Feature/hellodata 1292 advanced analytics #83
Conversation
| # configuration parameter. | ||
|
|
||
| # Spawn single-user servers as Docker containers | ||
| c.JupyterHub.spawner_class = "dockerspawner.DockerSpawner" |
There was a problem hiding this comment.
On Cloud deployment we use instead of DockerSpawner the KubeSpawner as profile_list see https://git.mgmtbi.ch/kubernetes/gitops/public/helm_charts/-/merge_requests/226/diffs#e5f164e73bbc0ee61aad9e18b87d28f6844d7bf1_0_54.
E.g. we use existing dockerimage either already existing such as jupyter/datascience-notebook:latest or jupyter/r-notebook:x86_64-python-3.11. or otherwise from Workspaces, where we already install all the necessary python libraries.
@wieczorslawo: What are you mapping into the dockerfile atm that we must have in the dockerimage?
E.g. append_csp.py, wouldn't that be something that needs to be inside the sidecar of Jupyterhub, where the Sidecar would set that "authentication settings" as well as the Postgres User.
There was a problem hiding this comment.
The sidecar doesn't know anything about spawned pods/containers so it's not possible. This config needs to go directly to the notebook image or its pod config via env var or what have you.
https://jupyter-notebook.readthedocs.io/en/5.7.1/public_server.html#embedding-the-notebook-in-another-website
https://discourse.jupyter.org/t/csp-jupyterlab4/20917/5
jupyter/notebook#886
There was a problem hiding this comment.
"What are you mapping into the dockerfile atm that we must have in the dockerimage"
- the CSP config,
- the Python script to connect to the database with a temp user
- the readme file on how to use the script
That's how it's made for the docker-compose
| import os | ||
| import requests | ||
| import psycopg2 | ||
| from requests.auth import HTTPBasicAuth |
There was a problem hiding this comment.
I think this script for connecting the database and other potential libraries hello_data_scripts.py would need to go into a PyPi Package. For a short term, should we add it to the side car as well, as we need to map external things into the jupyterhub already (postgres user, tornado/keycloak settings, etc)?
There was a problem hiding this comment.
The sidecar is a REST API service that provides a one-day temporary user creds.
If we can add a shared volume across all spawned pods the file can be a configMap here
We just have to remember to add it later to the PYTHONPATH, right?
No description provided.