1
0
mirror of https://github.com/open-telemetry/opentelemetry-go.git synced 2025-01-26 03:52:03 +02:00
David Ashpole 81b2a33e1b
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

195 lines
4.9 KiB
Go

// Copyright The OpenTelemetry Authors
// SPDX-License-Identifier: Apache-2.0
package aggregate // import "go.opentelemetry.io/otel/sdk/metric/internal/aggregate"
import (
"context"
"strconv"
"sync/atomic"
"testing"
"time"
"github.com/stretchr/testify/assert"
"go.opentelemetry.io/otel/attribute"
"go.opentelemetry.io/otel/sdk/metric/metricdata"
"go.opentelemetry.io/otel/sdk/metric/metricdata/metricdatatest"
)
var (
keyUser = "user"
userAlice = attribute.String(keyUser, "Alice")
userBob = attribute.String(keyUser, "Bob")
userCarol = attribute.String(keyUser, "Carol")
userDave = attribute.String(keyUser, "Dave")
adminTrue = attribute.Bool("admin", true)
adminFalse = attribute.Bool("admin", false)
alice = attribute.NewSet(userAlice, adminTrue)
bob = attribute.NewSet(userBob, adminFalse)
carol = attribute.NewSet(userCarol, adminFalse)
dave = attribute.NewSet(userDave, adminFalse)
// Filtered.
attrFltr = func(kv attribute.KeyValue) bool {
return kv.Key == attribute.Key(keyUser)
}
fltrAlice = attribute.NewSet(userAlice)
fltrBob = attribute.NewSet(userBob)
// Sat Jan 01 2000 00:00:00 GMT+0000.
y2k = time.Unix(946684800, 0)
)
// y2kPlus returns the timestamp at n seconds past Sat Jan 01 2000 00:00:00 GMT+0000.
func y2kPlus(n int64) time.Time {
d := time.Duration(n) * time.Second
return y2k.Add(d)
}
// clock is a test clock. It provides a predictable value for now() that can be
// reset.
type clock struct {
ticks atomic.Int64
}
// Now returns the mocked time starting at y2kPlus(0). Each call to Now will
// increment the returned value by one second.
func (c *clock) Now() time.Time {
old := c.ticks.Add(1) - 1
return y2kPlus(old)
}
// Reset resets the clock c to tick from y2kPlus(0).
func (c *clock) Reset() { c.ticks.Store(0) }
// Register registers clock c's Now method as the now var. It returns an
// unregister func that should be called to restore the original now value.
func (c *clock) Register() (unregister func()) {
orig := now
now = c.Now
return func() { now = orig }
}
func dropExemplars[N int64 | float64](attr attribute.Set) FilteredExemplarReservoir[N] {
return dropReservoir[N](attr)
}
func TestBuilderFilter(t *testing.T) {
t.Run("Int64", testBuilderFilter[int64]())
t.Run("Float64", testBuilderFilter[float64]())
}
func testBuilderFilter[N int64 | float64]() func(t *testing.T) {
return func(t *testing.T) {
t.Helper()
value, attr := N(1), alice
run := func(b Builder[N], wantF attribute.Set, wantD []attribute.KeyValue) func(*testing.T) {
return func(t *testing.T) {
t.Helper()
meas := b.filter(func(_ context.Context, v N, f attribute.Set, d []attribute.KeyValue) {
assert.Equal(t, value, v, "measured incorrect value")
assert.Equal(t, wantF, f, "measured incorrect filtered attributes")
assert.ElementsMatch(t, wantD, d, "measured incorrect dropped attributes")
})
meas(context.Background(), value, attr)
}
}
t.Run("NoFilter", run(Builder[N]{}, attr, nil))
t.Run("Filter", run(Builder[N]{Filter: attrFltr}, fltrAlice, []attribute.KeyValue{adminTrue}))
}
}
type arg[N int64 | float64] struct {
ctx context.Context
value N
attr attribute.Set
}
type output struct {
n int
agg metricdata.Aggregation
}
type teststep[N int64 | float64] struct {
input []arg[N]
expect output
}
func test[N int64 | float64](meas Measure[N], comp ComputeAggregation, steps []teststep[N]) func(*testing.T) {
return func(t *testing.T) {
t.Helper()
got := new(metricdata.Aggregation)
for i, step := range steps {
for _, args := range step.input {
meas(args.ctx, args.value, args.attr)
}
t.Logf("step: %d", i)
assert.Equal(t, step.expect.n, comp(got), "incorrect data size")
metricdatatest.AssertAggregationsEqual(t, step.expect.agg, *got)
}
}
}
func benchmarkAggregate[N int64 | float64](factory func() (Measure[N], ComputeAggregation)) func(*testing.B) {
counts := []int{1, 10, 100}
return func(b *testing.B) {
for _, n := range counts {
b.Run(strconv.Itoa(n), func(b *testing.B) {
benchmarkAggregateN(b, factory, n)
})
}
}
}
var bmarkRes metricdata.Aggregation
func benchmarkAggregateN[N int64 | float64](b *testing.B, factory func() (Measure[N], ComputeAggregation), count int) {
ctx := context.Background()
attrs := make([]attribute.Set, count)
for i := range attrs {
attrs[i] = attribute.NewSet(attribute.Int("value", i))
}
b.Run("Measure", func(b *testing.B) {
got := &bmarkRes
meas, comp := factory()
b.ReportAllocs()
b.ResetTimer()
for n := 0; n < b.N; n++ {
for _, attr := range attrs {
meas(ctx, 1, attr)
}
}
comp(got)
})
b.Run("ComputeAggregation", func(b *testing.B) {
comps := make([]ComputeAggregation, b.N)
for n := range comps {
meas, comp := factory()
for _, attr := range attrs {
meas(ctx, 1, attr)
}
comps[n] = comp
}
got := &bmarkRes
b.ReportAllocs()
b.ResetTimer()
for n := 0; n < b.N; n++ {
comps[n](got)
}
})
}