stats/opentelemetry: OpenTelemetry Semantic Convention compliance

[pellared]

1. Is https://pkg.go.dev/google.golang.org/grpc/stats/opentelemetry stable? It part of v1 release and there is nothing in the docs indicating that it is an experimental feature.
2. Is the plan of stat/opentelemetry to follow OpenTelemetry Semantic Conventions https://opentelemetry.io/docs/specs/semconv/rpc/grpc/? Currently it looks like the stats handler is supposed to follow https://grpc.io/docs/guides/opentelemetry-metrics/ instead that is not compliant with OpenTelemetry Semantic Conventions. What is the reason behind the differences?

[dfawley]

@pellared

We designed the OTel metrics for all gRPC languages in https://github.com/grpc/proposal/blob/master/A66-otel-stats.md. grpc-go should be matching that design - anything else would be considered a bug. @yashykt can provide more background on the design and why we chose to do things differently from how OTel originally defined them.

Edit: and yes, it is stable, since it is not marked experimental and was included in a 1.x release.

[yashykt]

Is the plan of stat/opentelemetry to follow OpenTelemetry Semantic Conventions https://opentelemetry.io/docs/specs/semconv/rpc/grpc/? Currently it looks like the stats handler is supposed to follow https://grpc.io/docs/guides/opentelemetry-metrics/ instead that is not compliant with OpenTelemetry Semantic Conventions. What is the reason behind the differences?

https://github.com/grpc/proposal/blob/master/A66-otel-stats.md#rationale talks about this. Additionally, we were able to build on top of this, and propose additional metrics that make sense for gRPC that would not necessarily make sense for other RPC systems. Refer

• https://github.com/grpc/proposal/blob/master/A79-non-per-call-metrics-architecture.md
• https://github.com/grpc/proposal/blob/master/A78-grpc-metrics-wrr-pf-xds.md
• https://github.com/grpc/proposal/blob/master/A89-backend-service-metric-label.md
• A96: OTel Metrics for Retries proposal#488
• A94: OTel metrics for Subchannels proposal#485
• A91: Outlier Detection Metrics proposal#478

That being said, there have been talks of trying to come together and come up with a set of conventions that might work. We could potentially explore ideas of running in a compatibility mode.

[dfawley]

I don't think there's anything left to answer, so I'm going to close the issue. Feel free to follow-up either here or offline with @yashykt or myself.