`test_0.yml` runtime is 1 hour 45 minutes

[kurtmckee]

Hello,

I've observed that test_0.yml takes an hour and 45 minutes to run [recent example, scroll to the bottom of the page to see the total time].

Would you be open to me submitting PRs to reduce the total runtime?

Some of the things that I'm seeing that are contributing to that long runtime include:

• Python 3.9 has to be downloaded and set up for ~35 jobs. 3.9 is not cached on the Ubuntu runners anymore, so each download and install contributes to the runtime.
• tox-uv is installed ~250 times (once per runner).
• The package has to be built (first as a .tar.gz file, and then converted to a wheel per tox defaults) for each run.

If you're open to it, I'd like to explore speeding up test_0.yml runtimes in the following ways:

1. Aggregating environments so that, say, py3{9,10,11,12,13,14}-test-instrumentation-aiopg all run on the same runner.
2. Building the opentelemetry wheel once, uploading it as an artifact, and then downloading it so that it doesn't have to be built for each tox environment
3. Using the setup-uv action and using uv pip install tox-uv to install tox, rather than pip install tox-uv, which should reduce the time required to install tox.

These are all examples of what's possible. I have a lot of experience speeding up CI runs, but it always involves reducing duplication, which can affect the shape of the jobs. If you're open to me submitting PRs for this, let me know! If you'd like to keep CI as it is, that's fine too. I'm interested in helping, but don't want to waste anybody's time, my own included. 😁

[xrmx]

Hello, thanks for your offer! I'd start with number 3 since it looks like the simpler one and then see how it goes.

[kurtmckee]

Super, I'll tackle that outside of work hours and post an initial PR. Thanks for the quick reply!

[emdneto]

yeah, I'm trying to optimize some piece of the CI, like this #4176. I would suggest start optimizing the jobs taking more > 1min as well