[Question] Why do collector metadata formats vary?

[jaydeluca]

While reviewing #897 i noticed that in some components, metrics are listed as a top level metadata signal:

https://github.com/open-telemetry/opentelemetry-collector-contrib/blob/main/receiver/activedirectorydsreceiver/metadata.yaml#L76

and others have it nested under a telemetry node:

https://github.com/open-telemetry/opentelemetry-collector/blob/main/processor/memorylimiterprocessor/metadata.yaml#L19-L20

Why is this, do they imply different meanings in some way? Does this indicate an issue upstream that should be fixed or should we just support both styles?

[jaydeluca]

looks liek the schemas define both:

https://github.com/open-telemetry/opentelemetry-ecosystem-explorer/blob/main/ecosystem-registry/collector/meta/schemas/053934807f90.yaml#L121-L123

https://github.com/open-telemetry/opentelemetry-ecosystem-explorer/blob/main/ecosystem-registry/collector/meta/schemas/053934807f90.yaml#L218-L221

🤔

[Adii-45]

I looked into this a bit more as well... It seems the two schema blocks are intentionally different rather than alternative formats!
From tracing the parser and the linked examples, the top-level metrics appear to describe the component's emitted telemetry, whereas telemetry.metrics is for the component's own internal/self-observability metrics... That also explains the parser-side inconsistency we were seeing in Explorer..

[jaydeluca]

interesting, I wonder if this is documented anywhere.

so we should probably account for differentiating them in the UI accordingly once we confirm

thanks for helping investigate this

[Adii-45]

Glad to help! I couldn't find anything documenting that distinction during my investigation either, but it'll be interesting to see if there's something upstream explaining it...