JupyterHub, Binder workshop in Oslo, Sept. 5-8

Technical things I am thinking about at the moment and probably still will be thinking about by the time of the workshop (aka topics I’ll try and make progress on):

  • federation redirection based on which repo is being launched, send the same repo to the same cluster as often as possible
  • custom k8s scheduler to assign build pods to nodes in a cluster. Goal is to send the same repo to the same node as often as possible
  • unified “design” of BinderHub’s web pages. Right now we have pages that have had stuff bolted on to them over time. Now that we know what all these things are, where we want to go and experience of how people use Binder I think creating a “fresh” design would be worth it.

I would love to listen to the talk. Excited to meet you in person too.


The talk would be perfect!

1 Like

I am not sure how much of this meeting I will be able to attend, but please keep me updated on the program!

Yay, thanks everyone!! :tada:

1 Like

Hi, things changed on our side. @arnim and me will most probably be there :slight_smile:

1 Like

Hi all,

I think this WS sounds fare to interesting not to be there and @bitnik and I would love to attend.

@minrk would it still be possible that we join the workshop? We just have filled the whenisgood form but its already somewhat after the end of July. If this works for you I’m looking forward to meeting you in Oslo.

I hope everything works out with our administration - since they are the one having the last say :wink:

1 Like

The venue is now set! I was unable to book the more scenic Holmenkollen venue, but we are at the Grand Hotel in downtown Oslo. I’ve booked lodging in the hotel for those whom I heard from (@willingc , @sgibson91 , @nuest , @betatim , @consideRatio ).

@arnim and @bitnik you are definitely welcome! I’ve already done the reservation, so you would need to book your own accommodations, but there’s room for a couple more in the workshop. There are lots of other hotels within easy walking distance.

I’ll update with more travel tips, but the hotel is very close to the nationaltheatret train station. You can take either the standard Vy (NSB) train or the more expensive Flytoget from Oslo airport. (Travel nitpick: Flytoget calls itself “airport express” but it’s not any faster than the regular train. What it is is more frequent and less crowded).

Grand Hotel Oslo
Karl Johans gate 31, 0159 Oslo
23 21 20 00


Thank you for organising @minrk! I booked my flights, I’ll be arriving around lunchtime on the 4th and leaving very early (:sleeping:) on the 9th. Hopefully the Grand Hotel do airport transfers?

1 Like

Thank you @minrk :wink:

Looking forward to meet you in Oslo


1 Like

I just learned of this and am interested in going (and have been waiting for it for a while)! I’ll have to talk tomorrow but possibly someone else on my team may come. Is it still possible for more to come and is there travel money still available?

I’ll update with what I’m interested in working on tomorrow.

1 Like

I think we do have one more spot available. I’m not sure about travel funding remaining, since I budgeted based on the earlier responses, but I’ll check.

I’m sure we can figure something out. The trains start at about 04:30 and take 30 minutes, if that’s early enough.

I’m definitely coming. Let me know if I am confirmed and what arrangements I should make myself.

The topics I am interested in (that I can think of now):

  • batchspawner/kubespawner/wrapspawner maintnance and relationship (I will make issues about my topics beforehand)
  • documentation
  • promiting JH as an interface to existing resources, not a separate platform
  • general JH hacking and development
1 Like

That sounds great! Definitely confirmed we have space for you, I just have to follow-up on funding, if that’s okay. Will do that via email.

1 Like

What’s the schedule? For now all I know is “Thursday to Sunday”. In particular, what are the earliest arrival times/latest departure times, and is there a difference in program each day or all hacking/social time that we’ll divide up depend on who is there and what we feel like? (Will it be empty the first/last days?)

I will arrive Friday around lunch time and be around till Sunday evening (my plane is Monday late afternoon). I think others I know of will arrive a bit earlier and stay till Sunday (afternoon).

I don’t think anyone has put together a formal agenda. Maybe we can spend some time on this issue working on one?? Would it be useful? Or maybe each morning do a brief “planning for the day” session? For me the most important part is “spend time with people I work with a lot online and haven’t ever/rarely seen in person”. This means it almost doesn’t matter what we spend time on related to the projects because it’ll all be interesting.

Of the three things I said I’d be interested in working on probably the third one is the one I’d start with because it benefits the most from in person communication and a shared white board. The other two are more technical in nature, more sit together with others and discover how to do this by trying to write some code.

I arrive Wednesday afternoon and leave Monday afternoon!

I’m currently very excited about @GeorgianaElena’s work on the Traefik proxy. The GitLab Helm chart comes with a Highly Available Redis deployment that I’d like to see if I can integrate with.

I read in the Jupyterhub Trafik Proxy docs here about the TKvProxy class that abstracts away the underlying mechanism on how the key value pairs are stored etc. I’m not very confident about the differences about using ETCD, Consul or other underlying KeyValue storage provides but eager to learn about Redis specifically because GitLab’s Helm chart comes with such deployment.

PS: I’m 0% confident about Redis vs ETCD etc but I like to not introduce additional containers running if possible.

1 Like

I’ll be there Thursday morning (8:10 at airport) to Sunday afternoon (16:10 from airport).

I’ll introduce myself: My background is as a computational researcher, and now I work for Aalto University (in Helsinki) in both the HPC group and Computer Science IT. For the last year, we’ve been doing more and more with Jupyter, and now we run two installations: one using batchspawner integrated to our HPC cluster for research, one using kubernetes on local hardware for teaching and general use. Through this time, I’ve ended up contributing to jupyterhub, batchspawner, nbgrader, and docker-stacks. Now we’ve got three other people working on Jupyter stuff for part of their time.

My vision is that Jupyter isn’t just another service, but a way to access services. I’ve gone through great efforts to make my JupyterHub deployments integrate to existing clusters, shell servers, data storage, and so on (easy with batchspawer, less obvious with kubernetes). My vision is that anywhere there’s an ssh server server or other ways of doing compute, Jupyter can exist in parallel and the old and new ways can support each other.

In the process I’ve worked on various add-on tools, such as envkernel, nbscript, and sjupyter. These generally fit the idea of “don’t redesign everything for Jupyter, but allow a seamless transition between Jupyter and existing systems”.


If we want it to run outside gitlab, we have to introduce a container for KV. I don’t really see any downsides to doing this, since it’s very small and a minor, hidden implementation detail where the lack of variability seems to outweigh the cost of one small container.