Add policy sampling processor - 1/2

[jpkrohling]

Signed-off-by: Juraci Paixão Kröhling juraci@kroehling.de

Description: This PR creates a new sampling processor based on the tail-based sampling, without the logic that is needed for the grouping of traces, which can be accomplished with the groupbytrace processor. Once this processor is merged, we can deprecate the tail-based sampler in favor of this + groupbytrace.

Testing: unit tests + manual tests (final solution).

Documentation: readme file.

[codecov]

Codecov Report

Merging #1786 (4579f6f) into master (d48fa8d) will increase coverage by 0.02%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #1786      +/-   ##
==========================================
+ Coverage   89.85%   89.87%   +0.02%     
==========================================
  Files         376      378       +2     
  Lines       18180    18216      +36     
==========================================
+ Hits        16336    16372      +36     
  Misses       1383     1383              
  Partials      461      461              

Flag	Coverage Δ
integration	69.77% <ø> (ø)
unit	88.57% <100.00%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
processor/policysamplingprocessor/factory.go	100.00% <100.00%> (ø)
processor/policysamplingprocessor/metrics.go	100.00% <100.00%> (ø)
processor/groupbytraceprocessor/event.go	95.96% <0.00%> (-0.81%)	⬇️
receiver/carbonreceiver/transport/tcp_server.go	67.00% <0.00%> (+1.00%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d48fa8d...4579f6f. Read the comment docs.

[kbrockhoff]

Do we want to restrict this to an enumerated list? If someone wants to create their own proprietary policy they will need to fork this repo and make changes.

[jpkrohling]

Thanks for the review! This part is inherited from the current tail-based sampling, but you are right. In fact, I think we could discuss making this go in another direction: instead of one "policysamplingprocessor", we could be having two or three processors: ratelimitingprocessor (rate_limiting from this PR) and attributesamplingprocessor (string and int policies from this PR). The always sampling would just disappear, as it's not really useful IMO.

What do you think?

[pmm-sumo]

As we discussed during the SIG, here's another attempt at making tail-sampling processor more robust: https://github.com/SumoLogic/opentelemetry-collector-contrib/tree/master/processor/cascadingfilterprocessor

[jpkrohling]

Closing this for now, as this will probably become two other processors. See #1797.