Simplify "Bring You Own Backend" deployments

[girodav]

Feature Request

It is already possible to use any otlphttp exporter following the documentation for Bring Your Own Backend for Kubernetes.

While this approach works, it would be great to simplify the process of deploying the OpenTelemetry demo with a custom backend using an OTLP HTTP exporter. Additionally, deploying with Docker Compose currently requires either adding a new custom collector configuration or modifying the existing one.

Describe the solution you'd like:

Being able to use any otlphttp exporter, with minimal configuration required.

For Kubernetes deployments, it would be great to add a new configuration option in the Helm charts that allows users to easily configure the OTLP HTTP endpoint. For example:

helm install otel-demo charts/opentelemetry-demo --set otlp-backend=true --set opentelemetry-collector.config.exporters.otlpbackend.endpoint="my-endpoint"

For Docker Compose deployments, it would be helpful to allow overriding OTLP endpoints and headers through environment variables. Additionally, it would be useful to enable users to select which exporters to use. One potential solution could be utilizing Docker Compose Profiles, enabling users to specify the desired profile with a command like:

PROFILE=byob make start

[rogercoll]

I support this idea, looks like it will be optional and won't affect the current deployment mechanism. The main benefit is that it will ease the Bring your won backend strategy, and testing different vendors/backends would be a matter of defining a different otlp endpoint.

Another option would be to use docker compose overrides, but Profiles seems more suitable for this use case (we can add as many otel-collector profiles as needed)

@open-telemetry/demo-approvers wdyt?

[julianocosta89]

Ohhhhh, TIL! 🤓

Profiles look great!
We could drop the minimal docker compose override and just configure the profiles.

I liked that!