Joshua MacDonald

Lightstep will work to submit their downstream fork of this code to the upstream repository. For now, see https://github.com/lightstep/otel-launcher-go/tree/main/lightstep/instrumentation

There appears to be a "single-writer principle" violation happening. I understand the desire to opt-out, but can we somehow address the underlying problem as well?

I'm not convinced that the implementation of `target_info` is correct. I believe the single-writer principle is broken by the current behavior. The single-writer principle is consistent with Prometheus requirements, and...

Thank you @dashpole. Sorry for mis-reading the code. I reviewed prom->otlp and otlp->prom specs and remembered 2381 (however, will add that the string "target_info" does not appear in the spec...

> Correct, we use consistent hashing to ensure a unique timeseries is only ever sent to one collector so we only see out of order samples on the target metric....

> We don't care about the sender's identity, only the overall percentile / histogram of request latencies. The metric sender will use consistent hashing to send each unique time series...

Here it is: https://github.com/open-telemetry/opentelemetry-specification/blob/main/specification/metrics/data-model.md#single-writer I don't want to make it sound like you're doing something wrong, I was just having trouble understanding how your series are unique when `instance` and...

That makes sense. Thank you for being specific and bearing with me. We've fully explained why only the `target_info` metric is a problem and your setup is a reason we...

> Note that I don't think we should restore these tests, instead write new ones as described in https://github.com/open-telemetry/opentelemetry-go-contrib/issues/2624 and https://github.com/open-telemetry/opentelemetry-go-contrib/issues/2625. I've drafted such tests in https://github.com/lightstep/otel-launcher-go/tree/main/lightstep/instrumentation/runtime.

This is blocked, we need an otel-go release. I'll set this as a draft.