1
0
mirror of https://github.com/open-telemetry/opentelemetry-go.git synced 2025-11-25 22:41:46 +02:00
Files
opentelemetry-go/sdk/metric/exemplar.go

78 lines
3.3 KiB
Go
Raw Normal View History

// Copyright The OpenTelemetry Authors
// SPDX-License-Identifier: Apache-2.0
package metric // import "go.opentelemetry.io/otel/sdk/metric"
import (
"runtime"
"go.opentelemetry.io/otel/attribute"
"go.opentelemetry.io/otel/sdk/metric/exemplar"
"go.opentelemetry.io/otel/sdk/metric/internal/aggregate"
)
Add selector of exemplar reservoir providers to metric.Stream configuration (#5861) Resolve https://github.com/open-telemetry/opentelemetry-go/issues/5249 ### Spec > exemplar_reservoir: A functional type that generates an exemplar reservoir a MeterProvider will use when storing exemplars. This functional type needs to be a factory or callback similar to aggregation selection functionality which allows different reservoirs to be chosen by the aggregation. > Users can provide an exemplar_reservoir, but it is up to their discretion. Therefore, the stream configuration parameter needs to be structured to accept an exemplar_reservoir, but MUST NOT obligate a user to provide one. If the user does not provide an exemplar_reservoir value, the MeterProvider MUST apply a [default exemplar reservoir](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#exemplar-defaults). Also, > the reservoir MUST be given the Attributes associated with its timeseries point either at construction so that additional sampling performed by the reservoir has access to all attributes from a measurement in the "offer" method. ### Changes In sdk/metric/exemplar, add: * `exemplar.ReservoirProvider` * `exemplar.HistogramReservoirProvider` * `exemplar.FixedSizeReservoirProvider` In sdk/metric, add: * `metric.ExemplarReservoirProviderSelector` (func Aggregation -> ReservoirProvider) * `metric.DefaultExemplarReservoirProviderSelector` (our default implementation) * `ExemplarReservoirProviderSelector` added to `metric.Stream` Note: the only required public types are `metric.ExemplarReservoirProviderSelector` and `ExemplarReservoirProviderSelector` in `metric.Stream`. Others are for convenience and readability. ### Alternatives considered #### Add ExemplarReservoirProvider directly to metric.Stream, instead of ExemplarReservoirProviderSelector This would mean users can configure a `func() exemplar.Reservoir` instead of a `func(Aggregation) func() exemplar.Reservoir`. I don't think this complies with the statement: `This functional type needs to be a factory or callback similar to aggregation selection functionality which allows different reservoirs to be chosen by the aggregation.`. I'm interpreting "allows different reservoirs to be chosen by the aggregation" as meaning "allows different reservoirs to be chosen **based on the** aggregation", rather than meaning that the aggregation is somehow choosing the reservoir. ### Future work There is some refactoring I plan to do after this to simplify the interaction between the internal/aggregate and exemplar package. I've omitted that from this PR to keep the diff smaller. --------- Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com> Co-authored-by: Robert Pająk <pellared@hotmail.com>
2024-10-18 09:05:10 -04:00
// ExemplarReservoirProviderSelector selects the
// [exemplar.ReservoirProvider] to use
// based on the [Aggregation] of the metric.
type ExemplarReservoirProviderSelector func(Aggregation) exemplar.ReservoirProvider
// reservoirFunc returns the appropriately configured exemplar reservoir
// creation func based on the passed InstrumentKind and filter configuration.
func reservoirFunc[N int64 | float64](
provider exemplar.ReservoirProvider,
filter exemplar.Filter,
) func(attribute.Set) aggregate.FilteredExemplarReservoir[N] {
return func(attrs attribute.Set) aggregate.FilteredExemplarReservoir[N] {
return aggregate.NewFilteredExemplarReservoir[N](filter, provider(attrs))
}
}
Add selector of exemplar reservoir providers to metric.Stream configuration (#5861) Resolve https://github.com/open-telemetry/opentelemetry-go/issues/5249 ### Spec > exemplar_reservoir: A functional type that generates an exemplar reservoir a MeterProvider will use when storing exemplars. This functional type needs to be a factory or callback similar to aggregation selection functionality which allows different reservoirs to be chosen by the aggregation. > Users can provide an exemplar_reservoir, but it is up to their discretion. Therefore, the stream configuration parameter needs to be structured to accept an exemplar_reservoir, but MUST NOT obligate a user to provide one. If the user does not provide an exemplar_reservoir value, the MeterProvider MUST apply a [default exemplar reservoir](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#exemplar-defaults). Also, > the reservoir MUST be given the Attributes associated with its timeseries point either at construction so that additional sampling performed by the reservoir has access to all attributes from a measurement in the "offer" method. ### Changes In sdk/metric/exemplar, add: * `exemplar.ReservoirProvider` * `exemplar.HistogramReservoirProvider` * `exemplar.FixedSizeReservoirProvider` In sdk/metric, add: * `metric.ExemplarReservoirProviderSelector` (func Aggregation -> ReservoirProvider) * `metric.DefaultExemplarReservoirProviderSelector` (our default implementation) * `ExemplarReservoirProviderSelector` added to `metric.Stream` Note: the only required public types are `metric.ExemplarReservoirProviderSelector` and `ExemplarReservoirProviderSelector` in `metric.Stream`. Others are for convenience and readability. ### Alternatives considered #### Add ExemplarReservoirProvider directly to metric.Stream, instead of ExemplarReservoirProviderSelector This would mean users can configure a `func() exemplar.Reservoir` instead of a `func(Aggregation) func() exemplar.Reservoir`. I don't think this complies with the statement: `This functional type needs to be a factory or callback similar to aggregation selection functionality which allows different reservoirs to be chosen by the aggregation.`. I'm interpreting "allows different reservoirs to be chosen by the aggregation" as meaning "allows different reservoirs to be chosen **based on the** aggregation", rather than meaning that the aggregation is somehow choosing the reservoir. ### Future work There is some refactoring I plan to do after this to simplify the interaction between the internal/aggregate and exemplar package. I've omitted that from this PR to keep the diff smaller. --------- Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com> Co-authored-by: Robert Pająk <pellared@hotmail.com>
2024-10-18 09:05:10 -04:00
// DefaultExemplarReservoirProviderSelector returns the default
// [exemplar.ReservoirProvider] for the
// provided [Aggregation].
//
// For explicit bucket histograms with more than 1 bucket, it uses the
// [exemplar.HistogramReservoirProvider].
// For exponential histograms, it uses the
// [exemplar.FixedSizeReservoirProvider]
// with a size of min(20, max_buckets).
// For all other aggregations, it uses the
// [exemplar.FixedSizeReservoirProvider]
// with a size equal to the number of CPUs.
//
// Exemplar default reservoirs MAY change in a minor version bump. No
// guarantees are made on the shape or statistical properties of returned
// exemplars.
func DefaultExemplarReservoirProviderSelector(agg Aggregation) exemplar.ReservoirProvider {
Move time.Now call into exemplar reservoir to improve performance (#5545) Part of addressing https://github.com/open-telemetry/opentelemetry-go/issues/5542. ### Motivation This removes the `time.Now()` call from filtered-out Exemplars by only invoking `time.Now()` after the filtering decision is made. This improvement is especially noticeable for measurements without any attributes. ``` goos: linux goarch: amd64 pkg: go.opentelemetry.io/otel/sdk/metric cpu: AMD EPYC 7B12 │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ SyncMeasure/NoView/Int64Counter/Attributes/0-24 158.20n ± 4% 99.83n ± 1% -36.90% (p=0.000 n=10) SyncMeasure/NoView/Int64Counter/Attributes/1-24 333.3n ± 4% 274.8n ± 1% -17.55% (p=0.000 n=10) SyncMeasure/NoView/Int64Counter/Attributes/10-24 1.640µ ± 1% 1.600µ ± 1% -2.41% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/0-24 159.0n ± 3% 101.3n ± 0% -36.27% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/1-24 340.0n ± 2% 272.0n ± 1% -20.00% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/10-24 1.661µ ± 1% 1.597µ ± 0% -3.85% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/0-24 159.8n ± 1% 103.1n ± 0% -35.50% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/1-24 339.5n ± 1% 273.1n ± 0% -19.57% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/10-24 1.656µ ± 0% 1.589µ ± 0% -4.05% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/0-24 159.3n ± 2% 100.8n ± 0% -36.74% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/1-24 337.9n ± 2% 271.8n ± 1% -19.55% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/10-24 1.657µ ± 0% 1.593µ ± 1% -3.83% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/0-24 144.65n ± 4% 89.38n ± 0% -38.21% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/1-24 235.7n ± 2% 183.5n ± 0% -22.15% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/10-24 900.8n ± 1% 836.8n ± 0% -7.10% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/0-24 145.60n ± 5% 93.48n ± 1% -35.80% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/1-24 240.9n ± 1% 183.0n ± 0% -24.06% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/10-24 905.6n ± 1% 826.3n ± 0% -8.76% (p=0.000 n=10) SyncMeasure/DropView/Int64Counter/Attributes/0-24 20.33n ± 0% 20.35n ± 0% ~ (p=0.302 n=10) SyncMeasure/DropView/Int64Counter/Attributes/1-24 26.46n ± 0% 26.45n ± 1% ~ (p=0.868 n=10) SyncMeasure/DropView/Int64Counter/Attributes/10-24 26.50n ± 0% 26.47n ± 0% ~ (p=0.208 n=10) SyncMeasure/DropView/Float64Counter/Attributes/0-24 20.34n ± 1% 20.27n ± 0% -0.34% (p=0.009 n=10) SyncMeasure/DropView/Float64Counter/Attributes/1-24 26.55n ± 0% 26.60n ± 1% ~ (p=0.109 n=10) SyncMeasure/DropView/Float64Counter/Attributes/10-24 26.59n ± 1% 26.57n ± 1% ~ (p=0.926 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/0-24 20.38n ± 1% 20.38n ± 0% ~ (p=0.725 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/1-24 26.39n ± 0% 26.44n ± 0% ~ (p=0.238 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/10-24 26.52n ± 0% 26.42n ± 0% -0.36% (p=0.049 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/0-24 20.30n ± 0% 20.25n ± 0% ~ (p=0.196 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/1-24 26.57n ± 0% 26.54n ± 1% ~ (p=0.540 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/10-24 26.57n ± 0% 26.51n ± 1% ~ (p=0.643 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/0-24 20.37n ± 0% 20.36n ± 1% ~ (p=1.000 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/1-24 26.41n ± 0% 26.50n ± 0% +0.32% (p=0.007 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/10-24 26.44n ± 0% 26.55n ± 1% +0.42% (p=0.012 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/0-24 20.30n ± 0% 20.45n ± 0% +0.74% (p=0.000 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/1-24 26.52n ± 0% 26.48n ± 0% ~ (p=0.127 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/10-24 26.55n ± 0% 26.48n ± 0% -0.26% (p=0.002 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/0-24 170.5n ± 2% 110.8n ± 0% -35.03% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/1-24 402.5n ± 1% 331.5n ± 1% -17.64% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/10-24 1.363µ ± 1% 1.281µ ± 1% -6.02% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/0-24 170.6n ± 1% 111.5n ± 1% -34.64% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/1-24 397.1n ± 1% 335.9n ± 0% -15.41% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/10-24 1.371µ ± 1% 1.279µ ± 1% -6.71% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/0-24 170.1n ± 1% 112.2n ± 0% -34.09% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/1-24 397.5n ± 1% 330.2n ± 0% -16.93% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/10-24 1.371µ ± 1% 1.289µ ± 1% -5.95% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/0-24 171.4n ± 2% 112.9n ± 0% -34.13% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/1-24 397.0n ± 3% 336.4n ± 0% -15.24% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/10-24 1.383µ ± 1% 1.305µ ± 1% -5.61% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Histogram/Attributes/0-24 157.30n ± 2% 98.58n ± 1% -37.33% (p=0.000 n=6+10) ``` ### Changes * Introduce `exemplar.Filter`, which is a filter function based on the context. It will not be user-facing, so we can always add other parameters later if needed. * Introduce `exemplar.FilteredReservoir`, which is similar to a reservoir, except it does not receive a timestamp. It gets the current time after the filter decision has been made. It uses generics to avoid the call to exemplar.NewValue(), since it is internal-only. * The `exemplar.Reservoir` is left as-is, so that it can be made public when exemplars are stable. It still includes a timestamp argument. * Unit tests are updated to expect a much lower number of calls to time.Now * `exemplar.Drop` is now an `exemplar.FilteredReservoir` instead of a `Reservoir`, since we don't need a Reservoir to store things in if the measurement is always dropped. Co-authored-by: Sam Xie <sam@samxie.me> Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
2024-07-01 12:36:11 -04:00
// https://github.com/open-telemetry/opentelemetry-specification/blob/d4b241f451674e8f611bb589477680341006ad2b/specification/metrics/sdk.md#exemplar-defaults
// Explicit bucket histogram aggregation with more than 1 bucket will
// use AlignedHistogramBucketExemplarReservoir.
a, ok := agg.(AggregationExplicitBucketHistogram)
if ok && len(a.Boundaries) > 0 {
Add selector of exemplar reservoir providers to metric.Stream configuration (#5861) Resolve https://github.com/open-telemetry/opentelemetry-go/issues/5249 ### Spec > exemplar_reservoir: A functional type that generates an exemplar reservoir a MeterProvider will use when storing exemplars. This functional type needs to be a factory or callback similar to aggregation selection functionality which allows different reservoirs to be chosen by the aggregation. > Users can provide an exemplar_reservoir, but it is up to their discretion. Therefore, the stream configuration parameter needs to be structured to accept an exemplar_reservoir, but MUST NOT obligate a user to provide one. If the user does not provide an exemplar_reservoir value, the MeterProvider MUST apply a [default exemplar reservoir](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#exemplar-defaults). Also, > the reservoir MUST be given the Attributes associated with its timeseries point either at construction so that additional sampling performed by the reservoir has access to all attributes from a measurement in the "offer" method. ### Changes In sdk/metric/exemplar, add: * `exemplar.ReservoirProvider` * `exemplar.HistogramReservoirProvider` * `exemplar.FixedSizeReservoirProvider` In sdk/metric, add: * `metric.ExemplarReservoirProviderSelector` (func Aggregation -> ReservoirProvider) * `metric.DefaultExemplarReservoirProviderSelector` (our default implementation) * `ExemplarReservoirProviderSelector` added to `metric.Stream` Note: the only required public types are `metric.ExemplarReservoirProviderSelector` and `ExemplarReservoirProviderSelector` in `metric.Stream`. Others are for convenience and readability. ### Alternatives considered #### Add ExemplarReservoirProvider directly to metric.Stream, instead of ExemplarReservoirProviderSelector This would mean users can configure a `func() exemplar.Reservoir` instead of a `func(Aggregation) func() exemplar.Reservoir`. I don't think this complies with the statement: `This functional type needs to be a factory or callback similar to aggregation selection functionality which allows different reservoirs to be chosen by the aggregation.`. I'm interpreting "allows different reservoirs to be chosen by the aggregation" as meaning "allows different reservoirs to be chosen **based on the** aggregation", rather than meaning that the aggregation is somehow choosing the reservoir. ### Future work There is some refactoring I plan to do after this to simplify the interaction between the internal/aggregate and exemplar package. I've omitted that from this PR to keep the diff smaller. --------- Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com> Co-authored-by: Robert Pająk <pellared@hotmail.com>
2024-10-18 09:05:10 -04:00
return exemplar.HistogramReservoirProvider(a.Boundaries)
}
Move time.Now call into exemplar reservoir to improve performance (#5545) Part of addressing https://github.com/open-telemetry/opentelemetry-go/issues/5542. ### Motivation This removes the `time.Now()` call from filtered-out Exemplars by only invoking `time.Now()` after the filtering decision is made. This improvement is especially noticeable for measurements without any attributes. ``` goos: linux goarch: amd64 pkg: go.opentelemetry.io/otel/sdk/metric cpu: AMD EPYC 7B12 │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ SyncMeasure/NoView/Int64Counter/Attributes/0-24 158.20n ± 4% 99.83n ± 1% -36.90% (p=0.000 n=10) SyncMeasure/NoView/Int64Counter/Attributes/1-24 333.3n ± 4% 274.8n ± 1% -17.55% (p=0.000 n=10) SyncMeasure/NoView/Int64Counter/Attributes/10-24 1.640µ ± 1% 1.600µ ± 1% -2.41% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/0-24 159.0n ± 3% 101.3n ± 0% -36.27% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/1-24 340.0n ± 2% 272.0n ± 1% -20.00% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/10-24 1.661µ ± 1% 1.597µ ± 0% -3.85% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/0-24 159.8n ± 1% 103.1n ± 0% -35.50% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/1-24 339.5n ± 1% 273.1n ± 0% -19.57% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/10-24 1.656µ ± 0% 1.589µ ± 0% -4.05% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/0-24 159.3n ± 2% 100.8n ± 0% -36.74% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/1-24 337.9n ± 2% 271.8n ± 1% -19.55% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/10-24 1.657µ ± 0% 1.593µ ± 1% -3.83% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/0-24 144.65n ± 4% 89.38n ± 0% -38.21% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/1-24 235.7n ± 2% 183.5n ± 0% -22.15% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/10-24 900.8n ± 1% 836.8n ± 0% -7.10% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/0-24 145.60n ± 5% 93.48n ± 1% -35.80% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/1-24 240.9n ± 1% 183.0n ± 0% -24.06% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/10-24 905.6n ± 1% 826.3n ± 0% -8.76% (p=0.000 n=10) SyncMeasure/DropView/Int64Counter/Attributes/0-24 20.33n ± 0% 20.35n ± 0% ~ (p=0.302 n=10) SyncMeasure/DropView/Int64Counter/Attributes/1-24 26.46n ± 0% 26.45n ± 1% ~ (p=0.868 n=10) SyncMeasure/DropView/Int64Counter/Attributes/10-24 26.50n ± 0% 26.47n ± 0% ~ (p=0.208 n=10) SyncMeasure/DropView/Float64Counter/Attributes/0-24 20.34n ± 1% 20.27n ± 0% -0.34% (p=0.009 n=10) SyncMeasure/DropView/Float64Counter/Attributes/1-24 26.55n ± 0% 26.60n ± 1% ~ (p=0.109 n=10) SyncMeasure/DropView/Float64Counter/Attributes/10-24 26.59n ± 1% 26.57n ± 1% ~ (p=0.926 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/0-24 20.38n ± 1% 20.38n ± 0% ~ (p=0.725 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/1-24 26.39n ± 0% 26.44n ± 0% ~ (p=0.238 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/10-24 26.52n ± 0% 26.42n ± 0% -0.36% (p=0.049 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/0-24 20.30n ± 0% 20.25n ± 0% ~ (p=0.196 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/1-24 26.57n ± 0% 26.54n ± 1% ~ (p=0.540 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/10-24 26.57n ± 0% 26.51n ± 1% ~ (p=0.643 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/0-24 20.37n ± 0% 20.36n ± 1% ~ (p=1.000 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/1-24 26.41n ± 0% 26.50n ± 0% +0.32% (p=0.007 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/10-24 26.44n ± 0% 26.55n ± 1% +0.42% (p=0.012 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/0-24 20.30n ± 0% 20.45n ± 0% +0.74% (p=0.000 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/1-24 26.52n ± 0% 26.48n ± 0% ~ (p=0.127 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/10-24 26.55n ± 0% 26.48n ± 0% -0.26% (p=0.002 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/0-24 170.5n ± 2% 110.8n ± 0% -35.03% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/1-24 402.5n ± 1% 331.5n ± 1% -17.64% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/10-24 1.363µ ± 1% 1.281µ ± 1% -6.02% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/0-24 170.6n ± 1% 111.5n ± 1% -34.64% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/1-24 397.1n ± 1% 335.9n ± 0% -15.41% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/10-24 1.371µ ± 1% 1.279µ ± 1% -6.71% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/0-24 170.1n ± 1% 112.2n ± 0% -34.09% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/1-24 397.5n ± 1% 330.2n ± 0% -16.93% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/10-24 1.371µ ± 1% 1.289µ ± 1% -5.95% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/0-24 171.4n ± 2% 112.9n ± 0% -34.13% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/1-24 397.0n ± 3% 336.4n ± 0% -15.24% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/10-24 1.383µ ± 1% 1.305µ ± 1% -5.61% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Histogram/Attributes/0-24 157.30n ± 2% 98.58n ± 1% -37.33% (p=0.000 n=6+10) ``` ### Changes * Introduce `exemplar.Filter`, which is a filter function based on the context. It will not be user-facing, so we can always add other parameters later if needed. * Introduce `exemplar.FilteredReservoir`, which is similar to a reservoir, except it does not receive a timestamp. It gets the current time after the filter decision has been made. It uses generics to avoid the call to exemplar.NewValue(), since it is internal-only. * The `exemplar.Reservoir` is left as-is, so that it can be made public when exemplars are stable. It still includes a timestamp argument. * Unit tests are updated to expect a much lower number of calls to time.Now * `exemplar.Drop` is now an `exemplar.FilteredReservoir` instead of a `Reservoir`, since we don't need a Reservoir to store things in if the measurement is always dropped. Co-authored-by: Sam Xie <sam@samxie.me> Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
2024-07-01 12:36:11 -04:00
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 = min(int(a.MaxSize), 20)
Move time.Now call into exemplar reservoir to improve performance (#5545) Part of addressing https://github.com/open-telemetry/opentelemetry-go/issues/5542. ### Motivation This removes the `time.Now()` call from filtered-out Exemplars by only invoking `time.Now()` after the filtering decision is made. This improvement is especially noticeable for measurements without any attributes. ``` goos: linux goarch: amd64 pkg: go.opentelemetry.io/otel/sdk/metric cpu: AMD EPYC 7B12 │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ SyncMeasure/NoView/Int64Counter/Attributes/0-24 158.20n ± 4% 99.83n ± 1% -36.90% (p=0.000 n=10) SyncMeasure/NoView/Int64Counter/Attributes/1-24 333.3n ± 4% 274.8n ± 1% -17.55% (p=0.000 n=10) SyncMeasure/NoView/Int64Counter/Attributes/10-24 1.640µ ± 1% 1.600µ ± 1% -2.41% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/0-24 159.0n ± 3% 101.3n ± 0% -36.27% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/1-24 340.0n ± 2% 272.0n ± 1% -20.00% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/10-24 1.661µ ± 1% 1.597µ ± 0% -3.85% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/0-24 159.8n ± 1% 103.1n ± 0% -35.50% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/1-24 339.5n ± 1% 273.1n ± 0% -19.57% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/10-24 1.656µ ± 0% 1.589µ ± 0% -4.05% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/0-24 159.3n ± 2% 100.8n ± 0% -36.74% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/1-24 337.9n ± 2% 271.8n ± 1% -19.55% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/10-24 1.657µ ± 0% 1.593µ ± 1% -3.83% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/0-24 144.65n ± 4% 89.38n ± 0% -38.21% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/1-24 235.7n ± 2% 183.5n ± 0% -22.15% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/10-24 900.8n ± 1% 836.8n ± 0% -7.10% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/0-24 145.60n ± 5% 93.48n ± 1% -35.80% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/1-24 240.9n ± 1% 183.0n ± 0% -24.06% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/10-24 905.6n ± 1% 826.3n ± 0% -8.76% (p=0.000 n=10) SyncMeasure/DropView/Int64Counter/Attributes/0-24 20.33n ± 0% 20.35n ± 0% ~ (p=0.302 n=10) SyncMeasure/DropView/Int64Counter/Attributes/1-24 26.46n ± 0% 26.45n ± 1% ~ (p=0.868 n=10) SyncMeasure/DropView/Int64Counter/Attributes/10-24 26.50n ± 0% 26.47n ± 0% ~ (p=0.208 n=10) SyncMeasure/DropView/Float64Counter/Attributes/0-24 20.34n ± 1% 20.27n ± 0% -0.34% (p=0.009 n=10) SyncMeasure/DropView/Float64Counter/Attributes/1-24 26.55n ± 0% 26.60n ± 1% ~ (p=0.109 n=10) SyncMeasure/DropView/Float64Counter/Attributes/10-24 26.59n ± 1% 26.57n ± 1% ~ (p=0.926 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/0-24 20.38n ± 1% 20.38n ± 0% ~ (p=0.725 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/1-24 26.39n ± 0% 26.44n ± 0% ~ (p=0.238 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/10-24 26.52n ± 0% 26.42n ± 0% -0.36% (p=0.049 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/0-24 20.30n ± 0% 20.25n ± 0% ~ (p=0.196 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/1-24 26.57n ± 0% 26.54n ± 1% ~ (p=0.540 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/10-24 26.57n ± 0% 26.51n ± 1% ~ (p=0.643 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/0-24 20.37n ± 0% 20.36n ± 1% ~ (p=1.000 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/1-24 26.41n ± 0% 26.50n ± 0% +0.32% (p=0.007 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/10-24 26.44n ± 0% 26.55n ± 1% +0.42% (p=0.012 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/0-24 20.30n ± 0% 20.45n ± 0% +0.74% (p=0.000 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/1-24 26.52n ± 0% 26.48n ± 0% ~ (p=0.127 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/10-24 26.55n ± 0% 26.48n ± 0% -0.26% (p=0.002 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/0-24 170.5n ± 2% 110.8n ± 0% -35.03% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/1-24 402.5n ± 1% 331.5n ± 1% -17.64% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/10-24 1.363µ ± 1% 1.281µ ± 1% -6.02% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/0-24 170.6n ± 1% 111.5n ± 1% -34.64% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/1-24 397.1n ± 1% 335.9n ± 0% -15.41% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/10-24 1.371µ ± 1% 1.279µ ± 1% -6.71% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/0-24 170.1n ± 1% 112.2n ± 0% -34.09% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/1-24 397.5n ± 1% 330.2n ± 0% -16.93% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/10-24 1.371µ ± 1% 1.289µ ± 1% -5.95% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/0-24 171.4n ± 2% 112.9n ± 0% -34.13% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/1-24 397.0n ± 3% 336.4n ± 0% -15.24% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/10-24 1.383µ ± 1% 1.305µ ± 1% -5.61% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Histogram/Attributes/0-24 157.30n ± 2% 98.58n ± 1% -37.33% (p=0.000 n=6+10) ``` ### Changes * Introduce `exemplar.Filter`, which is a filter function based on the context. It will not be user-facing, so we can always add other parameters later if needed. * Introduce `exemplar.FilteredReservoir`, which is similar to a reservoir, except it does not receive a timestamp. It gets the current time after the filter decision has been made. It uses generics to avoid the call to exemplar.NewValue(), since it is internal-only. * The `exemplar.Reservoir` is left as-is, so that it can be made public when exemplars are stable. It still includes a timestamp argument. * Unit tests are updated to expect a much lower number of calls to time.Now * `exemplar.Drop` is now an `exemplar.FilteredReservoir` instead of a `Reservoir`, since we don't need a Reservoir to store things in if the measurement is always dropped. Co-authored-by: Sam Xie <sam@samxie.me> Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
2024-07-01 12:36:11 -04:00
} 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.
//
// Use runtime.GOMAXPROCS instead of runtime.NumCPU to support
// containerized environments that may have less than the total number
// of logical CPUs available on the local machine allocated to it.
n = max(runtime.GOMAXPROCS(0), 1)
Move time.Now call into exemplar reservoir to improve performance (#5545) Part of addressing https://github.com/open-telemetry/opentelemetry-go/issues/5542. ### Motivation This removes the `time.Now()` call from filtered-out Exemplars by only invoking `time.Now()` after the filtering decision is made. This improvement is especially noticeable for measurements without any attributes. ``` goos: linux goarch: amd64 pkg: go.opentelemetry.io/otel/sdk/metric cpu: AMD EPYC 7B12 │ old.txt │ new.txt │ │ sec/op │ sec/op vs base │ SyncMeasure/NoView/Int64Counter/Attributes/0-24 158.20n ± 4% 99.83n ± 1% -36.90% (p=0.000 n=10) SyncMeasure/NoView/Int64Counter/Attributes/1-24 333.3n ± 4% 274.8n ± 1% -17.55% (p=0.000 n=10) SyncMeasure/NoView/Int64Counter/Attributes/10-24 1.640µ ± 1% 1.600µ ± 1% -2.41% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/0-24 159.0n ± 3% 101.3n ± 0% -36.27% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/1-24 340.0n ± 2% 272.0n ± 1% -20.00% (p=0.000 n=10) SyncMeasure/NoView/Float64Counter/Attributes/10-24 1.661µ ± 1% 1.597µ ± 0% -3.85% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/0-24 159.8n ± 1% 103.1n ± 0% -35.50% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/1-24 339.5n ± 1% 273.1n ± 0% -19.57% (p=0.000 n=10) SyncMeasure/NoView/Int64UpDownCounter/Attributes/10-24 1.656µ ± 0% 1.589µ ± 0% -4.05% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/0-24 159.3n ± 2% 100.8n ± 0% -36.74% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/1-24 337.9n ± 2% 271.8n ± 1% -19.55% (p=0.000 n=10) SyncMeasure/NoView/Float64UpDownCounter/Attributes/10-24 1.657µ ± 0% 1.593µ ± 1% -3.83% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/0-24 144.65n ± 4% 89.38n ± 0% -38.21% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/1-24 235.7n ± 2% 183.5n ± 0% -22.15% (p=0.000 n=10) SyncMeasure/NoView/Int64Histogram/Attributes/10-24 900.8n ± 1% 836.8n ± 0% -7.10% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/0-24 145.60n ± 5% 93.48n ± 1% -35.80% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/1-24 240.9n ± 1% 183.0n ± 0% -24.06% (p=0.000 n=10) SyncMeasure/NoView/Float64Histogram/Attributes/10-24 905.6n ± 1% 826.3n ± 0% -8.76% (p=0.000 n=10) SyncMeasure/DropView/Int64Counter/Attributes/0-24 20.33n ± 0% 20.35n ± 0% ~ (p=0.302 n=10) SyncMeasure/DropView/Int64Counter/Attributes/1-24 26.46n ± 0% 26.45n ± 1% ~ (p=0.868 n=10) SyncMeasure/DropView/Int64Counter/Attributes/10-24 26.50n ± 0% 26.47n ± 0% ~ (p=0.208 n=10) SyncMeasure/DropView/Float64Counter/Attributes/0-24 20.34n ± 1% 20.27n ± 0% -0.34% (p=0.009 n=10) SyncMeasure/DropView/Float64Counter/Attributes/1-24 26.55n ± 0% 26.60n ± 1% ~ (p=0.109 n=10) SyncMeasure/DropView/Float64Counter/Attributes/10-24 26.59n ± 1% 26.57n ± 1% ~ (p=0.926 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/0-24 20.38n ± 1% 20.38n ± 0% ~ (p=0.725 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/1-24 26.39n ± 0% 26.44n ± 0% ~ (p=0.238 n=10) SyncMeasure/DropView/Int64UpDownCounter/Attributes/10-24 26.52n ± 0% 26.42n ± 0% -0.36% (p=0.049 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/0-24 20.30n ± 0% 20.25n ± 0% ~ (p=0.196 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/1-24 26.57n ± 0% 26.54n ± 1% ~ (p=0.540 n=10) SyncMeasure/DropView/Float64UpDownCounter/Attributes/10-24 26.57n ± 0% 26.51n ± 1% ~ (p=0.643 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/0-24 20.37n ± 0% 20.36n ± 1% ~ (p=1.000 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/1-24 26.41n ± 0% 26.50n ± 0% +0.32% (p=0.007 n=10) SyncMeasure/DropView/Int64Histogram/Attributes/10-24 26.44n ± 0% 26.55n ± 1% +0.42% (p=0.012 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/0-24 20.30n ± 0% 20.45n ± 0% +0.74% (p=0.000 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/1-24 26.52n ± 0% 26.48n ± 0% ~ (p=0.127 n=10) SyncMeasure/DropView/Float64Histogram/Attributes/10-24 26.55n ± 0% 26.48n ± 0% -0.26% (p=0.002 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/0-24 170.5n ± 2% 110.8n ± 0% -35.03% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/1-24 402.5n ± 1% 331.5n ± 1% -17.64% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Counter/Attributes/10-24 1.363µ ± 1% 1.281µ ± 1% -6.02% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/0-24 170.6n ± 1% 111.5n ± 1% -34.64% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/1-24 397.1n ± 1% 335.9n ± 0% -15.41% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64Counter/Attributes/10-24 1.371µ ± 1% 1.279µ ± 1% -6.71% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/0-24 170.1n ± 1% 112.2n ± 0% -34.09% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/1-24 397.5n ± 1% 330.2n ± 0% -16.93% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64UpDownCounter/Attributes/10-24 1.371µ ± 1% 1.289µ ± 1% -5.95% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/0-24 171.4n ± 2% 112.9n ± 0% -34.13% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/1-24 397.0n ± 3% 336.4n ± 0% -15.24% (p=0.000 n=10) SyncMeasure/AttrFilterView/Float64UpDownCounter/Attributes/10-24 1.383µ ± 1% 1.305µ ± 1% -5.61% (p=0.000 n=10) SyncMeasure/AttrFilterView/Int64Histogram/Attributes/0-24 157.30n ± 2% 98.58n ± 1% -37.33% (p=0.000 n=6+10) ``` ### Changes * Introduce `exemplar.Filter`, which is a filter function based on the context. It will not be user-facing, so we can always add other parameters later if needed. * Introduce `exemplar.FilteredReservoir`, which is similar to a reservoir, except it does not receive a timestamp. It gets the current time after the filter decision has been made. It uses generics to avoid the call to exemplar.NewValue(), since it is internal-only. * The `exemplar.Reservoir` is left as-is, so that it can be made public when exemplars are stable. It still includes a timestamp argument. * Unit tests are updated to expect a much lower number of calls to time.Now * `exemplar.Drop` is now an `exemplar.FilteredReservoir` instead of a `Reservoir`, since we don't need a Reservoir to store things in if the measurement is always dropped. Co-authored-by: Sam Xie <sam@samxie.me> Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com>
2024-07-01 12:36:11 -04:00
}
Add selector of exemplar reservoir providers to metric.Stream configuration (#5861) Resolve https://github.com/open-telemetry/opentelemetry-go/issues/5249 ### Spec > exemplar_reservoir: A functional type that generates an exemplar reservoir a MeterProvider will use when storing exemplars. This functional type needs to be a factory or callback similar to aggregation selection functionality which allows different reservoirs to be chosen by the aggregation. > Users can provide an exemplar_reservoir, but it is up to their discretion. Therefore, the stream configuration parameter needs to be structured to accept an exemplar_reservoir, but MUST NOT obligate a user to provide one. If the user does not provide an exemplar_reservoir value, the MeterProvider MUST apply a [default exemplar reservoir](https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/sdk.md#exemplar-defaults). Also, > the reservoir MUST be given the Attributes associated with its timeseries point either at construction so that additional sampling performed by the reservoir has access to all attributes from a measurement in the "offer" method. ### Changes In sdk/metric/exemplar, add: * `exemplar.ReservoirProvider` * `exemplar.HistogramReservoirProvider` * `exemplar.FixedSizeReservoirProvider` In sdk/metric, add: * `metric.ExemplarReservoirProviderSelector` (func Aggregation -> ReservoirProvider) * `metric.DefaultExemplarReservoirProviderSelector` (our default implementation) * `ExemplarReservoirProviderSelector` added to `metric.Stream` Note: the only required public types are `metric.ExemplarReservoirProviderSelector` and `ExemplarReservoirProviderSelector` in `metric.Stream`. Others are for convenience and readability. ### Alternatives considered #### Add ExemplarReservoirProvider directly to metric.Stream, instead of ExemplarReservoirProviderSelector This would mean users can configure a `func() exemplar.Reservoir` instead of a `func(Aggregation) func() exemplar.Reservoir`. I don't think this complies with the statement: `This functional type needs to be a factory or callback similar to aggregation selection functionality which allows different reservoirs to be chosen by the aggregation.`. I'm interpreting "allows different reservoirs to be chosen by the aggregation" as meaning "allows different reservoirs to be chosen **based on the** aggregation", rather than meaning that the aggregation is somehow choosing the reservoir. ### Future work There is some refactoring I plan to do after this to simplify the interaction between the internal/aggregate and exemplar package. I've omitted that from this PR to keep the diff smaller. --------- Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com> Co-authored-by: Robert Pająk <pellared@hotmail.com>
2024-10-18 09:05:10 -04:00
return exemplar.FixedSizeReservoirProvider(n)
}