Parent-based tracing sampling has incorrect logic -- should behave differently on flag=0 vs Parent Span doesn't exist about dapr HOT 3 CLOSED

hi @yaron2 , I couldn't change this to a bug and it was a bit confusing w the changes I made as I dug in, so I created a new issue here: #7574
I should have linked it here.

from dapr.

I looked at the code and it looks like the code in DaprTraceSampler is close but not quite correct.

Looking at the DaprTraceSampler code:

/**
 * Decisions for the Dapr sampler are as follows:
 *  - parent has sample flag turned on -> sample
 *  - parent has sample turned off -> fall back to probability-based sampling
 */

I think this is incorrect and how how other OTel Parent samplers work.

The right behavior as I understand it (and how it is written in the .NET OTel library I checked):

• parent has sample flag turned on -> sample
• parent has sample turned off -> do not sample
• parent does not exist or no sample flag set -> fall back to probability-based sampling

from dapr.

@andrejpk did you mean to close this?

from dapr.