ITS Jupyter instance feedback

[andrewgait]

@douglowe has put me in touch with Tony Fargher who is part of the ITS team who develop/maintain https://jupyter.its.manchester.ac.uk. Tony would like information about the below:

The two main things I wanted to find out from everyone are: (1) how well has JupyterHub been working for you and (2) which specific tools and packages in JupyterHub do you use?

The reason for (2) is: we might be able to create smaller, more customised "images" to deliver better performance. This is something the tech team are looking into.

I am meeting with him on the afternoon of Monday 15th June and I'm happy to take any requests or issues that we have in the training team to him in this meeting.

As far as I know the only courses which have used this Jupyter instance are the Python courses (Intro, Intermediate), but if anyone else is using it feel free to comment as well.

Tagging @DrJonnyT @emmasimp @AnnAnnFryingPan @Scottan @BMatischen01 @jfecroftRSE who are I think the most recent trainers/helpers who have used the instance.

[DrJonnyT]

For intro to python:

(1) how well has JupyterHub been working for you

I think it's generally been great? We have a course of about 20-40 learners who use it and generally without it causing any issues. When everyone logs in at once it can be a bit slow to get them in. We did have an issue one time where the two trainers had logged in the day before, then something had happened overnight that meant it randomly stopped working for both trainers at the same time. Someone from the PET team sent us a link to the reset button and that sorted it. Hasn't happened since.

(2) which specific tools and packages in JupyterHub do you use?

Python, notebooks, a small amount of terminal. Numpy, matplotlib, glob.

Potential issues going forward
We use it for training because it's easy and we don't have to help them install python. People might then think that they can use it to do all the work- should they be doing it on their laptops instead? We tell them not to store data on there as there is no guarantee it will stay there.

[andrewgait]

I met with Tony today and fed back our specific information regarding what tools and packages we use. This will then feed back into their technical team's process for creating different "images" for our use (perhaps separate ones for each course, or perhaps an overall one for RIT (Python) training). I also spoke of our nascent plans for a possible split of the material in the intermediate python course into separate individual courses (see e.g. UoMResearchIT/Programming-With-Python#118).

I also asked on behalf of @DrAtbas for the technical team's opinion on whether we could run the Bash/Shell course via the Jupyter instance; this is not usually a problem for anyone who does that course, but we were thinking there could plausibly at some point be someone who has a technical issue that limits them to only using a browser for some reason, and that Jupyter might be a backup that they could use in this instance.

We use it for training because it's easy and we don't have to help them install python. People might then think that they can use it to do all the work- should they be doing it on their laptops instead? We tell them not to store data on there as there is no guarantee it will stay there.

I asked about this directly in the meeting, in terms of how permanent the data is on the Jupyter system; as Tony's not part of the technical team, he said he would find out the official position for us (or get them to come up with an official position) so that we can pass this information onto the learners.

Tony will be feeding information back to me on these processes, so when I receive that information I will add it to this issue.

[andrewgait]

I asked about this directly in the meeting, in terms of how permanent the data is on the Jupyter system; as Tony's not part of the technical team, he said he would find out the official position for us (or get them to come up with an official position) so that we can pass this information onto the learners.

I have had a reply on this from the tech team via Tony:

'Data is stored on EFS and is persistent. However, the data is scheduled to move to different tier archive storage after a while - not deleted. We do have back ups, for DR purposes. But we encourage users to save their work on source control tooling (Github etc.)'

As far as I can see, that means that the user's files will be retained - but the official advice is still to keep you own copy.

So this basically matches what we have been saying - there are better solutions for storing their work and we should encourage them to use these.