Collect trace_api.{requests,responses,errors} metrics in the background trace sender

[iamluc]

Description

Reviewer checklist

• Test coverage seems ok.
• Appropriate labels assigned.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 81.95489% with 24 lines in your changes are missing coverage. Please review.

Project coverage is 77.80%. Comparing base (f246102) to head (7bafb13).
Report is 3 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff            @@
##             master    #2672   +/-   ##
=========================================
  Coverage     77.79%   77.80%           
  Complexity     2223     2223           
=========================================
  Files           226      226           
  Lines         26276    26339   +63     
  Branches        988      988           
=========================================
+ Hits          20441    20492   +51     
- Misses         5309     5321   +12     
  Partials        526      526           

Flag	Coverage Δ
appsec-extension	69.13% <ø> (ø)
tracer-extension	78.67% <81.95%> (+0.01%)	⬆️
tracer-php	80.33% <ø> (ø)

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

Files	Coverage Δ
ext/ddtrace.c	74.54% <100.00%> (ø)
ext/ddtrace.h	62.50% <ø> (ø)
ext/sidecar.c	88.18% <86.20%> (+4.85%)	⬆️
ext/telemetry.c	93.84% <81.57%> (-5.07%)	⬇️
ext/coms.c	78.80% <80.00%> (-0.82%)	⬇️

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f246102...7bafb13. Read the comment docs.

[pr-commenter]

Benchmarks

Benchmark execution time: 2024-05-28 18:51:54

Comparing candidate commit 7bafb13 in PR branch luc/bgs-trace_api-metrics with baseline commit f246102 in branch master.

Found 8 performance improvements and 6 performance regressions! Performance is the same for 164 metrics, 0 unstable metrics.

scenario:ComposerTelemetryBench/benchTelemetryParsing-opcache

• 🟥 execution_time [+2.076µs; +4.324µs] or [+3.472%; +7.231%]

scenario:EmptyFileBench/benchEmptyFileOverhead

• 🟩 execution_time [-548.278µs; -392.942µs] or [-17.442%; -12.501%]

scenario:EmptyFileBench/benchEmptyFileOverhead-opcache

• 🟩 execution_time [-580.249µs; -433.171µs] or [-17.783%; -13.276%]

scenario:HookBench/benchHookOverheadInstallHookOnMethod

• 🟥 execution_time [+18.784µs; +52.465µs] or [+2.280%; +6.368%]

scenario:LaravelBench/benchLaravelOverhead

• 🟩 execution_time [-469.967µs; -324.573µs] or [-14.469%; -9.993%]

scenario:LaravelBench/benchLaravelOverhead-opcache

• 🟩 execution_time [-606.675µs; -444.065µs] or [-17.752%; -12.994%]

scenario:LogsInjectionBench/benchLogsInfoInjection-opcache

• 🟥 mem_peak [+87.431KB; +238.810KB] or [+4.301%; +11.747%]

scenario:MessagePackSerializationBench/benchMessagePackSerialization-opcache

• 🟩 execution_time [-6.397µs; -4.443µs] or [-4.106%; -2.852%]

scenario:PDOBench/benchPDOBaseline

• 🟥 execution_time [+15.894µs; +16.773µs] or [+9.071%; +9.573%]

scenario:PDOBench/benchPDOOverhead-opcache

• 🟥 execution_time [+13.009µs; +16.756µs] or [+4.658%; +5.999%]

scenario:PDOBench/benchPDOOverheadWithDBM

• 🟩 execution_time [-17.562µs; -13.130µs] or [-5.616%; -4.199%]

scenario:PDOBench/benchPDOOverheadWithDBM-opcache

• 🟥 execution_time [+14.123µs; +18.241µs] or [+4.582%; +5.918%]

scenario:SymfonyBench/benchSymfonyOverhead

• 🟩 execution_time [-613.423µs; -574.957µs] or [-8.821%; -8.268%]

scenario:SymfonyBench/benchSymfonyOverhead-opcache

• 🟩 execution_time [-584.712µs; -530.048µs] or [-8.271%; -7.497%]

[pierotibou]

LGTM funtionnaly speaking, just a comment about retries

[pierotibou]

Shouldn't we count the retries (in the for loop) as well ?

[iamluc]

I believe it's the same logic that in the sidecar. See DataDog/libdatadog#418 (comment)

[pierotibou]

Indeed, forgot about that, thanks for raising. Though, I'm not sure it's the right move to make... Indeed, if we want to be able to compare, we need to make sure this is how things have been implemented in other languages. The thing is, maybe all languages have implemented it differently, so maybe having a simple solution works and we just need to know the differences. Anyway, let's discuss that with everyone.

[bwoebi]

Things seem green now :-)