Python 3.11 profiling causes memory leaks

[jerr0328]

Summary of problem

This is possibly related to what's being fixed in #4895
Running latest release of dd-trace-py with Python 3.11 and having DD_PROFILING_ENABLED: "true" causes memory leaks.

Which version of dd-trace-py are you using?

1.7.0

Which version of pip are you using?

22.3.1

Which libraries and their versions are you using?

Let me know if you need more details on this, but I'm using Django, Django Rest Framework, and Celery (all latest versions as of this writing)

How can we reproduce your problem?

Run a web app as mentioned in the summary, and check memory usage.

What is the result that you get?

See image on the right side (after deploying with change to new version and enabling profiling, the deploy is in the middle), pod and processes inside pod were OOMing when hitting 100% of memory limit.

What is the result that you expected?

Flat memory usage (similar to left side of graph above with profiling disabled)

[edmundlam]

Hello, we encountered this leak in our app when testing upgrading to python-3.11 this week.

Will you be releasing a new version with the fix sometime soon? Thanks!

[Yun-Kim]

Hi @edmundlam and @jerr0328, the fix for the memory leak should now be included in 1.7.3! Feel free to reopen this issue if 1.7.3 doesn't resolve the problem.

[edmundlam]

Hello @Yun-Kim looks good in our tests, thanks!