How to guarantee events with same trace_id are sampled together when vector is distributed?

[nzxwang]

Hi team,

We would like to configure our distributed vector pipeline to batch all logs with the same trace id before sampling at some sample rate. The sample component's key_field does exactly this but since sample is a completely stateless component, I assume that logs with the same trace id flowing through different vector pipeline hosts are not guaranteed to be sampled together. Furthermore, in an ideal world, we would like to also batch traces too. I'd appreciate any insight on how we might achieve this.

Thank you