Binderbot and migrating workflows to > Python 3.10

[ktyle]

In recent weeks, various failures with nightly builds (and in some cases, PR's, e.g. ProjectPythia/HRRR-AWS-cookbook#39) are affecting some of our cookbooks. These look to be caused by environments that require Python >=3.11 (for example, zarr-python v3).

However, the version of Binderbot that we currently use will not run with Python > 3.10. This seems to be mostly related to deprecations in how binderbot uses async functions.

I have modified Binderbot to resolve this deprecation, although more testing and updating is certainly needed. However, it seems that we should reach out to the maintainers of Binderbot to request that we (Pythia and/or 2i2c) take over ownership of this repository. The last commit was 5 years ago.

What are your thoughts @brian-rose , @jmunroe , @erogluorhan and others?

[jmunroe]

I think 2i2c should support the maintenance of Binderbot (or a successor project that has the same outcomes with an updated design). This is work open source development work that can be well justified under 2i2c's portion of the Project Pythia collaborative grant (and that support should persist for the remaining duration of the NSF grant)

What I can't commit to, is a specific date when 2i2c can take on the binderbot work. We are currently focused on supporting Binderhub on Jetstream2 directly from our infrastructure.

FYI @yuvipanda for discussion at a later time.

[brian-rose]

However, it seems that we should reach out to the maintainers of Binderbot to request that we (Pythia and/or 2i2c) take over ownership of this repository. The last commit was 5 years ago.

I think the main issue here is that Binderbot does not currently have any maintainers. My understanding is that Ryan and Yuvi built it quickly to support the initial launch of the Pangeo Gallery, but further development (e.g. pangeo-gallery/binderbot#22) got sidetracked.

I think 2i2c should support the maintenance of Binderbot (or a successor project that has the same outcomes with an updated design).

I agree this makes sense, and this is one of the outcomes that I had in mind in initiating the Pythia - 2i2c funded collaboration.

In the shorter term, it seems that we might need to get some quick fixes merged. I'm guessing that @yuvipanda and @rabernat probably have push access to that repo, but we could always just use a Pythia-specific fork temporarily. The way we use use binderbot in the cookbook actions is hard coded here -- it would be trivial to point this to a different fork.