Errors in ddtrace collectors when using with flask + gunicorn + gevent

[mubin-tirsaiwala]

Description

We are hosting a flask app with gunicorn servers. Until now, we were using default synchronous workers and there were no ddtrace issues. Recently, we tried using gevent workers for async handling and ddtrace started throwing errors related to what seems like collector in profiling. Everything works perfectly well until there's a heavy load on the system. I could not reproduce this behavior locally but is happening in both our staging as well as production environments.

Note: The app still serves requests without a significant performance impact but the errors keep flowing

Running Environment

• Python Version: 3.10.15
• ddtrace version: 2.12.0
• gevent version: 24.10.3
• gunicorn version: 21.2.0
• Number of gevent workers = 12
• We are running the flask app in a kubernetes cluster.

What have I tried

• Bumping ddtrace version from 2.5.2 to 2.12.0. I tried bumping up the version to the latest that includes a fix regarding gevent in its release note i.e. v2.12

Expected behavior

• ddtrace works without producing errors and killing the workers.

Logs

Apologies for such limited logs, but this is all I got!

KeyError: 132388630494432
o = self.data[key]()
File "/usr/local/lib/python3.10/weakref.py", line 137, in __getitem__
File "ddtrace/profiling/collector/_task.pyx", line 125, in ddtrace.profiling.collector._task.list_tasks
File "ddtrace/profiling/collector/_task.pyx", line 102, in ddtrace.profiling.collector._task.list_tasks
File "ddtrace/profiling/collector/stack.pyx", line 321, in ddtrace.profiling.collector.stack.stack_collect
File "ddtrace/profiling/collector/stack.pyx", line 573, in ddtrace.profiling.collector.stack.StackCollector.collect
for events in self.collect():
File "/usr/local/lib/python3.10/site-packages/ddtrace/profiling/collector/__init__.py", line 43, in periodic
Traceback (most recent call last):

[mubin-tirsaiwala]

@taegyunkim Any updates on this?

[github-actions]

This issue has been automatically closed after a period of inactivity. If it's a
feature request, it has been added to the maintainers' internal backlog and will be
included in an upcoming round of feature prioritization. Please comment or reopen
if you think this issue was closed in error.