Skip to content

Comments

[Tracing] Add experimental support for exporting traces as OTLP (.NET 6+)#8211

Open
zacharycmontoya wants to merge 14 commits intomasterfrom
zach.montoya/otlp-export-refined
Open

[Tracing] Add experimental support for exporting traces as OTLP (.NET 6+)#8211
zacharycmontoya wants to merge 14 commits intomasterfrom
zach.montoya/otlp-export-refined

Conversation

@zacharycmontoya
Copy link
Contributor

@zacharycmontoya zacharycmontoya commented Feb 17, 2026 •
edited
Loading

Summary of changes

Adds experimental support for exporting traces using OTLP rather than the Datadog MessagePack protocols. This allows the DD SDK to send traces (and trace metrics) to an OTel collector rather a Datadog Trace Agent, with limited support for non-APM products.

This feature is enabled by setting OTEL_TRACES_EXPORTER=otlp.

Note: This feature is currently only supported for .NET 6+. In addition to sending sampled traces to the OTLP collector, we must also send trace metrics to the OTLP collector to power the full Datadog APM experience. To do this, we use our built-in OTLP Metrics support which is limited to our .NET 6+ build.

Configuration

Configuration Details
OTEL_TRACES_EXPORTER=otlp Enables the OTLP traces export
OTEL_EXPORTER_OTLP_TRACES_ENDPOINT See the OTLP Exporter Configuration docs
OTEL_EXPORTER_OTLP_TRACES_HEADERS See the OTLP Exporter Configuration docs
OTEL_EXPORTER_OTLP_TRACES_PROTOCOL See the OTLP Exporter Configuration docs
OTEL_EXPORTER_OTLP_TRACES_TIMEOUT See the OTLP Exporter Configuration docs

Reason for change

We are seeing an increasing number of scenarios were users have applications instrumented with the OTel SDK sending data to OTel collectors, and they would like to get additional features offered by the DD SDK without needing to update their OTel collector deployments. Although there will be follow-up work, this provides the ability for users to write vendor-neutral API instrumentation and emit vendor-neutral telemetry data so DD SDK users don't have to feel locked in when setting up Datadog APM.

Implementation details

  • Configuration: Read all of the OTLP Traces and OTLP Metrics environment variables from configuration sources in the ExporterSettings class and add them as properties there, so that we have the ability to refresh the exporters if there are changes to the configurations.
  • Enums: Create a new Datadog.Trace.Agent.TracesEncoding enum to indicate which Datadog Message protocol or OTLP protocol we are using.
  • Serialization:
    • Create a new ISpanBufferSerializer interface to abstract away the MessagePack and OTLP serialization implementations from the SpanBuffer class. Notably, to make the OTLP serialization work with our pre-emptive encoding of spans into the SpanBuffer, add a FinishBody method so that we can create valid JSON after appending spans multiple times to the the resource_spans[0].scope_spans[0].spans array.
    • Create a OtlpTracesJsonSerializer implementation which can serialize the TraceChunkModel (and its Span objects) to its corresponding OTLP JSON structure.
  • Datadog Semantics: The logic for adding OTel resource attributes or special Datadog tags is separated out into the OtlpMapper static class. This class also contains the logic for converting the contents of a StatsBuffer (and its DDSketch objects) to its corresponding OTLP Histogram. The specific names and attributes needed for the OTLP Histogram to be recognized for APM trace metrics is in the process of being defined and may change.
  • Stats Aggregation: This change maintains backwards compatibility for the existing behavior of client-side stats computation with Datadog traces (DD_TRACE_STATS_COMPUTATION_ENABLED=true) while also updating the StatsAggregationKey so we have distinct histograms for errors and top-level spans.
  • Export:
    • Traces: Once the OTLP traces are serialized into a SpanBuffer by the OtlpTracesJsonSerializer, the ApiOtlp class will receive them in the IApi.SendTracesAsync method and handle the HTTP request logic. This includes handling status codes as defined by the OTLP specification.
    • Metrics: When the ApiOtlp class receives the current StatsBuffer in the IApi.SendStatsAsync method, it calls into the OtlpMapper class to convert it into its corresponding OTLP Histogram then hands this off to the pre-existing OtlpExporter from our built-in OTel Metrics support to handle the HTTP request logic.

Test coverage

This adds a new integration test OpenTelemetrySdkTests.SubmitsOtlpTraces that sends OTLP traces to the dd-apm-test-agent, retrieves the payload, normalizes it, and does snapshot testing.

Other details

Since this is a first-pass and experimental implementation, there is some additional work needed to polish the feature for general use. That includes the following:

  • In our OpenTelemetrySdkTests.SubmitsOtlpTraces test, assert against the trace metrics (sent with OTLP)
  • Stop reading the OTLP metrics settings in TracerSettings now that they are being read in ExporterSettings
  • I suspect that the implementation is currently exporting ALL spans and not dropping P0 chunks. As the PR is large enough, I'd like to tackle this in a follow-up PR.

@zacharycmontoya zacharycmontoya force-pushed the zach.montoya/otlp-export-refined branch 4 times, most recently from 54892c0 to 771b40f Compare February 18, 2026 05:27
@pr-commenter
Copy link

pr-commenter bot commented Feb 18, 2026 •
edited
Loading

Benchmarks

Benchmark execution time: 2026-02-20 22:17:56

Comparing candidate commit fe6d013 in PR branch zach.montoya/otlp-export-refined with baseline commit 7ff6b28 in branch master.

Found 13 performance improvements and 12 performance regressions! Performance is the same for 152 metrics, 15 unstable metrics.

scenario:Benchmarks.Trace.AgentWriterBenchmark.WriteAndFlushEnrichedTraces net472

  • 🟥 throughput [-45.796op/s; -44.270op/s] or [-7.747%; -7.489%]
  • 🟩 execution_time [-11.726ms; -11.158ms] or [-5.415%; -5.153%]

scenario:Benchmarks.Trace.AgentWriterBenchmark.WriteAndFlushEnrichedTraces net6.0

  • 🟩 execution_time [-20.770ms; -20.656ms] or [-16.883%; -16.791%]

scenario:Benchmarks.Trace.Asm.AppSecBodyBenchmark.AllCycleSimpleBody net472

  • 🟥 throughput [-66829.177op/s; -64359.244op/s] or [-6.716%; -6.468%]

scenario:Benchmarks.Trace.Asm.AppSecBodyBenchmark.AllCycleSimpleBody net6.0

  • 🟩 execution_time [-27.648ms; -21.417ms] or [-12.438%; -9.635%]

scenario:Benchmarks.Trace.Asm.AppSecBodyBenchmark.ObjectExtractorSimpleBody netcoreapp3.1

  • 🟩 execution_time [-22.457ms; -16.360ms] or [-10.360%; -7.547%]

scenario:Benchmarks.Trace.Asm.AppSecEncoderBenchmark.EncodeArgs netcoreapp3.1

  • 🟥 execution_time [+14.181ms; +16.499ms] or [+7.483%; +8.707%]

scenario:Benchmarks.Trace.Asm.AppSecEncoderBenchmark.EncodeLegacyArgs net6.0

  • 🟥 execution_time [+24.965ms; +25.127ms] or [+13.958%; +14.049%]

scenario:Benchmarks.Trace.AspNetCoreBenchmark.SendRequest net6.0

  • 🟥 execution_time [+105.456ms; +110.456ms] or [+112.849%; +118.200%]

scenario:Benchmarks.Trace.CIVisibilityProtocolWriterBenchmark.WriteAndFlushEnrichedTraces net6.0

  • 🟥 execution_time [+49.955ms; +56.360ms] or [+31.625%; +35.679%]
  • 🟥 throughput [-258.909op/s; -206.542op/s] or [-17.015%; -13.574%]

scenario:Benchmarks.Trace.CIVisibilityProtocolWriterBenchmark.WriteAndFlushEnrichedTraces netcoreapp3.1

  • 🟩 throughput [+93.871op/s; +196.940op/s] or [+6.660%; +13.972%]

scenario:Benchmarks.Trace.CharSliceBenchmark.OptimizedCharSlice netcoreapp3.1

  • 🟩 execution_time [-161.061µs; -153.046µs] or [-5.569%; -5.292%]
  • 🟩 throughput [+19.349op/s; +20.358op/s] or [+5.596%; +5.888%]

scenario:Benchmarks.Trace.CharSliceBenchmark.OptimizedCharSliceWithPool net6.0

  • 🟩 execution_time [-77.075µs; -69.792µs] or [-7.098%; -6.427%]
  • 🟩 throughput [+63.582op/s; +69.986op/s] or [+6.904%; +7.599%]

scenario:Benchmarks.Trace.ElasticsearchBenchmark.CallElasticsearch netcoreapp3.1

  • 🟥 throughput [-33756.002op/s; -26657.830op/s] or [-6.739%; -5.322%]

scenario:Benchmarks.Trace.Iast.StringAspectsBenchmark.StringConcatAspectBenchmark net6.0

  • 🟩 throughput [+131.964op/s; +318.971op/s] or [+6.410%; +15.493%]

scenario:Benchmarks.Trace.Log4netBenchmark.EnrichedLog net472

  • 🟩 execution_time [-10.957ms; -10.627ms] or [-5.441%; -5.277%]

scenario:Benchmarks.Trace.Log4netBenchmark.EnrichedLog netcoreapp3.1

  • 🟥 execution_time [+9.891ms; +10.843ms] or [+5.239%; +5.743%]

scenario:Benchmarks.Trace.RedisBenchmark.SendReceive net472

  • 🟩 throughput [+19749.323op/s; +22705.170op/s] or [+5.551%; +6.382%]

scenario:Benchmarks.Trace.SpanBenchmark.StartFinishScope net6.0

  • 🟥 execution_time [+15.262ms; +18.325ms] or [+7.711%; +9.258%]

scenario:Benchmarks.Trace.SpanBenchmark.StartFinishSpan netcoreapp3.1

  • 🟩 execution_time [-16.515ms; -11.602ms] or [-7.709%; -5.416%]

scenario:Benchmarks.Trace.SpanBenchmark.StartFinishTwoScopes net6.0

  • 🟥 execution_time [+14.953ms; +18.335ms] or [+7.549%; +9.256%]

scenario:Benchmarks.Trace.SpanBenchmark.StartFinishTwoScopes netcoreapp3.1

  • 🟥 execution_time [+15.516ms; +17.897ms] or [+7.865%; +9.072%]

@zacharycmontoya
Add a new IApi.Ping method to separate the initial ping operation fro…
3e65620 
…m actually sending traces
@zacharycmontoya
Refactor SpanBuffer to encapsulate TraceChunk serialization into a sm…
eaa5827 
…all ISpanBufferSerializer/SpanBufferMessagePackSerializer class to enable the future OTLP serialization
@zacharycmontoya
Introduce TracesEncoding enum on IApi, which ultimately determines th…
188717a 
…e SpanBuffer serialization.
@zacharycmontoya
Add OTLP JSON implementation and incorporate it into AgentWriter/Span…
4c850fd 
…Buffer. This does not yet handle issuing the outbound HTTP requests through the Api classes, nor does it handle configuration to enable the feature"
@zacharycmontoya
Refactor OTLP metrics to accommodate trace stats use cases. This incl…
0e63f7a 
…udes:

- Making the explicit histogram bounds configurable
- Make property setters available internally so we can construct the final snapshots from our DD Sketches (rather than building up the snapshots over time and reading them at export time)
@zacharycmontoya
Update our client-side stats aggregation classes to account for the w…
5e187e9 
…ay we must distinguish the histogram timeseries when submitting trace stats using OTLP metrics.
@zacharycmontoya
Add OTLP trace exporter and OTLP metrics exporter (needed for emittin…
9ed8b78 
…g APM trace stats) configurations to dd-trace-dotnet and read them in ExporterSettings
@zacharycmontoya
Refactor OtlpExporter so it can be used directly by our OTLP metrics …
8dc33c9 
…submission, coming in a following commit
@zacharycmontoya
Create new IApi implementations to send trace stats via OTLP metrics …
5aa4cbf 
…and traces via OTLP traces, and plug them in to TracerManagerFactory.GetAgentWriter so we can use the new feature end-to-end
@zacharycmontoya
Update ExporterSettings and pass it into the OtlpExporter
8c7e634
@zacharycmontoya
Add snapshot tests for OTLP export in test OpenTelemetrySdkTests.Subm…
e729239 
…itsOtlpTraces. Currently only http/json for Traces and http/protobuf for Metrics is tested.
@zacharycmontoya zacharycmontoya force-pushed the zach.montoya/otlp-export-refined branch from 771b40f to e729239 Compare February 18, 2026 20:50
@dd-trace-dotnet-ci-bot
Copy link

dd-trace-dotnet-ci-bot bot commented Feb 18, 2026 •
edited
Loading

Execution-Time Benchmarks Report ⏱️

Execution-time results for samples comparing This PR (8211) and master.

⚠️ Potential regressions detected

HttpMessageHandler

Metric Master (Mean ± 95% CI) Current (Mean ± 95% CI) Change Status
.NET 6 - CallTarget+Inlining+NGEN
runtime.dotnet.mem.committed57.52 ± (57.36 - 57.69) MB63.02 ± (62.97 - 63.07) MB+9.6%❌⬆️
Full Metrics Comparison

FakeDbCommand

Metric Master (Mean ± 95% CI) Current (Mean ± 95% CI) Change Status
.NET Framework 4.8 - Baseline
duration76.29 ± (76.25 - 76.64) ms74.96 ± (74.84 - 75.11) ms-1.7%✅
.NET Framework 4.8 - Bailout
duration80.28 ± (80.11 - 80.50) ms79.89 ± (79.74 - 80.13) ms-0.5%✅
.NET Framework 4.8 - CallTarget+Inlining+NGEN
duration1096.91 ± (1095.68 - 1102.21) ms1092.37 ± (1093.61 - 1101.21) ms-0.4%✅
.NET Core 3.1 - Baseline
process.internal_duration_ms23.27 ± (23.22 - 23.32) ms22.93 ± (22.89 - 22.98) ms-1.5%✅
process.time_to_main_ms88.90 ± (88.67 - 89.12) ms86.80 ± (86.60 - 87.00) ms-2.4%✅
runtime.dotnet.exceptions.count0 ± (0 - 0)0 ± (0 - 0)+0.0%✅
runtime.dotnet.mem.committed10.93 ± (10.93 - 10.94) MB10.91 ± (10.91 - 10.92) MB-0.2%✅
runtime.dotnet.threads.count12 ± (12 - 12)12 ± (12 - 12)+0.0%✅
.NET Core 3.1 - Bailout
process.internal_duration_ms23.08 ± (23.02 - 23.14) ms22.78 ± (22.73 - 22.84) ms-1.3%✅
process.time_to_main_ms90.22 ± (90.04 - 90.40) ms87.99 ± (87.79 - 88.19) ms-2.5%✅
runtime.dotnet.exceptions.count0 ± (0 - 0)0 ± (0 - 0)+0.0%✅
runtime.dotnet.mem.committed10.95 ± (10.95 - 10.96) MB10.96 ± (10.95 - 10.96) MB+0.1%✅⬆️
runtime.dotnet.threads.count13 ± (13 - 13)13 ± (13 - 13)+0.0%✅
.NET Core 3.1 - CallTarget+Inlining+NGEN
process.internal_duration_ms233.04 ± (229.37 - 236.70) ms247.88 ± (243.92 - 251.84) ms+6.4%✅⬆️
process.time_to_main_ms504.89 ± (504.10 - 505.68) ms496.53 ± (495.84 - 497.22) ms-1.7%✅
runtime.dotnet.exceptions.count0 ± (0 - 0)0 ± (0 - 0)+0.0%✅
runtime.dotnet.mem.committed47.60 ± (47.58 - 47.62) MB47.73 ± (47.71 - 47.75) MB+0.3%✅⬆️
runtime.dotnet.threads.count28 ± (28 - 28)28 ± (28 - 28)+0.0%✅⬆️
.NET 6 - Baseline
process.internal_duration_ms22.01 ± (21.97 - 22.06) ms21.53 ± (21.49 - 21.58) ms-2.2%✅
process.time_to_main_ms78.03 ± (77.85 - 78.20) ms75.36 ± (75.18 - 75.54) ms-3.4%✅
runtime.dotnet.exceptions.count0 ± (0 - 0)0 ± (0 - 0)+0.0%✅
runtime.dotnet.mem.committed10.62 ± (10.62 - 10.62) MB10.63 ± (10.63 - 10.63) MB+0.1%✅⬆️
runtime.dotnet.threads.count10 ± (10 - 10)10 ± (10 - 10)+0.0%✅
.NET 6 - Bailout
process.internal_duration_ms21.77 ± (21.71 - 21.82) ms21.49 ± (21.44 - 21.55) ms-1.3%✅
process.time_to_main_ms78.75 ± (78.56 - 78.95) ms76.50 ± (76.30 - 76.70) ms-2.9%✅
runtime.dotnet.exceptions.count0 ± (0 - 0)0 ± (0 - 0)+0.0%✅
runtime.dotnet.mem.committed10.67 ± (10.66 - 10.67) MB10.72 ± (10.71 - 10.72) MB+0.5%✅⬆️
runtime.dotnet.threads.count11 ± (11 - 11)11 ± (11 - 11)+0.0%✅
.NET 6 - CallTarget+Inlining+NGEN
process.internal_duration_ms254.93 ± (251.33 - 258.54) ms1289.99 ± (1288.21 - 1291.77) ms+406.0%✅⬆️
process.time_to_main_ms482.71 ± (482.02 - 483.41) ms477.62 ± (476.55 - 478.70) ms-1.1%✅
runtime.dotnet.exceptions.count0 ± (0 - 0)18 ± (18 - 18)+0.0%✅
runtime.dotnet.mem.committed48.40 ± (48.38 - 48.42) MB50.02 ± (49.99 - 50.05) MB+3.3%✅⬆️
runtime.dotnet.threads.count28 ± (28 - 28)28 ± (28 - 28)+0.0%✅⬆️
.NET 8 - Baseline
process.internal_duration_ms20.25 ± (20.21 - 20.29) ms19.81 ± (19.77 - 19.85) ms-2.2%✅
process.time_to_main_ms76.87 ± (76.71 - 77.03) ms74.63 ± (74.46 - 74.80) ms-2.9%✅
runtime.dotnet.exceptions.count0 ± (0 - 0)0 ± (0 - 0)+0.0%✅
runtime.dotnet.mem.committed7.68 ± (7.68 - 7.69) MB7.69 ± (7.68 - 7.70) MB+0.1%✅⬆️
runtime.dotnet.threads.count10 ± (10 - 10)10 ± (10 - 10)+0.0%✅
.NET 8 - Bailout
process.internal_duration_ms20.15 ± (20.10 - 20.21) ms19.90 ± (19.86 - 19.94) ms-1.3%✅
process.time_to_main_ms77.50 ± (77.32 - 77.67) ms76.07 ± (75.88 - 76.25) ms-1.8%✅
runtime.dotnet.exceptions.count0 ± (0 - 0)0 ± (0 - 0)+0.0%✅
runtime.dotnet.mem.committed7.74 ± (7.73 - 7.75) MB7.72 ± (7.71 - 7.72) MB-0.2%✅
runtime.dotnet.threads.count11 ± (11 - 11)11 ± (11 - 11)+0.0%✅
.NET 8 - CallTarget+Inlining+NGEN
process.internal_duration_ms193.52 ± (192.88 - 194.16) ms1218.21 ± (1216.42 - 1219.99) ms+529.5%✅⬆️
process.time_to_main_ms462.15 ± (461.54 - 462.76) ms458.26 ± (457.02 - 459.50) ms-0.8%✅
runtime.dotnet.exceptions.count0 ± (0 - 0)18 ± (18 - 18)+0.0%✅
runtime.dotnet.mem.committed36.15 ± (36.11 - 36.18) MB37.49 ± (37.46 - 37.51) MB+3.7%✅⬆️
runtime.dotnet.threads.count27 ± (27 - 27)27 ± (26 - 27)-0.4%✅

HttpMessageHandler

Metric Master (Mean ± 95% CI) Current (Mean ± 95% CI) Change Status
.NET Framework 4.8 - Baseline
duration219.33 ± (219.40 - 220.80) ms214.65 ± (214.81 - 216.33) ms-2.1%✅
.NET Framework 4.8 - Bailout
duration224.47 ± (224.39 - 225.88) ms219.57 ± (220.38 - 222.57) ms-2.2%✅
.NET Framework 4.8 - CallTarget+Inlining+NGEN
duration1247.22 ± (1247.00 - 1255.89) ms1226.94 ± (1228.06 - 1235.65) ms-1.6%✅
.NET Core 3.1 - Baseline
process.internal_duration_ms213.38 ± (212.79 - 213.96) ms207.44 ± (206.84 - 208.05) ms-2.8%✅
process.time_to_main_ms92.32 ± (92.02 - 92.63) ms89.90 ± (89.64 - 90.17) ms-2.6%✅
runtime.dotnet.exceptions.count3 ± (3 - 3)3 ± (3 - 3)+0.0%✅
runtime.dotnet.mem.committed15.88 ± (15.86 - 15.90) MB15.90 ± (15.89 - 15.92) MB+0.1%✅⬆️
runtime.dotnet.threads.count20 ± (20 - 20)20 ± (20 - 20)-0.8%✅
.NET Core 3.1 - Bailout
process.internal_duration_ms213.72 ± (213.11 - 214.33) ms208.37 ± (207.80 - 208.94) ms-2.5%✅
process.time_to_main_ms94.54 ± (94.26 - 94.82) ms91.11 ± (90.88 - 91.33) ms-3.6%✅
runtime.dotnet.exceptions.count3 ± (3 - 3)3 ± (3 - 3)+0.0%✅
runtime.dotnet.mem.committed15.92 ± (15.90 - 15.93) MB15.89 ± (15.88 - 15.91) MB-0.2%✅
runtime.dotnet.threads.count21 ± (21 - 21)21 ± (21 - 21)-0.2%✅
.NET Core 3.1 - CallTarget+Inlining+NGEN
process.internal_duration_ms443.28 ± (439.66 - 446.91) ms445.91 ± (441.82 - 450.00) ms+0.6%✅⬆️
process.time_to_main_ms527.40 ± (526.25 - 528.55) ms517.82 ± (516.99 - 518.65) ms-1.8%✅
runtime.dotnet.exceptions.count3 ± (3 - 3)3 ± (3 - 3)+0.0%✅
runtime.dotnet.mem.committed58.06 ± (57.92 - 58.19) MB58.03 ± (57.89 - 58.17) MB-0.1%✅
runtime.dotnet.threads.count30 ± (30 - 30)30 ± (30 - 30)-0.5%✅
.NET 6 - Baseline
process.internal_duration_ms220.30 ± (219.42 - 221.19) ms212.71 ± (212.09 - 213.32) ms-3.4%✅
process.time_to_main_ms80.81 ± (80.52 - 81.10) ms77.63 ± (77.43 - 77.82) ms-3.9%✅
runtime.dotnet.exceptions.count4 ± (4 - 4)4 ± (4 - 4)+0.0%✅
runtime.dotnet.mem.committed16.13 ± (16.11 - 16.14) MB16.22 ± (16.20 - 16.23) MB+0.6%✅⬆️
runtime.dotnet.threads.count20 ± (19 - 20)19 ± (19 - 19)-1.0%✅
.NET 6 - Bailout
process.internal_duration_ms225.40 ± (224.03 - 226.76) ms212.99 ± (212.34 - 213.63) ms-5.5%✅
process.time_to_main_ms82.70 ± (82.39 - 83.01) ms79.23 ± (78.98 - 79.48) ms-4.2%✅
runtime.dotnet.exceptions.count4 ± (4 - 4)4 ± (4 - 4)+0.0%✅
runtime.dotnet.mem.committed16.14 ± (16.12 - 16.15) MB16.24 ± (16.22 - 16.26) MB+0.6%✅⬆️
runtime.dotnet.threads.count20 ± (20 - 21)20 ± (20 - 21)+0.0%✅⬆️
.NET 6 - CallTarget+Inlining+NGEN
process.internal_duration_ms478.46 ± (473.92 - 482.99) ms1550.64 ± (1543.90 - 1557.37) ms+224.1%✅⬆️
process.time_to_main_ms496.51 ± (495.42 - 497.59) ms488.97 ± (488.11 - 489.83) ms-1.5%✅
runtime.dotnet.exceptions.count4 ± (4 - 4)22 ± (22 - 22)+450.0%✅⬆️
runtime.dotnet.mem.committed57.52 ± (57.36 - 57.69) MB63.02 ± (62.97 - 63.07) MB+9.6%❌⬆️
runtime.dotnet.threads.count30 ± (30 - 30)30 ± (30 - 30)+1.6%✅⬆️
.NET 8 - Baseline
process.internal_duration_ms218.07 ± (217.25 - 218.88) ms211.73 ± (211.08 - 212.38) ms-2.9%✅
process.time_to_main_ms79.17 ± (78.91 - 79.43) ms76.78 ± (76.59 - 76.98) ms-3.0%✅
runtime.dotnet.exceptions.count4 ± (4 - 4)4 ± (4 - 4)+0.0%✅
runtime.dotnet.mem.committed11.47 ± (11.46 - 11.49) MB11.55 ± (11.54 - 11.57) MB+0.7%✅⬆️
runtime.dotnet.threads.count19 ± (19 - 19)19 ± (19 - 19)-0.0%✅
.NET 8 - Bailout
process.internal_duration_ms217.19 ± (216.45 - 217.93) ms211.12 ± (210.44 - 211.80) ms-2.8%✅
process.time_to_main_ms80.21 ± (79.97 - 80.45) ms77.97 ± (77.76 - 78.17) ms-2.8%✅
runtime.dotnet.exceptions.count4 ± (4 - 4)4 ± (4 - 4)+0.0%✅
runtime.dotnet.mem.committed11.51 ± (11.49 - 11.52) MB11.60 ± (11.58 - 11.61) MB+0.8%✅⬆️
runtime.dotnet.threads.count20 ± (20 - 20)20 ± (20 - 20)-0.1%✅
.NET 8 - CallTarget+Inlining+NGEN
process.internal_duration_ms492.94 ± (488.16 - 497.73) ms1528.58 ± (1525.79 - 1531.36) ms+210.1%✅⬆️
process.time_to_main_ms480.67 ± (479.84 - 481.50) ms472.40 ± (471.16 - 473.65) ms-1.7%✅
runtime.dotnet.exceptions.count4 ± (4 - 4)22 ± (22 - 22)+450.0%✅⬆️
runtime.dotnet.mem.committed49.80 ± (49.77 - 49.84) MB54.38 ± (54.34 - 54.41) MB+9.2%✅⬆️
runtime.dotnet.threads.count29 ± (29 - 29)29 ± (29 - 29)+0.7%✅⬆️
Comparison explanation

Execution-time benchmarks measure the whole time it takes to execute a program, and are intended to measure the one-off costs. Cases where the execution time results for the PR are worse than latest master results are highlighted in **red**. The following thresholds were used for comparing the execution times:

  • Welch test with statistical test for significance of 5%
  • Only results indicating a difference greater than 5% and 5 ms are considered.

Note that these results are based on a single point-in-time result for each branch. For full results, see the dashboard.

Graphs show the p99 interval based on the mean and StdDev of the test run, as well as the mean value of the run (shown as a diamond below the graph).

Duration charts
FakeDbCommand (.NET Framework 4.8) 
gantt
    title Execution time (ms) FakeDbCommand (.NET Framework 4.8)
    dateFormat  x
    axisFormat %Q
    todayMarker off
    section Baseline
    This PR (8211) - mean (75ms)  : 73, 77
    master - mean (76ms)  : 73, 79

    section Bailout
    This PR (8211) - mean (80ms)  : 78, 82
    master - mean (80ms)  : 78, 83

    section CallTarget+Inlining+NGEN
    This PR (8211) - mean (1,097ms)  : 1042, 1153
    master - mean (1,099ms)  : 1050, 1148
Loading
FakeDbCommand (.NET Core 3.1) 
gantt
    title Execution time (ms) FakeDbCommand (.NET Core 3.1)
    dateFormat  x
    axisFormat %Q
    todayMarker off
    section Baseline
    This PR (8211) - mean (117ms)  : 114, 120
    master - mean (119ms)  : 115, 123

    section Bailout
    This PR (8211) - mean (118ms)  : 115, 121
    master - mean (121ms)  : 117, 124

    section CallTarget+Inlining+NGEN
    This PR (8211) - mean (776ms)  : 716, 835
    master - mean (779ms)  : 720, 838
Loading
FakeDbCommand (.NET 6) 
gantt
    title Execution time (ms) FakeDbCommand (.NET 6)
    dateFormat  x
    axisFormat %Q
    todayMarker off
    section Baseline
    This PR (8211) - mean (104ms)  : 100, 107
    master - mean (107ms)  : 104, 110

    section Bailout
    This PR (8211) - mean (105ms)  : 102, 107
    master - mean (107ms)  : 105, 110

    section CallTarget+Inlining+NGEN
    This PR (8211) - mean (1,801ms)  : crit, 1782, 1820
    master - mean (771ms)  : 706, 836
Loading
FakeDbCommand (.NET 8) 
gantt
    title Execution time (ms) FakeDbCommand (.NET 8)
    dateFormat  x
    axisFormat %Q
    todayMarker off
    section Baseline
    This PR (8211) - mean (102ms)  : 99, 106
    master - mean (105ms)  : 102, 108

    section Bailout
    This PR (8211) - mean (104ms)  : 101, 107
    master - mean (106ms)  : 103, 108

    section CallTarget+Inlining+NGEN
    This PR (8211) - mean (1,714ms)  : crit, 1682, 1746
    master - mean (688ms)  : 663, 714
Loading
HttpMessageHandler (.NET Framework 4.8) 
gantt
    title Execution time (ms) HttpMessageHandler (.NET Framework 4.8)
    dateFormat  x
    axisFormat %Q
    todayMarker off
    section Baseline
    This PR (8211) - mean (216ms)  : 204, 227
    master - mean (220ms)  : 210, 231

    section Bailout
    This PR (8211) - mean (221ms)  : 205, 238
    master - mean (225ms)  : 214, 236

    section CallTarget+Inlining+NGEN
    This PR (8211) - mean (1,232ms)  : 1175, 1289
    master - mean (1,251ms)  : 1183, 1320
Loading
HttpMessageHandler (.NET Core 3.1) 
gantt
    title Execution time (ms) HttpMessageHandler (.NET Core 3.1)
    dateFormat  x
    axisFormat %Q
    todayMarker off
    section Baseline
    This PR (8211) - mean (307ms)  : 296, 319
    master - mean (316ms)  : 305, 327

    section Bailout
    This PR (8211) - mean (309ms)  : 297, 322
    master - mean (318ms)  : 306, 331

    section CallTarget+Inlining+NGEN
    This PR (8211) - mean (1,000ms)  : 930, 1070
    master - mean (1,015ms)  : 958, 1071
Loading
HttpMessageHandler (.NET 6) 
gantt
    title Execution time (ms) HttpMessageHandler (.NET 6)
    dateFormat  x
    axisFormat %Q
    todayMarker off
    section Baseline
    This PR (8211) - mean (300ms)  : 289, 311
    master - mean (311ms)  : 294, 328

    section Bailout
    This PR (8211) - mean (302ms)  : 287, 316
    master - mean (318ms)  : 293, 343

    section CallTarget+Inlining+NGEN
    This PR (8211) - mean (2,076ms)  : crit, 1976, 2175
    master - mean (1,017ms)  : 943, 1091
Loading
HttpMessageHandler (.NET 8) 
gantt
    title Execution time (ms) HttpMessageHandler (.NET 8)
    dateFormat  x
    axisFormat %Q
    todayMarker off
    section Baseline
    This PR (8211) - mean (299ms)  : 285, 314
    master - mean (308ms)  : 295, 322

    section Bailout
    This PR (8211) - mean (300ms)  : 286, 314
    master - mean (309ms)  : 295, 323

    section CallTarget+Inlining+NGEN
    This PR (8211) - mean (2,038ms)  : crit, 2001, 2075
    master - mean (1,010ms)  : 939, 1082
Loading

@zacharycmontoya zacharycmontoya changed the title [Tracing] Add support for exporting traces as OTLP [Tracing] Add support for exporting traces as OTLP (.NET 6+) Feb 18, 2026
@zacharycmontoya zacharycmontoya changed the title [Tracing] Add support for exporting traces as OTLP (.NET 6+) [Tracing] Add experimental support for exporting traces as OTLP (.NET 6+) Feb 19, 2026
@zacharycmontoya zacharycmontoya marked this pull request as ready for review February 19, 2026 01:13
@zacharycmontoya zacharycmontoya requested review from a team as code owners February 19, 2026 01:13
chatgpt-codex-connector[bot]
chatgpt-codex-connector bot reviewed Feb 19, 2026
View reviewed changes
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e729239afd

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

tracer/src/Datadog.Trace/Configuration/ExporterSettings.cs
AgentUri = traceSettings.AgentUri;

var otlpTraceSettings = GetOtlpTracesTransport(signalEndpoint: rawSettings.OtlpTracesEndpoint, generalEndpoint: rawSettings.OtlpEndpoint, signalProtocol: rawSettings.OtlpTracesProtocol, generalProtocol: rawSettings.OtlpProtocol);
TracesEncoding = otlpTraceSettings.Encoding;
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot Feb 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Gate OTLP transport behind OTEL_TRACES_EXPORTER

TracesEncoding is unconditionally overwritten from GetOtlpTracesTransport(...), so on .NET 6+ the tracer switches to ManagedApiOtlp even when OTEL_TRACES_EXPORTER is not set. This changes the default behavior from Datadog agent transport to OTLP defaults (localhost:4318), which can silently break trace delivery for existing deployments that only run the Datadog agent.

Useful? React with 👍 / 👎.

tracer/src/Datadog.Trace/Configuration/ExporterSettings.Shared.cs
}
else if (protocol != OtlpProtocol.Grpc)
{
if (signalEndpoint!.EndsWith("/") && TryGetAgentUriAndTransport($"{generalEndpoint!}{defaultHttpRelativePath}", origin, encoding, protocol, out var httpSettings))
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot Feb 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Handle OTLP base endpoint fallback without null dereference

In the generalEndpoint fallback path, the slash logic dereferences signalEndpoint!; if users configure only OTEL_EXPORTER_OTLP_ENDPOINT (without a signal-specific endpoint), this throws NullReferenceException during settings initialization. This makes a common OTLP configuration path fail before export starts.

Useful? React with 👍 / 👎.

tracer/src/Datadog.Trace/Agent/AgentWriter.cs
Comment on lines +101 to +102
TracesEncoding.OtlpJson => new OtlpTracesJsonSerializer(),
_ => new SpanBufferMessagePackSerializer(SpanFormatterResolver.Instance),
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot Feb 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Avoid MessagePack serialization for OTLP protobuf mode

When TracesEncoding is OtlpProtobuf (for grpc/http/protobuf protocol settings), this switch falls back to SpanBufferMessagePackSerializer instead of an OTLP serializer. That causes OTLP protobuf configurations to emit Datadog MessagePack payloads, which OTLP collectors will not parse correctly.

Useful? React with 👍 / 👎.

tracer/src/Datadog.Trace/Agent/ApiOtlp.cs
Comment on lines 49 to 52
_apiRequestFactory = apiRequestFactory;
_tracesEncoding = exporterSettings.TracesEncoding;
_tracesEndpoint = exporterSettings.OtlpTracesEndpoint;
_statsEndpoint = exporterSettings.OtlpMetricsEndpoint;
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot Feb 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Apply OTLP trace headers and timeout in ApiOtlp

The OTLP trace config keys for OTEL_EXPORTER_OTLP_TRACES_HEADERS and OTEL_EXPORTER_OTLP_TRACES_TIMEOUT are parsed, but ApiOtlp only consumes encoding and endpoints here and never wires trace headers/timeout into outgoing trace requests. In environments where the collector requires auth headers or tuned timeouts, trace export will fail despite the documented settings.

Useful? React with 👍 / 👎.

chatgpt-codex-connector[bot]
chatgpt-codex-connector bot reviewed Feb 19, 2026
View reviewed changes
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e729239afd

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

tracer/src/Datadog.Trace/Configuration/ExporterSettings.cs
AgentUri = traceSettings.AgentUri;

var otlpTraceSettings = GetOtlpTracesTransport(signalEndpoint: rawSettings.OtlpTracesEndpoint, generalEndpoint: rawSettings.OtlpEndpoint, signalProtocol: rawSettings.OtlpTracesProtocol, generalProtocol: rawSettings.OtlpProtocol);
TracesEncoding = otlpTraceSettings.Encoding;
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot Feb 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P0 Badge Gate OTLP trace encoding behind OTEL_TRACES_EXPORTER

This constructor always overwrites TracesEncoding with GetOtlpTracesTransport(...) regardless of OtelTracesExporter, so when OTEL_TRACES_EXPORTER is unset the default OTLP trace encoding still gets selected and .NET 6+ flows into ManagedApiOtlp instead of the Datadog trace agent path. That changes the default transport behavior for all users on supported runtimes, even though OTLP export is documented as opt-in.

Useful? React with 👍 / 👎.

tracer/src/Datadog.Trace/Configuration/ExporterSettings.Shared.cs
}
else if (protocol != OtlpProtocol.Grpc)
{
if (signalEndpoint!.EndsWith("/") && TryGetAgentUriAndTransport($"{generalEndpoint!}{defaultHttpRelativePath}", origin, encoding, protocol, out var httpSettings))
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot Feb 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Use general endpoint safely when deriving OTLP HTTP URL

When only OTEL_EXPORTER_OTLP_ENDPOINT is configured (without signal-specific endpoint), this branch dereferences signalEndpoint! to decide slash handling, causing a NullReferenceException during settings initialization. It also checks the wrong variable for trailing slash, so URL composition can be incorrect even when both endpoints are present.

Useful? React with 👍 / 👎.

tracer/src/Datadog.Trace/OpenTelemetry/Metrics/OtlpMetricsSerializer.cs
}

if (metric.SnapshotCount > 0)
if (metric.SnapshotCount > 0 & metric.SnapshotMin != double.NaN)
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot Feb 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Skip NaN histogram min/max when serializing OTLP metrics

The NaN guard here is ineffective because x != double.NaN is always true in .NET, so min/max are still serialized whenever SnapshotCount > 0. In this commit, OTLP trace-metric points set SnapshotMin/SnapshotMax to NaN, so the exporter emits NaN min/max fields instead of omitting them, which can break downstream handling of histogram points.

Useful? React with 👍 / 👎.

ida613
ida613 reviewed Feb 19, 2026
View reviewed changes
tracer/src/Datadog.Trace/Configuration/supported-configurations-docs.yaml
OTEL_EXPORTER_OTLP_TRACES_PROTOCOL: |
Configuration key to set the OTLP protocol for traces export.
Takes precedence over <see cref="ExporterOtlpProtocol"/>.
Valid values: http/json, defaults to http/json.
Copy link

@ida613 ida613 Feb 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

should the valid values be http/json, http/protobuf and grpc?

Copy link
Contributor Author

@zacharycmontoya zacharycmontoya Feb 19, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Eventually yes, but right now I've only implemented http/json so I haven't included http/protobuf or grpc

@zacharycmontoya zacharycmontoya mentioned this pull request Feb 23, 2026
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@ida613 ida613 ida613 left review comments

@chatgpt-codex-connector chatgpt-codex-connector[bot] chatgpt-codex-connector[bot] left review comments

At least 1 approving review is required to merge this pull request.

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@zacharycmontoya @ida613