How do we capture environments and examples from tutorial authors?

[mrocklin]

We need to crowdsource the specification of environments and the generation of notebooks from tutorial maintainers. What is the best way to do this? What information should we collect from them?

For example should we use binder's mechanism ?

cc @yuvipanda @choldgraf

[choldgraf]

Hmmm - I think that's a decent start (depending on whether they use a workflow that "works" in the context of Binder / repo2docker). Could you give a bit more detail about what you're trying to accomplish?

[mrocklin]

Ideally we would have a single (or several) JupyterHub instance(s) similar to the pangeo deployment that would serve the latter part of several tutorials. Each tutorial would have a notebook or two with exercises that students would go through, presumably engaging a distributed system like Dask along the way. Each tutorial is likely to have special software constraints that, I suspect, can probably fit into a requirements.txt file or an environment.yaml file, though I'm not sure.

One way to approach this would be that the student would log in, select the tutorial that they are in, and be taken to a Jupyter notebook server running the appropriate docker image.

I'm happy to provide more detail on my current thoughts if desired. I'm also curious to hear more from others. Also @choldgraf just in in case you haven't checked your Berkeley e-mail I also wrote to you about this there.

[choldgraf]

ah gotcha - yeah I think one thing we could do is just take the intersection of the environments for all of the tutorials etc, and then have one dockerfile that serves them all. This kinda depends on how specialized each environment gets however.

Another is to have a docker image per tutorial. The ability to select an image isn't exposed via UI right now (though perhaps it will be by the time scipy rolls around? It's definitely on the roadmap).

[choldgraf]

@yuvipanda may have a better idea for how feasible it'd be to serve multiple docker images via the same JupyterHub.

[mrocklin]

This has been an ongoing conversation with the pangeo project. There is also this issue here:

https://github.com/jupyterhub/zero-to-jupyterhub-k8s/issues/402

My general understanding is that this would require modest effort.

I think that the short term question is if this is something that we (probably meaning JHub devs) would be able to implement before scipy. If not then we would maybe go with a single mega-image? This decision would affect how we communicate with tutorial authors initially. Thoughts? Alternatives?

[yuvipanda]

You can probably use something like https://github.com/jupyterhub/kubespawner/pull/160

How does that sound?

On Thu, Apr 19, 2018 at 3:22 AM, Matthew Rocklin notifications@github.com wrote:

This has been an ongoing conversation with the pangeo project. There is also this issue here:

jupyterhub/zero-to-jupyterhub-k8s#402 https://github.com/jupyterhub/zero-to-jupyterhub-k8s/issues/402

My general understanding is that this would require modest effort.

I think that the short term question is if this is something that we (probably meaning JHub devs) would be able to implement before scipy. If not then we would maybe go with a single mega-image? This decision would affect how we communicate with tutorial authors initially. Thoughts? Alternatives?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/dask/scipy-tutorials-2018/issues/1#issuecomment-382685696, or mute the thread https://github.com/notifications/unsubscribe-auth/AAB23uSWBkwkMmB9acYIVkUelFjSQhgcks5tqGVcgaJpZM4TUcoj .

-- Yuvi Panda T http://yuvi.in/blog

[mrocklin]

That looks awesome to me!

On Thu, Apr 19, 2018 at 2:09 PM, Yuvi Panda notifications@github.com wrote:

You can probably use something like https://github.com/jupyterhub/kubespawner/pull/160

How does that sound?

On Thu, Apr 19, 2018 at 3:22 AM, Matthew Rocklin <notifications@github.com

wrote:

This has been an ongoing conversation with the pangeo project. There is also this issue here:

jupyterhub/zero-to-jupyterhub-k8s#402 https://github.com/jupyterhub/zero-to-jupyterhub-k8s/issues/402

My general understanding is that this would require modest effort.

I think that the short term question is if this is something that we (probably meaning JHub devs) would be able to implement before scipy. If not then we would maybe go with a single mega-image? This decision would affect how we communicate with tutorial authors initially. Thoughts? Alternatives?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/dask/scipy-tutorials-2018/issues/1# issuecomment-382685696, or mute the thread https://github.com/notifications/unsubscribe-auth/ AAB23uSWBkwkMmB9acYIVkUelFjSQhgcks5tqGVcgaJpZM4TUcoj .

-- Yuvi Panda T http://yuvi.in/blog

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/dask/scipy-tutorials-2018/issues/1#issuecomment-382829854, or mute the thread https://github.com/notifications/unsubscribe-auth/AASszA50GLqChGFS3hWKZP9iBltE22oSks5tqNK-gaJpZM4TUcoj .