Try to reduce prometheus port conflict test flake

[jack-berg]

Example flake here: https://github.com/open-telemetry/opentelemetry-java/actions/runs/13184973874/job/36804837095?pr=7055#step:6:1659

I believe I've only noticed these on windows. Not sure why. Maybe the way windows assigns random available port is more prone to race conditions / more deterministic than other OS. I.e. perhaps between the time we ask for a available port and initialize a prometheus server with that port, another test initializes a different prometheus server with that same port.

Trying to mitigate that by adjusting tests to favor setting the port to 0, causing the implementation to assign an available port. And then adjusting test code flow to make assertions of this port.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.84%. Comparing base (19650df) to head (ce77eb4).
Report is 19 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #7086      +/-   ##
============================================
- Coverage     89.86%   89.84%   -0.02%     
+ Complexity     6613     6612       -1     
============================================
  Files           740      740              
  Lines         19991    19991              
  Branches       1966     1966              
============================================
- Hits          17964    17961       -3     
- Misses         1437     1439       +2     
- Partials        590      591       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[trask]

heads up, I hit rerun, but it looks like it flaked beforehand: https://github.com/open-telemetry/opentelemetry-java/actions/runs/13207505490/job/36923855747?pr=7086

[jkwatson]

If this fails, I'm going to have questions.. ;)

[jack-berg]

heads up, I hit rerun, but it looks like it flaked beforehand

@trask what strategies does the instrumentation repo use for flaky tests? Ideally, I'd like to just be able to retry select tests a second time. The probability of failing twice in a row should be low enough to be tolerable.

[trask]

@trask what strategies does the instrumentation repo use for flaky tests? Ideally, I'd like to just be able to retry select tests a second time. The probability of failing twice in a row should be low enough to be tolerable.

https://github.com/open-telemetry/opentelemetry-java-instrumentation/blob/4395859daf0802344bb94e2b99dea67a6b69c01c/conventions/src/main/kotlin/otel.java-conventions.gradle.kts#L372-L375

[jack-berg]

Closing in favor of #7106