Add selector of exemplar reservoir providers to metric.Stream configu…

…ration (#5861)

Resolve #5249

### Spec

> exemplar_reservoir: A functional type that generates an exemplar
reservoir a MeterProvider will use when storing exemplars. This
functional type needs to be a factory or callback similar to aggregation
selection functionality which allows different reservoirs to be chosen
by the aggregation.

> Users can provide an exemplar_reservoir, but it is up to their
discretion. Therefore, the stream configuration parameter needs to be
structured to accept an exemplar_reservoir, but MUST NOT obligate a user
to provide one. If the user does not provide an exemplar_reservoir
value, the MeterProvider MUST apply a [default exemplar
reservoir](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#exemplar-defaults).

Also,

> the reservoir MUST be given the Attributes associated with its
timeseries point either at construction so that additional sampling
performed by the reservoir has access to all attributes from a
measurement in the "offer" method.

### Changes

In sdk/metric/exemplar, add:
* `exemplar.ReservoirProvider`
* `exemplar.HistogramReservoirProvider`
* `exemplar.FixedSizeReservoirProvider`

In sdk/metric, add:
* `metric.ExemplarReservoirProviderSelector` (func Aggregation ->
ReservoirProvider)
* `metric.DefaultExemplarReservoirProviderSelector` (our default
implementation)
* `ExemplarReservoirProviderSelector` added to `metric.Stream`

Note: the only required public types are
`metric.ExemplarReservoirProviderSelector` and
`ExemplarReservoirProviderSelector` in `metric.Stream`. Others are for
convenience and readability.

### Alternatives considered

#### Add ExemplarReservoirProvider directly to metric.Stream, instead of
ExemplarReservoirProviderSelector

This would mean users can configure a `func() exemplar.Reservoir`
instead of a `func(Aggregation) func() exemplar.Reservoir`.

I don't think this complies with the statement: `This functional type
needs to be a factory or callback similar to aggregation selection
functionality which allows different reservoirs to be chosen by the
aggregation.`. I'm interpreting "allows different reservoirs to be
chosen by the aggregation" as meaning "allows different reservoirs to be
chosen **based on the** aggregation", rather than meaning that the
aggregation is somehow choosing the reservoir.

### Future work

There is some refactoring I plan to do after this to simplify the
interaction between the internal/aggregate and exemplar package. I've
omitted that from this PR to keep the diff smaller.

---------

Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
Co-authored-by: Robert Pająk <pellared@hotmail.com>

CHANGELOG.md

@@ -12,6 +12,9 @@ This project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.htm
 
 - Add `go.opentelemetry.io/otel/sdk/metric/exemplar.AlwaysOffFilter`, which can be used to disable exemplar recording. (#5850)
 - Add `go.opentelemetry.io/otel/sdk/metric.WithExemplarFilter`, which can be used to configure the exemplar filter used by the metrics SDK. (#5850)
+- Add `ExemplarReservoirProviderSelector` and `DefaultExemplarReservoirProviderSelector` to `go.opentelemetry.io/otel/sdk/metric`, which defines the exemplar reservoir to use based on the aggregation of the metric. (#5861)
+- Add `ExemplarReservoirProviderSelector` to `go.opentelemetry.io/otel/sdk/metric.Stream` to allow using views to configure the exemplar reservoir to use for a metric. (#5861)
+- Add `ReservoirProvider`, `HistogramReservoirProvider` and `FixedSizeReservoirProvider` to `go.opentelemetry.io/otel/sdk/metric/exemplar` to make it convenient to use providers of Reservoirs. (#5861)
 
 <!-- Released section -->
 <!-- Don't change this section unless doing release -->

sdk/metric/example_test.go

@@ -242,6 +242,31 @@ func ExampleNewView_exponentialHistogram() {
 	)
 }
 
+func ExampleNewView_exemplarreservoirproviderselector() {
+	// Create a view that makes all metrics use a different exemplar reservoir.
+	view := metric.NewView(
+		metric.Instrument{Name: "*"},
+		metric.Stream{
+			ExemplarReservoirProviderSelector: func(agg metric.Aggregation) exemplar.ReservoirProvider {
+				// This example uses a fixed-size reservoir with a size of 10
+				// for explicit bucket histograms instead of the default
+				// bucket-aligned reservoir.
+				if _, ok := agg.(metric.AggregationExplicitBucketHistogram); ok {
+					return exemplar.FixedSizeReservoirProvider(10)
+				}
+				// Fall back to the default reservoir otherwise.
+				return metric.DefaultExemplarReservoirProviderSelector(agg)
+			},
+		},
+	)
+
+	// The created view can then be registered with the OpenTelemetry metric
+	// SDK using the WithView option.
+	_ = metric.NewMeterProvider(
+		metric.WithView(view),
+	)
+}
+
 func ExampleWithExemplarFilter_disabled() {
 	// Use exemplar.AlwaysOffFilter to disable exemplar collection.
 	_ = metric.NewMeterProvider(

sdk/metric/exemplar.go

@@ -5,25 +5,48 @@ package metric // import "go.opentelemetry.io/otel/sdk/metric"
 
 import (
 	"runtime"
-	"slices"
 
+	"go.opentelemetry.io/otel/attribute"
 	"go.opentelemetry.io/otel/sdk/metric/exemplar"
 	"go.opentelemetry.io/otel/sdk/metric/internal/aggregate"
 )
 
+// ExemplarReservoirProviderSelector selects the
+// [exemplar.ReservoirProvider] to use
+// based on the [Aggregation] of the metric.
+type ExemplarReservoirProviderSelector func(Aggregation) exemplar.ReservoirProvider
+
 // reservoirFunc returns the appropriately configured exemplar reservoir
 // creation func based on the passed InstrumentKind and filter configuration.
-func reservoirFunc[N int64 | float64](agg Aggregation, filter exemplar.Filter) func() aggregate.FilteredExemplarReservoir[N] {
+func reservoirFunc[N int64 | float64](provider exemplar.ReservoirProvider, filter exemplar.Filter) func(attribute.Set) aggregate.FilteredExemplarReservoir[N] {
+	return func(attrs attribute.Set) aggregate.FilteredExemplarReservoir[N] {
+		return aggregate.NewFilteredExemplarReservoir[N](filter, provider(attrs))
+	}
+}
+
+// DefaultExemplarReservoirProviderSelector returns the default
+// [exemplar.ReservoirProvider] for the
+// provided [Aggregation].
+//
+// For explicit bucket histograms with more than 1 bucket, it uses the
+// [exemplar.HistogramReservoirProvider].
+// For exponential histograms, it uses the
+// [exemplar.FixedSizeReservoirProvider]
+// with a size of min(20, max_buckets).
+// For all other aggregations, it uses the
+// [exemplar.FixedSizeReservoirProvider]
+// with a size equal to the number of CPUs.
+//
+// Exemplar default reservoirs MAY change in a minor version bump. No
+// guarantees are made on the shape or statistical properties of returned
+// exemplars.
+func DefaultExemplarReservoirProviderSelector(agg Aggregation) exemplar.ReservoirProvider {
 	// https://github.com/open-telemetry/opentelemetry-specification/blob/d4b241f451674e8f611bb589477680341006ad2b/specification/metrics/sdk.md#exemplar-defaults
 	// Explicit bucket histogram aggregation with more than 1 bucket will
 	// use AlignedHistogramBucketExemplarReservoir.
 	a, ok := agg.(AggregationExplicitBucketHistogram)
 	if ok && len(a.Boundaries) > 0 {
-		cp := slices.Clone(a.Boundaries)
-		return func() aggregate.FilteredExemplarReservoir[N] {
-			bounds := cp
-			return aggregate.NewFilteredExemplarReservoir[N](filter, exemplar.NewHistogramReservoir(bounds))
-		}
+		return exemplar.HistogramReservoirProvider(a.Boundaries)
 	}
 
 	var n int
@@ -50,7 +73,5 @@ func reservoirFunc[N int64 | float64](agg Aggregation, filter exemplar.Filter) f
 		}
 	}
 
-	return func() aggregate.FilteredExemplarReservoir[N] {
-		return aggregate.NewFilteredExemplarReservoir[N](filter, exemplar.NewFixedSizeReservoir(n))
-	}
+	return exemplar.FixedSizeReservoirProvider(n)
 }

sdk/metric/exemplar/fixed_size_reservoir.go

@@ -12,6 +12,13 @@ import (
 	"go.opentelemetry.io/otel/attribute"
 )
 
+// FixedSizeReservoirProvider returns a provider of [FixedSizeReservoir].
+func FixedSizeReservoirProvider(k int) ReservoirProvider {
+	return func(_ attribute.Set) Reservoir {
+		return NewFixedSizeReservoir(k)
+	}
+}
+
 // NewFixedSizeReservoir returns a [FixedSizeReservoir] that samples at most
 // k exemplars. If there are k or less measurements made, the Reservoir will
 // sample each one. If there are more than k, the Reservoir will then randomly

sdk/metric/exemplar/fixed_size_reservoir_test.go

@@ -15,12 +15,12 @@ import (
 )
 
 func TestNewFixedSizeReservoir(t *testing.T) {
-	t.Run("Int64", ReservoirTest[int64](func(n int) (Reservoir, int) {
-		return NewFixedSizeReservoir(n), n
+	t.Run("Int64", ReservoirTest[int64](func(n int) (ReservoirProvider, int) {
+		return FixedSizeReservoirProvider(n), n
 	}))
 
-	t.Run("Float64", ReservoirTest[float64](func(n int) (Reservoir, int) {
-		return NewFixedSizeReservoir(n), n
+	t.Run("Float64", ReservoirTest[float64](func(n int) (ReservoirProvider, int) {
+		return FixedSizeReservoirProvider(n), n
 	}))
 }
 

sdk/metric/exemplar/histogram_reservoir.go

@@ -12,13 +12,21 @@ import (
 	"go.opentelemetry.io/otel/attribute"
 )
 
+// HistogramReservoirProvider is a provider of [HistogramReservoir].
+func HistogramReservoirProvider(bounds []float64) ReservoirProvider {
+	cp := slices.Clone(bounds)
+	slices.Sort(cp)
+	return func(_ attribute.Set) Reservoir {
+		return NewHistogramReservoir(cp)
+	}
+}
+
 // NewHistogramReservoir returns a [HistogramReservoir] that samples the last
 // measurement that falls within a histogram bucket. The histogram bucket
 // upper-boundaries are define by bounds.
 //
-// The passed bounds will be sorted by this function.
+// The passed bounds must be sorted before calling this function.
 func NewHistogramReservoir(bounds []float64) *HistogramReservoir {
-	slices.Sort(bounds)
 	return &HistogramReservoir{
 		bounds:  bounds,
 		storage: newStorage(len(bounds) + 1),

sdk/metric/exemplar/histogram_reservoir_test.go

@@ -7,11 +7,11 @@ import "testing"
 
 func TestHist(t *testing.T) {
 	bounds := []float64{0, 100}
-	t.Run("Int64", ReservoirTest[int64](func(int) (Reservoir, int) {
-		return NewHistogramReservoir(bounds), len(bounds)
+	t.Run("Int64", ReservoirTest[int64](func(int) (ReservoirProvider, int) {
+		return HistogramReservoirProvider(bounds), len(bounds)
 	}))
 
-	t.Run("Float64", ReservoirTest[float64](func(int) (Reservoir, int) {
-		return NewHistogramReservoir(bounds), len(bounds)
+	t.Run("Float64", ReservoirTest[float64](func(int) (ReservoirProvider, int) {
+		return HistogramReservoirProvider(bounds), len(bounds)
 	}))
 }

sdk/metric/exemplar/reservoir.go

@@ -30,3 +30,11 @@ type Reservoir interface {
 	// The Reservoir state is preserved after this call.
 	Collect(dest *[]Exemplar)
 }
+
+// ReservoirProvider creates new [Reservoir]s.
+//
+// The attributes provided are attributes which are kept by the aggregation, and
+// are exclusive with attributes passed to Offer. The combination of these
+// attributes and the attributes passed to Offer is the complete set of
+// attributes a measurement was made with.
+type ReservoirProvider func(attr attribute.Set) Reservoir

sdk/metric/exemplar/reservoir_test.go

@@ -18,7 +18,7 @@ import (
 // Sat Jan 01 2000 00:00:00 GMT+0000.
 var staticTime = time.Unix(946684800, 0)
 
-type factory func(requestedCap int) (r Reservoir, actualCap int)
+type factory func(requestedCap int) (r ReservoirProvider, actualCap int)
 
 func ReservoirTest[N int64 | float64](f factory) func(*testing.T) {
 	return func(t *testing.T) {
@@ -29,10 +29,11 @@ func ReservoirTest[N int64 | float64](f factory) func(*testing.T) {
 		t.Run("CaptureSpanContext", func(t *testing.T) {
 			t.Helper()
 
-			r, n := f(1)
+			rp, n := f(1)
 			if n < 1 {
 				t.Skip("skipping, reservoir capacity less than 1:", n)
 			}
+			r := rp(*attribute.EmptySet())
 
 			tID, sID := trace.TraceID{0x01}, trace.SpanID{0x01}
 			sc := trace.NewSpanContext(trace.SpanContextConfig{
@@ -60,10 +61,11 @@ func ReservoirTest[N int64 | float64](f factory) func(*testing.T) {
 		t.Run("FilterAttributes", func(t *testing.T) {
 			t.Helper()
 
-			r, n := f(1)
+			rp, n := f(1)
 			if n < 1 {
 				t.Skip("skipping, reservoir capacity less than 1:", n)
 			}
+			r := rp(*attribute.EmptySet())
 
 			adminTrue := attribute.Bool("admin", true)
 			r.Offer(ctx, staticTime, NewValue(N(10)), []attribute.KeyValue{adminTrue})
@@ -83,10 +85,11 @@ func ReservoirTest[N int64 | float64](f factory) func(*testing.T) {
 		t.Run("CollectLessThanN", func(t *testing.T) {
 			t.Helper()
 
-			r, n := f(2)
+			rp, n := f(2)
 			if n < 2 {
 				t.Skip("skipping, reservoir capacity less than 2:", n)
 			}
+			r := rp(*attribute.EmptySet())
 
 			r.Offer(ctx, staticTime, NewValue(N(10)), nil)
 
@@ -99,10 +102,11 @@ func ReservoirTest[N int64 | float64](f factory) func(*testing.T) {
 		t.Run("MultipleOffers", func(t *testing.T) {
 			t.Helper()
 
-			r, n := f(3)
+			rp, n := f(3)
 			if n < 1 {
 				t.Skip("skipping, reservoir capacity less than 1:", n)
 			}
+			r := rp(*attribute.EmptySet())
 
 			for i := 0; i < n+1; i++ {
 				v := NewValue(N(i))
@@ -127,10 +131,11 @@ func ReservoirTest[N int64 | float64](f factory) func(*testing.T) {
 		t.Run("DropAll", func(t *testing.T) {
 			t.Helper()
 
-			r, n := f(0)
+			rp, n := f(0)
 			if n > 0 {
 				t.Skip("skipping, reservoir capacity greater than 0:", n)
 			}
+			r := rp(*attribute.EmptySet())
 
 			r.Offer(context.Background(), staticTime, NewValue(N(10)), nil)
 

sdk/metric/instrument.go

@@ -144,6 +144,12 @@ type Stream struct {
 	// Use NewAllowKeysFilter from "go.opentelemetry.io/otel/attribute" to
 	// provide an allow-list of attribute keys here.
 	AttributeFilter attribute.Filter
+	// ExemplarReservoirProvider selects the
+	// [go.opentelemetry.io/otel/sdk/metric/exemplar.ReservoirProvider] based
+	// on the [Aggregation].
+	//
+	// If unspecified, [DefaultExemplarReservoirProviderSelector] is used.
+	ExemplarReservoirProviderSelector ExemplarReservoirProviderSelector
 }
 
 // instID are the identifying properties of a instrument.

sdk/metric/internal/aggregate/aggregate.go

@@ -38,7 +38,7 @@ type Builder[N int64 | float64] struct {
 	//
 	// If this is not provided a default factory function that returns an
 	// dropReservoir reservoir will be used.
-	ReservoirFunc func() FilteredExemplarReservoir[N]
+	ReservoirFunc func(attribute.Set) FilteredExemplarReservoir[N]
 	// AggregationLimit is the cardinality limit of measurement attributes. Any
 	// measurement for new attributes once the limit has been reached will be
 	// aggregated into a single aggregate for the "otel.metric.overflow"
@@ -49,7 +49,7 @@ type Builder[N int64 | float64] struct {
 	AggregationLimit int
 }
 
-func (b Builder[N]) resFunc() func() FilteredExemplarReservoir[N] {
+func (b Builder[N]) resFunc() func(attribute.Set) FilteredExemplarReservoir[N] {
 	if b.ReservoirFunc != nil {
 		return b.ReservoirFunc
 	}

sdk/metric/internal/aggregate/aggregate_test.go

@@ -72,8 +72,8 @@ func (c *clock) Register() (unregister func()) {
 	return func() { now = orig }
 }
 
-func dropExemplars[N int64 | float64]() FilteredExemplarReservoir[N] {
-	return dropReservoir[N]()
+func dropExemplars[N int64 | float64](attr attribute.Set) FilteredExemplarReservoir[N] {
+	return dropReservoir[N](attr)
 }
 
 func TestBuilderFilter(t *testing.T) {

sdk/metric/internal/aggregate/drop.go

@@ -11,7 +11,9 @@ import (
 )
 
 // dropReservoir returns a [FilteredReservoir] that drops all measurements it is offered.
-func dropReservoir[N int64 | float64]() FilteredExemplarReservoir[N] { return &dropRes[N]{} }
+func dropReservoir[N int64 | float64](attribute.Set) FilteredExemplarReservoir[N] {
+	return &dropRes[N]{}
+}
 
 type dropRes[N int64 | float64] struct{}
 

sdk/metric/internal/aggregate/drop_test.go

@@ -8,6 +8,7 @@ import (
 
 	"github.com/stretchr/testify/assert"
 
+	"go.opentelemetry.io/otel/attribute"
 	"go.opentelemetry.io/otel/sdk/metric/exemplar"
 )
 
@@ -17,7 +18,7 @@ func TestDrop(t *testing.T) {
 }
 
 func testDropFiltered[N int64 | float64](t *testing.T) {
-	r := dropReservoir[N]()
+	r := dropReservoir[N](*attribute.EmptySet())
 
 	var dest []exemplar.Exemplar
 	r.Collect(&dest)

sdk/metric/internal/aggregate/exponential_histogram.go

@@ -283,7 +283,7 @@ func (b *expoBuckets) downscale(delta int32) {
 // newExponentialHistogram returns an Aggregator that summarizes a set of
 // measurements as an exponential histogram. Each histogram is scoped by attributes
 // and the aggregation cycle the measurements were made in.
-func newExponentialHistogram[N int64 | float64](maxSize, maxScale int32, noMinMax, noSum bool, limit int, r func() FilteredExemplarReservoir[N]) *expoHistogram[N] {
+func newExponentialHistogram[N int64 | float64](maxSize, maxScale int32, noMinMax, noSum bool, limit int, r func(attribute.Set) FilteredExemplarReservoir[N]) *expoHistogram[N] {
 	return &expoHistogram[N]{
 		noSum:    noSum,
 		noMinMax: noMinMax,
@@ -306,7 +306,7 @@ type expoHistogram[N int64 | float64] struct {
 	maxSize  int
 	maxScale int32
 
-	newRes   func() FilteredExemplarReservoir[N]
+	newRes   func(attribute.Set) FilteredExemplarReservoir[N]
 	limit    limiter[*expoHistogramDataPoint[N]]
 	values   map[attribute.Distinct]*expoHistogramDataPoint[N]
 	valuesMu sync.Mutex
@@ -327,7 +327,7 @@ func (e *expoHistogram[N]) measure(ctx context.Context, value N, fltrAttr attrib
 	v, ok := e.values[attr.Equivalent()]
 	if !ok {
 		v = newExpoHistogramDataPoint[N](attr, e.maxSize, e.maxScale, e.noMinMax, e.noSum)
-		v.res = e.newRes()
+		v.res = e.newRes(attr)
 
 		e.values[attr.Equivalent()] = v
 	}

sdk/metric/internal/aggregate/histogram.go

@@ -47,13 +47,13 @@ type histValues[N int64 | float64] struct {
 	noSum  bool
 	bounds []float64
 
-	newRes   func() FilteredExemplarReservoir[N]
+	newRes   func(attribute.Set) FilteredExemplarReservoir[N]
 	limit    limiter[*buckets[N]]
 	values   map[attribute.Distinct]*buckets[N]
 	valuesMu sync.Mutex
 }
 
-func newHistValues[N int64 | float64](bounds []float64, noSum bool, limit int, r func() FilteredExemplarReservoir[N]) *histValues[N] {
+func newHistValues[N int64 | float64](bounds []float64, noSum bool, limit int, r func(attribute.Set) FilteredExemplarReservoir[N]) *histValues[N] {
 	// The responsibility of keeping all buckets correctly associated with the
 	// passed boundaries is ultimately this type's responsibility. Make a copy
 	// here so we can always guarantee this. Or, in the case of failure, have
@@ -93,7 +93,7 @@ func (s *histValues[N]) measure(ctx context.Context, value N, fltrAttr attribute
 		//
 		//   buckets = (-∞, 0], (0, 5.0], (5.0, 10.0], (10.0, +∞)
 		b = newBuckets[N](attr, len(s.bounds)+1)
-		b.res = s.newRes()
+		b.res = s.newRes(attr)
 
 		// Ensure min and max are recorded values (not zero), for new buckets.
 		b.min, b.max = value, value
@@ -108,7 +108,7 @@ func (s *histValues[N]) measure(ctx context.Context, value N, fltrAttr attribute
 
 // newHistogram returns an Aggregator that summarizes a set of measurements as
 // an histogram.
-func newHistogram[N int64 | float64](boundaries []float64, noMinMax, noSum bool, limit int, r func() FilteredExemplarReservoir[N]) *histogram[N] {
+func newHistogram[N int64 | float64](boundaries []float64, noMinMax, noSum bool, limit int, r func(attribute.Set) FilteredExemplarReservoir[N]) *histogram[N] {
 	return &histogram[N]{
 		histValues: newHistValues[N](boundaries, noSum, limit, r),
 		noMinMax:   noMinMax,