[chore] Update usage of OTEL libraries (#2725)

* otel to 1.24
* prometheus exporter to 0.46
* bunotel to 1.1.17

Also:
* Use schemaless URL for metrics
* Add software version to tracing schema

35 files changed, 1402 insertions, 222 deletions

diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/cache.go b/vendor/go.opentelemetry.io/otel/sdk/metric/cache.go
index de9d6f001..e9c0b38d0 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/cache.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/cache.go
@@ -52,3 +52,43 @@ func (c *cache[K, V]) Lookup(key K, f func() V) V {
 	c.data[key] = val
 	return val
 }
+
+// HasKey returns true if Lookup has previously been called with that key
+//
+// HasKey is safe to call concurrently.
+func (c *cache[K, V]) HasKey(key K) bool {
+	c.Lock()
+	defer c.Unlock()
+	_, ok := c.data[key]
+	return ok
+}
+
+// cacheWithErr is a locking storage used to quickly return already computed values and an error.
+//
+// The zero value of a cacheWithErr is empty and ready to use.
+//
+// A cacheWithErr must not be copied after first use.
+//
+// All methods of a cacheWithErr are safe to call concurrently.
+type cacheWithErr[K comparable, V any] struct {
+	cache[K, valAndErr[V]]
+}
+
+type valAndErr[V any] struct {
+	val V
+	err error
+}
+
+// Lookup returns the value stored in the cacheWithErr with the associated key
+// if it exists. Otherwise, f is called and its returned value is set in the
+// cacheWithErr for key and returned.
+//
+// Lookup is safe to call concurrently. It will hold the cacheWithErr lock, so f
+// should not block excessively.
+func (c *cacheWithErr[K, V]) Lookup(key K, f func() (V, error)) (V, error) {
+	combined := c.cache.Lookup(key, func() valAndErr[V] {
+		val, err := f()
+		return valAndErr[V]{val: val, err: err}
+	})
+	return combined.val, combined.err
+}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/doc.go b/vendor/go.opentelemetry.io/otel/sdk/metric/doc.go
index 53f80c428..475d3e394 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/doc.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/doc.go
@@ -44,4 +44,7 @@
 //
 // See [go.opentelemetry.io/otel/metric] for more information about
 // the metric API.
+//
+// See [go.opentelemetry.io/otel/sdk/metric/internal/x] for information about
+// the experimental features.
 package metric // import "go.opentelemetry.io/otel/sdk/metric"
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/exemplar.go b/vendor/go.opentelemetry.io/otel/sdk/metric/exemplar.go
new file mode 100644
index 000000000..3f1ce9f1d
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/exemplar.go
@@ -0,0 +1,96 @@
+// Copyright The OpenTelemetry Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package metric // import "go.opentelemetry.io/otel/sdk/metric"
+
+import (
+	"os"
+	"runtime"
+
+	"go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
+	"go.opentelemetry.io/otel/sdk/metric/internal/x"
+)
+
+// reservoirFunc returns the appropriately configured exemplar reservoir
+// creation func based on the passed InstrumentKind and user defined
+// environment variables.
+//
+// Note: This will only return non-nil values when the experimental exemplar
+// feature is enabled and the OTEL_METRICS_EXEMPLAR_FILTER environment variable
+// is not set to always_off.
+func reservoirFunc[N int64 | float64](agg Aggregation) func() exemplar.Reservoir[N] {
+	if !x.Exemplars.Enabled() {
+		return nil
+	}
+
+	// https://github.com/open-telemetry/opentelemetry-specification/blob/d4b241f451674e8f611bb589477680341006ad2b/specification/metrics/sdk.md#exemplar-defaults
+	resF := func() func() exemplar.Reservoir[N] {
+		// Explicit bucket histogram aggregation with more than 1 bucket will
+		// use AlignedHistogramBucketExemplarReservoir.
+		a, ok := agg.(AggregationExplicitBucketHistogram)
+		if ok && len(a.Boundaries) > 0 {
+			cp := make([]float64, len(a.Boundaries))
+			copy(cp, a.Boundaries)
+			return func() exemplar.Reservoir[N] {
+				bounds := cp
+				return exemplar.Histogram[N](bounds)
+			}
+		}
+
+		var n int
+		if a, ok := agg.(AggregationBase2ExponentialHistogram); ok {
+			// Base2 Exponential Histogram Aggregation SHOULD use a
+			// SimpleFixedSizeExemplarReservoir with a reservoir equal to the
+			// smaller of the maximum number of buckets configured on the
+			// aggregation or twenty (e.g. min(20, max_buckets)).
+			n = int(a.MaxSize)
+			if n > 20 {
+				n = 20
+			}
+		} else {
+			// https://github.com/open-telemetry/opentelemetry-specification/blob/e94af89e3d0c01de30127a0f423e912f6cda7bed/specification/metrics/sdk.md#simplefixedsizeexemplarreservoir
+			//   This Exemplar reservoir MAY take a configuration parameter for
+			//   the size of the reservoir. If no size configuration is
+			//   provided, the default size MAY be the number of possible
+			//   concurrent threads (e.g. number of CPUs) to help reduce
+			//   contention. Otherwise, a default size of 1 SHOULD be used.
+			n = runtime.NumCPU()
+			if n < 1 {
+				// Should never be the case, but be defensive.
+				n = 1
+			}
+		}
+
+		return func() exemplar.Reservoir[N] {
+			return exemplar.FixedSize[N](n)
+		}
+	}
+
+	// https://github.com/open-telemetry/opentelemetry-specification/blob/d4b241f451674e8f611bb589477680341006ad2b/specification/configuration/sdk-environment-variables.md#exemplar
+	const filterEnvKey = "OTEL_METRICS_EXEMPLAR_FILTER"
+
+	switch os.Getenv(filterEnvKey) {
+	case "always_on":
+		return resF()
+	case "always_off":
+		return exemplar.Drop[N]
+	case "trace_based":
+		fallthrough
+	default:
+		newR := resF()
+		return func() exemplar.Reservoir[N] {
+			return exemplar.SampledFilter(newR())
+		}
+	}
+}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/instrument.go b/vendor/go.opentelemetry.io/otel/sdk/metric/instrument.go
index bb52f6ec7..a4cfcbb95 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/instrument.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/instrument.go
@@ -205,9 +205,6 @@ func (i *int64Inst) Record(ctx context.Context, val int64, opts ...metric.Record
 }
 
 func (i *int64Inst) aggregate(ctx context.Context, val int64, s attribute.Set) { // nolint:revive  // okay to shadow pkg with method.
-	if err := ctx.Err(); err != nil {
-		return
-	}
 	for _, in := range i.measures {
 		in(ctx, val, s)
 	}
@@ -238,9 +235,6 @@ func (i *float64Inst) Record(ctx context.Context, val float64, opts ...metric.Re
 }
 
 func (i *float64Inst) aggregate(ctx context.Context, val float64, s attribute.Set) {
-	if err := ctx.Err(); err != nil {
-		return
-	}
 	for _, in := range i.measures {
 		in(ctx, val, s)
 	}
@@ -270,9 +264,9 @@ var (
 	_ metric.Float64ObservableGauge         = float64Observable{}
 )
 
-func newFloat64Observable(m *meter, kind InstrumentKind, name, desc, u string, meas []aggregate.Measure[float64]) float64Observable {
+func newFloat64Observable(m *meter, kind InstrumentKind, name, desc, u string) float64Observable {
 	return float64Observable{
-		observable: newObservable(m, kind, name, desc, u, meas),
+		observable: newObservable[float64](m, kind, name, desc, u),
 	}
 }
 
@@ -291,9 +285,9 @@ var (
 	_ metric.Int64ObservableGauge         = int64Observable{}
 )
 
-func newInt64Observable(m *meter, kind InstrumentKind, name, desc, u string, meas []aggregate.Measure[int64]) int64Observable {
+func newInt64Observable(m *meter, kind InstrumentKind, name, desc, u string) int64Observable {
 	return int64Observable{
-		observable: newObservable(m, kind, name, desc, u, meas),
+		observable: newObservable[int64](m, kind, name, desc, u),
 	}
 }
 
@@ -301,11 +295,12 @@ type observable[N int64 | float64] struct {
 	metric.Observable
 	observablID[N]
 
-	meter    *meter
-	measures []aggregate.Measure[N]
+	meter           *meter
+	measures        measures[N]
+	dropAggregation bool
 }
 
-func newObservable[N int64 | float64](m *meter, kind InstrumentKind, name, desc, u string, meas []aggregate.Measure[N]) *observable[N] {
+func newObservable[N int64 | float64](m *meter, kind InstrumentKind, name, desc, u string) *observable[N] {
 	return &observable[N]{
 		observablID: observablID[N]{
 			name:        name,
@@ -314,14 +309,24 @@ func newObservable[N int64 | float64](m *meter, kind InstrumentKind, name, desc,
 			unit:        u,
 			scope:       m.scope,
 		},
-		meter:    m,
-		measures: meas,
+		meter: m,
 	}
 }
 
 // observe records the val for the set of attrs.
 func (o *observable[N]) observe(val N, s attribute.Set) {
-	for _, in := range o.measures {
+	o.measures.observe(val, s)
+}
+
+func (o *observable[N]) appendMeasures(meas []aggregate.Measure[N]) {
+	o.measures = append(o.measures, meas...)
+}
+
+type measures[N int64 | float64] []aggregate.Measure[N]
+
+// observe records the val for the set of attrs.
+func (m measures[N]) observe(val N, s attribute.Set) {
+	for _, in := range m {
 		in(context.Background(), val, s)
 	}
 }
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/aggregate.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/aggregate.go
index 8dec14237..4060a2f76 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/aggregate.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/aggregate.go
@@ -19,6 +19,7 @@ import (
 	"time"
 
 	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
 	"go.opentelemetry.io/otel/sdk/metric/metricdata"
 )
 
@@ -44,17 +45,43 @@ type Builder[N int64 | float64] struct {
 	// Filter is the attribute filter the aggregate function will use on the
 	// input of measurements.
 	Filter attribute.Filter
+	// ReservoirFunc is the factory function used by aggregate functions to
+	// create new exemplar reservoirs for a new seen attribute set.
+	//
+	// If this is not provided a default factory function that returns an
+	// exemplar.Drop reservoir will be used.
+	ReservoirFunc func() exemplar.Reservoir[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"
+	// attribute.
+	//
+	// If AggregationLimit is less than or equal to zero there will not be an
+	// aggregation limit imposed (i.e. unlimited attribute sets).
+	AggregationLimit int
+}
+
+func (b Builder[N]) resFunc() func() exemplar.Reservoir[N] {
+	if b.ReservoirFunc != nil {
+		return b.ReservoirFunc
+	}
+
+	return exemplar.Drop[N]
 }
 
-func (b Builder[N]) filter(f Measure[N]) Measure[N] {
+type fltrMeasure[N int64 | float64] func(ctx context.Context, value N, fltrAttr attribute.Set, droppedAttr []attribute.KeyValue)
+
+func (b Builder[N]) filter(f fltrMeasure[N]) Measure[N] {
 	if b.Filter != nil {
 		fltr := b.Filter // Copy to make it immutable after assignment.
 		return func(ctx context.Context, n N, a attribute.Set) {
-			fAttr, _ := a.Filter(fltr)
-			f(ctx, n, fAttr)
+			fAttr, dropped := a.Filter(fltr)
+			f(ctx, n, fAttr, dropped)
 		}
 	}
-	return f
+	return func(ctx context.Context, n N, a attribute.Set) {
+		f(ctx, n, a, nil)
+	}
 }
 
 // LastValue returns a last-value aggregate function input and output.
@@ -63,7 +90,7 @@ func (b Builder[N]) filter(f Measure[N]) Measure[N] {
 func (b Builder[N]) LastValue() (Measure[N], ComputeAggregation) {
 	// Delta temporality is the only temporality that makes semantic sense for
 	// a last-value aggregate.
-	lv := newLastValue[N]()
+	lv := newLastValue[N](b.AggregationLimit, b.resFunc())
 
 	return b.filter(lv.measure), func(dest *metricdata.Aggregation) int {
 		// Ignore if dest is not a metricdata.Gauge. The chance for memory
@@ -79,7 +106,7 @@ func (b Builder[N]) LastValue() (Measure[N], ComputeAggregation) {
 // PrecomputedSum returns a sum aggregate function input and output. The
 // arguments passed to the input are expected to be the precomputed sum values.
 func (b Builder[N]) PrecomputedSum(monotonic bool) (Measure[N], ComputeAggregation) {
-	s := newPrecomputedSum[N](monotonic)
+	s := newPrecomputedSum[N](monotonic, b.AggregationLimit, b.resFunc())
 	switch b.Temporality {
 	case metricdata.DeltaTemporality:
 		return b.filter(s.measure), s.delta
@@ -90,7 +117,7 @@ func (b Builder[N]) PrecomputedSum(monotonic bool) (Measure[N], ComputeAggregati
 
 // Sum returns a sum aggregate function input and output.
 func (b Builder[N]) Sum(monotonic bool) (Measure[N], ComputeAggregation) {
-	s := newSum[N](monotonic)
+	s := newSum[N](monotonic, b.AggregationLimit, b.resFunc())
 	switch b.Temporality {
 	case metricdata.DeltaTemporality:
 		return b.filter(s.measure), s.delta
@@ -102,7 +129,7 @@ func (b Builder[N]) Sum(monotonic bool) (Measure[N], ComputeAggregation) {
 // ExplicitBucketHistogram returns a histogram aggregate function input and
 // output.
 func (b Builder[N]) ExplicitBucketHistogram(boundaries []float64, noMinMax, noSum bool) (Measure[N], ComputeAggregation) {
-	h := newHistogram[N](boundaries, noMinMax, noSum)
+	h := newHistogram[N](boundaries, noMinMax, noSum, b.AggregationLimit, b.resFunc())
 	switch b.Temporality {
 	case metricdata.DeltaTemporality:
 		return b.filter(h.measure), h.delta
@@ -114,7 +141,7 @@ func (b Builder[N]) ExplicitBucketHistogram(boundaries []float64, noMinMax, noSu
 // ExponentialBucketHistogram returns a histogram aggregate function input and
 // output.
 func (b Builder[N]) ExponentialBucketHistogram(maxSize, maxScale int32, noMinMax, noSum bool) (Measure[N], ComputeAggregation) {
-	h := newExponentialHistogram[N](maxSize, maxScale, noMinMax, noSum)
+	h := newExponentialHistogram[N](maxSize, maxScale, noMinMax, noSum, b.AggregationLimit, b.resFunc())
 	switch b.Temporality {
 	case metricdata.DeltaTemporality:
 		return b.filter(h.measure), h.delta
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/exponential_histogram.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/exponential_histogram.go
index 98b7dc1e0..4139a6d15 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/exponential_histogram.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/exponential_histogram.go
@@ -23,6 +23,7 @@ import (
 
 	"go.opentelemetry.io/otel"
 	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
 	"go.opentelemetry.io/otel/sdk/metric/metricdata"
 )
 
@@ -40,6 +41,8 @@ const (
 
 // expoHistogramDataPoint is a single data point in an exponential histogram.
 type expoHistogramDataPoint[N int64 | float64] struct {
+	res exemplar.Reservoir[N]
+
 	count uint64
 	min   N
 	max   N
@@ -288,13 +291,15 @@ func (b *expoBuckets) downscale(delta int) {
 // 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) *expoHistogram[N] {
+func newExponentialHistogram[N int64 | float64](maxSize, maxScale int32, noMinMax, noSum bool, limit int, r func() exemplar.Reservoir[N]) *expoHistogram[N] {
 	return &expoHistogram[N]{
 		noSum:    noSum,
 		noMinMax: noMinMax,
 		maxSize:  int(maxSize),
 		maxScale: int(maxScale),
 
+		newRes: r,
+		limit:  newLimiter[*expoHistogramDataPoint[N]](limit),
 		values: make(map[attribute.Set]*expoHistogramDataPoint[N]),
 
 		start: now(),
@@ -309,27 +314,35 @@ type expoHistogram[N int64 | float64] struct {
 	maxSize  int
 	maxScale int
 
+	newRes   func() exemplar.Reservoir[N]
+	limit    limiter[*expoHistogramDataPoint[N]]
 	values   map[attribute.Set]*expoHistogramDataPoint[N]
 	valuesMu sync.Mutex
 
 	start time.Time
 }
 
-func (e *expoHistogram[N]) measure(_ context.Context, value N, attr attribute.Set) {
+func (e *expoHistogram[N]) measure(ctx context.Context, value N, fltrAttr attribute.Set, droppedAttr []attribute.KeyValue) {
 	// Ignore NaN and infinity.
 	if math.IsInf(float64(value), 0) || math.IsNaN(float64(value)) {
 		return
 	}
 
+	t := now()
+
 	e.valuesMu.Lock()
 	defer e.valuesMu.Unlock()
 
+	attr := e.limit.Attributes(fltrAttr, e.values)
 	v, ok := e.values[attr]
 	if !ok {
 		v = newExpoHistogramDataPoint[N](e.maxSize, e.maxScale, e.noMinMax, e.noSum)
+		v.res = e.newRes()
+
 		e.values[attr] = v
 	}
 	v.record(value)
+	v.res.Offer(ctx, t, value, droppedAttr)
 }
 
 func (e *expoHistogram[N]) delta(dest *metricdata.Aggregation) int {
@@ -362,6 +375,7 @@ func (e *expoHistogram[N]) delta(dest *metricdata.Aggregation) int {
 
 		hDPts[i].NegativeBucket.Offset = int32(b.negBuckets.startBin)
 		hDPts[i].NegativeBucket.Counts = reset(hDPts[i].NegativeBucket.Counts, len(b.negBuckets.counts), len(b.negBuckets.counts))
+		copy(hDPts[i].NegativeBucket.Counts, b.negBuckets.counts)
 
 		if !e.noSum {
 			hDPts[i].Sum = b.sum
@@ -371,6 +385,8 @@ func (e *expoHistogram[N]) delta(dest *metricdata.Aggregation) int {
 			hDPts[i].Max = metricdata.NewExtrema(b.max)
 		}
 
+		b.res.Collect(&hDPts[i].Exemplars)
+
 		delete(e.values, a)
 		i++
 	}
@@ -410,6 +426,7 @@ func (e *expoHistogram[N]) cumulative(dest *metricdata.Aggregation) int {
 
 		hDPts[i].NegativeBucket.Offset = int32(b.negBuckets.startBin)
 		hDPts[i].NegativeBucket.Counts = reset(hDPts[i].NegativeBucket.Counts, len(b.negBuckets.counts), len(b.negBuckets.counts))
+		copy(hDPts[i].NegativeBucket.Counts, b.negBuckets.counts)
 
 		if !e.noSum {
 			hDPts[i].Sum = b.sum
@@ -419,6 +436,8 @@ func (e *expoHistogram[N]) cumulative(dest *metricdata.Aggregation) int {
 			hDPts[i].Max = metricdata.NewExtrema(b.max)
 		}
 
+		b.res.Collect(&hDPts[i].Exemplars)
+
 		i++
 		// TODO (#3006): This will use an unbounded amount of memory if there
 		// are unbounded number of attribute sets being aggregated. Attribute
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/histogram.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/histogram.go
index 62ec51e1f..a9a4706bf 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/histogram.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/histogram.go
@@ -21,10 +21,13 @@ import (
 	"time"
 
 	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
 	"go.opentelemetry.io/otel/sdk/metric/metricdata"
 )
 
 type buckets[N int64 | float64] struct {
+	res exemplar.Reservoir[N]
+
 	counts   []uint64
 	count    uint64
 	total    N
@@ -54,11 +57,13 @@ type histValues[N int64 | float64] struct {
 	noSum  bool
 	bounds []float64
 
+	newRes   func() exemplar.Reservoir[N]
+	limit    limiter[*buckets[N]]
 	values   map[attribute.Set]*buckets[N]
 	valuesMu sync.Mutex
 }
 
-func newHistValues[N int64 | float64](bounds []float64, noSum bool) *histValues[N] {
+func newHistValues[N int64 | float64](bounds []float64, noSum bool, limit int, r func() exemplar.Reservoir[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
@@ -69,13 +74,15 @@ func newHistValues[N int64 | float64](bounds []float64, noSum bool) *histValues[
 	return &histValues[N]{
 		noSum:  noSum,
 		bounds: b,
+		newRes: r,
+		limit:  newLimiter[*buckets[N]](limit),
 		values: make(map[attribute.Set]*buckets[N]),
 	}
 }
 
 // Aggregate records the measurement value, scoped by attr, and aggregates it
 // into a histogram.
-func (s *histValues[N]) measure(_ context.Context, value N, attr attribute.Set) {
+func (s *histValues[N]) measure(ctx context.Context, value N, fltrAttr attribute.Set, droppedAttr []attribute.KeyValue) {
 	// This search will return an index in the range [0, len(s.bounds)], where
 	// it will return len(s.bounds) if value is greater than the last element
 	// of s.bounds. This aligns with the buckets in that the length of buckets
@@ -83,9 +90,12 @@ func (s *histValues[N]) measure(_ context.Context, value N, attr attribute.Set)
 	// (s.bounds[len(s.bounds)-1], +∞).
 	idx := sort.SearchFloat64s(s.bounds, float64(value))
 
+	t := now()
+
 	s.valuesMu.Lock()
 	defer s.valuesMu.Unlock()
 
+	attr := s.limit.Attributes(fltrAttr, s.values)
 	b, ok := s.values[attr]
 	if !ok {
 		// N+1 buckets. For example:
@@ -96,6 +106,8 @@ func (s *histValues[N]) measure(_ context.Context, value N, attr attribute.Set)
 		//
 		//   buckets = (-∞, 0], (0, 5.0], (5.0, 10.0], (10.0, +∞)
 		b = newBuckets[N](len(s.bounds) + 1)
+		b.res = s.newRes()
+
 		// Ensure min and max are recorded values (not zero), for new buckets.
 		b.min, b.max = value, value
 		s.values[attr] = b
@@ -104,13 +116,14 @@ func (s *histValues[N]) measure(_ context.Context, value N, attr attribute.Set)
 	if !s.noSum {
 		b.sum(value)
 	}
+	b.res.Offer(ctx, t, value, droppedAttr)
 }
 
 // newHistogram returns an Aggregator that summarizes a set of measurements as
 // an histogram.
-func newHistogram[N int64 | float64](boundaries []float64, noMinMax, noSum bool) *histogram[N] {
+func newHistogram[N int64 | float64](boundaries []float64, noMinMax, noSum bool, limit int, r func() exemplar.Reservoir[N]) *histogram[N] {
 	return &histogram[N]{
-		histValues: newHistValues[N](boundaries, noSum),
+		histValues: newHistValues[N](boundaries, noSum, limit, r),
 		noMinMax:   noMinMax,
 		start:      now(),
 	}
@@ -161,6 +174,8 @@ func (s *histogram[N]) delta(dest *metricdata.Aggregation) int {
 			hDPts[i].Max = metricdata.NewExtrema(b.max)
 		}
 
+		b.res.Collect(&hDPts[i].Exemplars)
+
 		// Unused attribute sets do not report.
 		delete(s.values, a)
 		i++
@@ -217,6 +232,9 @@ func (s *histogram[N]) cumulative(dest *metricdata.Aggregation) int {
 			hDPts[i].Min = metricdata.NewExtrema(b.min)
 			hDPts[i].Max = metricdata.NewExtrema(b.max)
 		}
+
+		b.res.Collect(&hDPts[i].Exemplars)
+
 		i++
 		// TODO (#3006): This will use an unbounded amount of memory if there
 		// are unbounded number of attribute sets being aggregated. Attribute
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/lastvalue.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/lastvalue.go
index 6af2d6061..5699e728f 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/lastvalue.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/lastvalue.go
@@ -20,6 +20,7 @@ import (
 	"time"
 
 	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
 	"go.opentelemetry.io/otel/sdk/metric/metricdata"
 )
 
@@ -27,24 +28,43 @@ import (
 type datapoint[N int64 | float64] struct {
 	timestamp time.Time
 	value     N
+	res       exemplar.Reservoir[N]
 }
 
-func newLastValue[N int64 | float64]() *lastValue[N] {
-	return &lastValue[N]{values: make(map[attribute.Set]datapoint[N])}
+func newLastValue[N int64 | float64](limit int, r func() exemplar.Reservoir[N]) *lastValue[N] {
+	return &lastValue[N]{
+		newRes: r,
+		limit:  newLimiter[datapoint[N]](limit),
+		values: make(map[attribute.Set]datapoint[N]),
+	}
 }
 
 // lastValue summarizes a set of measurements as the last one made.
 type lastValue[N int64 | float64] struct {
 	sync.Mutex
 
+	newRes func() exemplar.Reservoir[N]
+	limit  limiter[datapoint[N]]
 	values map[attribute.Set]datapoint[N]
 }
 
-func (s *lastValue[N]) measure(ctx context.Context, value N, attr attribute.Set) {
-	d := datapoint[N]{timestamp: now(), value: value}
+func (s *lastValue[N]) measure(ctx context.Context, value N, fltrAttr attribute.Set, droppedAttr []attribute.KeyValue) {
+	t := now()
+
 	s.Lock()
+	defer s.Unlock()
+
+	attr := s.limit.Attributes(fltrAttr, s.values)
+	d, ok := s.values[attr]
+	if !ok {
+		d.res = s.newRes()
+	}
+
+	d.timestamp = t
+	d.value = value
+	d.res.Offer(ctx, t, value, droppedAttr)
+
 	s.values[attr] = d
-	s.Unlock()
 }
 
 func (s *lastValue[N]) computeAggregation(dest *[]metricdata.DataPoint[N]) {
@@ -61,6 +81,7 @@ func (s *lastValue[N]) computeAggregation(dest *[]metricdata.DataPoint[N]) {
 		// ignored.
 		(*dest)[i].Time = v.timestamp
 		(*dest)[i].Value = v.value
+		v.res.Collect(&(*dest)[i].Exemplars)
 		// Do not report stale values.
 		delete(s.values, a)
 		i++
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/limit.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/limit.go
new file mode 100644
index 000000000..d3de84272
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/limit.go
@@ -0,0 +1,53 @@
+// Copyright The OpenTelemetry Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package aggregate // import "go.opentelemetry.io/otel/sdk/metric/internal/aggregate"
+
+import "go.opentelemetry.io/otel/attribute"
+
+// overflowSet is the attribute set used to record a measurement when adding
+// another distinct attribute set to the aggregate would exceed the aggregate
+// limit.
+var overflowSet = attribute.NewSet(attribute.Bool("otel.metric.overflow", true))
+
+// limiter limits aggregate values.
+type limiter[V any] struct {
+	// aggLimit is the maximum number of metric streams that can be aggregated.
+	//
+	// Any metric stream with attributes distinct from any set already
+	// aggregated once the aggLimit will be meet will instead be aggregated
+	// into an "overflow" metric stream. That stream will only contain the
+	// "otel.metric.overflow"=true attribute.
+	aggLimit int
+}
+
+// newLimiter returns a new Limiter with the provided aggregation limit.
+func newLimiter[V any](aggregation int) limiter[V] {
+	return limiter[V]{aggLimit: aggregation}
+}
+
+// Attributes checks if adding a measurement for attrs will exceed the
+// aggregation cardinality limit for the existing measurements. If it will,
+// overflowSet is returned. Otherwise, if it will not exceed the limit, or the
+// limit is not set (limit <= 0), attr is returned.
+func (l limiter[V]) Attributes(attrs attribute.Set, measurements map[attribute.Set]V) attribute.Set {
+	if l.aggLimit > 0 {
+		_, exists := measurements[attrs]
+		if !exists && len(measurements) >= l.aggLimit-1 {
+			return overflowSet
+		}
+	}
+
+	return attrs
+}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/sum.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/sum.go
index 1e52ff0d1..02de2483f 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/sum.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/aggregate/sum.go
@@ -20,31 +20,55 @@ import (
 	"time"
 
 	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
 	"go.opentelemetry.io/otel/sdk/metric/metricdata"
 )
 
+type sumValue[N int64 | float64] struct {
+	n   N
+	res exemplar.Reservoir[N]
+}
+
 // valueMap is the storage for sums.
 type valueMap[N int64 | float64] struct {
 	sync.Mutex
-	values map[attribute.Set]N
+	newRes func() exemplar.Reservoir[N]
+	limit  limiter[sumValue[N]]
+	values map[attribute.Set]sumValue[N]
 }
 
-func newValueMap[N int64 | float64]() *valueMap[N] {
-	return &valueMap[N]{values: make(map[attribute.Set]N)}
+func newValueMap[N int64 | float64](limit int, r func() exemplar.Reservoir[N]) *valueMap[N] {
+	return &valueMap[N]{
+		newRes: r,
+		limit:  newLimiter[sumValue[N]](limit),
+		values: make(map[attribute.Set]sumValue[N]),
+	}
 }
 
-func (s *valueMap[N]) measure(_ context.Context, value N, attr attribute.Set) {
+func (s *valueMap[N]) measure(ctx context.Context, value N, fltrAttr attribute.Set, droppedAttr []attribute.KeyValue) {
+	t := now()
+
 	s.Lock()
-	s.values[attr] += value
-	s.Unlock()
+	defer s.Unlock()
+
+	attr := s.limit.Attributes(fltrAttr, s.values)
+	v, ok := s.values[attr]
+	if !ok {
+		v.res = s.newRes()
+	}
+
+	v.n += value
+	v.res.Offer(ctx, t, value, droppedAttr)
+
+	s.values[attr] = v
 }
 
 // newSum returns an aggregator that summarizes a set of measurements as their
 // arithmetic sum. Each sum is scoped by attributes and the aggregation cycle
 // the measurements were made in.
-func newSum[N int64 | float64](monotonic bool) *sum[N] {
+func newSum[N int64 | float64](monotonic bool, limit int, r func() exemplar.Reservoir[N]) *sum[N] {
 	return &sum[N]{
-		valueMap:  newValueMap[N](),
+		valueMap:  newValueMap[N](limit, r),
 		monotonic: monotonic,
 		start:     now(),
 	}
@@ -74,11 +98,12 @@ func (s *sum[N]) delta(dest *metricdata.Aggregation) int {
 	dPts := reset(sData.DataPoints, n, n)
 
 	var i int
-	for attr, value := range s.values {
+	for attr, val := range s.values {
 		dPts[i].Attributes = attr
 		dPts[i].StartTime = s.start
 		dPts[i].Time = t
-		dPts[i].Value = value
+		dPts[i].Value = val.n
+		val.res.Collect(&dPts[i].Exemplars)
 		// Do not report stale values.
 		delete(s.values, attr)
 		i++
@@ -112,7 +137,8 @@ func (s *sum[N]) cumulative(dest *metricdata.Aggregation) int {
 		dPts[i].Attributes = attr
 		dPts[i].StartTime = s.start
 		dPts[i].Time = t
-		dPts[i].Value = value
+		dPts[i].Value = value.n
+		value.res.Collect(&dPts[i].Exemplars)
 		// TODO (#3006): This will use an unbounded amount of memory if there
 		// are unbounded number of attribute sets being aggregated. Attribute
 		// sets that become "stale" need to be forgotten so this will not
@@ -129,9 +155,9 @@ func (s *sum[N]) cumulative(dest *metricdata.Aggregation) int {
 // newPrecomputedSum returns an aggregator that summarizes a set of
 // observatrions as their arithmetic sum. Each sum is scoped by attributes and
 // the aggregation cycle the measurements were made in.
-func newPrecomputedSum[N int64 | float64](monotonic bool) *precomputedSum[N] {
+func newPrecomputedSum[N int64 | float64](monotonic bool, limit int, r func() exemplar.Reservoir[N]) *precomputedSum[N] {
 	return &precomputedSum[N]{
-		valueMap:  newValueMap[N](),
+		valueMap:  newValueMap[N](limit, r),
 		monotonic: monotonic,
 		start:     now(),
 	}
@@ -165,14 +191,15 @@ func (s *precomputedSum[N]) delta(dest *metricdata.Aggregation) int {
 
 	var i int
 	for attr, value := range s.values {
-		delta := value - s.reported[attr]
+		delta := value.n - s.reported[attr]
 
 		dPts[i].Attributes = attr
 		dPts[i].StartTime = s.start
 		dPts[i].Time = t
 		dPts[i].Value = delta
+		value.res.Collect(&dPts[i].Exemplars)
 
-		newReported[attr] = value
+		newReported[attr] = value.n
 		// Unused attribute sets do not report.
 		delete(s.values, attr)
 		i++
@@ -204,11 +231,12 @@ func (s *precomputedSum[N]) cumulative(dest *metricdata.Aggregation) int {
 	dPts := reset(sData.DataPoints, n, n)
 
 	var i int
-	for attr, value := range s.values {
+	for attr, val := range s.values {
 		dPts[i].Attributes = attr
 		dPts[i].StartTime = s.start
 		dPts[i].Time = t
-		dPts[i].Value = value
+		dPts[i].Value = val.n
+		val.res.Collect(&dPts[i].Exemplars)
 
 		// Unused attribute sets do not report.
 		delete(s.values, attr)
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/doc.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/doc.go
new file mode 100644
index 000000000..3caeb542c
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/doc.go
@@ -0,0 +1,17 @@
+// Copyright The OpenTelemetry Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package exemplar provides an implementation of the OpenTelemetry exemplar
+// reservoir to be used in metric collection pipelines.
+package exemplar // import "go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/drop.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/drop.go
new file mode 100644
index 000000000..39bf37b9e
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/drop.go
@@ -0,0 +1,36 @@
+// Copyright The OpenTelemetry Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package exemplar // import "go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
+
+import (
+	"context"
+	"time"
+
+	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/sdk/metric/metricdata"
+)
+
+// Drop returns a [Reservoir] that drops all measurements it is offered.
+func Drop[N int64 | float64]() Reservoir[N] { return &dropRes[N]{} }
+
+type dropRes[N int64 | float64] struct{}
+
+// Offer does nothing, all measurements offered will be dropped.
+func (r *dropRes[N]) Offer(context.Context, time.Time, N, []attribute.KeyValue) {}
+
+// Collect resets dest. No exemplars will ever be returned.
+func (r *dropRes[N]) Collect(dest *[]metricdata.Exemplar[N]) {
+	*dest = (*dest)[:0]
+}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/filter.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/filter.go
new file mode 100644
index 000000000..4f5946fb9
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/filter.go
@@ -0,0 +1,40 @@
+// Copyright The OpenTelemetry Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package exemplar // import "go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
+
+import (
+	"context"
+	"time"
+
+	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/trace"
+)
+
+// SampledFilter returns a [Reservoir] wrapping r that will only offer measurements
+// to r if the passed context associated with the measurement contains a sampled
+// [go.opentelemetry.io/otel/trace.SpanContext].
+func SampledFilter[N int64 | float64](r Reservoir[N]) Reservoir[N] {
+	return filtered[N]{Reservoir: r}
+}
+
+type filtered[N int64 | float64] struct {
+	Reservoir[N]
+}
+
+func (f filtered[N]) Offer(ctx context.Context, t time.Time, n N, a []attribute.KeyValue) {
+	if trace.SpanContextFromContext(ctx).IsSampled() {
+		f.Reservoir.Offer(ctx, t, n, a)
+	}
+}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/hist.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/hist.go
new file mode 100644
index 000000000..6f4fe5524
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/hist.go
@@ -0,0 +1,47 @@
+// Copyright The OpenTelemetry Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package exemplar // import "go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
+
+import (
+	"context"
+	"sort"
+	"time"
+
+	"go.opentelemetry.io/otel/attribute"
+)
+
+// Histogram returns a [Reservoir] 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.
+func Histogram[N int64 | float64](bounds []float64) Reservoir[N] {
+	sort.Float64s(bounds)
+	return &histRes[N]{
+		bounds:  bounds,
+		storage: newStorage[N](len(bounds) + 1),
+	}
+}
+
+type histRes[N int64 | float64] struct {
+	*storage[N]
+
+	// bounds are bucket bounds in ascending order.
+	bounds []float64
+}
+
+func (r *histRes[N]) Offer(ctx context.Context, t time.Time, n N, a []attribute.KeyValue) {
+	r.store[sort.SearchFloat64s(r.bounds, float64(n))] = newMeasurement(ctx, t, n, a)
+}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/rand.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/rand.go
new file mode 100644
index 000000000..7f9fda5b4
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/rand.go
@@ -0,0 +1,195 @@
+// Copyright The OpenTelemetry Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package exemplar // import "go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
+
+import (
+	"context"
+	"math"
+	"math/rand"
+	"time"
+
+	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/sdk/metric/metricdata"
+)
+
+// rng is used to make sampling decisions.
+//
+// Do not use crypto/rand. There is no reason for the decrease in performance
+// given this is not a security sensitive decision.
+var rng = rand.New(rand.NewSource(time.Now().UnixNano()))
+
+// random returns, as a float64, a uniform pseudo-random number in the open
+// interval (0.0,1.0).
+func random() float64 {
+	// TODO: This does not return a uniform number. rng.Float64 returns a
+	// uniformly random int in [0,2^53) that is divided by 2^53. Meaning it
+	// returns multiples of 2^-53, and not all floating point numbers between 0
+	// and 1 (i.e. for values less than 2^-4 the 4 last bits of the significand
+	// are always going to be 0).
+	//
+	// An alternative algorithm should be considered that will actually return
+	// a uniform number in the interval (0,1). For example, since the default
+	// rand source provides a uniform distribution for Int63, this can be
+	// converted following the prototypical code of Mersenne Twister 64 (Takuji
+	// Nishimura and Makoto Matsumoto:
+	// http://www.math.sci.hiroshima-u.ac.jp/m-mat/MT/VERSIONS/C-LANG/mt19937-64.c)
+	//
+	//   (float64(rng.Int63()>>11) + 0.5) * (1.0 / 4503599627370496.0)
+	//
+	// There are likely many other methods to explore here as well.
+
+	f := rng.Float64()
+	for f == 0 {
+		f = rng.Float64()
+	}
+	return f
+}
+
+// FixedSize returns a [Reservoir] 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 sample all
+// additional measurement with a decreasing probability.
+func FixedSize[N int64 | float64](k int) Reservoir[N] {
+	r := &randRes[N]{storage: newStorage[N](k)}
+	r.reset()
+	return r
+}
+
+type randRes[N int64 | float64] struct {
+	*storage[N]
+
+	// count is the number of measurement seen.
+	count int64
+	// next is the next count that will store a measurement at a random index
+	// once the reservoir has been filled.
+	next int64
+	// w is the largest random number in a distribution that is used to compute
+	// the next next.
+	w float64
+}
+
+func (r *randRes[N]) Offer(ctx context.Context, t time.Time, n N, a []attribute.KeyValue) {
+	// The following algorithm is "Algorithm L" from Li, Kim-Hung (4 December
+	// 1994). "Reservoir-Sampling Algorithms of Time Complexity
+	// O(n(1+log(N/n)))". ACM Transactions on Mathematical Software. 20 (4):
+	// 481–493 (https://dl.acm.org/doi/10.1145/198429.198435).
+	//
+	// A high-level overview of "Algorithm L":
+	//   0) Pre-calculate the random count greater than the storage size when
+	//      an exemplar will be replaced.
+	//   1) Accept all measurements offered until the configured storage size is
+	//      reached.
+	//   2) Loop:
+	//      a) When the pre-calculate count is reached, replace a random
+	//         existing exemplar with the offered measurement.
+	//      b) Calculate the next random count greater than the existing one
+	//         which will replace another exemplars
+	//
+	// The way a "replacement" count is computed is by looking at `n` number of
+	// independent random numbers each corresponding to an offered measurement.
+	// Of these numbers the smallest `k` (the same size as the storage
+	// capacity) of them are kept as a subset. The maximum value in this
+	// subset, called `w` is used to weight another random number generation
+	// for the next count that will be considered.
+	//
+	// By weighting the next count computation like described, it is able to
+	// perform a uniformly-weighted sampling algorithm based on the number of
+	// samples the reservoir has seen so far. The sampling will "slow down" as
+	// more and more samples are offered so as to reduce a bias towards those
+	// offered just prior to the end of the collection.
+	//
+	// This algorithm is preferred because of its balance of simplicity and
+	// performance. It will compute three random numbers (the bulk of
+	// computation time) for each item that becomes part of the reservoir, but
+	// it does not spend any time on items that do not. In particular it has an
+	// asymptotic runtime of O(k(1 + log(n/k)) where n is the number of
+	// measurements offered and k is the reservoir size.
+	//
+	// See https://en.wikipedia.org/wiki/Reservoir_sampling for an overview of
+	// this and other reservoir sampling algorithms. See
+	// https://github.com/MrAlias/reservoir-sampling for a performance
+	// comparison of reservoir sampling algorithms.
+
+	if int(r.count) < cap(r.store) {
+		r.store[r.count] = newMeasurement(ctx, t, n, a)
+	} else {
+		if r.count == r.next {
+			// Overwrite a random existing measurement with the one offered.
+			idx := int(rng.Int63n(int64(cap(r.store))))
+			r.store[idx] = newMeasurement(ctx, t, n, a)
+			r.advance()
+		}
+	}
+	r.count++
+}
+
+// reset resets r to the initial state.
+func (r *randRes[N]) reset() {
+	// This resets the number of exemplars known.
+	r.count = 0
+	// Random index inserts should only happen after the storage is full.
+	r.next = int64(cap(r.store))
+
+	// Initial random number in the series used to generate r.next.
+	//
+	// This is set before r.advance to reset or initialize the random number
+	// series. Without doing so it would always be 0 or never restart a new
+	// random number series.
+	//
+	// This maps the uniform random number in (0,1) to a geometric distribution
+	// over the same interval. The mean of the distribution is inversely
+	// proportional to the storage capacity.
+	r.w = math.Exp(math.Log(random()) / float64(cap(r.store)))
+
+	r.advance()
+}
+
+// advance updates the count at which the offered measurement will overwrite an
+// existing exemplar.
+func (r *randRes[N]) advance() {
+	// Calculate the next value in the random number series.
+	//
+	// The current value of r.w is based on the max of a distribution of random
+	// numbers (i.e. `w = max(u_1,u_2,...,u_k)` for `k` equal to the capacity
+	// of the storage and each `u` in the interval (0,w)). To calculate the
+	// next r.w we use the fact that when the next exemplar is selected to be
+	// included in the storage an existing one will be dropped, and the
+	// corresponding random number in the set used to calculate r.w will also
+	// be replaced. The replacement random number will also be within (0,w),
+	// therefore the next r.w will be based on the same distribution (i.e.
+	// `max(u_1,u_2,...,u_k)`). Therefore, we can sample the next r.w by
+	// computing the next random number `u` and take r.w as `w * u^(1/k)`.
+	r.w *= math.Exp(math.Log(random()) / float64(cap(r.store)))
+	// Use the new random number in the series to calculate the count of the
+	// next measurement that will be stored.
+	//
+	// Given 0 < r.w < 1, each iteration will result in subsequent r.w being
+	// smaller. This translates here into the next next being selected against
+	// a distribution with a higher mean (i.e. the expected value will increase
+	// and replacements become less likely)
+	//
+	// Important to note, the new r.next will always be at least 1 more than
+	// the last r.next.
+	r.next += int64(math.Log(random())/math.Log(1-r.w)) + 1
+}
+
+func (r *randRes[N]) Collect(dest *[]metricdata.Exemplar[N]) {
+	r.storage.Collect(dest)
+	// Call reset here even though it will reset r.count and restart the random
+	// number series. This will persist any old exemplars as long as no new
+	// measurements are offered, but it will also prioritize those new
+	// measurements that are made over the older collection cycle ones.
+	r.reset()
+}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/reservoir.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/reservoir.go
new file mode 100644
index 000000000..7d5276a34
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/reservoir.go
@@ -0,0 +1,44 @@
+// Copyright The OpenTelemetry Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package exemplar // import "go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
+
+import (
+	"context"
+	"time"
+
+	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/sdk/metric/metricdata"
+)
+
+// Reservoir holds the sampled exemplar of measurements made.
+type Reservoir[N int64 | float64] interface {
+	// Offer accepts the parameters associated with a measurement. The
+	// parameters will be stored as an exemplar if the Reservoir decides to
+	// sample the measurement.
+	//
+	// The passed ctx needs to contain any baggage or span that were active
+	// when the measurement was made. This information may be used by the
+	// Reservoir in making a sampling decision.
+	//
+	// The time t is the time when the measurement was made. The val and attr
+	// parameters are the value and dropped (filtered) attributes of the
+	// measurement respectively.
+	Offer(ctx context.Context, t time.Time, val N, attr []attribute.KeyValue)
+
+	// Collect returns all the held exemplars.
+	//
+	// The Reservoir state is preserved after this call.
+	Collect(dest *[]metricdata.Exemplar[N])
+}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/storage.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/storage.go
new file mode 100644
index 000000000..e2c2b90a3
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/exemplar/storage.go
@@ -0,0 +1,107 @@
+// Copyright The OpenTelemetry Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package exemplar // import "go.opentelemetry.io/otel/sdk/metric/internal/exemplar"
+
+import (
+	"context"
+	"time"
+
+	"go.opentelemetry.io/otel/attribute"
+	"go.opentelemetry.io/otel/sdk/metric/metricdata"
+	"go.opentelemetry.io/otel/trace"
+)
+
+// storage is an exemplar storage for [Reservoir] implementations.
+type storage[N int64 | float64] struct {
+	// store are the measurements sampled.
+	//
+	// This does not use []metricdata.Exemplar because it potentially would
+	// require an allocation for trace and span IDs in the hot path of Offer.
+	store []measurement[N]
+}
+
+func newStorage[N int64 | float64](n int) *storage[N] {
+	return &storage[N]{store: make([]measurement[N], n)}
+}
+
+// Collect returns all the held exemplars.
+//
+// The Reservoir state is preserved after this call.
+func (r *storage[N]) Collect(dest *[]metricdata.Exemplar[N]) {
+	*dest = reset(*dest, len(r.store), len(r.store))
+	var n int
+	for _, m := range r.store {
+		if !m.valid {
+			continue
+		}
+
+		m.Exemplar(&(*dest)[n])
+		n++
+	}
+	*dest = (*dest)[:n]
+}
+
+// measurement is a measurement made by a telemetry system.
+type measurement[N int64 | float64] struct {
+	// FilteredAttributes are the attributes dropped during the measurement.
+	FilteredAttributes []attribute.KeyValue
+	// Time is the time when the measurement was made.
+	Time time.Time
+	// Value is the value of the measurement.
+	Value N
+	// SpanContext is the SpanContext active when a measurement was made.
+	SpanContext trace.SpanContext
+
+	valid bool
+}
+
+// newMeasurement returns a new non-empty Measurement.
+func newMeasurement[N int64 | float64](ctx context.Context, ts time.Time, v N, droppedAttr []attribute.KeyValue) measurement[N] {
+	return measurement[N]{
+		FilteredAttributes: droppedAttr,
+		Time:               ts,
+		Value:              v,
+		SpanContext:        trace.SpanContextFromContext(ctx),
+		valid:              true,
+	}
+}
+
+// Exemplar returns m as a [metricdata.Exemplar].
+func (m measurement[N]) Exemplar(dest *metricdata.Exemplar[N]) {
+	dest.FilteredAttributes = m.FilteredAttributes
+	dest.Time = m.Time
+	dest.Value = m.Value
+
+	if m.SpanContext.HasTraceID() {
+		traceID := m.SpanContext.TraceID()
+		dest.TraceID = traceID[:]
+	} else {
+		dest.TraceID = dest.TraceID[:0]
+	}
+
+	if m.SpanContext.HasSpanID() {
+		spanID := m.SpanContext.SpanID()
+		dest.SpanID = spanID[:]
+	} else {
+		dest.SpanID = dest.SpanID[:0]
+	}
+}
+
+func reset[T any](s []T, length, capacity int) []T {
+	if cap(s) < capacity {
+		return make([]T, length, capacity)
+	}
+	return s[:length]
+}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/x/README.md b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/x/README.md
new file mode 100644
index 000000000..aba69d654
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/x/README.md
@@ -0,0 +1,112 @@
+# Experimental Features
+
+The metric SDK contains features that have not yet stabilized in the OpenTelemetry specification.
+These features are added to the OpenTelemetry Go metric SDK prior to stabilization in the specification so that users can start experimenting with them and provide feedback.
+
+These feature may change in backwards incompatible ways as feedback is applied.
+See the [Compatibility and Stability](#compatibility-and-stability) section for more information.
+
+## Features
+
+- [Cardinality Limit](#cardinality-limit)
+- [Exemplars](#exemplars)
+
+### Cardinality Limit
+
+The cardinality limit is the hard limit on the number of metric streams that can be collected for a single instrument.
+
+This experimental feature can be enabled by setting the `OTEL_GO_X_CARDINALITY_LIMIT` environment value.
+The value must be an integer value.
+All other values are ignored.
+
+If the value set is less than or equal to `0`, no limit will be applied.
+
+#### Examples
+
+Set the cardinality limit to 2000.
+
+```console
+export OTEL_GO_X_CARDINALITY_LIMIT=2000
+```
+
+Set an infinite cardinality limit (functionally equivalent to disabling the feature).
+
+```console
+export OTEL_GO_X_CARDINALITY_LIMIT=-1
+```
+
+Disable the cardinality limit.
+
+```console
+unset OTEL_GO_X_CARDINALITY_LIMIT
+```
+
+### Exemplars
+
+A sample of measurements made may be exported directly as a set of exemplars.
+
+This experimental feature can be enabled by setting the `OTEL_GO_X_EXEMPLAR` environment variable.
+The value of must be the case-insensitive string of `"true"` to enable the feature.
+All other values are ignored.
+
+Exemplar filters are a supported.
+The exemplar filter applies to all measurements made.
+They filter these measurements, only allowing certain measurements to be passed to the underlying exemplar reservoir.
+
+To change the exemplar filter from the default `"trace_based"` filter set the `OTEL_METRICS_EXEMPLAR_FILTER` environment variable.
+The value must be the case-sensitive string defined by the [OpenTelemetry specification].
+
+- `"always_on"`: allows all measurements
+- `"always_off"`: denies all measurements
+- `"trace_based"`: allows only sampled measurements
+
+All values other than these will result in the default, `"trace_based"`, exemplar filter being used.
+
+[OpenTelemetry specification]: https://github.com/open-telemetry/opentelemetry-specification/blob/a6ca2fd484c9e76fe1d8e1c79c99f08f4745b5ee/specification/configuration/sdk-environment-variables.md#exemplar
+
+#### Examples
+
+Enable exemplars to be exported.
+
+```console
+export OTEL_GO_X_EXEMPLAR=true
+```
+
+Disable exemplars from being exported.
+
+```console
+unset OTEL_GO_X_EXEMPLAR
+```
+
+Set the exemplar filter to allow all measurements.
+
+```console
+export OTEL_METRICS_EXEMPLAR_FILTER=always_on
+```
+
+Set the exemplar filter to deny all measurements.
+
+```console
+export OTEL_METRICS_EXEMPLAR_FILTER=always_off
+```
+
+Set the exemplar filter to only allow sampled measurements.
+
+```console
+export OTEL_METRICS_EXEMPLAR_FILTER=trace_based
+```
+
+Revert to the default exemplar filter (`"trace_based"`)
+
+```console
+unset OTEL_METRICS_EXEMPLAR_FILTER
+```
+
+## Compatibility and Stability
+
+Experimental features do not fall within the scope of the OpenTelemetry Go versioning and stability [policy](../../../../VERSIONING.md).
+These features may be removed or modified in successive version releases, including patch versions.
+
+When an experimental feature is promoted to a stable feature, a migration path will be included in the changelog entry of the release.
+There is no guarantee that any environment variable feature flags that enabled the experimental feature will be supported by the stable version.
+If they are supported, they may be accompanied with a deprecation notice stating a timeline for the removal of that support.
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/internal/x/x.go b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/x/x.go
new file mode 100644
index 000000000..541160f94
--- /dev/null
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/internal/x/x.go
@@ -0,0 +1,96 @@
+// Copyright The OpenTelemetry Authors
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+// Package x contains support for OTel metric SDK experimental features.
+//
+// This package should only be used for features defined in the specification.
+// It should not be used for experiments or new project ideas.
+package x // import "go.opentelemetry.io/otel/sdk/metric/internal/x"
+
+import (
+	"os"
+	"strconv"
+	"strings"
+)
+
+var (
+	// Exemplars is an experimental feature flag that defines if exemplars
+	// should be recorded for metric data-points.
+	//
+	// To enable this feature set the OTEL_GO_X_EXEMPLAR environment variable
+	// to the case-insensitive string value of "true" (i.e. "True" and "TRUE"
+	// will also enable this).
+	Exemplars = newFeature("EXEMPLAR", func(v string) (string, bool) {
+		if strings.ToLower(v) == "true" {
+			return v, true
+		}
+		return "", false
+	})
+
+	// CardinalityLimit is an experimental feature flag that defines if
+	// cardinality limits should be applied to the recorded metric data-points.
+	//
+	// To enable this feature set the OTEL_GO_X_CARDINALITY_LIMIT environment
+	// variable to the integer limit value you want to use.
+	//
+	// Setting OTEL_GO_X_CARDINALITY_LIMIT to a value less than or equal to 0
+	// will disable the cardinality limits.
+	CardinalityLimit = newFeature("CARDINALITY_LIMIT", func(v string) (int, bool) {
+		n, err := strconv.Atoi(v)
+		if err != nil {
+			return 0, false
+		}
+		return n, true
+	})
+)
+
+// Feature is an experimental feature control flag. It provides a uniform way
+// to interact with these feature flags and parse their values.
+type Feature[T any] struct {
+	key   string
+	parse func(v string) (T, bool)
+}
+
+func newFeature[T any](suffix string, parse func(string) (T, bool)) Feature[T] {
+	const envKeyRoot = "OTEL_GO_X_"
+	return Feature[T]{
+		key:   envKeyRoot + suffix,
+		parse: parse,
+	}
+}
+
+// Key returns the environment variable key that needs to be set to enable the
+// feature.
+func (f Feature[T]) Key() string { return f.key }
+
+// Lookup returns the user configured value for the feature and true if the
+// user has enabled the feature. Otherwise, if the feature is not enabled, a
+// zero-value and false are returned.
+func (f Feature[T]) Lookup() (v T, ok bool) {
+	// https://github.com/open-telemetry/opentelemetry-specification/blob/62effed618589a0bec416a87e559c0a9d96289bb/specification/configuration/sdk-environment-variables.md#parsing-empty-value
+	//
+	// > The SDK MUST interpret an empty value of an environment variable the
+	// > same way as when the variable is unset.
+	vRaw := os.Getenv(f.key)
+	if vRaw == "" {
+		return v, ok
+	}
+	return f.parse(vRaw)
+}
+
+// Enabled returns if the feature is enabled.
+func (f Feature[T]) Enabled() bool {
+	_, ok := f.Lookup()
+	return ok
+}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/meter.go b/vendor/go.opentelemetry.io/otel/sdk/metric/meter.go
index 7f51ec512..beb7876ec 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/meter.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/meter.go
@@ -23,6 +23,7 @@ import (
 	"go.opentelemetry.io/otel/metric"
 	"go.opentelemetry.io/otel/metric/embedded"
 	"go.opentelemetry.io/otel/sdk/instrumentation"
+
 	"go.opentelemetry.io/otel/sdk/metric/internal/aggregate"
 )
 
@@ -40,6 +41,11 @@ type meter struct {
 	scope instrumentation.Scope
 	pipes pipelines
 
+	int64Insts             *cacheWithErr[instID, *int64Inst]
+	float64Insts           *cacheWithErr[instID, *float64Inst]
+	int64ObservableInsts   *cacheWithErr[instID, int64Observable]
+	float64ObservableInsts *cacheWithErr[instID, float64Observable]
+
 	int64Resolver   resolver[int64]
 	float64Resolver resolver[float64]
 }
@@ -49,11 +55,20 @@ func newMeter(s instrumentation.Scope, p pipelines) *meter {
 	// meter is asked to create are logged to the user.
 	var viewCache cache[string, instID]
 
+	var int64Insts cacheWithErr[instID, *int64Inst]
+	var float64Insts cacheWithErr[instID, *float64Inst]
+	var int64ObservableInsts cacheWithErr[instID, int64Observable]
+	var float64ObservableInsts cacheWithErr[instID, float64Observable]
+
 	return &meter{
-		scope:           s,
-		pipes:           p,
-		int64Resolver:   newResolver[int64](p, &viewCache),
-		float64Resolver: newResolver[float64](p, &viewCache),
+		scope:                  s,
+		pipes:                  p,
+		int64Insts:             &int64Insts,
+		float64Insts:           &float64Insts,
+		int64ObservableInsts:   &int64ObservableInsts,
+		float64ObservableInsts: &float64ObservableInsts,
+		int64Resolver:          newResolver[int64](p, &viewCache),
+		float64Resolver:        newResolver[float64](p, &viewCache),
 	}
 }
 
@@ -104,20 +119,62 @@ func (m *meter) Int64Histogram(name string, options ...metric.Int64HistogramOpti
 	return i, validateInstrumentName(name)
 }
 
+// int64ObservableInstrument returns a new observable identified by the Instrument.
+// It registers callbacks for each reader's pipeline.
+func (m *meter) int64ObservableInstrument(id Instrument, callbacks []metric.Int64Callback) (int64Observable, error) {
+	key := instID{
+		Name:        id.Name,
+		Description: id.Description,
+		Unit:        id.Unit,
+		Kind:        id.Kind,
+	}
+	if m.int64ObservableInsts.HasKey(key) && len(callbacks) > 0 {
+		warnRepeatedObservableCallbacks(id)
+	}
+	return m.int64ObservableInsts.Lookup(key, func() (int64Observable, error) {
+		inst := newInt64Observable(m, id.Kind, id.Name, id.Description, id.Unit)
+		for _, insert := range m.int64Resolver.inserters {
+			// Connect the measure functions for instruments in this pipeline with the
+			// callbacks for this pipeline.
+			in, err := insert.Instrument(id, insert.readerDefaultAggregation(id.Kind))
+			if err != nil {
+				return inst, err
+			}
+			// Drop aggregation
+			if len(in) == 0 {
+				inst.dropAggregation = true
+				continue
+			}
+			inst.appendMeasures(in)
+			for _, cback := range callbacks {
+				inst := int64Observer{measures: in}
+				fn := cback
+				insert.addCallback(func(ctx context.Context) error { return fn(ctx, inst) })
+			}
+		}
+		return inst, validateInstrumentName(id.Name)
+	})
+}
+
 // Int64ObservableCounter returns a new instrument identified by name and
 // configured with options. The instrument is used to asynchronously record
 // increasing int64 measurements once per a measurement collection cycle.
 // Only the measurements recorded during the collection cycle are exported.
+//
+// If Int64ObservableCounter is invoked repeatedly with the same Name,
+// Description, and Unit, only the first set of callbacks provided are used.
+// Use meter.RegisterCallback and Registration.Unregister to manage callbacks
+// if instrumentation can be created multiple times with different callbacks.
 func (m *meter) Int64ObservableCounter(name string, options ...metric.Int64ObservableCounterOption) (metric.Int64ObservableCounter, error) {
 	cfg := metric.NewInt64ObservableCounterConfig(options...)
-	const kind = InstrumentKindObservableCounter
-	p := int64ObservProvider{m}
-	inst, err := p.lookup(kind, name, cfg.Description(), cfg.Unit())
-	if err != nil {
-		return nil, err
+	id := Instrument{
+		Name:        name,
+		Description: cfg.Description(),
+		Unit:        cfg.Unit(),
+		Kind:        InstrumentKindObservableCounter,
+		Scope:       m.scope,
 	}
-	p.registerCallbacks(inst, cfg.Callbacks())
-	return inst, validateInstrumentName(name)
+	return m.int64ObservableInstrument(id, cfg.Callbacks())
 }
 
 // Int64ObservableUpDownCounter returns a new instrument identified by name and
@@ -126,14 +183,14 @@ func (m *meter) Int64ObservableCounter(name string, options ...metric.Int64Obser
 // measurements recorded during the collection cycle are exported.
 func (m *meter) Int64ObservableUpDownCounter(name string, options ...metric.Int64ObservableUpDownCounterOption) (metric.Int64ObservableUpDownCounter, error) {
 	cfg := metric.NewInt64ObservableUpDownCounterConfig(options...)
-	const kind = InstrumentKindObservableUpDownCounter
-	p := int64ObservProvider{m}
-	inst, err := p.lookup(kind, name, cfg.Description(), cfg.Unit())
-	if err != nil {
-		return nil, err
+	id := Instrument{
+		Name:        name,
+		Description: cfg.Description(),
+		Unit:        cfg.Unit(),
+		Kind:        InstrumentKindObservableUpDownCounter,
+		Scope:       m.scope,
 	}
-	p.registerCallbacks(inst, cfg.Callbacks())
-	return inst, validateInstrumentName(name)
+	return m.int64ObservableInstrument(id, cfg.Callbacks())
 }
 
 // Int64ObservableGauge returns a new instrument identified by name and
@@ -142,14 +199,14 @@ func (m *meter) Int64ObservableUpDownCounter(name string, options ...metric.Int6
 // Only the measurements recorded during the collection cycle are exported.
 func (m *meter) Int64ObservableGauge(name string, options ...metric.Int64ObservableGaugeOption) (metric.Int64ObservableGauge, error) {
 	cfg := metric.NewInt64ObservableGaugeConfig(options...)
-	const kind = InstrumentKindObservableGauge
-	p := int64ObservProvider{m}
-	inst, err := p.lookup(kind, name, cfg.Description(), cfg.Unit())
-	if err != nil {
-		return nil, err
+	id := Instrument{
+		Name:        name,
+		Description: cfg.Description(),
+		Unit:        cfg.Unit(),
+		Kind:        InstrumentKindObservableGauge,
+		Scope:       m.scope,
 	}
-	p.registerCallbacks(inst, cfg.Callbacks())
-	return inst, validateInstrumentName(name)
+	return m.int64ObservableInstrument(id, cfg.Callbacks())
 }
 
 // Float64Counter returns a new instrument identified by name and configured
@@ -196,20 +253,62 @@ func (m *meter) Float64Histogram(name string, options ...metric.Float64Histogram
 	return i, validateInstrumentName(name)
 }
 
+// float64ObservableInstrument returns a new observable identified by the Instrument.
+// It registers callbacks for each reader's pipeline.
+func (m *meter) float64ObservableInstrument(id Instrument, callbacks []metric.Float64Callback) (float64Observable, error) {
+	key := instID{
+		Name:        id.Name,
+		Description: id.Description,
+		Unit:        id.Unit,
+		Kind:        id.Kind,
+	}
+	if m.int64ObservableInsts.HasKey(key) && len(callbacks) > 0 {
+		warnRepeatedObservableCallbacks(id)
+	}
+	return m.float64ObservableInsts.Lookup(key, func() (float64Observable, error) {
+		inst := newFloat64Observable(m, id.Kind, id.Name, id.Description, id.Unit)
+		for _, insert := range m.float64Resolver.inserters {
+			// Connect the measure functions for instruments in this pipeline with the
+			// callbacks for this pipeline.
+			in, err := insert.Instrument(id, insert.readerDefaultAggregation(id.Kind))
+			if err != nil {
+				return inst, err
+			}
+			// Drop aggregation
+			if len(in) == 0 {
+				inst.dropAggregation = true
+				continue
+			}
+			inst.appendMeasures(in)
+			for _, cback := range callbacks {
+				inst := float64Observer{measures: in}
+				fn := cback
+				insert.addCallback(func(ctx context.Context) error { return fn(ctx, inst) })
+			}
+		}
+		return inst, validateInstrumentName(id.Name)
+	})
+}
+
 // Float64ObservableCounter returns a new instrument identified by name and
 // configured with options. The instrument is used to asynchronously record
 // increasing float64 measurements once per a measurement collection cycle.
 // Only the measurements recorded during the collection cycle are exported.
+//
+// If Float64ObservableCounter is invoked repeatedly with the same Name,
+// Description, and Unit, only the first set of callbacks provided are used.
+// Use meter.RegisterCallback and Registration.Unregister to manage callbacks
+// if instrumentation can be created multiple times with different callbacks.
 func (m *meter) Float64ObservableCounter(name string, options ...metric.Float64ObservableCounterOption) (metric.Float64ObservableCounter, error) {
 	cfg := metric.NewFloat64ObservableCounterConfig(options...)
-	const kind = InstrumentKindObservableCounter
-	p := float64ObservProvider{m}
-	inst, err := p.lookup(kind, name, cfg.Description(), cfg.Unit())
-	if err != nil {
-		return nil, err
+	id := Instrument{
+		Name:        name,
+		Description: cfg.Description(),
+		Unit:        cfg.Unit(),
+		Kind:        InstrumentKindObservableCounter,
+		Scope:       m.scope,
 	}
-	p.registerCallbacks(inst, cfg.Callbacks())
-	return inst, validateInstrumentName(name)
+	return m.float64ObservableInstrument(id, cfg.Callbacks())
 }
 
 // Float64ObservableUpDownCounter returns a new instrument identified by name
@@ -218,14 +317,14 @@ func (m *meter) Float64ObservableCounter(name string, options ...metric.Float64O
 // measurements recorded during the collection cycle are exported.
 func (m *meter) Float64ObservableUpDownCounter(name string, options ...metric.Float64ObservableUpDownCounterOption) (metric.Float64ObservableUpDownCounter, error) {
 	cfg := metric.NewFloat64ObservableUpDownCounterConfig(options...)
-	const kind = InstrumentKindObservableUpDownCounter
-	p := float64ObservProvider{m}
-	inst, err := p.lookup(kind, name, cfg.Description(), cfg.Unit())
-	if err != nil {
-		return nil, err
+	id := Instrument{
+		Name:        name,
+		Description: cfg.Description(),
+		Unit:        cfg.Unit(),
+		Kind:        InstrumentKindObservableUpDownCounter,
+		Scope:       m.scope,
 	}
-	p.registerCallbacks(inst, cfg.Callbacks())
-	return inst, validateInstrumentName(name)
+	return m.float64ObservableInstrument(id, cfg.Callbacks())
 }
 
 // Float64ObservableGauge returns a new instrument identified by name and
@@ -234,14 +333,14 @@ func (m *meter) Float64ObservableUpDownCounter(name string, options ...metric.Fl
 // Only the measurements recorded during the collection cycle are exported.
 func (m *meter) Float64ObservableGauge(name string, options ...metric.Float64ObservableGaugeOption) (metric.Float64ObservableGauge, error) {
 	cfg := metric.NewFloat64ObservableGaugeConfig(options...)
-	const kind = InstrumentKindObservableGauge
-	p := float64ObservProvider{m}
-	inst, err := p.lookup(kind, name, cfg.Description(), cfg.Unit())
-	if err != nil {
-		return nil, err
+	id := Instrument{
+		Name:        name,
+		Description: cfg.Description(),
+		Unit:        cfg.Unit(),
+		Kind:        InstrumentKindObservableGauge,
+		Scope:       m.scope,
 	}
-	p.registerCallbacks(inst, cfg.Callbacks())
-	return inst, validateInstrumentName(name)
+	return m.float64ObservableInstrument(id, cfg.Callbacks())
 }
 
 func validateInstrumentName(name string) error {
@@ -273,6 +372,16 @@ func isAlphanumeric(c rune) bool {
 	return isAlpha(c) || ('0' <= c && c <= '9')
 }
 
+func warnRepeatedObservableCallbacks(id Instrument) {
+	inst := fmt.Sprintf(
+		"Instrument{Name: %q, Description: %q, Kind: %q, Unit: %q}",
+		id.Name, id.Description, "InstrumentKind"+id.Kind.String(), id.Unit,
+	)
+	global.Warn("Repeated observable instrument creation with callbacks. Ignoring new callbacks. Use meter.RegisterCallback and Registration.Unregister to manage callbacks.",
+		"instrument", inst,
+	)
+}
+
 // RegisterCallback registers f to be called each collection cycle so it will
 // make observations for insts during those cycles.
 //
@@ -389,12 +498,14 @@ func (r observer) ObserveFloat64(o metric.Float64Observable, v float64, opts ...
 	}
 
 	if _, registered := r.float64[oImpl.observablID]; !registered {
-		global.Error(errUnregObserver, "failed to record",
-			"name", oImpl.name,
-			"description", oImpl.description,
-			"unit", oImpl.unit,
-			"number", fmt.Sprintf("%T", float64(0)),
-		)
+		if !oImpl.dropAggregation {
+			global.Error(errUnregObserver, "failed to record",
+				"name", oImpl.name,
+				"description", oImpl.description,
+				"unit", oImpl.unit,
+				"number", fmt.Sprintf("%T", float64(0)),
+			)
+		}
 		return
 	}
 	c := metric.NewObserveConfig(opts)
@@ -422,12 +533,14 @@ func (r observer) ObserveInt64(o metric.Int64Observable, v int64, opts ...metric
 	}
 
 	if _, registered := r.int64[oImpl.observablID]; !registered {
-		global.Error(errUnregObserver, "failed to record",
-			"name", oImpl.name,
-			"description", oImpl.description,
-			"unit", oImpl.unit,
-			"number", fmt.Sprintf("%T", int64(0)),
-		)
+		if !oImpl.dropAggregation {
+			global.Error(errUnregObserver, "failed to record",
+				"name", oImpl.name,
+				"description", oImpl.description,
+				"unit", oImpl.unit,
+				"number", fmt.Sprintf("%T", int64(0)),
+			)
+		}
 		return
 	}
 	c := metric.NewObserveConfig(opts)
@@ -474,14 +587,28 @@ func (p int64InstProvider) histogramAggs(name string, cfg metric.Int64HistogramC
 
 // lookup returns the resolved instrumentImpl.
 func (p int64InstProvider) lookup(kind InstrumentKind, name, desc, u string) (*int64Inst, error) {
-	aggs, err := p.aggs(kind, name, desc, u)
-	return &int64Inst{measures: aggs}, err
+	return p.meter.int64Insts.Lookup(instID{
+		Name:        name,
+		Description: desc,
+		Unit:        u,
+		Kind:        kind,
+	}, func() (*int64Inst, error) {
+		aggs, err := p.aggs(kind, name, desc, u)
+		return &int64Inst{measures: aggs}, err
+	})
 }
 
 // lookupHistogram returns the resolved instrumentImpl.
 func (p int64InstProvider) lookupHistogram(name string, cfg metric.Int64HistogramConfig) (*int64Inst, error) {
-	aggs, err := p.histogramAggs(name, cfg)
-	return &int64Inst{measures: aggs}, err
+	return p.meter.int64Insts.Lookup(instID{
+		Name:        name,
+		Description: cfg.Description(),
+		Unit:        cfg.Unit(),
+		Kind:        InstrumentKindHistogram,
+	}, func() (*int64Inst, error) {
+		aggs, err := p.histogramAggs(name, cfg)
+		return &int64Inst{measures: aggs}, err
+	})
 }
 
 // float64InstProvider provides float64 OpenTelemetry instruments.
@@ -518,42 +645,33 @@ func (p float64InstProvider) histogramAggs(name string, cfg metric.Float64Histog
 
 // lookup returns the resolved instrumentImpl.
 func (p float64InstProvider) lookup(kind InstrumentKind, name, desc, u string) (*float64Inst, error) {
-	aggs, err := p.aggs(kind, name, desc, u)
-	return &float64Inst{measures: aggs}, err
+	return p.meter.float64Insts.Lookup(instID{
+		Name:        name,
+		Description: desc,
+		Unit:        u,
+		Kind:        kind,
+	}, func() (*float64Inst, error) {
+		aggs, err := p.aggs(kind, name, desc, u)
+		return &float64Inst{measures: aggs}, err
+	})
 }
 
 // lookupHistogram returns the resolved instrumentImpl.
 func (p float64InstProvider) lookupHistogram(name string, cfg metric.Float64HistogramConfig) (*float64Inst, error) {
-	aggs, err := p.histogramAggs(name, cfg)
-	return &float64Inst{measures: aggs}, err
-}
-
-type int64ObservProvider struct{ *meter }
-
-func (p int64ObservProvider) lookup(kind InstrumentKind, name, desc, u string) (int64Observable, error) {
-	aggs, err := (int64InstProvider)(p).aggs(kind, name, desc, u)
-	return newInt64Observable(p.meter, kind, name, desc, u, aggs), err
-}
-
-func (p int64ObservProvider) registerCallbacks(inst int64Observable, cBacks []metric.Int64Callback) {
-	if inst.observable == nil || len(inst.measures) == 0 {
-		// Drop aggregator.
-		return
-	}
-
-	for _, cBack := range cBacks {
-		p.pipes.registerCallback(p.callback(inst, cBack))
-	}
-}
-
-func (p int64ObservProvider) callback(i int64Observable, f metric.Int64Callback) func(context.Context) error {
-	inst := int64Observer{int64Observable: i}
-	return func(ctx context.Context) error { return f(ctx, inst) }
+	return p.meter.float64Insts.Lookup(instID{
+		Name:        name,
+		Description: cfg.Description(),
+		Unit:        cfg.Unit(),
+		Kind:        InstrumentKindHistogram,
+	}, func() (*float64Inst, error) {
+		aggs, err := p.histogramAggs(name, cfg)
+		return &float64Inst{measures: aggs}, err
+	})
 }
 
 type int64Observer struct {
 	embedded.Int64Observer
-	int64Observable
+	measures[int64]
 }
 
 func (o int64Observer) Observe(val int64, opts ...metric.ObserveOption) {
@@ -561,32 +679,9 @@ func (o int64Observer) Observe(val int64, opts ...metric.ObserveOption) {
 	o.observe(val, c.Attributes())
 }
 
-type float64ObservProvider struct{ *meter }
-
-func (p float64ObservProvider) lookup(kind InstrumentKind, name, desc, u string) (float64Observable, error) {
-	aggs, err := (float64InstProvider)(p).aggs(kind, name, desc, u)
-	return newFloat64Observable(p.meter, kind, name, desc, u, aggs), err
-}
-
-func (p float64ObservProvider) registerCallbacks(inst float64Observable, cBacks []metric.Float64Callback) {
-	if inst.observable == nil || len(inst.measures) == 0 {
-		// Drop aggregator.
-		return
-	}
-
-	for _, cBack := range cBacks {
-		p.pipes.registerCallback(p.callback(inst, cBack))
-	}
-}
-
-func (p float64ObservProvider) callback(i float64Observable, f metric.Float64Callback) func(context.Context) error {
-	inst := float64Observer{float64Observable: i}
-	return func(ctx context.Context) error { return f(ctx, inst) }
-}
-
 type float64Observer struct {
 	embedded.Float64Observer
-	float64Observable
+	measures[float64]
 }
 
 func (o float64Observer) Observe(val float64, opts ...metric.ObserveOption) {
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/metricdata/data.go b/vendor/go.opentelemetry.io/otel/sdk/metric/metricdata/data.go
index 995d42b38..32c17934f 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/metricdata/data.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/metricdata/data.go
@@ -15,6 +15,7 @@
 package metricdata // import "go.opentelemetry.io/otel/sdk/metric/metricdata"
 
 import (
+	"encoding/json"
 	"time"
 
 	"go.opentelemetry.io/otel/attribute"
@@ -211,6 +212,19 @@ type Extrema[N int64 | float64] struct {
 	valid bool
 }
 
+// MarshalText converts the Extrema value to text.
+func (e Extrema[N]) MarshalText() ([]byte, error) {
+	if !e.valid {
+		return json.Marshal(nil)
+	}
+	return json.Marshal(e.value)
+}
+
+// MarshalJSON converts the Extrema value to JSON number.
+func (e *Extrema[N]) MarshalJSON() ([]byte, error) {
+	return e.MarshalText()
+}
+
 // NewExtrema returns an Extrema set to v.
 func NewExtrema[N int64 | float64](v N) Extrema[N] {
 	return Extrema[N]{value: v, valid: true}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/pipeline.go b/vendor/go.opentelemetry.io/otel/sdk/metric/pipeline.go
index 48abcc8a7..da39ab961 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/pipeline.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/pipeline.go
@@ -29,6 +29,7 @@ import (
 	"go.opentelemetry.io/otel/sdk/instrumentation"
 	"go.opentelemetry.io/otel/sdk/metric/internal"
 	"go.opentelemetry.io/otel/sdk/metric/internal/aggregate"
+	"go.opentelemetry.io/otel/sdk/metric/internal/x"
 	"go.opentelemetry.io/otel/sdk/metric/metricdata"
 	"go.opentelemetry.io/otel/sdk/resource"
 )
@@ -93,14 +94,6 @@ func (p *pipeline) addSync(scope instrumentation.Scope, iSync instrumentSync) {
 	p.aggregations[scope] = append(p.aggregations[scope], iSync)
 }
 
-// addCallback registers a single instrument callback to be run when
-// `produce()` is called.
-func (p *pipeline) addCallback(cback func(context.Context) error) {
-	p.Lock()
-	defer p.Unlock()
-	p.callbacks = append(p.callbacks, cback)
-}
-
 type multiCallback func(context.Context) error
 
 // addMultiCallback registers a multi-instrument callback to be run when
@@ -281,6 +274,14 @@ func (i *inserter[N]) Instrument(inst Instrument, readerAggregation Aggregation)
 	return measures, errs.errorOrNil()
 }
 
+// addCallback registers a single instrument callback to be run when
+// `produce()` is called.
+func (i *inserter[N]) addCallback(cback func(context.Context) error) {
+	i.pipeline.Lock()
+	defer i.pipeline.Unlock()
+	i.pipeline.callbacks = append(i.pipeline.callbacks, cback)
+}
+
 var aggIDCount uint64
 
 // aggVal is the cached value in an aggregators cache.
@@ -358,9 +359,16 @@ func (i *inserter[N]) cachedAggregator(scope instrumentation.Scope, kind Instrum
 	normID := id.normalize()
 	cv := i.aggregators.Lookup(normID, func() aggVal[N] {
 		b := aggregate.Builder[N]{
-			Temporality: i.pipeline.reader.temporality(kind),
+			Temporality:   i.pipeline.reader.temporality(kind),
+			ReservoirFunc: reservoirFunc[N](stream.Aggregation),
 		}
 		b.Filter = stream.AttributeFilter
+		// A value less than or equal to zero will disable the aggregation
+		// limits for the builder (an all the created aggregates).
+		// CardinalityLimit.Lookup returns 0 by default if unset (or
+		// unrecognized input). Use that value directly.
+		b.AggregationLimit, _ = x.CardinalityLimit.Lookup()
+
 		in, out, err := i.aggregateFunc(b, stream.Aggregation, kind)
 		if err != nil {
 			return aggVal[N]{0, nil, err}
@@ -557,12 +565,6 @@ func newPipelines(res *resource.Resource, readers []Reader, views []View) pipeli
 	return pipes
 }
 
-func (p pipelines) registerCallback(cback func(context.Context) error) {
-	for _, pipe := range p {
-		pipe.addCallback(cback)
-	}
-}
-
 func (p pipelines) registerMultiCallback(c multiCallback) metric.Registration {
 	unregs := make([]func(), len(p))
 	for i, pipe := range p {
diff --git a/vendor/go.opentelemetry.io/otel/sdk/metric/version.go b/vendor/go.opentelemetry.io/otel/sdk/metric/version.go
index 4437747f2..310fa5a53 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/metric/version.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/metric/version.go
@@ -16,5 +16,5 @@ package metric // import "go.opentelemetry.io/otel/sdk/metric"
 
 // version is the current release version of the metric SDK in use.
 func version() string {
-	return "1.20.0"
+	return "1.24.0"
 }
diff --git a/vendor/go.opentelemetry.io/otel/sdk/resource/auto.go b/vendor/go.opentelemetry.io/otel/sdk/resource/auto.go
index 4279013be..aed756c5e 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/resource/auto.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/resource/auto.go
@@ -41,8 +41,20 @@ type Detector interface {
 	// must never be done outside of a new major release.
 }
 
-// Detect calls all input detectors sequentially and merges each result with the previous one.
-// It returns the merged error too.
+// Detect returns a new [Resource] merged from all the Resources each of the
+// detectors produces. Each of the detectors are called sequentially, in the
+// order they are passed, merging the produced resource into the previous.
+//
+// This may return a partial Resource along with an error containing
+// [ErrPartialResource] if that error is returned from a detector. It may also
+// return a merge-conflicting Resource along with an error containing
+// [ErrSchemaURLConflict] if merging Resources from different detectors results
+// in a schema URL conflict. It is up to the caller to determine if this
+// returned Resource should be used or not.
+//
+// If one of the detectors returns an error that is not [ErrPartialResource],
+// the resource produced by the detector will not be merged and the returned
+// error will wrap that detector's error.
 func Detect(ctx context.Context, detectors ...Detector) (*Resource, error) {
 	r := new(Resource)
 	return r, detect(ctx, r, detectors)
@@ -50,6 +62,10 @@ func Detect(ctx context.Context, detectors ...Detector) (*Resource, error) {
 
 // detect runs all detectors using ctx and merges the result into res. This
 // assumes res is allocated and not nil, it will panic otherwise.
+//
+// If the detectors or merging resources produces any errors (i.e.
+// [ErrPartialResource] [ErrSchemaURLConflict]), a single error wrapping all of
+// these errors will be returned. Otherwise, nil is returned.
 func detect(ctx context.Context, res *Resource, detectors []Detector) error {
 	var (
 		r    *Resource
@@ -78,6 +94,11 @@ func detect(ctx context.Context, res *Resource, detectors []Detector) error {
 	if len(errs) == 0 {
 		return nil
 	}
+	if errors.Is(errs, ErrSchemaURLConflict) {
+		// If there has been a merge conflict, ensure the resource has no
+		// schema URL.
+		res.schemaURL = ""
+	}
 	return errs
 }
 
diff --git a/vendor/go.opentelemetry.io/otel/sdk/resource/builtin.go b/vendor/go.opentelemetry.io/otel/sdk/resource/builtin.go
index c63a0dd1f..6a2c08293 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/resource/builtin.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/resource/builtin.go
@@ -22,7 +22,7 @@ import (
 
 	"go.opentelemetry.io/otel/attribute"
 	"go.opentelemetry.io/otel/sdk"
-	semconv "go.opentelemetry.io/otel/semconv/v1.21.0"
+	semconv "go.opentelemetry.io/otel/semconv/v1.24.0"
 )
 
 type (
diff --git a/vendor/go.opentelemetry.io/otel/sdk/resource/container.go b/vendor/go.opentelemetry.io/otel/sdk/resource/container.go
index 3d5362282..c1b47193f 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/resource/container.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/resource/container.go
@@ -22,14 +22,14 @@ import (
 	"os"
 	"regexp"
 
-	semconv "go.opentelemetry.io/otel/semconv/v1.21.0"
+	semconv "go.opentelemetry.io/otel/semconv/v1.24.0"
 )
 
 type containerIDProvider func() (string, error)
 
 var (
 	containerID         containerIDProvider = getContainerIDFromCGroup
-	cgroupContainerIDRe                     = regexp.MustCompile(`^.*/(?:.*-)?([0-9a-f]+)(?:\.|\s*$)`)
+	cgroupContainerIDRe                     = regexp.MustCompile(`^.*/(?:.*[-:])?([0-9a-f]+)(?:\.|\s*$)`)
 )
 
 type cgroupContainerIDDetector struct{}
diff --git a/vendor/go.opentelemetry.io/otel/sdk/resource/env.go b/vendor/go.opentelemetry.io/otel/sdk/resource/env.go
index e29ae563a..be4cbe423 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/resource/env.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/resource/env.go
@@ -23,7 +23,7 @@ import (
 
 	"go.opentelemetry.io/otel"
 	"go.opentelemetry.io/otel/attribute"
-	semconv "go.opentelemetry.io/otel/semconv/v1.21.0"
+	semconv "go.opentelemetry.io/otel/semconv/v1.24.0"
 )
 
 const (
diff --git a/vendor/go.opentelemetry.io/otel/sdk/resource/host_id.go b/vendor/go.opentelemetry.io/otel/sdk/resource/host_id.go
index fb1ebf2ca..f579329c2 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/resource/host_id.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/resource/host_id.go
@@ -19,7 +19,7 @@ import (
 	"errors"
 	"strings"
 
-	semconv "go.opentelemetry.io/otel/semconv/v1.21.0"
+	semconv "go.opentelemetry.io/otel/semconv/v1.24.0"
 )
 
 type hostIDProvider func() (string, error)
diff --git a/vendor/go.opentelemetry.io/otel/sdk/resource/os.go b/vendor/go.opentelemetry.io/otel/sdk/resource/os.go
index 0cbd55973..8fbf071c1 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/resource/os.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/resource/os.go
@@ -19,7 +19,7 @@ import (
 	"strings"
 
 	"go.opentelemetry.io/otel/attribute"
-	semconv "go.opentelemetry.io/otel/semconv/v1.21.0"
+	semconv "go.opentelemetry.io/otel/semconv/v1.24.0"
 )
 
 type osDescriptionProvider func() (string, error)
diff --git a/vendor/go.opentelemetry.io/otel/sdk/resource/process.go b/vendor/go.opentelemetry.io/otel/sdk/resource/process.go
index ecdd11dd7..739ea4512 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/resource/process.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/resource/process.go
@@ -22,7 +22,7 @@ import (
 	"path/filepath"
 	"runtime"
 
-	semconv "go.opentelemetry.io/otel/semconv/v1.21.0"
+	semconv "go.opentelemetry.io/otel/semconv/v1.24.0"
 )
 
 type (
diff --git a/vendor/go.opentelemetry.io/otel/sdk/resource/resource.go b/vendor/go.opentelemetry.io/otel/sdk/resource/resource.go
index 176ff1066..cb1ee0a9c 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/resource/resource.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/resource/resource.go
@@ -17,6 +17,7 @@ package resource // import "go.opentelemetry.io/otel/sdk/resource"
 import (
 	"context"
 	"errors"
+	"fmt"
 	"sync"
 
 	"go.opentelemetry.io/otel"
@@ -40,9 +41,20 @@ var (
 	defaultResourceOnce sync.Once
 )
 
-var errMergeConflictSchemaURL = errors.New("cannot merge resource due to conflicting Schema URL")
+// ErrSchemaURLConflict is an error returned when two Resources are merged
+// together that contain different, non-empty, schema URLs.
+var ErrSchemaURLConflict = errors.New("conflicting Schema URL")
 
-// New returns a Resource combined from the user-provided detectors.
+// New returns a [Resource] built using opts.
+//
+// This may return a partial Resource along with an error containing
+// [ErrPartialResource] if options that provide a [Detector] are used and that
+// error is returned from one or more of the Detectors. It may also return a
+// merge-conflict Resource along with an error containing
+// [ErrSchemaURLConflict] if merging Resources from the opts results in a
+// schema URL conflict (see [Resource.Merge] for more information). It is up to
+// the caller to determine if this returned Resource should be used or not
+// based on these errors.
 func New(ctx context.Context, opts ...Option) (*Resource, error) {
 	cfg := config{}
 	for _, opt := range opts {
@@ -98,7 +110,7 @@ func (r *Resource) String() string {
 	return r.attrs.Encoded(attribute.DefaultEncoder())
 }
 
-// MarshalLog is the marshaling function used by the logging system to represent this exporter.
+// MarshalLog is the marshaling function used by the logging system to represent this Resource.
 func (r *Resource) MarshalLog() interface{} {
 	return struct {
 		Attributes attribute.Set
@@ -146,16 +158,29 @@ func (r *Resource) Equal(eq *Resource) bool {
 	return r.Equivalent() == eq.Equivalent()
 }
 
-// Merge creates a new resource by combining resource a and b.
+// Merge creates a new [Resource] by merging a and b.
+//
+// If there are common keys between a and b, then the value from b will
+// overwrite the value from a, even if b's value is empty.
 //
-// If there are common keys between resource a and b, then the value
-// from resource b will overwrite the value from resource a, even
-// if resource b's value is empty.
+// The SchemaURL of the resources will be merged according to the
+// [OpenTelemetry specification rules]:
 //
-// The SchemaURL of the resources will be merged according to the spec rules:
-// https://github.com/open-telemetry/opentelemetry-specification/blob/v1.20.0/specification/resource/sdk.md#merge
-// If the resources have different non-empty schemaURL an empty resource and an error
-// will be returned.
+//   - If a's schema URL is empty then the returned Resource's schema URL will
+//     be set to the schema URL of b,
+//   - Else if b's schema URL is empty then the returned Resource's schema URL
+//     will be set to the schema URL of a,
+//   - Else if the schema URLs of a and b are the same then that will be the
+//     schema URL of the returned Resource,
+//   - Else this is a merging error. If the resources have different,
+//     non-empty, schema URLs an error containing [ErrSchemaURLConflict] will
+//     be returned with the merged Resource. The merged Resource will have an
+//     empty schema URL. It may be the case that some unintended attributes
+//     have been overwritten or old semantic conventions persisted in the
+//     returned Resource. It is up to the caller to determine if this returned
+//     Resource should be used or not.
+//
+// [OpenTelemetry specification rules]: https://github.com/open-telemetry/opentelemetry-specification/blob/v1.20.0/specification/resource/sdk.md#merge
 func Merge(a, b *Resource) (*Resource, error) {
 	if a == nil && b == nil {
 		return Empty(), nil
@@ -167,19 +192,6 @@ func Merge(a, b *Resource) (*Resource, error) {
 		return a, nil
 	}
 
-	// Merge the schema URL.
-	var schemaURL string
-	switch true {
-	case a.schemaURL == "":
-		schemaURL = b.schemaURL
-	case b.schemaURL == "":
-		schemaURL = a.schemaURL
-	case a.schemaURL == b.schemaURL:
-		schemaURL = a.schemaURL
-	default:
-		return Empty(), errMergeConflictSchemaURL
-	}
-
 	// Note: 'b' attributes will overwrite 'a' with last-value-wins in attribute.Key()
 	// Meaning this is equivalent to: append(a.Attributes(), b.Attributes()...)
 	mi := attribute.NewMergeIterator(b.Set(), a.Set())
@@ -187,8 +199,23 @@ func Merge(a, b *Resource) (*Resource, error) {
 	for mi.Next() {
 		combine = append(combine, mi.Attribute())
 	}
-	merged := NewWithAttributes(schemaURL, combine...)
-	return merged, nil
+
+	switch {
+	case a.schemaURL == "":
+		return NewWithAttributes(b.schemaURL, combine...), nil
+	case b.schemaURL == "":
+		return NewWithAttributes(a.schemaURL, combine...), nil
+	case a.schemaURL == b.schemaURL:
+		return NewWithAttributes(a.schemaURL, combine...), nil
+	}
+	// Return the merged resource with an appropriate error. It is up to
+	// the user to decide if the returned resource can be used or not.
+	return NewSchemaless(combine...), fmt.Errorf(
+		"%w: %s and %s",
+		ErrSchemaURLConflict,
+		a.schemaURL,
+		b.schemaURL,
+	)
 }
 
 // Empty returns an instance of Resource with no attributes. It is
diff --git a/vendor/go.opentelemetry.io/otel/sdk/trace/batch_span_processor.go b/vendor/go.opentelemetry.io/otel/sdk/trace/batch_span_processor.go
index c9c7effbf..fca26f2e7 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/trace/batch_span_processor.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/trace/batch_span_processor.go
@@ -406,7 +406,7 @@ func (bsp *batchSpanProcessor) enqueueDrop(ctx context.Context, sd ReadOnlySpan)
 	return false
 }
 
-// MarshalLog is the marshaling function used by the logging system to represent this exporter.
+// MarshalLog is the marshaling function used by the logging system to represent this Span Processor.
 func (bsp *batchSpanProcessor) MarshalLog() interface{} {
 	return struct {
 		Type         string
diff --git a/vendor/go.opentelemetry.io/otel/sdk/trace/provider.go b/vendor/go.opentelemetry.io/otel/sdk/trace/provider.go
index 7d46c4b48..b1ac60846 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/trace/provider.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/trace/provider.go
@@ -55,7 +55,7 @@ type tracerProviderConfig struct {
 	resource *resource.Resource
 }
 
-// MarshalLog is the marshaling function used by the logging system to represent this exporter.
+// MarshalLog is the marshaling function used by the logging system to represent this Provider.
 func (cfg tracerProviderConfig) MarshalLog() interface{} {
 	return struct {
 		SpanProcessors  []SpanProcessor
diff --git a/vendor/go.opentelemetry.io/otel/sdk/trace/span.go b/vendor/go.opentelemetry.io/otel/sdk/trace/span.go
index 36dbf6776..85bc702a0 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/trace/span.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/trace/span.go
@@ -30,7 +30,7 @@ import (
 	"go.opentelemetry.io/otel/sdk/instrumentation"
 	"go.opentelemetry.io/otel/sdk/internal"
 	"go.opentelemetry.io/otel/sdk/resource"
-	semconv "go.opentelemetry.io/otel/semconv/v1.21.0"
+	semconv "go.opentelemetry.io/otel/semconv/v1.24.0"
 	"go.opentelemetry.io/otel/trace"
 	"go.opentelemetry.io/otel/trace/embedded"
 )
@@ -208,6 +208,16 @@ func (s *recordingSpan) SetStatus(code codes.Code, description string) {
 	s.status = status
 }
 
+// ensureAttributesCapacity inlines functionality from slices.Grow
+// so that we can avoid needing to import golang.org/x/exp for go1.20.
+// Once support for go1.20 is dropped, we can use slices.Grow available since go1.21 instead.
+// Tracking issue: https://github.com/open-telemetry/opentelemetry-go/issues/4819.
+func (s *recordingSpan) ensureAttributesCapacity(minCapacity int) {
+	if n := minCapacity - cap(s.attributes); n > 0 {
+		s.attributes = append(s.attributes[:cap(s.attributes)], make([]attribute.KeyValue, n)...)[:len(s.attributes)]
+	}
+}
+
 // SetAttributes sets attributes of this span.
 //
 // If a key from attributes already exists the value associated with that key
@@ -242,6 +252,7 @@ func (s *recordingSpan) SetAttributes(attributes ...attribute.KeyValue) {
 
 	// Otherwise, add without deduplication. When attributes are read they
 	// will be deduplicated, optimizing the operation.
+	s.ensureAttributesCapacity(len(s.attributes) + len(attributes))
 	for _, a := range attributes {
 		if !a.Valid() {
 			// Drop all invalid attributes.
@@ -277,6 +288,12 @@ func (s *recordingSpan) addOverCapAttrs(limit int, attrs []attribute.KeyValue) {
 
 	// Now that s.attributes is deduplicated, adding unique attributes up to
 	// the capacity of s will not over allocate s.attributes.
+	if sum := len(attrs) + len(s.attributes); sum < limit {
+		// After support for go1.20 is dropped, simplify if-else to min(sum, limit).
+		s.ensureAttributesCapacity(sum)
+	} else {
+		s.ensureAttributesCapacity(limit)
+	}
 	for _, a := range attrs {
 		if !a.Valid() {
 			// Drop all invalid attributes.
diff --git a/vendor/go.opentelemetry.io/otel/sdk/version.go b/vendor/go.opentelemetry.io/otel/sdk/version.go
index 7048c788e..42de0b9a7 100644
--- a/vendor/go.opentelemetry.io/otel/sdk/version.go
+++ b/vendor/go.opentelemetry.io/otel/sdk/version.go
@@ -16,5 +16,5 @@ package sdk // import "go.opentelemetry.io/otel/sdk"
 
 // Version is the current release version of the OpenTelemetry SDK in use.
 func Version() string {
-	return "1.20.0"
+	return "1.24.0"
 }