feat/backpressure with retryable codes

[bmiguel-teixeira]

Description

Change the Exporter returns codes so it provides proper information to the OTEL Pipeline.
In this case, it provides backpressure information for retryable codes.

I have followed the conventions from https://opentelemetry.io/docs/specs/otlp/#failures-1
With the exception of making ALL 5xx retryable, since the same logic is applied for current built-in retry mechanism of the exporter.

Why this change?

In some scenarios in the OTEL Config pipelines, we run the prometheusremotewriteexporter in a "failfast" configuration.
This means the exporter tries to deliver, and if it fails, it moves on immediately.

However, in certain scenarios, id like for this "exit code" to be propagated back to the original client. This behaviour is natively implemented in OTLP Exporters.

Currently: ANY error is always cataloged as permanent error to the OTEL pipeline.
Now: Most OTLP Compliant codes + ALL 5xx are catalog as retryable, allowing the error to due backpressure to the OTEL pipeline and allowing final error to report back to the upstream.

P.S: Unsure if this is considered a breaking change?

Cheers,

Example of fail fast configuration

receivers:
  otlp:
    protocols:
      http:
        endpoint: "0.0.0.0:4318"

processors: {}

exporters:
  prometheusremotewrite:
    endpoint: http://localhost:9090/api/v1/write
    remote_write_queue:
      enabled: false
    retry_on_failure:
      enabled: false

service:
  extensions: []
  pipelines:
    metrics:
      receivers: [otlp]
      processors: []
      exporters: [prometheusremotewrite]

Link to tracking issue

N/A

Testing

Unit tests, and some local testing.

Documentation

TBD

[bmiguel-teixeira]

I will take care of the linting and failling if this initial approach is valid!

[github-actions]

This PR was marked stale due to lack of activity. It will be closed in 14 days.

[bmiguel-teixeira]

Any news here?

[github-actions]

This PR was marked stale due to lack of activity. It will be closed in 14 days.

[atoulme]

@ArthurSens please take a look