victoriametrics/sd_configs.md

@@ -1923,6 +1923,9 @@ scrape_configs:
   # Example values:
   # - "10MiB" - 10 * 1024 * 1024 bytes
   # - "100MB" - 100 * 1000 * 1000 bytes
+  # The max_scrape_size can be set on a per-target basis by specifying `__max_scrape_size__`
+  # label during target relabeling phase.
+  # See https://docs.victoriametrics.com/victoriametrics/relabeling/
   #
   # max_scrape_size: <size>
 

docs/victoriametrics/vmagent.md

@@ -127,7 +127,8 @@ See [these docs](#how-to-collect-metrics-in-prometheus-format) for details.
 
 `vmagent` can be used as an alternative to [StatsD](https://github.com/statsd/statsd)
 when [stream aggregation](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/) is enabled.
-See [these docs](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#statsd-alternative) for details.
+See [these docs](https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#statsd-alternative) for details,
+or the blog post on [how vmagent's stream aggregation works](https://victoriametrics.com/blog/vmagent-key-features-explained/#stream-aggregation).
 
 ### Flexible metrics relay
 
@@ -137,7 +138,7 @@ to other remote storage systems that support Prometheus `remote_write` protocol
 
 ### Replication and high availability
 
-`vmagent` replicates the collected metrics among multiple remote storage instances configured via `-remoteWrite.url` args.
+`vmagent` [replicates the collected metrics](https://victoriametrics.com/blog/vmagent-how-it-works/#step-4-sharding--replication) among multiple remote storage instances configured via `-remoteWrite.url` args.
 If a single remote storage instance is temporarily unavailable, the collected data remains available on the other remote storage instances.
 `vmagent` buffers the collected data in files at `-remoteWrite.tmpDataPath` until the remote storage becomes available again.
 Then it sends the buffered data to the remote storage in order to prevent data gaps.
@@ -150,7 +151,8 @@ See [these docs](https://docs.victoriametrics.com/victoriametrics/cluster-victor
 
 `vmagent` can add, remove, or update labels on the collected data before sending it to the remote storage.
 It can filter scrape targets or remove unwanted samples via Prometheus-like relabeling.
-Please see the [Relabeling cookbook](https://docs.victoriametrics.com/victoriametrics/relabeling/) for details.
+Please see the [Relabeling cookbook](https://docs.victoriametrics.com/victoriametrics/relabeling/) for details,
+or the blog post on [how vmagent applies global relabeling to reduce cardinality](https://victoriametrics.com/blog/vmagent-how-it-works/#step-2-global-relabeling-cardinality-reduction).
 
 ### Sharding among remote storages
 
@@ -158,6 +160,7 @@ By default, `vmagent` replicates data to remote storage systems via the `-remote
 If the `-remoteWrite.shardByURL` command-line flag is set, then `vmagent` spreads
 the outgoing [time series](https://docs.victoriametrics.com/victoriametrics/keyconcepts/#time-series) evenly among all the remote storage
 systems listed in `-remoteWrite.url`.
+See the blog post on [how vmagent shards data across remote storage systems](https://victoriametrics.com/blog/vmagent-key-features-explained/#sharding) for the details.
 
 It is possible to replicate samples among remote storage systems by passing `-remoteWrite.shardByURLReplicas=N`
 to `vmagent` in addition to the `-remoteWrite.shardByURL` command-line flag.
@@ -268,7 +271,9 @@ for the collected samples. Examples:
   ```sh
   ./vmagent -remoteWrite.url=http://remote-storage/api/v1/write -streamAggr.dropInputLabels=replica -streamAggr.dedupInterval=60s
   ```
-  
+
+See the blog post on [how vmagent performs global deduplication and stream aggregation](https://victoriametrics.com/blog/vmagent-how-it-works/#step-3-global-deduplication--stream-aggregation) for the details.
+
 ### Monitoring Data eXchange
 
 The Monitoring Data eXchange (MDX){{% available_from "v1.147.0" %}} feature allows `vmagent` to forward only VictoriaMetrics metrics to selected `-remoteWrite.url` destinations while dropping metrics from non-VictoriaMetrics services.
@@ -357,6 +362,8 @@ in addition to the pull-based Prometheus-compatible targets' scraping:
 * Prometheus exposition format via `http://<vmagent>:8429/api/v1/import/prometheus`. See [these docs](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#how-to-import-data-in-prometheus-exposition-format) for details.
 * Arbitrary CSV data via `http://<vmagent>:8429/api/v1/import/csv`. See [these docs](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/#how-to-import-csv-data).
 
+See the blog post on [how vmagent receives data via push APIs or scraping](https://victoriametrics.com/blog/vmagent-how-it-works/#step-1-receiving-data-via-api-or-scrape) for the details.
+
 ## How to collect metrics in Prometheus format
 
 Specify the path to the `prometheus.yml` file via the `-promscrape.config` command-line flag. `vmagent` takes into account the following
@@ -490,15 +497,13 @@ SRV URLs are useful when HTTP services run on different TCP ports or when their
 
 When comparing the remote protocols between VictoriaMetrics and Prometheus, VictoriaMetrics provides the following benefits:
 
-* Reduced network bandwidth usage by 2x-5x. This allows saving network bandwidth usage costs when `vmagent` and
+* Reduced network bandwidth usage by 2x-5x. This allows [saving network bandwidth usage costs](https://victoriametrics.com/blog/victoriametrics-remote-write/) when `vmagent` and
   the configured remote storage systems are located in different datacenters, availability zones, or regions.
 
 * Reduced disk read/write IO and disk space usage at `vmagent` when the remote storage is temporarily unavailable.
   In this case, `vmagent` buffers incoming data to disk using the VictoriaMetrics remote write format.
   This reduces disk read/write IO and disk space usage by 2x-5x compared to the Prometheus remote write format.
 
-> See blogpost [Save network costs with VictoriaMetrics remote write protocol](https://victoriametrics.com/blog/victoriametrics-remote-write/).
-
 `vmagent` uses VictoriaMetrics remote write protocol by default {{% available_from "v1.116.0" %}} when it sends data to VictoriaMetrics components such as other `vmagent` instances,
 [single-node VictoriaMetrics](https://docs.victoriametrics.com/victoriametrics/single-server-victoriametrics/)
 or `vminsert` at [cluster version](https://docs.victoriametrics.com/victoriametrics/cluster-victoriametrics/). If needed, it can automatically downgrade to a Prometheus protocol at runtime.
@@ -835,7 +840,7 @@ as soon as it is parsed in stream parsing mode.
 ## Scraping big number of targets
 
 A single `vmagent` instance can scrape tens of thousands of scrape targets. Sometimes this isn't enough due to limitations on CPU, network, RAM, etc.
-In this case, scrape targets can be split among multiple `vmagent` instances (aka `vmagent` horizontal scaling, sharding, and clustering).
+In this case, scrape targets can be split among multiple `vmagent` instances (aka [`vmagent` horizontal scaling](https://victoriametrics.com/blog/vmagent-key-features-explained/#scaling-vmagent), sharding, and clustering).
 The number of `vmagent` instances in the cluster must be passed to the `-promscrape.cluster.membersCount` command-line flag.
 Each `vmagent` instance in the cluster must use identical `-promscrape.config` files with distinct `-promscrape.cluster.memberNum` values
 in the range `0 ... N-1`, where `N` is the number of `vmagent` instances in the cluster specified via `-promscrape.cluster.membersCount`.
@@ -948,7 +953,7 @@ scrape_configs:
 
 ## On-disk persistence
 
-`vmagent` stores pending data that cannot be sent to the configured remote storage systems in a timely manner.
+`vmagent` [stores pending data that cannot be sent to the configured remote storage systems](https://victoriametrics.com/blog/vmagent-key-features-explained/#persistent-disk-for-remote-write) in a timely manner.
 By default, `vmagent` writes all the pending data to the folder configured via `-remoteWrite.tmpDataPath` cmd-line flag
 until this data is sent to the configured `-remoteWrite.url` systems or until the folder becomes full.
 The maximum data size that can be saved to `-remoteWrite.tmpDataPath` per every configured `-remoteWrite.url` can be
@@ -996,7 +1001,7 @@ moment it becomes visible at the remote storage.
 
 This behavior can be changed with the `-remoteWrite.inmemoryQueues` {{% available_from "v1.146.0" %}} command-line flag.
 When set to a non-zero value, vmagent starts the given number of additional workers,
-which send only recently ingested data from the in-memory queue, while the workers configured via `-remoteWrite.queues` drain the file-based backlog concurrently.
+which send only recently ingested data from the [in-memory queue](https://victoriametrics.com/blog/vmagent-how-it-works/#in-memory-queue), while the workers configured via `-remoteWrite.queues` drain the file-based backlog concurrently.
 This reduces the delivery lag for fresh samples after remote storage outages or slowdowns. The flag can be set individually per each `-remoteWrite.url`.
 
 Note that these workers are started in addition to the workers configured via `-remoteWrite.queues`, so the total number of concurrent connections to

docs/victoriametrics/vmalert.md

@@ -209,6 +209,8 @@ rules:
 [ debug: <bool> | default = false ]
 ```
 
+See a blogpost about [how rule groups work](https://victoriametrics.com/blog/alerting-recording-rules-alertmanager/#groups).
+
 ### Rules
 
 Every rule contains an `expr` field for the expression to evaluate against the configured datasource.
@@ -285,6 +287,8 @@ annotations:
   [ <labelname>: <tmpl_string> ]
 ```
 
+See a blogpost about [how alerting rules work](https://victoriametrics.com/blog/alerting-recording-rules-alertmanager/#alerting-rules).
+
 #### Recording rules
 
 The syntax for recording rules is the following:
@@ -322,10 +326,13 @@ labels:
 
 For recording rules to work `-remoteWrite.url` must be specified.
 
+See a blogpost about [how recording rules work](https://victoriametrics.com/blog/alerting-recording-rules-alertmanager/#recording-rules).
+
 ## Templating
 
 It is allowed to use [Go templating](https://golang.org/pkg/text/template/) in annotations and labels(with limited support) to format data, iterate over
 or execute expressions.
+See a blogpost about [templating alerts](https://victoriametrics.com/blog/alerting-recording-rules-alertmanager/#templates).
 The following variables are available in templating:
 
 | Variable                           | Description                                                                                                                                           | Example                                                                                                                                                                                |
@@ -928,7 +935,7 @@ See full description for these flags in `./vmalert -help`.
 
 ## Unit Testing for Rules
 
-You can use `vmalert-tool` to test your alerting and recording rules like [promtool does](https://prometheus.io/docs/prometheus/latest/configuration/unit_testing_rules/).
+You can use `vmalert-tool` to [test your alerting and recording rules](https://victoriametrics.com/blog/alerting-best-practices/#testing-alerts) like [promtool does](https://prometheus.io/docs/prometheus/latest/configuration/unit_testing_rules/).
 See more details in [vmalert-tool](https://docs.victoriametrics.com/victoriametrics/vmalert-tool/#unit-testing-for-rules).
 
 ## Monitoring
@@ -982,6 +989,8 @@ Try the following tips to avoid common issues:
     In that case, the default step will be used (`-datasource.queryStep`) and may cause unexpected results compared to
     executing this query in vmui/Grafana, where step is adjusted differently.
 
+See a blogpost about [reducing alert noise](https://victoriametrics.com/blog/alerting-best-practices/#reducing-noise).
+
 ### Rule state
 
 vmalert keeps the last `-rule.updateEntriesLimit` updates (or `update_entries_limit` [per-rule config](https://docs.victoriametrics.com/victoriametrics/vmalert/#alerting-rules))
@@ -1037,6 +1046,8 @@ Sometimes, it's hard to understand why a specific alert fired or not. Keep in mi
 If evaluation returns error (i.e. datasource is unavailable), alert state doesn't change.
 If at least one evaluation returns no data, then alert's `for` state resets.
 
+See a blogpost about [tuning the `for` param](https://victoriametrics.com/blog/alerting-best-practices/#the-for-param).
+
 > Note: The alert state is tracked separately for each time series returned during evaluation.
 > For example, if the 1st evaluation returns series A and B, and the 2nd evaluation returns only B – the alert will remain active **only for B**.
 
@@ -1110,6 +1121,8 @@ How to reduce the chance for a rule to flap:
 
 See [common mistakes](#common-mistakes) for rules config.
 
+See a blogpost about [tuning `keep_firing_for`](https://victoriametrics.com/blog/alerting-best-practices/#the-keep_firing_for-param).
+
 ### Never-firing alerts
 
 vmalert can detect {{% available_from "v1.91.0" %}} if alert's expression doesn't match any time series in runtime.
@@ -1360,6 +1373,8 @@ The list of configured or discovered Notifiers can be explored via [UI](#web).
 If Alertmanager runs in cluster mode then all its URLs needs to be available during discovery
 to ensure [high availability](https://github.com/prometheus/alertmanager#high-availability).
 
+See a blogpost about [how Alertmanager processes alerts](https://victoriametrics.com/blog/alerting-recording-rules-alertmanager/#alertmanager).
+
 The configuration file [specification](https://github.com/VictoriaMetrics/VictoriaMetrics/blob/master/app/vmalert/notifier/config.go)
 is the following:
 

go.mod

@@ -11,7 +11,7 @@ require (
 	github.com/VictoriaMetrics/easyproto v1.2.0
 	github.com/VictoriaMetrics/fastcache v1.13.3
 	github.com/VictoriaMetrics/metrics v1.44.0
-	github.com/VictoriaMetrics/metricsql v0.87.2
+	github.com/VictoriaMetrics/metricsql v0.87.3
 	github.com/aws/aws-sdk-go-v2 v1.42.0
 	github.com/aws/aws-sdk-go-v2/config v1.32.25
 	github.com/aws/aws-sdk-go-v2/feature/s3/manager v1.22.27
@@ -36,7 +36,7 @@ require (
 	github.com/valyala/quicktemplate v1.8.0
 	golang.org/x/net v0.56.0
 	golang.org/x/oauth2 v0.36.0
-	golang.org/x/sys v0.46.0
+	golang.org/x/sys v0.47.0
 	google.golang.org/api v0.284.0
 	gopkg.in/yaml.v2 v2.4.0
 )

go.sum

@@ -62,6 +62,8 @@ github.com/VictoriaMetrics/metrics v1.44.0 h1:Fr8yqQSV+ZfYaDD/anqk1E8e9YPgfleSle
 github.com/VictoriaMetrics/metrics v1.44.0/go.mod h1:xDM82ULLYCYdFRgQ2JBxi8Uf1+8En1So9YUwlGTOqTc=
 github.com/VictoriaMetrics/metricsql v0.87.2 h1:7OsrcDBWREWKqqpnFyIUEOM4FNv2qHvCoww2GYz3Tc0=
 github.com/VictoriaMetrics/metricsql v0.87.2/go.mod h1:d4EisFO6ONP/HIGDYTAtwrejJBBeKGQYiRl095bS4QQ=
+github.com/VictoriaMetrics/metricsql v0.87.3 h1:JU4JnVKSC5Vp3b4AvogXyOAjkz1iFF9n1KBMphS4WO8=
+github.com/VictoriaMetrics/metricsql v0.87.3/go.mod h1:d4EisFO6ONP/HIGDYTAtwrejJBBeKGQYiRl095bS4QQ=
 github.com/VividCortex/ewma v1.2.0 h1:f58SaIzcDXrSy3kWaHNvuJgJ3Nmz59Zji6XoJR/q1ow=
 github.com/VividCortex/ewma v1.2.0/go.mod h1:nz4BbCtbLyFDeC9SUHbtcT5644juEuWfUAUnGx7j5l4=
 github.com/alecthomas/units v0.0.0-20240927000941-0f3dac36c52b h1:mimo19zliBX/vSQ6PWWSL9lK8qwHozUj03+zLoEB8O0=
@@ -544,6 +546,8 @@ golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f/go.mod h1:h1NjWce9XRLGQEsW7w
 golang.org/x/sys v0.1.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.46.0 h1:noSf2Fq6F8DBgS+LysIkx7rIExoNHJsxOAtPp4rthXw=
 golang.org/x/sys v0.46.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
+golang.org/x/sys v0.47.0 h1:o7XGOvZQCADBQQ4Y7VNq2dRWQR7JmOUW8Kxx4ZsNgWs=
+golang.org/x/sys v0.47.0/go.mod h1:4GL1E5IUh+htKOUEOaiffhrAeqysfVGipDYzABqnCmw=
 golang.org/x/term v0.44.0 h1:0rLvDRCtNj0gZkyIXhCyOb2OAzEhLVqc4B+hrsBhrmc=
 golang.org/x/term v0.44.0/go.mod h1:7ze4MdzUzLXpSAoFP1H0bOI9aXDqveSvatT5vKcFh2Y=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=

lib/backup/s3remote/s3.go

@@ -164,6 +164,14 @@ func (fs *FS) Init(ctx context.Context) error {
 						// when using EKS Pod Identity or similar.
 						// See: https://github.com/VictoriaMetrics/VictoriaMetrics/issues/9280
 						"ExpiredToken": {},
+						// S3-compatible stores may return TooManyRequests instead of TooManyRequestsException.
+						// See: https://github.com/VictoriaMetrics/VictoriaMetrics/issues/11218
+						"TooManyRequests": {},
+					},
+				}, retry.RetryableHTTPStatusCode{
+					Codes: map[int]struct{}{
+						// S3-compatible stores may return HTTP 429 for rate limiting instead of HTTP 503.
+						429: {},
 					},
 				}, retry.IsErrorRetryableFunc(func(err error) aws.Ternary {
 					if errors.Is(err, io.ErrUnexpectedEOF) {

lib/mergeset/part.go

@@ -78,7 +78,8 @@ type part struct {
 
 	size uint64
 
-	mrs []metaindexRow
+	mrs                []metaindexRow
+	metaindexSizeBytes uint64
 
 	indexFile fs.MustReadAtCloser
 	itemsFile fs.MustReadAtCloser
@@ -131,6 +132,7 @@ func newPart(ph *partHeader, path string, size uint64, metaindexReader filestrea
 	p.path = path
 	p.size = size
 	p.mrs = mrs
+	p.metaindexSizeBytes = metaindexSizeBytes(mrs)
 
 	p.indexFile = indexFile
 	p.itemsFile = itemsFile
@@ -155,6 +157,14 @@ func (p *part) MustClose() {
 	ibSparseCache.RemoveBlocksForPart(p)
 }
 
+func metaindexSizeBytes(mrs []metaindexRow) uint64 {
+	n := uint64(cap(mrs)) * uint64(unsafe.Sizeof(metaindexRow{}))
+	for i := range mrs {
+		n += uint64(cap(mrs[i].firstItem))
+	}
+	return n
+}
+
 type indexBlock struct {
 	bhs []blockHeader
 

lib/mergeset/table.go

@@ -584,6 +584,8 @@ type TableMetrics struct {
 	PartsRefCount uint64
 
 	TooLongItemsDroppedTotal uint64
+
+	MetaindexSizeBytes uint64
 }
 
 // TotalItemsCount returns the total number of items in the table.
@@ -617,6 +619,7 @@ func (tb *Table) UpdateMetrics(m *TableMetrics) {
 		m.InmemoryBlocksCount += p.ph.blocksCount
 		m.InmemoryItemsCount += p.ph.itemsCount
 		m.InmemorySizeBytes += p.size
+		m.MetaindexSizeBytes += p.metaindexSizeBytes
 		m.PartsRefCount += uint64(pw.refCount.Load())
 	}
 
@@ -626,6 +629,7 @@ func (tb *Table) UpdateMetrics(m *TableMetrics) {
 		m.FileBlocksCount += p.ph.blocksCount
 		m.FileItemsCount += p.ph.itemsCount
 		m.FileSizeBytes += p.size
+		m.MetaindexSizeBytes += p.metaindexSizeBytes
 		m.PartsRefCount += uint64(pw.refCount.Load())
 	}
 	tb.partsLock.Unlock()

lib/persistentqueue/persistentqueue.go

@@ -52,6 +52,7 @@ type queue struct {
 	writerFlushedOffset uint64
 
 	lastMetainfoFlushTime uint64
+	hasDataToFlush        bool
 
 	blocksDropped *metrics.Counter
 	bytesDropped  *metrics.Counter
@@ -84,6 +85,7 @@ func (q *queue) mustResetFiles() {
 	}
 	q.reader.MustClose()
 	q.writer.MustClose()
+	q.hasDataToFlush = false
 	fs.MustRemovePath(q.readerPath)
 
 	q.writerOffset = 0
@@ -318,6 +320,7 @@ func tryOpeningQueue(path, name string, chunkFileSize, maxBlockSize, maxPendingB
 func (q *queue) MustClose() {
 	// Close writer.
 	q.writer.MustClose()
+	q.hasDataToFlush = false
 	q.writer = nil
 
 	// Close reader.
@@ -414,7 +417,7 @@ func (q *queue) writeBlock(block []byte) error {
 	}
 	q.blocksWritten.Inc()
 	q.bytesWritten.Add(len(block))
-	return q.flushWriterMetainfoIfNeeded()
+	return q.flushBufAndMetainfoIfNeeded()
 }
 
 var writeDurationSeconds = metrics.NewFloatCounter(`vm_persistentqueue_write_duration_seconds_total`)
@@ -422,6 +425,7 @@ var writeDurationSeconds = metrics.NewFloatCounter(`vm_persistentqueue_write_dur
 func (q *queue) nextChunkFileForWrite() error {
 	// Finalize the current chunk and start new one.
 	q.writer.MustClose()
+	q.hasDataToFlush = false
 	// There is no need to do fs.MustSyncPath(q.writerPath) here,
 	// since MustClose already does this.
 	if n := q.writerOffset % q.chunkFileSize; n > 0 {
@@ -513,7 +517,7 @@ again:
 	}
 	q.blocksRead.Inc()
 	q.bytesRead.Add(int(blockLen))
-	if err := q.flushReaderMetainfoIfNeeded(); err != nil {
+	if err := q.flushBufAndMetainfoIfNeeded(); err != nil {
 		return dst, err
 	}
 	return dst, nil
@@ -566,6 +570,7 @@ func (q *queue) write(buf []byte) error {
 	}
 	q.writerLocalOffset += bufLen
 	q.writerOffset += bufLen
+	q.hasDataToFlush = true
 	return nil
 }
 
@@ -595,24 +600,16 @@ func (q *queue) checkReaderWriterOffsets() error {
 	return nil
 }
 
-func (q *queue) flushReaderMetainfoIfNeeded() error {
+func (q *queue) flushBufAndMetainfoIfNeeded() error {
 	t := fasttime.UnixTimestamp()
 	if t == q.lastMetainfoFlushTime {
 		return nil
 	}
-	if err := q.flushMetainfo(); err != nil {
-		return fmt.Errorf("cannot flush metainfo: %w", err)
+	if q.hasDataToFlush {
+		q.writer.MustFlush(true)
+		q.writerFlushedOffset = q.writerOffset
+		q.hasDataToFlush = false
 	}
-	q.lastMetainfoFlushTime = t
-	return nil
-}
-
-func (q *queue) flushWriterMetainfoIfNeeded() error {
-	t := fasttime.UnixTimestamp()
-	if t == q.lastMetainfoFlushTime {
-		return nil
-	}
-	q.writer.MustFlush(true)
 	if err := q.flushMetainfo(); err != nil {
 		return fmt.Errorf("cannot flush metainfo: %w", err)
 	}

lib/persistentqueue/persistentqueue_synctest_test.go

@@ -0,0 +1,60 @@
+//go:build synctest
+
+package persistentqueue
+
+import (
+	"testing"
+	"testing/synctest"
+	"time"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
+)
+
+func TestFlushReaderMetainfoFlushesPendingWriterData(t *testing.T) {
+	synctest.Test(t, func(t *testing.T) {
+		path := "queue-flush-reader-metainfo"
+		fs.MustRemoveDir(path)
+		q := mustOpen(path, "foobar", 0)
+		defer func() {
+			q.MustClose()
+			fs.MustRemoveDir(path)
+		}()
+
+		block := []byte("foobar")
+		data := encoding.MarshalUint64(nil, uint64(len(block)))
+		data = append(data, block...)
+		// it will call `flushBufAndMetainfoIfNeeded` internally to flush the data and metadata.
+		err := q.writeBlock(data)
+		if err != nil {
+			t.Fatalf("unexpected error when writing data to queue: %s", err)
+		}
+		// the second call will update the writeOffset in memory without flushing the data and metadata,
+		// because the last flush was performed less than 1 second ago.
+		err = q.writeBlock(data)
+		if err != nil {
+			t.Fatalf("unexpected error when writing data to queue: %s", err)
+		}
+		time.Sleep(2 * time.Second)
+
+		// it will call `flushBufAndMetainfoIfNeeded` internally to flush the data and metadata.
+		if _, err = q.readBlock(nil); err != nil {
+			t.Fatalf("unexpected error when flushing reader metainfo: %s", err)
+		}
+
+		if fileSize := fs.MustFileSize(q.writerPath); fileSize != q.writerOffset {
+			t.Fatalf("unexpected writer file size after flushing reader metainfo; got %d bytes; want %d bytes", fileSize, q.writerOffset)
+		}
+		var mi metainfo
+		if err := mi.ReadFromFile(q.metainfoPath()); err != nil {
+			t.Fatalf("cannot read metainfo: %s", err)
+		}
+		if mi.ReaderOffset != q.readerOffset {
+			t.Fatalf("unexpected ReaderOffset in metainfo; got %d; want %d", mi.ReaderOffset, q.readerOffset)
+		}
+		if mi.WriterOffset != q.writerOffset {
+			t.Fatalf("unexpected WriterOffset in metainfo; got %d; want %d", mi.WriterOffset, q.writerOffset)
+		}
+
+	})
+}

lib/promscrape/config.go

@@ -367,6 +367,9 @@ func (sc *ScrapeConfig) mustStart(baseDir string) {
 	for i := range sc.KubernetesSDConfigs {
 		sc.KubernetesSDConfigs[i].MustStart(baseDir, swosFunc)
 	}
+	for i := range sc.HTTPSDConfigs {
+		sc.HTTPSDConfigs[i].MustStart(baseDir)
+	}
 }
 
 func (sc *ScrapeConfig) mustStop() {
@@ -1271,6 +1274,17 @@ func (swc *scrapeWorkConfig) getScrapeWork(target string, extraLabels, metaLabel
 		}
 		scrapeTimeout = d
 	}
+	// Read max_scrape_size option from __max_scrape_size__ label.
+	targetMaxScrapeSize := swc.maxScrapeSize
+	if s := labels.Get("__max_scrape_size__"); len(s) > 0 {
+		n, err := flagutil.ParseBytes(s)
+		if err != nil {
+			return nil, fmt.Errorf("cannot parse __max_scrape_size__=%q: %w", s, err)
+		}
+		if n > 0 {
+			targetMaxScrapeSize = n
+		}
+	}
 	// Read series_limit option from __series_limit__ label.
 	// See https://docs.victoriametrics.com/victoriametrics/vmagent/#cardinality-limiter
 	seriesLimit := swc.seriesLimit
@@ -1333,7 +1347,7 @@ func (swc *scrapeWorkConfig) getScrapeWork(target string, extraLabels, metaLabel
 		ScrapeURL:            scrapeURL,
 		ScrapeInterval:       scrapeInterval,
 		ScrapeTimeout:        scrapeTimeout,
-		MaxScrapeSize:        swc.maxScrapeSize,
+		MaxScrapeSize:        targetMaxScrapeSize,
 		HonorLabels:          swc.honorLabels,
 		HonorTimestamps:      swc.honorTimestamps,
 		DenyRedirects:        swc.denyRedirects,

lib/promscrape/config_test.go

@@ -1153,6 +1153,57 @@ scrape_configs:
 	})
 	f(`
 scrape_configs:
+- job_name: foo
+  max_scrape_size: 8MiB
+  relabel_configs:
+  - source_labels: [__address__]
+    regex: foo1:.*
+    target_label: __max_scrape_size__
+    replacement: 2.5MiB
+  - source_labels: [__address__]
+    regex: foo2:.*
+    target_label: __max_scrape_size__
+    replacement: -1
+  static_configs:
+  - targets: ["foo1:1234", "foo2:1234", "foo3:1234"]
+`, []*ScrapeWork{
+		{
+			ScrapeURL:      "http://foo1:1234/metrics",
+			ScrapeInterval: defaultScrapeInterval,
+			ScrapeTimeout:  defaultScrapeTimeout,
+			MaxScrapeSize:  2.5 * 1024 * 1024,
+			Labels: promutil.NewLabelsFromMap(map[string]string{
+				"instance": "foo1:1234",
+				"job":      "foo",
+			}),
+			jobNameOriginal: "foo",
+		},
+		// invalid __max_scrape_size__ will be ignored
+		{
+			ScrapeURL:      "http://foo2:1234/metrics",
+			ScrapeInterval: defaultScrapeInterval,
+			ScrapeTimeout:  defaultScrapeTimeout,
+			MaxScrapeSize:  8 * 1024 * 1024,
+			Labels: promutil.NewLabelsFromMap(map[string]string{
+				"instance": "foo2:1234",
+				"job":      "foo",
+			}),
+			jobNameOriginal: "foo",
+		},
+		{
+			ScrapeURL:      "http://foo3:1234/metrics",
+			ScrapeInterval: defaultScrapeInterval,
+			ScrapeTimeout:  defaultScrapeTimeout,
+			MaxScrapeSize:  8 * 1024 * 1024,
+			Labels: promutil.NewLabelsFromMap(map[string]string{
+				"instance": "foo3:1234",
+				"job":      "foo",
+			}),
+			jobNameOriginal: "foo",
+		},
+	})
+	f(`
+scrape_configs:
 - job_name: foo
   static_configs:
   - targets: ["foo.bar:1234"]

lib/promscrape/discovery/http/api.go

@@ -1,25 +1,41 @@
 package http
 
 import (
+	"bytes"
 	"encoding/json"
 	"fmt"
 	"net/http"
 	"net/url"
 	"strconv"
+	"sync"
+	"sync/atomic"
+	"time"
+
+	"github.com/VictoriaMetrics/metrics"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discoveryutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutil"
-	"github.com/VictoriaMetrics/metrics"
 )
 
-var configMap = discoveryutil.NewConfigMap()
-
 type apiConfig struct {
-	client *discoveryutil.Client
-	path   string
+	client        *discoveryutil.Client
+	path          string
+	sourceURL     string
+	checkInterval time.Duration
 
 	fetchErrors *metrics.Counter
 	parseErrors *metrics.Counter
+
+	initOnce        sync.Once
+	prevAPIResponse atomic.Pointer[[]byte]
+	targetLabels    atomic.Pointer[targetLabelsResult]
+
+	wg sync.WaitGroup
+}
+
+type targetLabelsResult struct {
+	labels []*promutil.Labels
+	err    error
 }
 
 // httpGroupTarget represent prometheus GroupTarget
@@ -49,37 +65,89 @@ func newAPIConfig(sdc *SDConfig, baseDir string) (*apiConfig, error) {
 		return nil, fmt.Errorf("cannot create HTTP client for %q: %w", apiServer, err)
 	}
 	cfg := &apiConfig{
-		client:      client,
-		path:        parsedURL.RequestURI(),
-		fetchErrors: metrics.GetOrCreateCounter(fmt.Sprintf(`promscrape_discovery_http_errors_total{type="fetch",url=%q}`, sdc.URL)),
-		parseErrors: metrics.GetOrCreateCounter(fmt.Sprintf(`promscrape_discovery_http_errors_total{type="parse",url=%q}`, sdc.URL)),
+		client:        client,
+		path:          parsedURL.RequestURI(),
+		sourceURL:     sdc.URL,
+		checkInterval: max(*SDCheckInterval/2, time.Second),
+		fetchErrors:   metrics.GetOrCreateCounter(fmt.Sprintf(`promscrape_discovery_http_errors_total{type="fetch",url=%q}`, sdc.URL)),
+		parseErrors:   metrics.GetOrCreateCounter(fmt.Sprintf(`promscrape_discovery_http_errors_total{type="parse",url=%q}`, sdc.URL)),
 	}
+	cfg.wg.Go(func() {
+		cfg.run()
+	})
 	return cfg, nil
 }
 
-func getAPIConfig(sdc *SDConfig, baseDir string) (*apiConfig, error) {
-	v, err := configMap.Get(sdc, func() (any, error) { return newAPIConfig(sdc, baseDir) })
-	if err != nil {
-		return nil, err
-	}
-	return v.(*apiConfig), nil
+func (cfg *apiConfig) init() {
+	cfg.initOnce.Do(func() {
+		cfg.refreshTargetsIfNeeded()
+	})
 }
 
-func getHTTPTargets(cfg *apiConfig) ([]httpGroupTarget, error) {
+func (cfg *apiConfig) run() {
+	cfg.init()
+
+	ticker := time.NewTicker(cfg.checkInterval)
+	defer ticker.Stop()
+	stopCh := cfg.client.Context().Done()
+	for {
+		select {
+		case <-ticker.C:
+			cfg.refreshTargetsIfNeeded()
+		case <-stopCh:
+			return
+		}
+	}
+}
+
+func (cfg *apiConfig) refreshTargetsIfNeeded() {
+	apiResponse, err := cfg.getAPIResponseData()
+	if err != nil {
+		cfg.targetLabels.Store(&targetLabelsResult{err: err})
+		cfg.prevAPIResponse.Store(nil)
+		return
+	}
+	prevAPIResponse := cfg.prevAPIResponse.Load()
+	if prevAPIResponse != nil && bytes.Equal(apiResponse, *prevAPIResponse) {
+		return
+	}
+	hts, err := parseAPIResponse(apiResponse, cfg.path)
+	if err != nil {
+		cfg.prevAPIResponse.Store(nil)
+		cfg.parseErrors.Inc()
+		cfg.targetLabels.Store(&targetLabelsResult{err: err})
+		return
+	}
+	newTargets := addHTTPTargetLabels(hts, cfg.sourceURL)
+	cfg.targetLabels.Store(&targetLabelsResult{labels: newTargets})
+	cfg.prevAPIResponse.Store(&apiResponse)
+}
+
+func (cfg *apiConfig) getAPIResponseData() ([]byte, error) {
 	data, err := cfg.client.GetAPIResponseWithReqParams(cfg.path, func(request *http.Request) {
-		request.Header.Set("X-Prometheus-Refresh-Interval-Seconds", strconv.FormatFloat(SDCheckInterval.Seconds(), 'f', 0, 64))
+		request.Header.Set("X-Prometheus-Refresh-Interval-Seconds", strconv.FormatFloat(cfg.checkInterval.Seconds(), 'f', 0, 64))
 		request.Header.Set("Accept", "application/json")
 	})
 	if err != nil {
 		cfg.fetchErrors.Inc()
 		return nil, fmt.Errorf("cannot read http_sd api response: %w", err)
 	}
-	tg, err := parseAPIResponse(data, cfg.path)
-	if err != nil {
-		cfg.parseErrors.Inc()
-		return nil, err
+	return data, nil
+}
+
+func (cfg *apiConfig) getLabels() ([]*promutil.Labels, error) {
+	cfg.init()
+
+	tlr := cfg.targetLabels.Load()
+	if tlr.err != nil {
+		return nil, tlr.err
 	}
-	return tg, nil
+	return tlr.labels, nil
+}
+
+func (cfg *apiConfig) mustStop() {
+	cfg.client.Stop()
+	cfg.wg.Wait()
 }
 
 func parseAPIResponse(data []byte, path string) ([]httpGroupTarget, error) {

lib/promscrape/discovery/http/http.go

@@ -23,19 +23,35 @@ type SDConfig struct {
 	HTTPClientConfig  promauth.HTTPClientConfig  `yaml:",inline"`
 	ProxyURL          *proxy.URL                 `yaml:"proxy_url,omitempty"`
 	ProxyClientConfig promauth.ProxyClientConfig `yaml:",inline"`
+
+	cfg      *apiConfig
+	startErr error
+}
+
+// MustStart initializes sdc before its usage.
+func (sdc *SDConfig) MustStart(baseDir string) {
+	cfg, err := newAPIConfig(sdc, baseDir)
+	if err != nil {
+		sdc.startErr = fmt.Errorf("cannot create API config for http_sd: %w", err)
+		return
+	}
+	sdc.cfg = cfg
 }
 
 // GetLabels returns http service discovery labels according to sdc.
 func (sdc *SDConfig) GetLabels(baseDir string) ([]*promutil.Labels, error) {
-	cfg, err := getAPIConfig(sdc, baseDir)
-	if err != nil {
-		return nil, fmt.Errorf("cannot get API config: %w", err)
+	if sdc.cfg == nil {
+		return nil, sdc.startErr
 	}
-	hts, err := getHTTPTargets(cfg)
-	if err != nil {
-		return nil, err
+	return sdc.cfg.getLabels()
+}
+
+// MustStop stops further usage for sdc.
+func (sdc *SDConfig) MustStop() {
+	if sdc.cfg == nil {
+		return
 	}
-	return addHTTPTargetLabels(hts, sdc.URL), nil
+	sdc.cfg.mustStop()
 }
 
 func addHTTPTargetLabels(src []httpGroupTarget, sourceURL string) []*promutil.Labels {
@@ -54,12 +70,3 @@ func addHTTPTargetLabels(src []httpGroupTarget, sourceURL string) []*promutil.La
 	}
 	return ms
 }
-
-// MustStop stops further usage for sdc.
-func (sdc *SDConfig) MustStop() {
-	v := configMap.Delete(sdc)
-	if v != nil {
-		cfg := v.(*apiConfig)
-		cfg.client.Stop()
-	}
-}

lib/promscrape/discovery/http/http_test.go

@@ -1,6 +1,9 @@
 package http
 
 import (
+	"net/http"
+	"net/http/httptest"
+	"sync/atomic"
 	"testing"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promscrape/discoveryutil"
@@ -38,3 +41,121 @@ func TestAddHTTPTargetLabels(t *testing.T) {
 	}
 	f(src, labelssExpected)
 }
+
+func TestSDConfigGetLabels(t *testing.T) {
+
+	type apiResponse struct {
+		statusCode int
+		body       string
+	}
+	var currentResponse atomic.Pointer[apiResponse]
+
+	// add initial non-empty response
+	currentResponse.Store(&apiResponse{
+		body:       `[{"targets":["10.0.0.2:9100"],"labels":{"job":"node"}}]`,
+		statusCode: http.StatusOK,
+	})
+
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		resp := currentResponse.Load()
+		w.WriteHeader(resp.statusCode)
+		_, _ = w.Write([]byte(resp.body))
+	}))
+	defer srv.Close()
+
+	sdc := &SDConfig{
+		URL: srv.URL,
+	}
+	sdc.MustStart(".")
+	defer sdc.MustStop()
+
+	assertLabelss := func(expectedLabelss []*promutil.Labels) {
+		t.Helper()
+		got, err := sdc.GetLabels(".")
+		if err != nil {
+			t.Fatalf("unexpected GetLabels error: %s", err)
+		}
+		if len(got) == 0 && len(expectedLabelss) == 0 {
+			return
+		}
+		discoveryutil.TestEqualLabelss(t, got, expectedLabelss)
+	}
+
+	// check initial state, it must be non-empty
+	// it also inits apiConfig below
+	assertLabelss([]*promutil.Labels{
+		promutil.NewLabelsFromMap(map[string]string{
+			"__address__": "10.0.0.2:9100",
+			"job":         "node",
+			"__meta_url":  srv.URL,
+		}),
+	})
+
+	updateAPIResponse := func(response apiResponse) {
+		currentResponse.Store(&response)
+		sdc.cfg.refreshTargetsIfNeeded()
+
+	}
+
+	// change response to empty
+	updateAPIResponse(apiResponse{
+		statusCode: http.StatusOK,
+		body:       `[]`,
+	})
+	assertLabelss([]*promutil.Labels{})
+
+	// change response to non-empty
+	updateAPIResponse(apiResponse{
+		statusCode: http.StatusOK,
+		body:       `[{"targets":["10.0.0.1:9100"],"labels":{"job":"node"}},{"targets":["10.0.0.5:8429"],"labels":{"job":"vmagent"}}]`,
+	})
+	assertLabelss([]*promutil.Labels{
+		promutil.NewLabelsFromMap(map[string]string{
+			"__address__": "10.0.0.1:9100",
+			"job":         "node",
+			"__meta_url":  srv.URL,
+		}),
+		promutil.NewLabelsFromMap(map[string]string{
+			"__address__": "10.0.0.5:8429",
+			"job":         "vmagent",
+			"__meta_url":  srv.URL,
+		}),
+	})
+
+	// change response to error
+	updateAPIResponse(apiResponse{
+		statusCode: http.StatusServiceUnavailable,
+		body:       `Internal Server Error`,
+	})
+	_, err := sdc.GetLabels(".")
+	if err == nil {
+		t.Fatalf("unexpected empty error")
+	}
+
+	// transit back to correct api response
+	updateAPIResponse(apiResponse{
+		statusCode: http.StatusOK,
+		body:       `[{"targets":["10.0.0.1:9100"],"labels":{"job":"node"}},{"targets":["10.0.0.5:8429"],"labels":{"job":"vmagent"}}]`,
+	})
+	assertLabelss([]*promutil.Labels{
+		promutil.NewLabelsFromMap(map[string]string{
+			"__address__": "10.0.0.1:9100",
+			"job":         "node",
+			"__meta_url":  srv.URL,
+		}),
+		promutil.NewLabelsFromMap(map[string]string{
+			"__address__": "10.0.0.5:8429",
+			"job":         "vmagent",
+			"__meta_url":  srv.URL,
+		}),
+	})
+
+	// make sure that api response is properly cached
+	before := sdc.cfg.targetLabels.Load()
+	updateAPIResponse(apiResponse{statusCode: http.StatusOK,
+		body: `[{"targets":["10.0.0.1:9100"],"labels":{"job":"node"}},{"targets":["10.0.0.5:8429"],"labels":{"job":"vmagent"}}]`})
+
+	if sdc.cfg.targetLabels.Load() != before {
+		t.Fatalf("expected identical response to be deduplicated")
+	}
+}

lib/storage/index_db.go

@@ -1242,12 +1242,9 @@ func (db *indexDB) GetSeriesCount(deadline uint64) (uint64, error) {
 func (is *indexSearch) getSeriesCount() (uint64, error) {
 	ts := &is.ts
 	kb := &is.kb
-	mp := &is.mp
 	loopsPaceLimiter := 0
 	var metricIDsLen uint64
-	// Extract the number of series from ((__name__=value): metricIDs) rows
-	kb.B = is.marshalCommonPrefix(kb.B[:0], nsPrefixTagToMetricIDs)
-	kb.B = marshalTagValue(kb.B, nil)
+	kb.B = is.marshalCommonPrefix(kb.B[:0], nsPrefixMetricIDToTSID)
 	ts.Seek(kb.B)
 	for ts.NextItem() {
 		if loopsPaceLimiter&paceLimiterFastIterationsMask == 0 {
@@ -1260,19 +1257,10 @@ func (is *indexSearch) getSeriesCount() (uint64, error) {
 		if !bytes.HasPrefix(item, kb.B) {
 			break
 		}
-		tail := item[len(kb.B):]
-		n := bytes.IndexByte(tail, tagSeparatorChar)
-		if n < 0 {
-			return 0, fmt.Errorf("invalid tag->metricIDs line %q: cannot find tagSeparatorChar %d", item, tagSeparatorChar)
-		}
-		tail = tail[n+1:]
-		if err := mp.InitOnlyTail(item, tail); err != nil {
-			return 0, err
-		}
 		// Take into account deleted timeseries too.
 		// It is OK if series can be counted multiple times in rare cases -
 		// the returned number is an estimation.
-		metricIDsLen += uint64(mp.MetricIDsLen())
+		metricIDsLen++
 	}
 	if err := ts.Error(); err != nil {
 		return 0, fmt.Errorf("error when counting unique timeseries: %w", err)

lib/storage/index_db_test.go

@@ -1949,10 +1949,11 @@ func newTestStorage() *Storage {
 	s := &Storage{
 		cachePath: "test-storage-cache",
 
-		metricIDCache:   workingsetcache.New(1234),
-		metricNameCache: workingsetcache.New(1234),
-		tsidCache:       workingsetcache.New(1234),
-		retentionMsecs:  retentionMax.Milliseconds(),
+		metricIDCache:       workingsetcache.New(1234),
+		metricNameCache:     workingsetcache.New(1234),
+		tsidCache:           workingsetcache.New(1234),
+		retentionMsecs:      retentionMax.Milliseconds(),
+		maxBackfillAgeMsecs: retentionMax.Milliseconds(),
 	}
 	return s
 }

lib/storage/part.go

@@ -42,7 +42,8 @@ type part struct {
 	valuesFile     fs.MustReadAtCloser
 	indexFile      fs.MustReadAtCloser
 
-	metaindex []metaindexRow
+	metaindex          []metaindexRow
+	metaindexSizeBytes uint64
 }
 
 // mustOpenFilePart opens file-based part from the given path.
@@ -102,6 +103,7 @@ func newPart(ph *partHeader, path string, size uint64, metaindexReader filestrea
 	p.valuesFile = valuesFile
 	p.indexFile = indexFile
 	p.metaindex = metaindex
+	p.metaindexSizeBytes = metaindexSizeBytes(metaindex)
 
 	return &p
 }
@@ -128,6 +130,10 @@ func (p *part) MustClose() {
 	ibCache.RemoveBlocksForPart(p)
 }
 
+func metaindexSizeBytes(metaindex []metaindexRow) uint64 {
+	return uint64(cap(metaindex)) * uint64(unsafe.Sizeof(metaindexRow{}))
+}
+
 type indexBlock struct {
 	bhs []blockHeader
 }

lib/storage/partition.go

@@ -11,9 +11,7 @@ import (
 	"sync"
 	"sync/atomic"
 	"time"
-	"unsafe"
 
-	"github.com/VictoriaMetrics/VictoriaMetrics/lib/atomicutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/encoding"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
@@ -40,14 +38,6 @@ const maxInmemoryParts = 60
 // See appendPartsToMerge tests for details.
 const defaultPartsToMerge = 15
 
-// The number of shards for rawRow entries per partition.
-//
-// Higher number of shards reduces CPU contention and increases the max bandwidth on multi-core systems.
-var rawRowsShardsPerPartition = cgroup.AvailableCPUs()
-
-// The interval for flushing buffered rows into parts, so they become visible to search.
-const pendingRowsFlushInterval = 2 * time.Second
-
 // The interval for guaranteed flush of recently ingested data from memory to on-disk parts, so they survive process crash.
 var dataFlushInterval = 5 * time.Second
 
@@ -66,11 +56,6 @@ func SetDataFlushInterval(d time.Duration) {
 	dataFlushInterval = d
 }
 
-// The maximum number of rawRow items in rawRowsShard.
-//
-// Limit the maximum shard size to 8Mb, since this gives the lowest CPU usage under high ingestion rate.
-const maxRawRowsPerShard = (8 << 20) / int(unsafe.Sizeof(rawRow{}))
-
 // partition represents a partition.
 type partition struct {
 	activeInmemoryMerges atomic.Int64
@@ -334,9 +319,10 @@ type partitionMetrics struct {
 	IndexBlocksCacheRequests     uint64
 	IndexBlocksCacheMisses       uint64
 
-	InmemorySizeBytes uint64
-	SmallSizeBytes    uint64
-	BigSizeBytes      uint64
+	InmemorySizeBytes  uint64
+	SmallSizeBytes     uint64
+	BigSizeBytes       uint64
+	MetaindexSizeBytes uint64
 
 	InmemoryRowsCount uint64
 	SmallRowsCount    uint64
@@ -397,6 +383,7 @@ func (pt *partition) UpdateMetrics(m *partitionMetrics) {
 		m.InmemoryRowsCount += p.ph.RowsCount
 		m.InmemoryBlocksCount += p.ph.BlocksCount
 		m.InmemorySizeBytes += p.size
+		m.MetaindexSizeBytes += p.metaindexSizeBytes
 		m.InmemoryPartsRefCount += uint64(pw.refCount.Load())
 		if isDedupScheduled {
 			m.ScheduledDownsamplingPartitionsSize += p.size
@@ -407,6 +394,7 @@ func (pt *partition) UpdateMetrics(m *partitionMetrics) {
 		m.SmallRowsCount += p.ph.RowsCount
 		m.SmallBlocksCount += p.ph.BlocksCount
 		m.SmallSizeBytes += p.size
+		m.MetaindexSizeBytes += p.metaindexSizeBytes
 		m.SmallPartsRefCount += uint64(pw.refCount.Load())
 		if isDedupScheduled {
 			m.ScheduledDownsamplingPartitionsSize += p.size
@@ -417,6 +405,7 @@ func (pt *partition) UpdateMetrics(m *partitionMetrics) {
 		m.BigRowsCount += p.ph.RowsCount
 		m.BigBlocksCount += p.ph.BlocksCount
 		m.BigSizeBytes += p.size
+		m.MetaindexSizeBytes += p.metaindexSizeBytes
 		m.BigPartsRefCount += uint64(pw.refCount.Load())
 		if isDedupScheduled {
 			m.ScheduledDownsamplingPartitionsSize += p.size
@@ -476,150 +465,29 @@ func (pt *partition) AddRows(rows []rawRow) {
 		}
 	}
 
-	pt.rawRows.addRows(pt, rows)
+	pt.rawRows.addRows(pt.flushRowssToInmemoryParts, rows)
 }
 
 var isDebug = false
 
-type rawRowsShards struct {
-	flushDeadlineMs atomic.Int64
-
-	shardIdx atomic.Uint32
-
-	// Shards reduce lock contention when adding rows on multi-CPU systems.
-	shards []rawRowsShard
-
-	rowssToFlushLock sync.Mutex
-	rowssToFlush     [][]rawRow
-}
-
-func (rrss *rawRowsShards) init() {
-	rrss.shards = make([]rawRowsShard, rawRowsShardsPerPartition)
-}
-
-func (rrss *rawRowsShards) addRows(pt *partition, rows []rawRow) {
-	shards := rrss.shards
-	shardsLen := uint32(len(shards))
-	for len(rows) > 0 {
-		n := rrss.shardIdx.Add(1)
-		idx := n % shardsLen
-		tailRows, rowsToFlush := shards[idx].addRows(rows)
-		rrss.addRowsToFlush(pt, rowsToFlush)
-		rows = tailRows
-	}
-}
-
-func (rrss *rawRowsShards) addRowsToFlush(pt *partition, rowsToFlush []rawRow) {
-	if len(rowsToFlush) == 0 {
-		return
-	}
-
-	var rowssToMerge [][]rawRow
-
-	rrss.rowssToFlushLock.Lock()
-	if len(rrss.rowssToFlush) == 0 {
-		rrss.updateFlushDeadline()
-	}
-	rrss.rowssToFlush = append(rrss.rowssToFlush, rowsToFlush)
-	if len(rrss.rowssToFlush) >= defaultPartsToMerge {
-		rowssToMerge = rrss.rowssToFlush
-		rrss.rowssToFlush = nil
-	}
-	rrss.rowssToFlushLock.Unlock()
-
-	pt.flushRowssToInmemoryParts(rowssToMerge)
-}
-
-func (rrss *rawRowsShards) Len() int {
-	n := 0
-	for i := range rrss.shards[:] {
-		n += rrss.shards[i].Len()
-	}
-
-	rrss.rowssToFlushLock.Lock()
-	for _, rows := range rrss.rowssToFlush {
-		n += len(rows)
-	}
-	rrss.rowssToFlushLock.Unlock()
-
-	return n
-}
-
-func (rrss *rawRowsShards) updateFlushDeadline() {
-	rrss.flushDeadlineMs.Store(time.Now().Add(pendingRowsFlushInterval).UnixMilli())
-}
-
-type rawRowsShardNopad struct {
-	flushDeadlineMs atomic.Int64
-
-	mu   sync.Mutex
-	rows []rawRow
-}
-
-type rawRowsShard struct {
-	rawRowsShardNopad
-
-	// The padding prevents false sharing
-	_ [atomicutil.CacheLineSize - unsafe.Sizeof(rawRowsShardNopad{})%atomicutil.CacheLineSize]byte
-}
-
-func (rrs *rawRowsShard) Len() int {
-	rrs.mu.Lock()
-	n := len(rrs.rows)
-	rrs.mu.Unlock()
-	return n
-}
-
-func (rrs *rawRowsShard) addRows(rows []rawRow) ([]rawRow, []rawRow) {
-	var rowsToFlush []rawRow
-
-	rrs.mu.Lock()
-	if cap(rrs.rows) == 0 {
-		rrs.rows = newRawRows()
-	}
-	if len(rrs.rows) == 0 {
-		rrs.updateFlushDeadline()
-	}
-	n := copy(rrs.rows[len(rrs.rows):cap(rrs.rows)], rows)
-	rrs.rows = rrs.rows[:len(rrs.rows)+n]
-	rows = rows[n:]
-	if len(rows) > 0 {
-		rowsToFlush = rrs.rows
-		rrs.rows = newRawRows()
-		rrs.updateFlushDeadline()
-		n = copy(rrs.rows[:cap(rrs.rows)], rows)
-		rrs.rows = rrs.rows[:n]
-		rows = rows[n:]
-	}
-	rrs.mu.Unlock()
-
-	return rows, rowsToFlush
-}
-
-func newRawRows() []rawRow {
-	return make([]rawRow, 0, maxRawRowsPerShard)
-}
-
 func (pt *partition) flushRowssToInmemoryParts(rowss [][]rawRow) {
-	if len(rowss) == 0 {
+	var nonEmptyRowss [][]rawRow
+	for _, rows := range rowss {
+		if len(rows) > 0 {
+			nonEmptyRowss = append(nonEmptyRowss, rows)
+		}
+	}
+	if len(nonEmptyRowss) == 0 {
 		return
 	}
 
 	// Convert rowss into in-memory parts.
-	var pwsLock sync.Mutex
-	pws := make([]*partWrapper, 0, len(rowss))
+	pws := make([]*partWrapper, len(nonEmptyRowss))
 	wg := getWaitGroup()
-	for _, rows := range rowss {
+	for i, rows := range nonEmptyRowss {
 		inmemoryPartsConcurrencyCh <- struct{}{}
-
 		wg.Go(func() {
-			pw := pt.createInmemoryPart(rows)
-			if pw != nil {
-				pwsLock.Lock()
-				pws = append(pws, pw)
-				pwsLock.Unlock()
-			}
-
+			pws[i] = pt.mustCreateInmemoryPart(rows)
 			<-inmemoryPartsConcurrencyCh
 		})
 	}
@@ -874,9 +742,14 @@ func (pt *partition) mustMergeInmemoryPartsFinal(pws []*partWrapper) *partWrappe
 	return newPartWrapperFromInmemoryPart(mpDst, flushToDiskDeadline)
 }
 
-func (pt *partition) createInmemoryPart(rows []rawRow) *partWrapper {
+// mustCreateInmemoryPart creates a new in-memory part from rawRows.
+//
+// The number of rawRows cannot be zero. Otherwise the method will panic.
+//
+// The returned value is always a non-nil partWrapper.
+func (pt *partition) mustCreateInmemoryPart(rows []rawRow) *partWrapper {
 	if len(rows) == 0 {
-		return nil
+		logger.Panicf("BUG: a part cannot be created from 0 rawRows")
 	}
 	mp := getInmemoryPart()
 	mp.InitFromRows(rows)
@@ -1138,7 +1011,7 @@ func (pt *partition) pendingRowsFlusher() {
 }
 
 func (pt *partition) flushPendingRows(isFinal bool) {
-	pt.rawRows.flush(pt, isFinal)
+	pt.rawRows.flush(pt.flushRowssToInmemoryParts, isFinal)
 }
 
 func (pt *partition) flushInmemoryRowsToFiles() {
@@ -1164,66 +1037,6 @@ func (pt *partition) flushInmemoryPartsToFiles(isFinal bool) {
 	}
 }
 
-func (rrss *rawRowsShards) flush(pt *partition, isFinal bool) {
-	var dst [][]rawRow
-
-	currentTimeMs := time.Now().UnixMilli()
-	flushDeadlineMs := rrss.flushDeadlineMs.Load()
-	if isFinal || currentTimeMs >= flushDeadlineMs {
-		rrss.rowssToFlushLock.Lock()
-		dst = rrss.rowssToFlush
-		rrss.rowssToFlush = nil
-		rrss.rowssToFlushLock.Unlock()
-	}
-
-	for i := range rrss.shards {
-		dst = rrss.shards[i].appendRawRowsToFlush(dst, currentTimeMs, isFinal)
-	}
-
-	pt.flushRowssToInmemoryParts(dst)
-}
-
-func (rrs *rawRowsShard) appendRawRowsToFlush(dst [][]rawRow, currentTimeMs int64, isFinal bool) [][]rawRow {
-	flushDeadlineMs := rrs.flushDeadlineMs.Load()
-	if !isFinal && currentTimeMs < flushDeadlineMs {
-		// Fast path - nothing to flush
-		return dst
-	}
-
-	// Slow path - move rrs.rows to dst.
-	rrs.mu.Lock()
-	dst = appendRawRowss(dst, rrs.rows)
-	rrs.rows = rrs.rows[:0]
-	rrs.mu.Unlock()
-
-	return dst
-}
-
-func (rrs *rawRowsShard) updateFlushDeadline() {
-	rrs.flushDeadlineMs.Store(time.Now().Add(pendingRowsFlushInterval).UnixMilli())
-}
-
-func appendRawRowss(dst [][]rawRow, src []rawRow) [][]rawRow {
-	if len(src) == 0 {
-		return dst
-	}
-	if len(dst) == 0 {
-		dst = append(dst, newRawRows())
-	}
-	prows := &dst[len(dst)-1]
-	n := copy((*prows)[len(*prows):cap(*prows)], src)
-	*prows = (*prows)[:len(*prows)+n]
-	src = src[n:]
-	for len(src) > 0 {
-		rows := newRawRows()
-		n := copy(rows[:cap(rows)], src)
-		rows = rows[:len(rows)+n]
-		src = src[n:]
-		dst = append(dst, rows)
-	}
-	return dst
-}
-
 func (pt *partition) mergePartsToFiles(pws []*partWrapper, stopCh <-chan struct{}, concurrencyCh chan struct{}, useSparseCache bool) error {
 	pwsLen := len(pws)
 

lib/storage/raw_row.go

@@ -3,11 +3,29 @@ package storage
 import (
 	"sort"
 	"sync"
+	"sync/atomic"
+	"time"
+	"unsafe"
 
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/atomicutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/cgroup"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/decimal"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 )
 
+// The number of shards for rawRow entries.
+//
+// Higher number of shards reduces CPU contention and increases the max bandwidth on multi-core systems.
+var numRawRowsShards = cgroup.AvailableCPUs()
+
+// The interval for flushing buffered rows into parts, so they become visible to search.
+const pendingRowsFlushInterval = 2 * time.Second
+
+// The maximum number of rawRow items in rawRowsShard.
+//
+// Limit the maximum shard size to 8Mb, since this gives the lowest CPU usage under high ingestion rate.
+const maxRawRowsPerShard = (8 << 20) / int(unsafe.Sizeof(rawRow{}))
+
 // rawRow represents raw timeseries row.
 type rawRow struct {
 	// TSID is time series id.
@@ -149,3 +167,182 @@ func putRawRowsMarshaler(rrm *rawRowsMarshaler) {
 }
 
 var rrmPool sync.Pool
+
+type rawRowsShards struct {
+	flushDeadlineMs atomic.Int64
+
+	shardIdx atomic.Uint32
+
+	// Shards reduce lock contention when adding rows on multi-CPU systems.
+	shards []rawRowsShard
+
+	rowssToFlushLock sync.Mutex
+	rowssToFlush     [][]rawRow
+}
+
+func (rrss *rawRowsShards) init() {
+	rrss.shards = make([]rawRowsShard, numRawRowsShards)
+}
+
+func (rrss *rawRowsShards) Len() int {
+	n := 0
+	for i := range rrss.shards[:] {
+		n += rrss.shards[i].Len()
+	}
+
+	rrss.rowssToFlushLock.Lock()
+	for _, rows := range rrss.rowssToFlush {
+		n += len(rows)
+	}
+	rrss.rowssToFlushLock.Unlock()
+
+	return n
+}
+
+func (rrss *rawRowsShards) addRows(flush func([][]rawRow), rows []rawRow) {
+	shards := rrss.shards
+	shardsLen := uint32(len(shards))
+	for len(rows) > 0 {
+		n := rrss.shardIdx.Add(1)
+		idx := n % shardsLen
+		tailRows, rowsToFlush := shards[idx].addRows(rows)
+		rrss.addRowsToFlush(flush, rowsToFlush)
+		rows = tailRows
+	}
+}
+
+func (rrss *rawRowsShards) addRowsToFlush(flush func([][]rawRow), rowsToFlush []rawRow) {
+	if len(rowsToFlush) == 0 {
+		return
+	}
+
+	var rowssToMerge [][]rawRow
+
+	rrss.rowssToFlushLock.Lock()
+	if len(rrss.rowssToFlush) == 0 {
+		rrss.updateFlushDeadline()
+	}
+	rrss.rowssToFlush = append(rrss.rowssToFlush, rowsToFlush)
+	if len(rrss.rowssToFlush) >= defaultPartsToMerge {
+		rowssToMerge = rrss.rowssToFlush
+		rrss.rowssToFlush = nil
+	}
+	rrss.rowssToFlushLock.Unlock()
+
+	flush(rowssToMerge)
+}
+
+func (rrss *rawRowsShards) updateFlushDeadline() {
+	rrss.flushDeadlineMs.Store(time.Now().Add(pendingRowsFlushInterval).UnixMilli())
+}
+
+func (rrss *rawRowsShards) flush(flush func(rrs [][]rawRow), isFinal bool) {
+	var dst [][]rawRow
+
+	currentTimeMs := time.Now().UnixMilli()
+	flushDeadlineMs := rrss.flushDeadlineMs.Load()
+	if isFinal || currentTimeMs >= flushDeadlineMs {
+		rrss.rowssToFlushLock.Lock()
+		dst = rrss.rowssToFlush
+		rrss.rowssToFlush = nil
+		rrss.rowssToFlushLock.Unlock()
+	}
+
+	for i := range rrss.shards {
+		dst = rrss.shards[i].appendRawRowsToFlush(dst, currentTimeMs, isFinal)
+	}
+
+	flush(dst)
+}
+
+type rawRowsShardNopad struct {
+	flushDeadlineMs atomic.Int64
+
+	mu   sync.Mutex
+	rows []rawRow
+}
+
+type rawRowsShard struct {
+	rawRowsShardNopad
+
+	// The padding prevents false sharing
+	_ [atomicutil.CacheLineSize - unsafe.Sizeof(rawRowsShardNopad{})%atomicutil.CacheLineSize]byte
+}
+
+func (rrs *rawRowsShard) Len() int {
+	rrs.mu.Lock()
+	n := len(rrs.rows)
+	rrs.mu.Unlock()
+	return n
+}
+
+func (rrs *rawRowsShard) addRows(rows []rawRow) ([]rawRow, []rawRow) {
+	var rowsToFlush []rawRow
+
+	rrs.mu.Lock()
+	if cap(rrs.rows) == 0 {
+		rrs.rows = newRawRows()
+	}
+	if len(rrs.rows) == 0 {
+		rrs.updateFlushDeadline()
+	}
+	n := copy(rrs.rows[len(rrs.rows):cap(rrs.rows)], rows)
+	rrs.rows = rrs.rows[:len(rrs.rows)+n]
+	rows = rows[n:]
+	if len(rows) > 0 {
+		rowsToFlush = rrs.rows
+		rrs.rows = newRawRows()
+		rrs.updateFlushDeadline()
+		n = copy(rrs.rows[:cap(rrs.rows)], rows)
+		rrs.rows = rrs.rows[:n]
+		rows = rows[n:]
+	}
+	rrs.mu.Unlock()
+
+	return rows, rowsToFlush
+}
+
+func newRawRows() []rawRow {
+	return make([]rawRow, 0, maxRawRowsPerShard)
+}
+
+func (rrs *rawRowsShard) updateFlushDeadline() {
+	rrs.flushDeadlineMs.Store(time.Now().Add(pendingRowsFlushInterval).UnixMilli())
+}
+
+func (rrs *rawRowsShard) appendRawRowsToFlush(dst [][]rawRow, currentTimeMs int64, isFinal bool) [][]rawRow {
+	flushDeadlineMs := rrs.flushDeadlineMs.Load()
+	if !isFinal && currentTimeMs < flushDeadlineMs {
+		// Fast path - nothing to flush
+		return dst
+	}
+
+	// Slow path - move rrs.rows to dst.
+	rrs.mu.Lock()
+	dst = appendRawRowss(dst, rrs.rows)
+	rrs.rows = rrs.rows[:0]
+	rrs.mu.Unlock()
+
+	return dst
+}
+
+func appendRawRowss(dst [][]rawRow, src []rawRow) [][]rawRow {
+	if len(src) == 0 {
+		return dst
+	}
+	if len(dst) == 0 {
+		dst = append(dst, newRawRows())
+	}
+	prows := &dst[len(dst)-1]
+	n := copy((*prows)[len(*prows):cap(*prows)], src)
+	*prows = (*prows)[:len(*prows)+n]
+	src = src[n:]
+	for len(src) > 0 {
+		rows := newRawRows()
+		n := copy(rows[:cap(rows)], src)
+		rows = rows[:len(rows)+n]
+		src = src[n:]
+		dst = append(dst, rows)
+	}
+	return dst
+}

lib/storage/storage.go

@@ -65,6 +65,7 @@ type Storage struct {
 	cachePath                   string
 	retentionMsecs              int64
 	futureRetentionMsecs        int64
+	maxBackfillAgeMsecs         int64
 	denyQueriesOutsideRetention bool
 
 	// lock file for exclusive access to the storage on the given path.
@@ -164,6 +165,7 @@ type Storage struct {
 type OpenOptions struct {
 	Retention                   time.Duration
 	FutureRetention             time.Duration
+	MaxBackfillAge              time.Duration
 	DenyQueriesOutsideRetention bool
 	MaxHourlySeries             int
 	MaxDailySeries              int
@@ -187,6 +189,10 @@ func MustOpenStorage(path string, opts OpenOptions) *Storage {
 		retention = retentionMax
 	}
 	futureRetention := max(opts.FutureRetention, retention2Days)
+	maxBackfillAge := opts.MaxBackfillAge
+	if maxBackfillAge <= 0 || maxBackfillAge > retention {
+		maxBackfillAge = retention
+	}
 	idbPrefillStart := opts.IDBPrefillStart
 	if idbPrefillStart <= 0 {
 		idbPrefillStart = time.Hour
@@ -196,6 +202,7 @@ func MustOpenStorage(path string, opts OpenOptions) *Storage {
 		cachePath:                   filepath.Join(path, cacheDirname),
 		retentionMsecs:              retention.Milliseconds(),
 		futureRetentionMsecs:        futureRetention.Milliseconds(),
+		maxBackfillAgeMsecs:         maxBackfillAge.Milliseconds(),
 		denyQueriesOutsideRetention: opts.DenyQueriesOutsideRetention,
 		stopCh:                      make(chan struct{}),
 		idbPrefillStartSeconds:      idbPrefillStart.Milliseconds() / 1000,
@@ -1236,7 +1243,7 @@ func (s *Storage) checkTimeRange(tr TimeRange) error {
 		return nil
 	}
 
-	minTimestamp, maxTimestamp := s.tb.getMinMaxTimestamps()
+	minTimestamp, maxTimestamp := s.tb.getMinMaxRetentionTimestamps()
 	if minTimestamp <= tr.MinTimestamp && tr.MaxTimestamp <= maxTimestamp {
 		return nil
 	}
@@ -1896,7 +1903,7 @@ func (s *Storage) add(rows []rawRow, dstMrs []*MetricRow, mrs []MetricRow, preci
 	var newSeriesCount uint64
 	var seriesRepopulated uint64
 
-	minTimestamp, maxTimestamp := s.tb.getMinMaxTimestamps()
+	minTimestamp, maxTimestamp := s.tb.getMinMaxIngestionTimestamps()
 
 	var lTSID legacyTSID
 	var ptw *partitionWrapper
@@ -1918,11 +1925,11 @@ func (s *Storage) add(rows []rawRow, dstMrs []*MetricRow, mrs []MetricRow, preci
 			}
 		}
 		if mr.Timestamp < minTimestamp {
-			// Skip rows with too small timestamps outside the retention.
+			// Skip rows with too small timestamps outside the retention or -maxBackfillAge.
 			if firstWarn == nil {
 				metricName := getUserReadableMetricName(mr.MetricNameRaw)
-				firstWarn = fmt.Errorf("cannot insert row with too small timestamp %d outside the retention; minimum allowed timestamp is %d; "+
-					"probably you need updating -retentionPeriod command-line flag; metricName: %s",
+				firstWarn = fmt.Errorf("cannot insert row with too small timestamp %d; minimum allowed timestamp is %d; "+
+					"probably you need updating -retentionPeriod or -maxBackfillAge command-line flags; metricName: %s",
 					mr.Timestamp, minTimestamp, metricName)
 			}
 			s.tooSmallTimestampRows.Add(1)

lib/storage/storage_synctest_test.go

@@ -1410,3 +1410,80 @@ func TestStorage_denyQueriesOutsideRetention(t *testing.T) {
 
 	})
 }
+
+func TestStorageAddRows_MaxBackfillAge(t *testing.T) {
+	defer testRemoveAll(t)
+
+	mn := MetricName{
+		MetricGroup: []byte("metric"),
+	}
+	mr := MetricRow{
+		MetricNameRaw: mn.marshalRaw(nil),
+		Value:         123,
+	}
+
+	f := func(s *Storage, age time.Duration, want uint64) {
+		t.Helper()
+
+		mr.Timestamp = time.Now().UTC().Add(-age).UnixMilli()
+		s.AddRows([]MetricRow{mr}, defaultPrecisionBits)
+		s.DebugFlush()
+		if got := s.tooSmallTimestampRows.Load(); got != want {
+			t.Fatalf("unexpected number of tooSmallTimestampRows: got %d, want %d", got, want)
+		}
+	}
+
+	synctest.Test(t, func(t *testing.T) {
+		// synctest time begins at 2000-01-01T00:00:00Z.
+
+		retention1y := 365 * 24 * time.Hour
+		var s *Storage
+
+		s = MustOpenStorage(t.Name(), OpenOptions{
+			Retention: retention1y,
+			// By default MaxBackfillAge must be the same as Retention
+		})
+		// Verify that the sample with timestamp 1ms older than retention is
+		// rejected.
+		f(s, retention1y+time.Millisecond, 1)
+		// Verify that the sample with timestamp which is exactly at retention
+		// boundary is accepted.
+		f(s, retention1y, 1)
+
+		// Restart storage with negative MaxBackfillAge. In this case,
+		// MaxBackfillAge must be the same as Retention.
+		// Also advance time a bit so that the storage will not use the same
+		// nanosecond for creating a new part for storing the samples.
+		s.MustClose()
+		time.Sleep(time.Nanosecond)
+		s = MustOpenStorage(t.Name(), OpenOptions{
+			Retention:      retention1y,
+			MaxBackfillAge: -1,
+		})
+		f(s, retention1y+time.Millisecond, 1)
+		f(s, retention1y, 1)
+
+		// Restart storage with MaxBackfillAge bigger than Retention. In this
+		// case, MaxBackfillAge must be the same as Retention.
+		s.MustClose()
+		time.Sleep(time.Nanosecond)
+		s = MustOpenStorage(t.Name(), OpenOptions{
+			Retention:      retention1y,
+			MaxBackfillAge: retention1y + time.Millisecond,
+		})
+		f(s, retention1y+time.Millisecond, 1)
+		f(s, retention1y, 1)
+
+		// Restart storage with MaxBackfillAge smaller than Retention.
+		s.MustClose()
+		time.Sleep(time.Nanosecond)
+		s = MustOpenStorage(t.Name(), OpenOptions{
+			Retention:      retention1y,
+			MaxBackfillAge: retention1y - time.Millisecond,
+		})
+		f(s, retention1y, 1)
+		f(s, retention1y-time.Millisecond, 1)
+
+		s.MustClose()
+	})
+}

lib/storage/storage_test.go

@@ -1807,18 +1807,19 @@ func TestStorageRowsNotAdded(t *testing.T) {
 	defer testRemoveAll(t)
 
 	type options struct {
-		name        string
-		retention   time.Duration
-		mrs         []MetricRow
-		tr          TimeRange
-		wantMetrics *Metrics
+		name           string
+		retention      time.Duration
+		maxBackfillAge time.Duration
+		mrs            []MetricRow
+		tr             TimeRange
+		wantMetrics    *Metrics
 	}
 	f := func(opts *options) {
 		t.Helper()
 
 		var gotMetrics Metrics
 		path := fmt.Sprintf("%s/%s", t.Name(), opts.name)
-		s := MustOpenStorage(path, OpenOptions{Retention: opts.retention})
+		s := MustOpenStorage(path, OpenOptions{Retention: opts.retention, MaxBackfillAge: opts.maxBackfillAge})
 		defer s.MustClose()
 		s.AddRows(opts.mrs, defaultPrecisionBits)
 		s.DebugFlush()
@@ -1890,6 +1891,22 @@ func TestStorageRowsNotAdded(t *testing.T) {
 		},
 	})
 
+	retention = retentionMax
+	maxBackfillAge := 48 * time.Hour
+	minTimestamp = time.Now().Add(-maxBackfillAge - time.Hour).UnixMilli()
+	maxTimestamp = minTimestamp + 1000
+	f(&options{
+		name:           "TooSmallTimestampsForMaxBackfillAge",
+		retention:      retention,
+		maxBackfillAge: maxBackfillAge,
+		mrs:            testGenerateMetricRows(rng, numRows, minTimestamp, maxTimestamp),
+		tr:             TimeRange{minTimestamp, maxTimestamp},
+		wantMetrics: &Metrics{
+			RowsReceivedTotal:     numRows,
+			TooSmallTimestampRows: numRows,
+		},
+	})
+
 	minTimestamp = time.Now().UnixMilli()
 	maxTimestamp = minTimestamp + 1000
 	mrs = testGenerateMetricRows(rng, numRows, minTimestamp, maxTimestamp)

lib/storage/table.go

@@ -368,7 +368,7 @@ func (tb *table) MustAddRows(rows []rawRow) {
 	// The slowest path - there are rows that don't fit any existing partition.
 	// Create new partitions for these rows.
 	// Do this under tb.ptwsLock.
-	minTimestamp, maxTimestamp := tb.getMinMaxTimestamps()
+	minTimestamp, maxTimestamp := tb.getMinMaxIngestionTimestamps()
 	tb.ptwsLock.Lock()
 	for i := range missingRows {
 		r := &missingRows[i]
@@ -407,9 +407,28 @@ func (tb *table) MustGetIndexDBIDByHour(hour uint64) uint64 {
 	return ptw.pt.idb.id
 }
 
-func (tb *table) getMinMaxTimestamps() (int64, int64) {
+// getMinMaxRetentionTimestamps returns the minimum and maximum timestamps
+// allowed by the configured -retentionPeriod and -futureRetention.
+//
+// It is used for checking whether the given time range is fully covered
+// by the retention, e.g. for -denyQueriesOutsideRetention.
+func (tb *table) getMinMaxRetentionTimestamps() (int64, int64) {
+	return tb.getMinMaxTimestampsForAge(tb.s.retentionMsecs)
+}
+
+// getMinMaxIngestionTimestamps returns the minimum and maximum timestamps
+// allowed for newly ingested rows.
+//
+// The minimum timestamp is bound by -maxBackfillAge instead of -retentionPeriod,
+// since -maxBackfillAge can be configured to reject backfilled rows with historical
+// timestamps stricter than the full -retentionPeriod window.
+func (tb *table) getMinMaxIngestionTimestamps() (int64, int64) {
+	return tb.getMinMaxTimestampsForAge(tb.s.maxBackfillAgeMsecs)
+}
+
+func (tb *table) getMinMaxTimestampsForAge(minAgeMsecs int64) (int64, int64) {
 	now := int64(fasttime.UnixTimestamp() * 1000)
-	minTimestamp := now - tb.s.retentionMsecs
+	minTimestamp := now - minAgeMsecs
 	if minTimestamp < 0 {
 		// Negative timestamps aren't supported by the storage.
 		minTimestamp = 0

vendor/github.com/VictoriaMetrics/metricsql/binary_op.go

@@ -73,6 +73,16 @@ func isBinaryOp(op string) bool {
 	return binaryOps[op]
 }
 
+func isSetOperator(op string) bool {
+	op = strings.ToLower(op)
+	switch op {
+	case "and", "or", "unless", "if", "ifnot", "default":
+		return true
+	default:
+		return false
+	}
+}
+
 func binaryOpPriority(op string) int {
 	op = strings.ToLower(op)
 	return binaryOpPriorities[op]
@@ -123,6 +133,16 @@ func isBinaryOpBoolModifier(s string) bool {
 	return s == "bool"
 }
 
+func isBinaryOpFillModifier(s string) bool {
+	s = strings.ToLower(s)
+	switch s {
+	case "fill", "fill_left", "fill_right":
+		return true
+	default:
+		return false
+	}
+}
+
 // IsBinaryOpCmp returns true if op is comparison operator such as '==', '!=', etc.
 func IsBinaryOpCmp(op string) bool {
 	switch op {

vendor/github.com/VictoriaMetrics/metricsql/optimizer.go

@@ -39,6 +39,9 @@ func canOptimize(e Expr) bool {
 			}
 		}
 	case *BinaryOpExpr:
+		if t.FillLeft != nil && t.FillRight != nil {
+			return canOptimize(t.Left) || canOptimize(t.Right)
+		}
 		return true
 	}
 	return false
@@ -66,8 +69,23 @@ func optimizeInplace(e Expr) {
 	case *BinaryOpExpr:
 		optimizeInplace(t.Left)
 		optimizeInplace(t.Right)
-		lfs := getCommonLabelFilters(t)
-		pushdownBinaryOpFiltersInplace(lfs, t)
+		switch {
+		case t.FillLeft != nil && t.FillRight != nil:
+			// for two-sided fill, no cross-side pushdown.
+		case t.FillLeft != nil:
+			// for fill_left(), only propagate filters from right to left.
+			lfs := getCommonLabelFilters(t.Right)
+			lfs = TrimFiltersByGroupModifier(lfs, t)
+			pushdownBinaryOpFiltersInplace(lfs, t.Left)
+		case t.FillRight != nil:
+			// for fill_right(), only propagate filters from left to right.
+			lfs := getCommonLabelFilters(t.Left)
+			lfs = TrimFiltersByGroupModifier(lfs, t)
+			pushdownBinaryOpFiltersInplace(lfs, t.Right)
+		default:
+			lfs := getCommonLabelFilters(t)
+			pushdownBinaryOpFiltersInplace(lfs, t)
+		}
 	}
 }
 

vendor/github.com/VictoriaMetrics/metricsql/parser.go

@@ -402,6 +402,15 @@ func (p *parser) parseExpr() (Expr, error) {
 				}
 			}
 		}
+		for {
+			ok, err := p.parseFillModifier(&be)
+			if err != nil {
+				return nil, err
+			}
+			if !ok {
+				break
+			}
+		}
 		e2, err := p.parseSingleExpr()
 		if err != nil {
 			return nil, err
@@ -506,6 +515,65 @@ func (p *parser) parseSingleExprWithoutRollupSuffix() (Expr, error) {
 	}
 }
 
+func (p *parser) parseFillModifier(be *BinaryOpExpr) (bool, error) {
+	if !isBinaryOpFillModifier(p.lex.Token) {
+		return false, nil
+	}
+	if isSetOperator(be.Op) {
+		return false, fmt.Errorf(`fill modifier cannot be applied to %q`, be.Op)
+	}
+	op := strings.ToLower(p.lex.Token)
+	if err := p.lex.Next(); err != nil {
+		return false, err
+	}
+	if p.lex.Token != "(" {
+		p.lex.Prev()
+		return false, nil
+	}
+	if err := p.lex.Next(); err != nil {
+		return false, err
+	}
+	ne, err := p.parseFillValue()
+	if err != nil {
+		return false, fmt.Errorf("cannot parse %s fill value: %w", op, err)
+	}
+	if p.lex.Token != ")" {
+		return false, fmt.Errorf(`%s: unexpected token %q; want ")"`, op, p.lex.Token)
+	}
+	if err := p.lex.Next(); err != nil {
+		return false, err
+	}
+	switch op {
+	case "fill":
+		be.FillLeft = ne
+		be.FillRight = ne
+	case "fill_left":
+		be.FillLeft = ne
+	case "fill_right":
+		be.FillRight = ne
+	}
+	return true, nil
+}
+
+func (p *parser) parseFillValue() (*NumberExpr, error) {
+	neg := false
+	if p.lex.Token == "-" {
+		neg = true
+		if err := p.lex.Next(); err != nil {
+			return nil, err
+		}
+	}
+	ne, err := p.parsePositiveNumberExpr()
+	if err != nil {
+		return nil, err
+	}
+	if neg {
+		ne.N = -ne.N
+		ne.s = "-" + ne.s
+	}
+	return ne, nil
+}
+
 func (p *parser) parsePositiveNumberExpr() (*NumberExpr, error) {
 	if !isPositiveNumberPrefix(p.lex.Token) && !isInfOrNaN(p.lex.Token) {
 		return nil, fmt.Errorf(`positiveNumberExpr: unexpected token %q; want "number"`, p.lex.Token)
@@ -1896,6 +1964,12 @@ type BinaryOpExpr struct {
 	// If KeepMetricNames is set to true, then the operation should keep metric names.
 	KeepMetricNames bool
 
+	// FillLeft contains the fill value for fill_left() or fill() modifier.
+	FillLeft *NumberExpr
+
+	// FillRight contains the fill value for fill_right() or fill() modifier.
+	FillRight *NumberExpr
+
 	// Left contains left arg for the `left op right` expression.
 	Left Expr
 
@@ -1971,6 +2045,22 @@ func (be *BinaryOpExpr) appendModifiers(dst []byte) []byte {
 			dst = prefix.AppendString(dst)
 		}
 	}
+	if be.FillLeft != nil && be.FillLeft == be.FillRight {
+		dst = append(dst, " fill("...)
+		dst = be.FillLeft.AppendString(dst)
+		dst = append(dst, ')')
+	} else {
+		if be.FillLeft != nil {
+			dst = append(dst, " fill_left("...)
+			dst = be.FillLeft.AppendString(dst)
+			dst = append(dst, ')')
+		}
+		if be.FillRight != nil {
+			dst = append(dst, " fill_right("...)
+			dst = be.FillRight.AppendString(dst)
+			dst = append(dst, ')')
+		}
+	}
 	return dst
 }
 
@@ -1989,7 +2079,7 @@ func needBinaryOpArgParens(arg Expr) bool {
 }
 
 func isReservedBinaryOpIdent(s string) bool {
-	return isBinaryOpGroupModifier(s) || isBinaryOpJoinModifier(s) || isBinaryOpBoolModifier(s) || isPrefixModifier(s)
+	return isBinaryOpGroupModifier(s) || isBinaryOpJoinModifier(s) || isBinaryOpBoolModifier(s) || isPrefixModifier(s) || isBinaryOpFillModifier(s)
 }
 
 func isPrefixModifier(s string) bool {

vendor/golang.org/x/sys/cpu/parse.go

@@ -6,38 +6,50 @@ package cpu
 
 import "strconv"
 
-// parseRelease parses a dot-separated version number. It follows the semver
-// syntax, but allows the minor and patch versions to be elided.
+// parseRelease parses a dot-separated version number from the prefix
+// of rel. It returns ok=true only if at least the major and minor
+// components were successfully parsed; the patch component is
+// best-effort. Trailing vendor or build suffixes such as
+// "-generic", "+", "_hi3535", or "-rc1" are ignored.
 //
 // This is a copy of the Go runtime's parseRelease from
-// https://golang.org/cl/209597.
+// https://golang.org/cl/209597, updated in https://golang.org/cl/781800.
 func parseRelease(rel string) (major, minor, patch int, ok bool) {
-	// Strip anything after a dash or plus.
-	for i := range len(rel) {
-		if rel[i] == '-' || rel[i] == '+' {
-			rel = rel[:i]
-			break
+	// next consumes a run of decimal digits from the front of rel,
+	// returning the parsed value. If the digits are followed by a
+	// '.', it is consumed and more is set so the caller knows to
+	// parse another component; otherwise scanning terminates and
+	// the rest of rel is discarded.
+	next := func() (n int, more, ok bool) {
+		i := 0
+		for i < len(rel) && rel[i] >= '0' && rel[i] <= '9' {
+			i++
 		}
+		if i == 0 {
+			return 0, false, false
+		}
+		n, err := strconv.Atoi(rel[:i])
+		if err != nil {
+			return 0, false, false
+		}
+		if i < len(rel) && rel[i] == '.' {
+			rel = rel[i+1:]
+			return n, true, true
+		}
+		rel = ""
+		return n, false, true
 	}
 
-	next := func() (int, bool) {
-		for i := range len(rel) {
-			if rel[i] == '.' {
-				ver, err := strconv.Atoi(rel[:i])
-				rel = rel[i+1:]
-				return ver, err == nil
-			}
-		}
-		ver, err := strconv.Atoi(rel)
-		rel = ""
-		return ver, err == nil
+	var more bool
+	if major, more, ok = next(); !ok || !more {
+		return 0, 0, 0, false
 	}
-	if major, ok = next(); !ok || rel == "" {
-		return
+	if minor, more, ok = next(); !ok {
+		return 0, 0, 0, false
 	}
-	if minor, ok = next(); !ok || rel == "" {
-		return
+	if !more {
+		return major, minor, 0, true
 	}
-	patch, ok = next()
-	return
+	patch, _, _ = next()
+	return major, minor, patch, true
 }

vendor/golang.org/x/sys/unix/syscall_linux.go

@@ -1874,6 +1874,7 @@ func Dup2(oldfd, newfd int) error {
 //sys	Dup3(oldfd int, newfd int, flags int) (err error)
 //sysnb	EpollCreate1(flag int) (fd int, err error)
 //sysnb	EpollCtl(epfd int, op int, fd int, event *EpollEvent) (err error)
+//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) = SYS_EPOLL_PWAIT
 //sys	Eventfd(initval uint, flags int) (fd int, err error) = SYS_EVENTFD2
 //sys	Exit(code int) = SYS_EXIT_GROUP
 //sys	Fallocate(fd int, mode uint32, off int64, len int64) (err error)

vendor/golang.org/x/sys/unix/syscall_linux_386.go

@@ -20,7 +20,6 @@ func setTimeval(sec, usec int64) Timeval {
 
 // 64-bit file system and 32-bit uid calls
 // (386 default is 32-bit file system and 16-bit uid).
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error)
 //sys	Fadvise(fd int, offset int64, length int64, advice int) (err error) = SYS_FADVISE64_64
 //sys	Fchown(fd int, uid int, gid int) (err error) = SYS_FCHOWN32
 //sys	Fstat(fd int, stat *Stat_t) (err error) = SYS_FSTAT64

vendor/golang.org/x/sys/unix/syscall_linux_amd64.go

@@ -6,7 +6,6 @@
 
 package unix
 
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error)
 //sys	Fadvise(fd int, offset int64, length int64, advice int) (err error) = SYS_FADVISE64
 //sys	Fchown(fd int, uid int, gid int) (err error)
 //sys	Fstat(fd int, stat *Stat_t) (err error)

vendor/golang.org/x/sys/unix/syscall_linux_arm.go

@@ -44,7 +44,6 @@ func Seek(fd int, offset int64, whence int) (newoffset int64, err error) {
 
 // 64-bit file system and 32-bit uid calls
 // (16-bit uid calls are not always supported in newer kernels)
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error)
 //sys	Fchown(fd int, uid int, gid int) (err error) = SYS_FCHOWN32
 //sys	Fstat(fd int, stat *Stat_t) (err error) = SYS_FSTAT64
 //sys	Fstatat(dirfd int, path string, stat *Stat_t, flags int) (err error) = SYS_FSTATAT64

vendor/golang.org/x/sys/unix/syscall_linux_arm64.go

@@ -8,7 +8,6 @@ package unix
 
 import "unsafe"
 
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) = SYS_EPOLL_PWAIT
 //sys	Fadvise(fd int, offset int64, length int64, advice int) (err error) = SYS_FADVISE64
 //sys	Fchown(fd int, uid int, gid int) (err error)
 //sys	Fstat(fd int, stat *Stat_t) (err error)

vendor/golang.org/x/sys/unix/syscall_linux_loong64.go

@@ -8,7 +8,6 @@ package unix
 
 import "unsafe"
 
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) = SYS_EPOLL_PWAIT
 //sys	Fadvise(fd int, offset int64, length int64, advice int) (err error) = SYS_FADVISE64
 //sys	Fchown(fd int, uid int, gid int) (err error)
 //sys	Fstatfs(fd int, buf *Statfs_t) (err error)

vendor/golang.org/x/sys/unix/syscall_linux_mips64x.go

@@ -6,7 +6,6 @@
 
 package unix
 
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error)
 //sys	Fadvise(fd int, offset int64, length int64, advice int) (err error) = SYS_FADVISE64
 //sys	Fchown(fd int, uid int, gid int) (err error)
 //sys	Fstatfs(fd int, buf *Statfs_t) (err error)

vendor/golang.org/x/sys/unix/syscall_linux_mipsx.go

@@ -13,7 +13,6 @@ import (
 
 func Syscall9(trap, a1, a2, a3, a4, a5, a6, a7, a8, a9 uintptr) (r1, r2 uintptr, err syscall.Errno)
 
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error)
 //sys	Fadvise(fd int, offset int64, length int64, advice int) (err error) = SYS_FADVISE64
 //sys	Fchown(fd int, uid int, gid int) (err error)
 //sys	Ftruncate(fd int, length int64) (err error) = SYS_FTRUNCATE64

vendor/golang.org/x/sys/unix/syscall_linux_ppc.go

@@ -11,7 +11,6 @@ import (
 	"unsafe"
 )
 
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error)
 //sys	Fchown(fd int, uid int, gid int) (err error)
 //sys	Fstat(fd int, stat *Stat_t) (err error) = SYS_FSTAT64
 //sys	Fstatat(dirfd int, path string, stat *Stat_t, flags int) (err error) = SYS_FSTATAT64

vendor/golang.org/x/sys/unix/syscall_linux_ppc64x.go

@@ -6,7 +6,6 @@
 
 package unix
 
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error)
 //sys	Fadvise(fd int, offset int64, length int64, advice int) (err error) = SYS_FADVISE64
 //sys	Fchown(fd int, uid int, gid int) (err error)
 //sys	Fstat(fd int, stat *Stat_t) (err error)

vendor/golang.org/x/sys/unix/syscall_linux_riscv64.go

@@ -8,7 +8,6 @@ package unix
 
 import "unsafe"
 
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) = SYS_EPOLL_PWAIT
 //sys	Fadvise(fd int, offset int64, length int64, advice int) (err error) = SYS_FADVISE64
 //sys	Fchown(fd int, uid int, gid int) (err error)
 //sys	Fstat(fd int, stat *Stat_t) (err error)

vendor/golang.org/x/sys/unix/syscall_linux_s390x.go

@@ -10,7 +10,6 @@ import (
 	"unsafe"
 )
 
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error)
 //sys	Fadvise(fd int, offset int64, length int64, advice int) (err error) = SYS_FADVISE64
 //sys	Fchown(fd int, uid int, gid int) (err error)
 //sys	Fstat(fd int, stat *Stat_t) (err error)

vendor/golang.org/x/sys/unix/syscall_linux_sparc64.go

@@ -6,7 +6,6 @@
 
 package unix
 
-//sys	EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error)
 //sys	Fadvise(fd int, offset int64, length int64, advice int) (err error) = SYS_FADVISE64
 //sys	Fchown(fd int, uid int, gid int) (err error)
 //sys	Fstat(fd int, stat *Stat_t) (err error)

vendor/golang.org/x/sys/unix/zerrors_linux.go

@@ -1359,6 +1359,7 @@ const (
 	FAN_UNLIMITED_MARKS                         = 0x20
 	FAN_UNLIMITED_QUEUE                         = 0x10
 	FD_CLOEXEC                                  = 0x1
+	FD_PIDFS_ROOT                               = -0x2712
 	FD_SETSIZE                                  = 0x400
 	FF0                                         = 0x0
 	FIB_RULE_DEV_DETACHED                       = 0x8
@@ -1970,6 +1971,8 @@ const (
 	MADV_DONTNEED                               = 0x4
 	MADV_DONTNEED_LOCKED                        = 0x18
 	MADV_FREE                                   = 0x8
+	MADV_GUARD_INSTALL                          = 0x66
+	MADV_GUARD_REMOVE                           = 0x67
 	MADV_HUGEPAGE                               = 0xe
 	MADV_HWPOISON                               = 0x64
 	MADV_KEEPONFORK                             = 0x13
@@ -2114,7 +2117,7 @@ const (
 	MS_NOSEC                                    = 0x10000000
 	MS_NOSUID                                   = 0x2
 	MS_NOSYMFOLLOW                              = 0x100
-	MS_NOUSER                                   = -0x80000000
+	MS_NOUSER                                   = 0x80000000
 	MS_POSIXACL                                 = 0x10000
 	MS_PRIVATE                                  = 0x40000
 	MS_RDONLY                                   = 0x1
@@ -3786,6 +3789,9 @@ const (
 	TCPOPT_TIMESTAMP                            = 0x8
 	TCPOPT_TSTAMP_HDR                           = 0x101080a
 	TCPOPT_WINDOW                               = 0x3
+	TCP_AO_KEYF_EXCLUDE_OPT                     = 0x2
+	TCP_AO_KEYF_IFINDEX                         = 0x1
+	TCP_AO_MAXKEYLEN                            = 0x50
 	TCP_CC_INFO                                 = 0x1a
 	TCP_CM_INQ                                  = 0x24
 	TCP_CONGESTION                              = 0xd

vendor/golang.org/x/sys/unix/zsyscall_linux.go

@@ -700,6 +700,23 @@ func EpollCtl(epfd int, op int, fd int, event *EpollEvent) (err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
+func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
+	var _p0 unsafe.Pointer
+	if len(events) > 0 {
+		_p0 = unsafe.Pointer(&events[0])
+	} else {
+		_p0 = unsafe.Pointer(&_zero)
+	}
+	r0, _, e1 := Syscall6(SYS_EPOLL_PWAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
+	n = int(r0)
+	if e1 != 0 {
+		err = errnoErr(e1)
+	}
+	return
+}
+
+// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
+
 func Eventfd(initval uint, flags int) (fd int, err error) {
 	r0, _, e1 := Syscall(SYS_EVENTFD2, uintptr(initval), uintptr(flags), 0)
 	fd = int(r0)

vendor/golang.org/x/sys/unix/zsyscall_linux_386.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64_64, uintptr(fd), uintptr(offset), uintptr(offset>>32), uintptr(length), uintptr(length>>32), uintptr(advice))
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_amd64.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64, uintptr(fd), uintptr(offset), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_arm.go

@@ -213,23 +213,6 @@ func sendmsg(s int, msg *Msghdr, flags int) (n int, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fchown(fd int, uid int, gid int) (err error) {
 	_, _, e1 := Syscall(SYS_FCHOWN32, uintptr(fd), uintptr(uid), uintptr(gid))
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_arm64.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_PWAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64, uintptr(fd), uintptr(offset), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_loong64.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_PWAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64, uintptr(fd), uintptr(offset), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_mips.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall9(SYS_FADVISE64, uintptr(fd), 0, uintptr(offset>>32), uintptr(offset), uintptr(length>>32), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_mips64.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64, uintptr(fd), uintptr(offset), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_mips64le.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64, uintptr(fd), uintptr(offset), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_mipsle.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall9(SYS_FADVISE64, uintptr(fd), 0, uintptr(offset), uintptr(offset>>32), uintptr(length), uintptr(length>>32), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_ppc.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fchown(fd int, uid int, gid int) (err error) {
 	_, _, e1 := Syscall(SYS_FCHOWN, uintptr(fd), uintptr(uid), uintptr(gid))
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_ppc64.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64, uintptr(fd), uintptr(offset), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_ppc64le.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64, uintptr(fd), uintptr(offset), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_riscv64.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_PWAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64, uintptr(fd), uintptr(offset), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_s390x.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64, uintptr(fd), uintptr(offset), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/unix/zsyscall_linux_sparc64.go

@@ -45,23 +45,6 @@ func Tee(rfd int, wfd int, len int, flags int) (n int64, err error) {
 
 // THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
 
-func EpollWait(epfd int, events []EpollEvent, msec int) (n int, err error) {
-	var _p0 unsafe.Pointer
-	if len(events) > 0 {
-		_p0 = unsafe.Pointer(&events[0])
-	} else {
-		_p0 = unsafe.Pointer(&_zero)
-	}
-	r0, _, e1 := Syscall6(SYS_EPOLL_WAIT, uintptr(epfd), uintptr(_p0), uintptr(len(events)), uintptr(msec), 0, 0)
-	n = int(r0)
-	if e1 != 0 {
-		err = errnoErr(e1)
-	}
-	return
-}
-
-// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT
-
 func Fadvise(fd int, offset int64, length int64, advice int) (err error) {
 	_, _, e1 := Syscall6(SYS_FADVISE64, uintptr(fd), uintptr(offset), uintptr(length), uintptr(advice), 0, 0)
 	if e1 != 0 {

vendor/golang.org/x/sys/windows/security_windows.go

@@ -1109,17 +1109,53 @@ const (
 )
 
 // This type is the union inside of TRUSTEE and must be created using one of the TrusteeValueFrom* functions.
+//
+// Go pointers stored in a TrusteeValue must be pinned using [runtime.Pinner]
+// for the lifetime of the TrusteeValue.
 type TrusteeValue uintptr
 
+// TrusteeValueFromString is unsafe and should not be used.
+//
+// It returns a uintptr containing a reference to newly-allocated memory
+// which will be freed by the garbage collector.
+// There is no way for the caller to safely reference this memory.
+//
+// To create a [TrusteeValue] from a string, use:
+//
+//	p, err := windows.UTF16PtrFromString(s)
+//	if err != nil {
+//		// handle error
+//	}
+//
+//	// Pin the string for as long as it is used.
+//	var pinner runtime.Pinner
+//	pinner.Pin(p)
+//	defer pinner.Unpin()
+//
+//	tv := TrusteeValue(unsafe.Pointer(p))
+//
+// Deprecated: TrusteeValueFromString is unsafe and should not be used.
 func TrusteeValueFromString(str string) TrusteeValue {
 	return TrusteeValue(unsafe.Pointer(StringToUTF16Ptr(str)))
 }
+
+// TrusteeValueFromSID returns a [TrusteeValue] referencing sid.
+//
+// The caller must pin sid using a [runtime.Pinner] for the lifetime of the TrusteeValue.
 func TrusteeValueFromSID(sid *SID) TrusteeValue {
 	return TrusteeValue(unsafe.Pointer(sid))
 }
+
+// TrusteeValueFromObjectsAndSid returns a [TrusteeValue] referencing objectsAndSid.
+//
+// The caller must pin objectsAndSid using a [runtime.Pinner] for the lifetime of the TrusteeValue.
 func TrusteeValueFromObjectsAndSid(objectsAndSid *OBJECTS_AND_SID) TrusteeValue {
 	return TrusteeValue(unsafe.Pointer(objectsAndSid))
 }
+
+// TrusteeValueFromObjectsAndName returns a [TrusteeValue] referencing objectsAndName.
+//
+// The caller must pin objectsAndName using a [runtime.Pinner] for the lifetime of the TrusteeValue.
 func TrusteeValueFromObjectsAndName(objectsAndName *OBJECTS_AND_NAME) TrusteeValue {
 	return TrusteeValue(unsafe.Pointer(objectsAndName))
 }

vendor/golang.org/x/sys/windows/syscall_windows.go

@@ -1728,11 +1728,15 @@ func (s *NTUnicodeString) String() string {
 // the more common *uint16 string type.
 func NewNTString(s string) (*NTString, error) {
 	var nts NTString
-	s8, err := BytePtrFromString(s)
+	s8, err := ByteSliceFromString(s)
 	if err != nil {
 		return nil, err
 	}
-	RtlInitString(&nts, s8)
+	// The source string plus its terminating NUL must fit within MAX_USHORT.
+	if len(s8) > MAX_USHORT {
+		return nil, syscall.EINVAL
+	}
+	RtlInitString(&nts, &s8[0])
 	return &nts, nil
 }
 

vendor/golang.org/x/sys/windows/types_windows.go

@@ -169,6 +169,7 @@ const (
 	FORMAT_MESSAGE_ARGUMENT_ARRAY  = 8192
 	FORMAT_MESSAGE_MAX_WIDTH_MASK  = 255
 
+	MAX_USHORT    = 0xffff
 	MAX_PATH      = 260
 	MAX_LONG_PATH = 32768
 

vendor/modules.txt

@@ -146,7 +146,7 @@ github.com/VictoriaMetrics/fastcache
 # github.com/VictoriaMetrics/metrics v1.44.0
 ## explicit; go 1.24.0
 github.com/VictoriaMetrics/metrics
-# github.com/VictoriaMetrics/metricsql v0.87.2
+# github.com/VictoriaMetrics/metricsql v0.87.3
 ## explicit; go 1.24.2
 github.com/VictoriaMetrics/metricsql
 github.com/VictoriaMetrics/metricsql/binaryop
@@ -902,7 +902,7 @@ golang.org/x/oauth2/jwt
 ## explicit; go 1.25.0
 golang.org/x/sync/errgroup
 golang.org/x/sync/semaphore
-# golang.org/x/sys v0.46.0
+# golang.org/x/sys v0.47.0
 ## explicit; go 1.25.0
 golang.org/x/sys/cpu
 golang.org/x/sys/plan9