otel spans not flushed in _after?

[vin]

Expected Behavior

Spans from uncaught exceptions should still reliably reach datadog.

Actual Behavior

Many spans from uncaught errors are lost before they can be sent.

This needs further investigation, but I suspect it's because datadog_wrapper's _after method specifically flushes ddtrace-generated spans but not opentelemetry-api ones? Maybe it needs to also call TracerProvider.force_flush if applicable?

Specifications

• Datadog Lambda Layer version: 86
• Python version: 3.11.14

Stacktrace

Paste here

[purple4reina]

Hey @vin, thanks for alerting us to this issue. I'd love to take a closer look at it. Can you provide more details, and ideally a code example, about "spans from uncaught exceptions"? How are you creating these otel spans? How exactly are you instrumenting the app to initialize both datadog and otel tracers? Did you follow our recommended setup?

With the information I have and the testing I've done to try to reproduce this, my first guess is that the problem is that you're trying to mix both the datadog python tracer and the otel python tracer. Does the issue still exist when you remove the datadog tracer?