Could logfire use decoupled processor on AWS lambda?

[bensi94]

Question

Hi,

I love logfire, thank you for your fantastic work.

I have recently been using it with AWS lambda and it works great with the logfire.instrument_aws_lambda.

However, one downside I’ve noticed is that traces and metrics are flushed during function execution, adding a noticeable performance overhead (as shown in this AWS X-Ray screenshot).

I’ve been wondering if it could be integrated with something like the OpenTelemetry Lambda Collector using a decouple processor to offload the processing and reduce the overhead. In that case, logfire could push logs to this as an alternative backend.

I haven’t fully tested this yet, and I’m not entirely sure how to set it up. Do you think this could work in theory? If so, how would the exporter authenticate with logfire, given that it doesn’t natively handle the logfire token?

Alternatively, instead of relying on the OpenTelemetry Lambda Collector, could logfire implement a similar approach internally with this kind of decoupling as the collector does it instead?

Would love to hear your thoughts!

Thanks!

[alexmojaki]

Hi @bensi94, thanks for sharing, this looks very interesting. It looks like you currently know more about this area than we do. So I think it would be best for you to try it out. Yes, I do think it could work. The relevant collector config looks something like this:

    exporters:
      otlphttp:
        endpoint: "https://logfire-api.pydantic.dev"
        headers:
          Authorization: "Bearer ${env:LOGFIRE_TOKEN}"

(based on https://logfire.pydantic.dev/docs/how-to-guides/otel-collector/#collecting-system-logs)

Alternatively, instead of relying on the OpenTelemetry Lambda Collector, could logfire implement a similar approach internally with this kind of decoupling as the collector does it instead?

In https://github.com/open-telemetry/opentelemetry-lambda/blob/main/collector/processor/decoupleprocessor/README.md it says:

The processor is aware of the Lambda lifecycle and will prevent the environment from being frozen or shutdown until any pending traces/metrics/logs have been exported.

I don't know how that works, but it sounds like it relies on using a Lambda extension. So I don't think the SDK could do this on its own, but maybe there could be a logfire extension? I really don't know.

[bensi94]

Thanks @alexmojaki! I will check this out.