Is there a way to emulate a login storm? I want to test what happens when 40 to 100 students login at the start of a lecture. Is it possible to script this?
It is definitely possible to test this. You can write your own tests using the JupyterHub REST API, but this is what hubtraf was written to test, I believe. You can use it directly, or just for inspiration and reference to write your own scripts.
Another approach is using general purpose tools like locust, again generally exciting REST APIs, but can also throw a couple selenium sessions in there to do the full thing. Coupling this with opentelemetry at a couple key places can give a very good picture of what is really happening When Things Go Wrong.
There is also GitHub - IBM/jupyter-tools: Collection of tools for working with JupyterHub and notebooks to simulate users.
With the proposed changes, it runs without issues.
proposed changes
Welp, having waited my share of time for IBMRedHat PRs to get reviewed… without a customer success advocate, might be waiting a long time for that to get prioritized.
For testing and performance evaluation, this is why general purpose, more broadly-maintained tools can be more robust to keep running over time.
Thanks. Your fork of the jupyter-tools works great!