NH-11236 sampler creates Service Entry Internal KVs

[tammy-baylis-swi]

Here the custom Sampler creates Service Entry Internal KVs by setting them as attributes only for the root span:

1. BucketCapacity
2. BucketRate
3. SampleRate
4. SampleSource

I've also tidied the calculate_attributes a little bit to avoid ambiguity in new attributes assignment.

I've done some manual testing, with resulting exported traces with those KVs recorded on this doc: https://swicloud.atlassian.net/wiki/spaces/NIT/pages/2893414559/2022-04-13+service+entry+internal+KVs+testing It doesn't really cover the different SampleSource results possible though. Would that be liboboe scope or in this PR's scope?

This PR is to merge into add-custom-propagator branch at the moment because this one hasn't been merged yet #11 But I can change the destination if this one needs lots of work too! Please let me know and/or if I need to do more testing.

[cheempz]

LGTM, thanks @tammy-baylis-swi! I think this is ready w/o further SampleSource testing since that's all implemented by liboboe. In fact the test results you provided has helped highlight a bug in liboboe -- the returned SampleSource and SampleRate should be -1 for trigger traces per spec, but liboboe is setting to 0. I'll submit a PR to get that fixed :)

[tammy-baylis-swi]

Hi @cheempz ! Based on our separate chat, in 23a12a2 I've updated the Sampler's calculate_attributes method to set the service entry KVs (bucket/sample) only if liboboe returns non -1 values -- instead of every root span as before. I also merged your liboboe updates into my add-otel-oboe-features branch and rebuilt. Test traces are are on the lower half of the table on this doc, including a test of Acceptance 5 to double-check downstream of new decision-making: https://swicloud.atlassian.net/wiki/spaces/NIT/pages/2893414559/2022-04-13+to+18+service+entry+internal+KVs+testing#At-23a12a2-and-liboboe-merge%3A

I wasn't sure if the identification of a continued decision should be based on all four liboboe outputs. Currently, if even one is not -1 then it is considered not a continued trace. If all four are -1 then it is a continued trace. That helper is here: https://github.com/appoptics/opentelemetry-python-instrumentation-custom-distro/blob/23a12a255f8d0e159b5726cc3da7bb710756b216/opentelemetry_distro_solarwinds/sampler.py#L128-L141

Please let me know your thoughts! In the meantime I'll try to resolve the merge conflicts. I made some commits to the big PR this morning which ended up refactoring calculate_attributes quite a bit... I think I did them in the wrong order 😅

[cheempz]

LGTM, thanks @tammy-baylis-swi! Theoretically if sample rate and bucket rate are -1 it would indicate a continued decision, but I think checking all four decision values is the most correct and accessing two more dictionary shouldn't cost that much ;)