Staff use of JupyterHub
The JupyterHub is a teaching resource, and there are a couple of ways you can customise the experience for your students.
Users see an examples/
folder in their home directory. This points
to a collection of example resources, managed by a class- or lab-head.
If your class needs this (and we expect most classes do), then talk to
the JupyterHub admins. Once this is in place, you can manage it by
ssh-ing to brutha0.physics.gla.ac.uk
and looking in
/local/examples
.
Gotchas regarding the examples/
folder:
-
When users are still unfamiliar with the JupterHub interface, they can end up in the examples folder, and trying to save files there. They're not permitted to do this, obviously, but the error message isn't terribly clear, and it might be worth priming lab demonstrators to look out for this.
-
It might be worth creating a little script which unpacks a writable copy of the examples in the student's own space. The JupyterHub admins can advise about this.
-
In some circumstances, students have occasionally managed to accidentally delete the link to the
examples/
folder. If this happens, a lab demonstrator can help the student patch things up by getting the student to open a terminal (ie, a certain amount of handholding is required here), andln -s /local/examples
in their home directory. If that's problematic, the JupyterHub admins can sort it out.
The initial allocation of disk quota is fairly mean – only 4GB. We
can increase this if suggested by the lab head or a demonstrator.
Users can see their quota at https://it.physics.gla.ac.uk/identity.
If someone runs out of quota, they can't write. Unfortunately, the
JupyterHub UI doesn't deal with this at all gracefully, and produces a
rather apocalyptic ‘server error’ message. If this happens to one of
your students, then checking the quota is possibly the first thing to
do (staff users can also check users' quotas in a terminal,
by saying brutha-mgt quota <username>
).
For brief forays into the terminal, recall that JupyterHub has a
Terminal
button on the Launcher page; you don't have to ssh to
brutha.physics
to get a shell (though you can if you want to).
JupyterHub allows users to select a ‘kernel’, which is an environment containing a certain selection of libraries (it's really just the JupyterHub name for Python ‘virtual environments’). You may want to customise this, or add new ones for particular classes: talk to the JupyterHub admins about this if you think this will be useful or necessary. We don't generally want too many different kernels, since this is potentially confusing to users.