Skip to content

llm-d observability: finish remaining OpenTelemetry GenAI semconv and telemetry refactoring - #2516

Open
PlateauGao wants to merge 1 commit into
llm-d:mainfrom
PlateauGao:feature/otel-semconv-proxy-and-remaining-llmd
Open

llm-d observability: finish remaining OpenTelemetry GenAI semconv and telemetry refactoring#2516
PlateauGao wants to merge 1 commit into
llm-d:mainfrom
PlateauGao:feature/otel-semconv-proxy-and-remaining-llmd

Conversation

@PlateauGao

@PlateauGao PlateauGao commented Aug 23, 2026

Copy link
Copy Markdown
Contributor

What type of PR is this?
/kind cleanup

What this PR does / why we need it:

This is a follow-up PR to #2503 (implemented subtasks 1&2 of #708), completing the migration of all remaining hardcoded llm_d.* and gen_ai.* OpenTelemetry attributes across the codebase to centralized & type-safe semantic conventions in pkg/common/observability/semconv

Summary of Changes:

  1. Semantic Conventions (pkg/common/observability/semconv):
    • Expanded llm_d.go with typed attribute.Key definitions and constructors for remaining legacy usages
    • Added unit test

Verification:

  • local go test
  • local make lint

Which issue(s) this PR fixes:
#708

Release note (write NONE if no user-facing change):

NONE

@github-actions github-actions Bot added size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files. area/sidecar area/epp area/telemetry area/scheduling area/kvcache labels Aug 23, 2026
@PlateauGao PlateauGao changed the title Feature/otel semconv part2 for llmd telemetry Draft : Feature/otel semconv part2 for llmd telemetry Aug 23, 2026
@PlateauGao
PlateauGao force-pushed the feature/otel-semconv-proxy-and-remaining-llmd branch from a7ddf64 to ee2107f Compare September 9, 2026 14:32
@PlateauGao PlateauGao changed the title Draft : Feature/otel semconv part2 for llmd telemetry llm-d observability: finish remaining OpenTelemetry GenAI semconv and telemetry refactoring Sep 9, 2026
@PlateauGao
PlateauGao marked this pull request as ready for review September 9, 2026 14:47
@PlateauGao

Copy link
Copy Markdown
Contributor Author

/assign @gyliu513

Comment thread pkg/kvevents/pool.go Outdated
Comment thread pkg/common/observability/semconv/llm_d.go Outdated
Comment thread pkg/epp/scheduling/scheduler_profile_picker_tracing_test.go Outdated
@PlateauGao
PlateauGao force-pushed the feature/otel-semconv-proxy-and-remaining-llmd branch from ee2107f to 29ec668 Compare September 10, 2026 19:46
Complete migration of all remaining llm_d.* and gen_ai.* OpenTelemetry
attributes across llm-d-router to centralized semantic conventions in
pkg/common/observability/semconv:
- Expand semconv/llm_d.go with typed keys and constructors for all EPP
  scheduling/filtering/scoring, token producer, KV cache indexing/lookup,
  KV events pub/sub, and sidecar proxy attributes.
- Migrate EPP scheduling, profile handler, data producers, and tests to
  semconv keys and constructors.
- Migrate kvcache indexer, prefix match, traced scorer, and kvblock to
  semconv.
- Migrate kvevents pool and ZeroMQ subscriber to semconv.
- Migrate sidecar proxy connectors (nixlv2, p2p, decode, pd_common,
  chat completions) to semconv.
- Add unit tests for all new attribute constructors in semconv/llm_d_test.go.

Signed-off-by: Yuan Gao <plateau451@gmail.com>
@PlateauGao
PlateauGao force-pushed the feature/otel-semconv-proxy-and-remaining-llmd branch from 29ec668 to a98cd2a Compare September 10, 2026 20:31
@gyliu513

Copy link
Copy Markdown
Member

lgtm, thanks @PlateauGao !

@ahg-g can you help approve? Thanks

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants