Repository navigation
What Python installation/platform? #26
Description
Last time we used a VM with Python already installed but this might not be an option in the future?
Decisions to make:
- Do we (a) get everyone to install it on their own machines or (b) provide a platform with it preinstalled?
- If (a), what installation method, works for all OSs, need extra pre-workshop setup session, extra helpers to start
- If (b), what service, who/how hosting, who pays?
Metadata
Metadata
Assignees
Labels
No labels
Activity
[-]What Python installation/PaaS?[/-][+]What Python installation/platform?[/+]SJaffa commentedon Feb 25, 2025
@douglowe what solution did we use on the course in the end and was it suitable? Is that what we want to go with long term?
douglowe commentedon Feb 27, 2025
We used the VM that we have used for the last 3-4 years.
We should investigate the jupyter lab servers more over the next 2 months, and see if they are suitable though. It might be nice for us to use a central solution instead.
SJaffa commentedon May 7, 2025
Current solution: ITS-owned Jupyter Hub located at https://jupyter.its.manchester.ac.uk/
We requested an increase in memory allocation per user which should allow people to run the conda installations, it might be slow if they all do it at once but we will see on the day.
We need to give them get them to download the data and upload it to the jupyter hub, adn test if we can create and use environments in the jupyter notebook.
SJaffa commentedon May 7, 2025
In the terminal in Jupyter hub:
listing and creating conda environments works, but before acivating you need to do
conda init bash. This advises you to close the terminal and open it which loses the bashrc, so instead you can keep the terminal open and runsource ~/.bashrcto initialise conda, then activate itSJaffa commentedon May 7, 2025
Could run scripts form the terminal with
conda run -n py38 python my_script.py(source)SJaffa commentedon May 7, 2025
All the libraries we use in the course are preinstalled in the Jupyter Hub (I did Ctrl-f "import" and checked they all run), so we don't need to get the environment working for the notebook, it just seems a bit disjointed to teach them and then never use it...
douglowe commentedon May 7, 2025
I agree it is now disjointed - the central team were a bit too helpful in installing all the libraries we needed for this course.
Ideally we'll split out the packaging episode into a dedicated packaging lesson at some point. For now perhaps you could simply move this lesson to the end of the day, so that it doesn't disrupt the flow of the rest of the episodes.
andrewgait commentedon Apr 16, 2026
I agree with @douglowe that the temporary solution here should be to move the
condachapter to the end of the day.andrewgait commentedon May 1, 2026
We used the central Jupyter instance with no issues yesterday (as far as I could see, the biggest issue was it took a while to load up initially for some people). I would probably argue that this issue can be closed, and if we need to reconsider what we use in the future then a new issue for that should be made. @Scottan any thoughts?
Scottan commentedon May 1, 2026
Yeah, I think this can be closed. The Jupyter server worked well, the main problem was that it seemed to struggle with lots of people logging on at the same time at the start of the course, but that was largely solved by waiting a bit then refreshing.
If we need to revisit this we can open a new issue. There might need be an action in the main training repo to ask for more resource for the server from infra to see if that helps with its performance? (it might just need a clean-out as it has been used for a number of course already)