feat(otlp): semantic registry provider for embedded or live registries

[webern]

Human Summary

Adds a SemanticRegistryProvider to provide either the embedded or live semantic registry. This PR doesn't actually wire up the provider to the remote configuration client yet, so it would always return embedded. In the next PR we wire it up to return the registry updates.

#2719 adds the semantic RC decoder and subscription method; #2724 connects that subscription during ADP startup.

AI Summary

A span's HTTP status may appear as http.response.status_code or http.status_code. The semantic registry maps concepts like HTTP status code to ordered attribute-name fallbacks, expected types, and conditions. OTLP trace translation uses these mappings to read and normalize spans; APM stats uses the registry's peer-attribute names to group stats by the service or resource a span calls. The embedded mappings.json provides the default registry. Remote Configuration can supply a complete replacement, as in the Agent’s semantic registry.

This PR gives those consumers a SemanticRegistryProvider. It follows Saluki's pattern of passing providers through component builders: the OTLP source, OTLP decoder, and APM stats builder pass a handle to the translator or stats concentrator instead of having either consumer find a process-wide registry. The handle wraps an existing Remote Configuration Subscription, not a separate registry store or background task. The RC client publishes accepted snapshots to the subscription; provider clones see the same publications. With RC off (an inert subscription), or before any snapshot is accepted, snapshot() returns the embedded registry. A rejected update leaves the last accepted registry in place.

• The translator pins one Arc<Registry> per ResourceSpans and passes &Registry through the span lookups. An update arriving mid-batch takes effect on a later batch, without a subscription read per span.
• Before each input event buffer, APM stats checks the provider and updates its peer-tag key set if the registry changed. That set combines the registry's peer-attribute names with configured custom tags; it determines which span attributes distinguish stats groups. Existing aggregated stats are not rewritten.
• The embedded file must parse without skipped entries. Remote JSON keeps the concepts and fallbacks this binary understands and reports skipped entries; malformed documents or missing required registry metadata are rejected. As in the Agent, a fingerprint of the payload bytes tracks changes, rather than the producer-supplied metadata.content_hash label.

This PR makes the consumers ready for a subscribed provider; decoding APM_SEMANTIC_CORE_DD and wiring the subscription into ADP follow separately. Lookups that still use literal attribute keys, such as env and container ID, are unchanged.

Change Type

• New feature

How did you test this PR?

• make fmt, cargo check --workspace, cargo check --workspace --tests, make check-docs, make check-clippy at the stack tip.
• Filtered saluki-components tests: 549 passed at the stack tip.

References

• Closes: Traces(otlp+dd): Make the semantic registry replaceable at runtime #2432
• Related: feat(remote-config): add a remote configuration client #2713

[pr-commenter]

Binary Size Analysis (Agent Data Plane)

Baseline: 216e499 · Comparison: b286fb2 · diff
Analysis Configuration: stripped binaries · Pass/Fail Threshold: +5%
Sizes: 38.27 MiB (baseline) vs 38.24 MiB (comparison)
Size Change: -28.50 KiB (-0.07%)

✅ Binary size difference within threshold

Changes by Module

Module	File Size	Symbols
core	+69.92 KiB	8990
tonic_prost	+38.64 KiB	42
prost	-38.46 KiB	311
&mut serde_json	-28.30 KiB	94
saluki_metrics::mapped::MappedMetric<H>	-24.51 KiB	6
axum	-24.42 KiB	404
saluki_components::destinations::dogstatsd_client_telemetry	+22.75 KiB	31
serde_json	+16.46 KiB	459
anon.7bcce2bce1d7289ca2b67a9dabbae1a7.13.llvm.8132447645249006792	+16.18 KiB	1
anon.87e7c38bfd928d17fa28187081c51bf0.2.llvm.240572210426274404	-16.09 KiB	1
anon.87e7c38bfd928d17fa28187081c51bf0.768.llvm.240572210426274404	-14.62 KiB	1
anon.7bcce2bce1d7289ca2b67a9dabbae1a7.477.llvm.8132447645249006792	+14.62 KiB	1
otlp_protos::otlp_include::opentelemetry	-14.03 KiB	178
saluki_components::sources::otlp	-13.08 KiB	267
anon.87e7c38bfd928d17fa28187081c51bf0.766.llvm.240572210426274404	-12.84 KiB	1
anon.7bcce2bce1d7289ca2b67a9dabbae1a7.6.llvm.8132447645249006792	+12.83 KiB	1
quick_cache	+12.40 KiB	109
alloc	-12.18 KiB	1425
anon.9d1057ec6821611763596b0232be6ef0.693.llvm.6678353967787269721	-11.13 KiB	1
tonic	+11.12 KiB	324

Detailed Symbol Changes

    FILE SIZE        VM SIZE    
 --------------  -------------- 
  [NEW] +61.1Ki  [NEW] +61.0Ki    saluki_components::common::datadog::io::run_endpoint_io_loop::_{{closure}}::ha38a49b72e023545
  [NEW] +42.2Ki  [NEW] +42.0Ki    _<saluki_components::forwarders::otlp::OtlpForwarder as saluki_core::components::forwarders::Forwarder>::run::_{{closure}}::hb872765f165820b0
  [NEW] +40.6Ki  [NEW] +40.5Ki    agent_data_plane::cli::run::create_topology::_{{closure}}::h75562c7ea09bcbff
  [NEW] +37.4Ki  [NEW] +37.2Ki    _<saluki_components::transforms::apm_stats::ApmStats as saluki_core::components::transforms::Transform>::run::_{{closure}}::h79ca62828ea1b270
  [NEW] +34.5Ki  [NEW] +34.4Ki    agent_data_plane::cli::run::handle_run_command::_{{closure}}::h0f7a307e9cc0a16a
  [NEW] +31.0Ki  [NEW] +30.8Ki    agent_data_plane::cli::dogstatsd::run_dogstatsd_command::_{{closure}}::hdfcd4810bc0a3053
  [NEW] +29.0Ki  [NEW] +28.8Ki    datadog_agent_commons::ipc::client::RemoteAgentClient::connect::_{{closure}}::_{{closure}}::_{{closure}}::h0a198a9356f00b58
  [NEW] +28.6Ki  [NEW] +28.5Ki    agent_data_plane::dogstatsd_contexts::artifact::for_each_record::h1b4e32864ddc5f08
  [NEW] +26.4Ki  [NEW] +26.3Ki    saluki_components::destinations::dogstatsd_client_telemetry::DogStatsDClientTelemetry::record_metric::h6e61a932a14c058b
  [NEW] +25.6Ki  [NEW] +25.4Ki    agent_data_plane::internal::remote_agent::run_remote_agent_registration_loop::_{{closure}}::h044bf8d66771780c
  [DEL] -25.7Ki  [DEL] -25.6Ki    core::ops::function::FnOnce::call_once::h4d6ad556f476b8c9
  [DEL] -28.4Ki  [DEL] -28.2Ki    datadog_agent_commons::ipc::client::RemoteAgentClient::connect::_{{closure}}::_{{closure}}::_{{closure}}::h1bd6a026eeafe819
  [DEL] -28.7Ki  [DEL] -28.6Ki    agent_data_plane::dogstatsd_contexts::artifact::for_each_record::hb60d5262ffff75a0
  [DEL] -31.4Ki  [DEL] -31.2Ki    agent_data_plane::cli::dogstatsd::run_dogstatsd_command::_{{closure}}::h2f5246c806e7f392
  [DEL] -34.9Ki  [DEL] -34.7Ki    agent_data_plane::cli::run::handle_run_command::_{{closure}}::haa79196f741be08a
  [DEL] -35.8Ki  [DEL] -35.6Ki    _<saluki_components::transforms::apm_stats::ApmStats as saluki_core::components::transforms::Transform>::run::_{{closure}}::h8bffda5d06bacbfd
  [DEL] -40.5Ki  [DEL] -40.4Ki    agent_data_plane::cli::run::create_topology::_{{closure}}::h02ed82f16b54dbc0
  [DEL] -44.7Ki  [DEL] -44.5Ki    _<saluki_components::forwarders::otlp::OtlpForwarder as saluki_core::components::forwarders::Forwarder>::run::_{{closure}}::h252ca254b106e4c3
  [DEL] -58.6Ki  [DEL] -58.5Ki    saluki_components::common::datadog::io::run_endpoint_io_loop::_{{closure}}::h15a965fc1b5a6f1e
  -0.4% -56.3Ki  -0.7% -71.7Ki    [28806 Others]
  [ = ]       0 -87.5% -1.75Mi    _rjem_je_arena_emap_global
  -0.1% -28.5Ki  -5.5% -1.79Mi    TOTAL

[pr-commenter]

Regression Detector (Agent Data Plane)

Run ID: eef15432-dfd0-459d-84e4-5095cabeaa15
Baseline: 216e499a · Comparison: b286fb2f · diff

Optimization Goals: ✅ No significant changes detected

Fine details of change detection per experiment (5)

Experiments configured erratic: true are tagged (ignored) and skipped when determining which experiments regressed or improved. Experiments which are detected as erratic at runtime are tagged (erratic) to flag that the run's sample dispersion was high, but their regression / improvement signal still counts.

experiment	goal	Δ mean %	links
quality_gates_rss_dsd_ultraheavy	memory	⚪ +0.59	metrics profiles logs
quality_gates_rss_dsd_low	memory	⚪ +0.26	metrics profiles logs
quality_gates_rss_dsd_medium	memory	⚪ -0.00	metrics profiles logs
quality_gates_rss_dsd_heavy	memory	⚪ -0.07	metrics profiles logs
quality_gates_rss_idle	memory	⚪ -0.62	metrics profiles logs

Bounds Checks: ✅ Passed (5)

experiment	check	replicates	observed	links
quality_gates_rss_dsd_heavy	memory_usage	10/10	✅ 230 MiB ≤ 250 MiB	metrics profiles logs
quality_gates_rss_dsd_low	memory_usage	10/10	✅ 51.8 MiB ≤ 60 MiB	metrics profiles logs
quality_gates_rss_dsd_medium	memory_usage	10/10	✅ 91.3 MiB ≤ 100 MiB	metrics profiles logs
quality_gates_rss_dsd_ultraheavy	memory_usage	10/10	✅ 386 MiB ≤ 420 MiB	metrics profiles logs
quality_gates_rss_idle	memory_usage	10/10	✅ 33.2 MiB ≤ 40 MiB	metrics profiles logs

Explanation

A change is flagged as a regression when |Δ mean %| > 5.00% in the regressing direction for its optimization goal AND SMP marks the experiment as a regression (is_regression: true). Improvements use the matching criteria for the improving direction. Experiments configured erratic: true (tagged (ignored)) are skipped outright; experiments detected as erratic at runtime (tagged (erratic)) still count, since that flag describes sample dispersion rather than directional certainty. The Δ mean % cell is colored accordingly: 🟢 = improvement, 🔴 = regression, ⚪ = neutral. Reduction in CPU or memory is an improvement; reduction in ingress throughput is a regression. Experiments tagged (no analysis) show ⚠️ n/a: SMP ran them but produced no analysis, usually because a replicate failed and exhausted its retries. Check the SMP report for that experiment's replicate failures.

[datadog-datadog-prod-us1-2]

More details

The translator snapshots the live registry per ResourceSpans batch, while APM stats refresh peer-tag keys per event buffer; the completed static review found no concrete regression in these replacement boundaries.

Was this helpful? React 👍 or 👎

_{Open Bits AI session}

_{🤖 Bits Code Review · Commit 028e9d9 · @DataDog review to ask questions}

[datadog-datadog-prod-us1-2]

An Unimplemented response can leave previously published remote configuration active indefinitely after the Agent disables Remote Configuration.

_{Open Bits AI session}

_{🤖 Bits Code Review · Commit 3a398d6 · @DataDog review to ask questions}

---

Findings that could not be posted inline

lib/datadog-agent/remote-config/src/worker.rs:152

Withdraw assignments when Remote Configuration is disabled

After configurations have been published, an Unimplemented response—such as when Remote Configuration is disabled at runtime—only changes the retry schedule. The repository retains its files and subscribers continue using the last snapshot indefinitely instead of receiving an empty assignment, leaving disabled configuration active.

Was this helpful? React 👍 or 👎
_{🤖 Bits Code Review · @DataDog review to ask questions · Open Bits AI session}

[webern]

AI

Re the Bits finding "Withdraw assignments when Remote Configuration is disabled" (lib/datadog-agent/remote-config/src/worker.rs:152): I think this is intended, and it matches the Agent's own client. On an update error, including Unimplemented, the Go client keeps its configurations and only tells listeners that connectivity changed (pkg/config/remote/client/client.go:468-479 at 17ecddf4e3e); it withdraws nothing. The trace-agent therefore keeps its last applied rates too. Keeping the last accepted snapshot when the Agent stops serving is the client's last-known-good rule. The client only withdraws configurations when the Agent reports its cache as expired, which is the case the Agent itself withdraws on. The client is also no longer part of this PR's diff: it merged in #2713.

[datadog-datadog-prod-us1-2]

More details

Registry snapshots remain consistent within each trace batch, while APM stats refreshes peer-tag keys between input buffers; no concrete regression was identified.

Was this helpful? React 👍 or 👎

_{Open Bits AI session}

_{🤖 Bits Code Review · Commit b286fb2 · @DataDog review to ask questions}