Span ID always null in Logs | Logs ID injection

[lb1mg]

Describe the bug: I am using logging support which injects correlation ids (trace, transaction, span). However even after running with 100% sample rate I've not been able to get a single log with a Span ID!
Also "maybe" this could be the reason that I am not seeing any logs in trace waterfall

To Reproduce

1. Install elastic apm agent version
2. Simply allow logging integration (default)
3. Check in your APM

Environment (please complete the following information)

• OS: Linux and Mac
• Python version: 3.9/10
• Framework and version [e.g. Django 2.1]: Should not matter but Sanic v22
• APM Server version: Proprietary Solution
• Agent version: elastic-apm==6.7.1

Additional context
N/A

[basepi]

Would you mind trying to put together a reproduction case with a minimal sanic app? I'm wondering if something is breaking down with the contextvar propagation or something, but I haven't seen this behavior.

[lb1mg]

Thanks for such quick response!! Sure, pls allow me some time to get a simple app. I would also love to help in fixing anything broken

[lb1mg]

I have two questions:

1. Can modifying the asyncio task factory using loop.set_task_factory() affect elastic apm agent in any way? we use this
2. I have a complex running service, sharing it is not possible and would not be helpful how can I debug this issue or what can I share so someone here might be able to help?

Thanks!

[basepi]

Can modifying the asyncio task factory using loop.set_task_factory() affect elastic apm agent in any way?

It's possible. The most likely culprit would be changes to contextvars propagation.

You could try the following code snippet from within your app to see if you can properly fetch the current transaction and span:

from elasticapm.traces import execution_context
t = execution_context.get_transaction()
s = execution_context.get_span()