Skip to content

perf(query): keep compaction, rollups and gates ahead of sustained ingest #312

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
vishr merged 2 commits into from
Oct 15, 2026
Merged

perf(query): keep compaction, rollups and gates ahead of sustained ingest #312

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
1a33cf0
perf(query): keep compaction, rollups and gates ahead of sustained in…
vishr Oct 15, 2026
54fda80
test(query): synchronize the maintenance schedule test's recorded cycles
vishr Oct 15, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
164 changes: 164 additions & 0 deletions internal/ingest/conversion_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,164 @@
package ingest

import (
"bytes"
"context"
"encoding/json"
"os"
"testing"

"github.com/labstack/fanout/internal/config"
telemetrystore "github.com/labstack/fanout/internal/telemetry/store"
collectorlogs "go.opentelemetry.io/proto/otlp/collector/logs/v1"
collectormetrics "go.opentelemetry.io/proto/otlp/collector/metrics/v1"
collectortrace "go.opentelemetry.io/proto/otlp/collector/trace/v1"
common "go.opentelemetry.io/proto/otlp/common/v1"
logspb "go.opentelemetry.io/proto/otlp/logs/v1"
metricspb "go.opentelemetry.io/proto/otlp/metrics/v1"
resourcepb "go.opentelemetry.io/proto/otlp/resource/v1"
tracepb "go.opentelemetry.io/proto/otlp/trace/v1"
)

type conversionSubmitter func(context.Context, telemetrystore.Batch) error

func (f conversionSubmitter) Submit(ctx context.Context, batch telemetrystore.Batch) error {
return f(ctx, batch)
}

func conversionExports(n int) (*collectortrace.ExportTraceServiceRequest, *collectorlogs.ExportLogsServiceRequest, *collectormetrics.ExportMetricsServiceRequest) {
traces := &collectortrace.ExportTraceServiceRequest{}
logs := &collectorlogs.ExportLogsServiceRequest{}
metrics := &collectormetrics.ExportMetricsServiceRequest{}
text := func(k, v string) *common.KeyValue {
return &common.KeyValue{Key: k, Value: &common.AnyValue{Value: &common.AnyValue_StringValue{StringValue: v}}}
}
attrs := []*common.KeyValue{
text("http.method", "POST"),
text("http.request.method", "GET"),
text("http.route", "/orders"),
{Key: "http.status_code", Value: &common.AnyValue{Value: &common.AnyValue_IntValue{IntValue: 503}}},
{Key: "flag", Value: &common.AnyValue{Value: &common.AnyValue_BoolValue{BoolValue: true}}},
{Key: "bytes", Value: &common.AnyValue{Value: &common.AnyValue_BytesValue{BytesValue: []byte{0, 255}}}},
}
for r := 0; r < 2; r++ {
resource := &resourcepb.Resource{Attributes: []*common.KeyValue{text("service.name", []string{"api", "worker"}[r]), text("service.version", []string{"v1", "v2"}[r]), text("deployment.environment", []string{"prod", "test"}[r]), {Key: "nested", Value: &common.AnyValue{Value: &common.AnyValue_ArrayValue{ArrayValue: &common.ArrayValue{Values: []*common.AnyValue{{Value: &common.AnyValue_IntValue{IntValue: 7}}, {Value: &common.AnyValue_BoolValue{BoolValue: false}}}}}}}}}
if r == 0 {
resource.Attributes = append(resource.Attributes, text("service.namespace", "shop"))
}
rs := &tracepb.ResourceSpans{Resource: resource}
rl := &logspb.ResourceLogs{Resource: resource}
rm := &metricspb.ResourceMetrics{Resource: resource}
for scope := 0; scope < 2; scope++ {
info := &common.InstrumentationScope{Name: []string{"http", "jobs"}[scope], Version: []string{"1.0", "2.0"}[scope]}
ss := &tracepb.ScopeSpans{Scope: info}
sl := &logspb.ScopeLogs{Scope: info}
sm := &metricspb.ScopeMetrics{Scope: info}
for i := 0; i < n; i++ {
ss.Spans = append(ss.Spans, &tracepb.Span{
TraceId: []byte{1, 2, byte(r)}, SpanId: []byte{3, byte(scope), byte(i)}, ParentSpanId: []byte{4},
Name: "operation", Kind: tracepb.Span_SPAN_KIND_CLIENT,
StartTimeUnixNano: 1790856000000000017, EndTimeUnixNano: 1790856000001000019,
Status: &tracepb.Status{Code: tracepb.Status_STATUS_CODE_ERROR, Message: "failed"},
Attributes: attrs, TraceState: "vendor=state", Flags: 257,
Events: []*tracepb.Span_Event{{
Name: "exception", TimeUnixNano: 1790856000000000020,
Attributes: []*common.KeyValue{text("exception.type", "Timeout"), text("exception.message", "late")},
}},
Links: []*tracepb.Span_Link{{
TraceId: []byte{5}, SpanId: []byte{6}, TraceState: "a=b",
Attributes: []*common.KeyValue{text("link", "value")},
}},
})
sl.LogRecords = append(sl.LogRecords, &logspb.LogRecord{TimeUnixNano: 1790856000000000017, ObservedTimeUnixNano: 1790856000000000019, SeverityNumber: logspb.SeverityNumber_SEVERITY_NUMBER_ERROR, Body: &common.AnyValue{Value: &common.AnyValue_StringValue{StringValue: "order 123 failed"}}, TraceId: []byte{1, 2}, SpanId: []byte{3, 4}, Flags: 1, Attributes: attrs})
}
point := &metricspb.NumberDataPoint{TimeUnixNano: 1790856000000000017, StartTimeUnixNano: 1790855940000000009, Value: &metricspb.NumberDataPoint_AsInt{AsInt: 12}, Attributes: attrs}
points := make([]*metricspb.NumberDataPoint, n)
for i := range points {
points[i] = point
}
switch r*2 + scope {
case 0:
sm.Metrics = []*metricspb.Metric{{Name: "gauge", Description: "description", Unit: "ms", Data: &metricspb.Metric_Gauge{Gauge: &metricspb.Gauge{DataPoints: points}}}, {Name: "sum", Data: &metricspb.Metric_Sum{Sum: &metricspb.Sum{IsMonotonic: true, AggregationTemporality: metricspb.AggregationTemporality_AGGREGATION_TEMPORALITY_DELTA, DataPoints: points}}}}
case 1:
sm.Metrics = []*metricspb.Metric{{Name: "histogram", Data: &metricspb.Metric_Histogram{Histogram: &metricspb.Histogram{AggregationTemporality: metricspb.AggregationTemporality_AGGREGATION_TEMPORALITY_CUMULATIVE, DataPoints: []*metricspb.HistogramDataPoint{{TimeUnixNano: 1790856000000000017, Count: 2, ExplicitBounds: []float64{1}, BucketCounts: []uint64{1, 1}, Attributes: attrs}}}}}}
case 2:
sm.Metrics = []*metricspb.Metric{{Name: "exponential", Data: &metricspb.Metric_ExponentialHistogram{ExponentialHistogram: &metricspb.ExponentialHistogram{AggregationTemporality: metricspb.AggregationTemporality_AGGREGATION_TEMPORALITY_DELTA, DataPoints: []*metricspb.ExponentialHistogramDataPoint{{TimeUnixNano: 1790856000000000017, Count: 1, Positive: &metricspb.ExponentialHistogramDataPoint_Buckets{BucketCounts: []uint64{1}}, Attributes: attrs}}}}}}
case 3:
sm.Metrics = []*metricspb.Metric{{Name: "summary", Data: &metricspb.Metric_Summary{Summary: &metricspb.Summary{DataPoints: []*metricspb.SummaryDataPoint{{TimeUnixNano: 1790856000000000017, Count: 1, Sum: 3, QuantileValues: []*metricspb.SummaryDataPoint_ValueAtQuantile{{Quantile: 0.5, Value: 3}}, Attributes: attrs}}}}}}
}
rs.ScopeSpans = append(rs.ScopeSpans, ss)
rl.ScopeLogs = append(rl.ScopeLogs, sl)
rm.ScopeMetrics = append(rm.ScopeMetrics, sm)
}
traces.ResourceSpans = append(traces.ResourceSpans, rs)
logs.ResourceLogs = append(logs.ResourceLogs, rl)
metrics.ResourceMetrics = append(metrics.ResourceMetrics, rm)
}
return traces, logs, metrics
}

func TestExportConversionRowsMatchBaseline(t *testing.T) {
traces, logs, metrics := conversionExports(1)
var got telemetrystore.Batch
srv := NewServer(config.Config{DefaultNamespace: "fallback"}, conversionSubmitter(func(_ context.Context, b telemetrystore.Batch) error {
got.Spans = append(got.Spans, b.Spans...)
got.Logs = append(got.Logs, b.Logs...)
got.Metrics = append(got.Metrics, b.Metrics...)
return nil
}))
if _, err := srv.exportTraces(t.Context(), traces); err != nil {
t.Fatal(err)
}
if _, err := srv.exportLogs(t.Context(), logs); err != nil {
t.Fatal(err)
}
if _, err := srv.exportMetrics(t.Context(), metrics); err != nil {
t.Fatal(err)
}
for i := range got.Spans {
got.Spans[i].IngestedAt = 0
}
for i := range got.Logs {
got.Logs[i].IngestedAt = 0
}
for i := range got.Metrics {
got.Metrics[i].IngestedAt = 0
}
encoded, err := json.MarshalIndent(got, "", " ")
if err != nil {
t.Fatal(err)
}
encoded = append(encoded, '\n')
path := "testdata/conversion_rows.json"

want, err := os.ReadFile(path)
if err != nil {
t.Fatal(err)
}
if !bytes.Equal(encoded, want) {
t.Fatalf("converted rows changed:\n%s", encoded)
}
}
func BenchmarkExportConversion(b *testing.B) {
traces, logs, metrics := conversionExports(250)
srv := NewServer(config.Config{DefaultNamespace: "fallback"}, conversionSubmitter(func(context.Context, telemetrystore.Batch) error { return nil }))
for _, name := range []string{"spans", "logs", "metrics"} {
b.Run(name, func(b *testing.B) {
b.ReportAllocs()
for b.Loop() {
var err error
switch name {
case "spans":
_, err = srv.exportTraces(b.Context(), traces)
case "logs":
_, err = srv.exportLogs(b.Context(), logs)
case "metrics":
_, err = srv.exportMetrics(b.Context(), metrics)
}
if err != nil {
b.Fatal(err)
}
}
})
}
}
46 changes: 44 additions & 2 deletions internal/ingest/server.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,7 +72,16 @@ func (ts *traceService) Export(ctx context.Context, req *collectortrace.ExportTr
func (s *Server) exportTraces(ctx context.Context, req *collectortrace.ExportTraceServiceRequest) (*collectortrace.ExportTraceServiceResponse, error) {
cfg := s.cfg
now := time.Now().UnixNano()
count := 0
for _, rs := range req.ResourceSpans {
for _, ss := range rs.ScopeSpans {
count += len(ss.Spans)
}
}
batch := telemetrystore.Batch{}
if count > 0 {
batch.Spans = make([]telemetry.Span, 0, count)
}
for _, rs := range req.ResourceSpans {
resources, err := resourceAttributes(rs.Resource)
if err != nil {
Expand All @@ -83,6 +92,8 @@ func (s *Server) exportTraces(ctx context.Context, req *collectortrace.ExportTra
if namespace == "" {
namespace = cfg.DefaultNamespace
}
serviceVersion := getResourceAttr(rs.Resource, "service.version")
deploymentEnv := getResourceAttr(rs.Resource, "deployment.environment")
for _, ss := range rs.ScopeSpans {
scopeName, scopeVer := scopeInfo(ss.Scope)
for _, sp := range ss.Spans {
Expand Down Expand Up @@ -124,8 +135,8 @@ func (s *Server) exportTraces(ctx context.Context, req *collectortrace.ExportTra
RPCMethod: spanAttr(sp.Attributes, "rpc.method"),
RPCService: spanAttr(sp.Attributes, "rpc.service"),
PeerService: spanAttr(sp.Attributes, "peer.service", "server.address"),
ServiceVersion: getResourceAttr(rs.Resource, "service.version"),
DeploymentEnv: getResourceAttr(rs.Resource, "deployment.environment"),
ServiceVersion: serviceVersion,
DeploymentEnv: deploymentEnv,
}
// Pre-extracted exception info from span events.
excType, excMsg := extractException(sp.Events)
Expand Down Expand Up @@ -161,7 +172,16 @@ func (ls *logsService) Export(ctx context.Context, req *collectorlogs.ExportLogs
func (s *Server) exportLogs(ctx context.Context, req *collectorlogs.ExportLogsServiceRequest) (*collectorlogs.ExportLogsServiceResponse, error) {
cfg := s.cfg
now := time.Now().UnixNano()
count := 0
for _, rl := range req.ResourceLogs {
for _, sl := range rl.ScopeLogs {
count += len(sl.LogRecords)
}
}
batch := telemetrystore.Batch{}
if count > 0 {
batch.Logs = make([]telemetry.Log, 0, count)
}
for _, rl := range req.ResourceLogs {
resources, err := resourceAttributes(rl.Resource)
if err != nil {
Expand Down Expand Up @@ -218,7 +238,29 @@ func (ms *metricsService) Export(ctx context.Context, req *collectormetrics.Expo
func (s *Server) exportMetrics(ctx context.Context, req *collectormetrics.ExportMetricsServiceRequest) (*collectormetrics.ExportMetricsServiceResponse, error) {
cfg := s.cfg
now := time.Now().UnixNano()
count := 0
for _, rm := range req.ResourceMetrics {
for _, sm := range rm.ScopeMetrics {
for _, m := range sm.Metrics {
switch d := m.Data.(type) {
case *metricspb.Metric_Gauge:
count += len(d.Gauge.DataPoints)
case *metricspb.Metric_Sum:
count += len(d.Sum.DataPoints)
case *metricspb.Metric_Histogram:
count += len(d.Histogram.DataPoints)
case *metricspb.Metric_ExponentialHistogram:
count += len(d.ExponentialHistogram.DataPoints)
case *metricspb.Metric_Summary:
count += len(d.Summary.DataPoints)
}
}
}
}
batch := telemetrystore.Batch{}
if count > 0 {
batch.Metrics = make([]telemetry.Metric, 0, count)
}
for _, rm := range req.ResourceMetrics {
resources, err := resourceAttributes(rm.Resource)
if err != nil {
Expand Down
Loading
Loading