Resource attribute keys are lowercased in OTLP output

[laitm]

Question: Resource attribute keys are lowercased in OTLP output

Description

While testing the otlp output in gNMIc, I noticed that resource attribute keys appear to be converted to lowercase in the debug output and resulting OTLP data.

Configuration

Example configuration:

outputs:
  otlp:
    type: otlp
    endpoint: otel-collector:4317
    strings-as-attributes: false
    resource-attributes:
      - HostName
      - SiteID

Expected behavior

Resource attribute keys preserve their original case as defined in the configuration:

HostName
SiteID

Actual behavior

The attribute keys appear lowercased in debug output and exported telemetry:

hostname
siteid

This suggests the attribute keys may be normalized to lowercase somewhere in the gNMIc event or OTLP export pipeline.

Questions

1. Is the lowercasing of attribute keys intentional behavior in gNMIc?
2. If so, where in the pipeline does this occur (event model, processors, OTLP exporter)?
3. Is there a configuration option to preserve the original case of attribute keys?

Additional context

The goal is to preserve case for compatibility with an existing telemetry schema that relies on case-sensitive attribute names.

Thanks for any clarification.

[karimra]

gNMIc does not intentionally convert attribute names to lowercase.
I see an issue with your config snippet, I think you mean to use resource-tag-keys instead of resource-attributes:

    # list of tag keys to place as OTLP Resource attributes.
    # these tags are excluded from data point attributes.
    # defaults to empty (all tags become data point attributes).
    resource-tag-keys:
      # - device
      # - vendor
      # - model
      # - site
      # - source

    # map of string:string, additional static attributes to add to the OTLP Resource.
    resource-attributes:
      # key: value

that being said, OTLP recommends using lowercase snake_case naming: https://opentelemetry.io/docs/specs/semconv/general/naming/