JupyterHub and BinderHub Team Meeting - September#

20 September 2018

Welcome to the Team Meeting#


If you are joining the team video meeting, sign in below so we know who was here. Roll call:

  • name / institution / GitHub handle

  • min / simula / @minrk

  • tim / wild tree tech / @betatim

  • erik / sandvik / @consideratio

  • sylvain / QuantStack / @SylvainCorlay

  • zach / cal poly / @Zsailer

  • carol / cal poly / @willingc

Agenda items#

  • state of binder-for-AI with AWS grant (needs Brian or Jess?)

  • burn rate (needs Chris?)

    • Yes we need Chris

  • no-cost-extension

    • Fernando is working on it

    • probably will cover 6-8months of cloud bill

    • We have a no-cost extension on our grant that we can milk until probably 6-8 months from now. Should start looking for $$$ now.

  • releasing BinderHub v0.1?

    • just do it!

  • publish a call for hosting mybinder.org

    • not super urgent so let’s punt on it for now

    • people costs are higher so focus on that

  • report from conversation with kaggle

  • OVH hosting / sponsorship discussion.

    • Really interested in Binder

    • They’re offering to host a Binder for free

    • first federated binderhub sounds good

    • opportunity to grow our team to have a EuroBinder

    • offer a kubernetes cluster before they have it in their official product offering

    • having a second deployed live binderhub makes sure we can run on clouds beyond google

    • let’s setup a meeting to get a better feeling for each other!

      • put your name here if you want to join the meeting so we can find a good time and date

        • Chris H

        • Tim H

  • create and publicise the “point of contact” for reaching the binder team (Answering the question: How do I reach the Binder team in an official capacity?)

    • there is a public list

    • there is no private “contact the team” channel

    • create a new google group for this purpose

  • Zach and Brian: ongoing work related to users groups and auth

    • how to deal with stale user information

    • let’s collect needs from various sides and then get implementing

  • Merged PRs z2jh: update about k8s 1.11 and overprovisioning (erik, short)

    • in repo todo: document how this works (requires helm k8s 1.11, which requires 2.11)

    • make this auto-scaling behaviour the recommended way in the z2jh guide

    • no more waiting for new nodes!!

    • write a blog post on this to advertise it with next chart release

  • EU proposal for “European Open Science Cloud (EOSC)” http://ec.europa.eu/research/participants/portal/desktop/en/opportunities/h2020/topics/infraeosc-02-2019.html

    • “The European Open Science Cloud (EOSC) is a European Commission project to provide a public data repository which conforms to open science values. The project began in 2015 with the plan that its organizers finish it by 2020.[1] A European Union committee on research endorsed a plan for the cloud’s development in May 2018.[2]”

  • Funding Binder Sustainability as an extension from Moore

    • a grant proposing to work on federation and admin things around making binder sustainable

    • seems like a low effort thing compared to EU proposals


  • Closed 50 open issues on JupyterHub repo

    • 🎉

From Carol – I went through and changed jupyterhub, zero-to-jupyterhub, kubespawner’s doc build to use pip instead of conda on RTD. There is some error with Sphinx and/or RTD that isn’t handling latex correctly for Sphinx or honoring the dependency pin for sphinx when using environment.yml. Downside of not doing this for other repos: the docs builds will fail until a new Sphinx release (hopefully) fixes the problem.

  • Is this something Jessica can do?

  • should revive Carol’s dashboard to see which RTD builds have failed

Tim (and Jess) would be interested in writing a short blog post to introduce repo2docker to the world. Who should we ask before pressing publish?

Notes from kaggle meeting#

Tim and Erik just had a super cool call with the Kaggle people (Anthony Goldbloom, Meghan O’Connell, Rachel Tatman) who asked if we could update our data a bit so they can poke around in terms of making a nice story for GCP to help make the case for the cloud bill to go away. I told them the same factoids as I had in my JupyterCon talk as well as some of the stories about who uses it.

(For kaggle itself having access to our analytics is probably pretty useful for their kaggle kernels product. But we put openness on our flag, so we should stand by that I guess.)

They seemed positive and serious about helping us out. It is early days as they’ve never championed a project internally, so let’s see what happens. They were also not sure what GCP would ask in return. Best guess a logo/link? They will do some work internally to move this along. I promised Rachel an email when we updated our data.

We had a chat about where mybinder.org came from, how Erik and I got involved, encouraged them to check out repo2docker and that we’d be interested in making r2d something used in many places, cost of cloud bill and cost of devops support, “federation super light” - basically could kaggle kernels be a backup for mybinder.org (I commented that federation in the sense of sending traffic could take some time to discuss in the team, but if kaggle kernels used/understood repo2docker then it would be simple for users to use it as a backup).