Skip to content

Commit

Permalink
Adds upper bound to latency tail sampling as per review feedback
Browse files Browse the repository at this point in the history
  • Loading branch information
garry-cairns committed Sep 15, 2023
1 parent f34c6a8 commit c72e96a
Show file tree
Hide file tree
Showing 8 changed files with 106 additions and 30 deletions.
1 change: 1 addition & 0 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -38,3 +38,4 @@ integration-coverage.html

go.work*

/result
8 changes: 1 addition & 7 deletions processor/tailsamplingprocessor/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -29,7 +29,6 @@ The following configuration options are required:
Multiple policies exist today and it is straight forward to add more. These include:
- `always_sample`: Sample all traces
- `latency`: Sample based on the duration of the trace. The duration is determined by looking at the earliest start time and latest end time, without taking into consideration what happened in between.
- `duration`: Sample based on the *bounded* duration of the trace. Otherwise identical to latency
- `numeric_attribute`: Sample based on number attributes (resource and record)
- `probabilistic`: Sample a percentage of traces. Read [a comparison with the Probabilistic Sampling Processor](#probabilistic-sampling-processor-compared-to-the-tail-sampling-processor-with-the-probabilistic-policy).
- `status_code`: Sample based upon the status code (`OK`, `ERROR` or `UNSET`)
Expand Down Expand Up @@ -77,12 +76,7 @@ processors:
{
name: test-policy-2,
type: latency,
latency: {threshold_ms: 5000}
},
{
name: test-policy-2a,
type: duration,
latency: {lower_bound_ms: 5000, upper_bound_ms: 10000}
latency: {threshold_ms: 5000, upper_threshold_ms: 10000}
},
{
name: test-policy-3,
Expand Down
2 changes: 1 addition & 1 deletion processor/tailsamplingprocessor/and_helper_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ func TestAndHelper(t *testing.T) {
require.NoError(t, err)

expected := sampling.NewAnd(zap.NewNop(), []sampling.PolicyEvaluator{
sampling.NewLatency(componenttest.NewNopTelemetrySettings(), 100),
sampling.NewLatency(componenttest.NewNopTelemetrySettings(), 100, 0),
})
assert.Equal(t, expected, actual)
})
Expand Down
4 changes: 2 additions & 2 deletions processor/tailsamplingprocessor/composite_helper_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -50,11 +50,11 @@ func TestCompositeHelper(t *testing.T) {

expected := sampling.NewComposite(zap.NewNop(), 1000, []sampling.SubPolicyEvalParams{
{
Evaluator: sampling.NewLatency(componenttest.NewNopTelemetrySettings(), 100),
Evaluator: sampling.NewLatency(componenttest.NewNopTelemetrySettings(), 100, 0),
MaxSpansPerSecond: 250,
},
{
Evaluator: sampling.NewLatency(componenttest.NewNopTelemetrySettings(), 200),
Evaluator: sampling.NewLatency(componenttest.NewNopTelemetrySettings(), 200, 0),
MaxSpansPerSecond: 500,
},
}, sampling.MonotonicClock{})
Expand Down
15 changes: 2 additions & 13 deletions processor/tailsamplingprocessor/config.go
Original file line number Diff line number Diff line change
Expand Up @@ -17,8 +17,6 @@ const (
AlwaysSample PolicyType = "always_sample"
// Latency sample traces that are longer than a given threshold.
Latency PolicyType = "latency"
// Duration sample traces that are longer than a given threshold.
Duration PolicyType = "duration"
// NumericAttribute sample traces that have a given numeric attribute in a specified
// range, e.g.: attribute "http.status_code" >= 399 and <= 999.
NumericAttribute PolicyType = "numeric_attribute"
Expand Down Expand Up @@ -56,8 +54,6 @@ type sharedPolicyCfg struct {
Type PolicyType `mapstructure:"type"`
// Configs for latency filter sampling policy evaluator.
LatencyCfg LatencyCfg `mapstructure:"latency"`
// Configs for duration filter sampling policy evaluator.
DurationCfg DurationCfg `mapstructure:"duration"`
// Configs for numeric attribute filter sampling policy evaluator.
NumericAttributeCfg NumericAttributeCfg `mapstructure:"numeric_attribute"`
// Configs for probabilistic sampling policy evaluator.
Expand Down Expand Up @@ -130,17 +126,10 @@ type PolicyCfg struct {
// LatencyCfg holds the configurable settings to create a latency filter sampling policy
// evaluator
type LatencyCfg struct {
// ThresholdMs in milliseconds.
// Lower bound in milliseconds. Retaining original name for compatibility
ThresholdMs int64 `mapstructure:"threshold_ms"`
}

// DurationCfg holds the configurable settings to create a duration filter sampling policy
// evaluator, which is essentially a latency sampler with two bounds
type DurationCfg struct {
// Lower bound in milliseconds.
LowerDurationMs int64 `mapstructure:"lower_bound_ms"`
// Upper bound in milliseconds.
UpperDurationMs int64 `mapstructure:"upper_bound_ms"`
UpperThresholdmsMs int64 `mapstructure:"upper_threshold_ms"`
}

// NumericAttributeCfg holds the configurable settings to create a numeric attribute filter
Expand Down
15 changes: 10 additions & 5 deletions processor/tailsamplingprocessor/internal/sampling/latency.go
Original file line number Diff line number Diff line change
Expand Up @@ -13,17 +13,19 @@ import (
)

type latency struct {
logger *zap.Logger
logger *zap.Logger
thresholdMs int64
upperThresholdMs int64
}

var _ PolicyEvaluator = (*latency)(nil)

// NewLatency creates a policy evaluator sampling traces with a duration higher than a configured threshold
func NewLatency(settings component.TelemetrySettings, thresholdMs int64) PolicyEvaluator {
func NewLatency(settings component.TelemetrySettings, thresholdMs int64, upperThresholdMs int64) PolicyEvaluator {
return &latency{
logger: settings.Logger,
thresholdMs: thresholdMs,
logger: settings.Logger,
thresholdMs: thresholdMs,
upperThresholdMs: upperThresholdMs,
}
}

Expand All @@ -47,6 +49,9 @@ func (l *latency) Evaluate(_ context.Context, _ pcommon.TraceID, traceData *Trac
}

duration := maxTime.AsTime().Sub(minTime.AsTime())
return duration.Milliseconds() >= l.thresholdMs
if l.upperThresholdMs == 0 {
return duration.Milliseconds() >= l.thresholdMs
}
return (l.thresholdMs < duration.Milliseconds() && duration.Milliseconds() <= l.upperThresholdMs)
}), nil
}
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,7 @@ import (
)

func TestEvaluate_Latency(t *testing.T) {
filter := NewLatency(componenttest.NewNopTelemetrySettings(), 5000)
filter := NewLatency(componenttest.NewNopTelemetrySettings(), 5000, 0)

traceID := pcommon.TraceID([16]byte{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16})
now := time.Now()
Expand Down Expand Up @@ -71,6 +71,93 @@ func TestEvaluate_Latency(t *testing.T) {
}
}

func TestEvaluate_Bounded_Latency(t *testing.T) {
filter := NewLatency(componenttest.NewNopTelemetrySettings(), 5000, 10000)

traceID := pcommon.TraceID([16]byte{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16})
now := time.Now()

cases := []struct {
Desc string
Spans []spanWithTimeAndDuration
Decision Decision
}{
{
"trace duration shorter than lower bound",
[]spanWithTimeAndDuration{
{
StartTime: now,
Duration: 4500 * time.Millisecond,
},
},
NotSampled,
},
{
"trace duration is equal to lower bound",
[]spanWithTimeAndDuration{
{
StartTime: now,
Duration: 5000 * time.Millisecond,
},
},
NotSampled,
},
{
"trace duration is within lower and upper bounds",
[]spanWithTimeAndDuration{
{
StartTime: now,
Duration: 5001 * time.Millisecond,
},
},
Sampled,
},
{
"trace duration is above upper bound",
[]spanWithTimeAndDuration{
{
StartTime: now,
Duration: 10001 * time.Millisecond,
},
},
NotSampled,
},
{
"trace duration equals upper bound",
[]spanWithTimeAndDuration{
{
StartTime: now,
Duration: 10000 * time.Millisecond,
},
},
Sampled,
},
{
"total trace duration is longer than threshold but every single span is shorter",
[]spanWithTimeAndDuration{
{
StartTime: now,
Duration: 3000 * time.Millisecond,
},
{
StartTime: now.Add(2500 * time.Millisecond),
Duration: 3000 * time.Millisecond,
},
},
Sampled,
},
}

for _, c := range cases {
t.Run(c.Desc, func(t *testing.T) {
decision, err := filter.Evaluate(context.Background(), traceID, newTraceWithSpans(c.Spans))

assert.NoError(t, err)
assert.Equal(t, decision, c.Decision)
})
}
}

type spanWithTimeAndDuration struct {
StartTime time.Time
Duration time.Duration
Expand Down
2 changes: 1 addition & 1 deletion processor/tailsamplingprocessor/processor.go
Original file line number Diff line number Diff line change
Expand Up @@ -133,7 +133,7 @@ func getSharedPolicyEvaluator(settings component.TelemetrySettings, cfg *sharedP
return sampling.NewAlwaysSample(settings), nil
case Latency:
lfCfg := cfg.LatencyCfg
return sampling.NewLatency(settings, lfCfg.ThresholdMs), nil
return sampling.NewLatency(settings, lfCfg.ThresholdMs, lfCfg.UpperThresholdmsMs), nil
case NumericAttribute:
nafCfg := cfg.NumericAttributeCfg
return sampling.NewNumericAttributeFilter(settings, nafCfg.Key, nafCfg.MinValue, nafCfg.MaxValue, nafCfg.InvertMatch), nil
Expand Down

0 comments on commit c72e96a

Please sign in to comment.