How to configure a custom span processor on top of python auto-instrumentation

[bjrara]

My request is very similar to the one asked in Java: Best way to add a custom span processor on top of auto-configuration, but this is in Python's version.

I have a custom span processor that adds additional attributes into spans based on the parent context. Is there any way to add it without user invoking add_span_processor in their codes?

Thanks!

[srikanthccv]

Is there any way to add it without user invoking add_span_processor in their codes?

No, this cannot be done at the moment, but it's also not that difficult to add this capability. Add an entry point for the span processor with the default being the "batch" and let the user configure the span processor of their choice to be used in auto instrumentation like custom exporter here

[bjrara]

Thanks @srikanthccv for the confirmation, and the code sharing.

If users don't use auto-instrumentation SDK, we can provide customized tracer provider and ask users to configure it using OTEL_PYTHON_TRACER_PROVIDER env var. However, according to the link you sent to me, when using auto-instrumentation, this env var is not respected, and the tracer provider defaults to opentelemetry.sdk.trace.TracerProvider.

I'm wondering if this is discussed before whether to allow customers to wire a different tracer provider into auto-instrumentation. In Java auto-instrumentation, this feature is supported.

[kokikathir]

@bjrara Have you found any solution to add a custom processor with auto instrumentation?