The method load_profiles can download a given ConfigMap from OpenShift/Kubernetes and extract the configuration from data section. This allows for dynamic changes to how (with what environment variables) singleuser servers are deployed - by updating the configuration in the ConfigMap, changes can be automatically pulled on singleuser server (re)start.
By default, the load_profiles method attempts to read profile data from a ConfigMap named jupyter-singleuser-profiles. In addition, it will also search for any ConfigMap with a label that matches jupyterhub=singleuser-profiles and load profile data from a key therein named jupyterhub-singleuser-profiles.yaml. This allows for dynamic customization of the JupyterHub deployment. These ConfigMaps will be loaded alphabetically, where configuration in later ConfigMaps will override earlier ones.
This approach requires c.KubeSpawner.modify_pod_hook option to point to a function similar to this - it simply calls out to SingleuserProfiles.apply_pod_profile which takes a pod and updates its configuration based on the profile and returns new pod manifest.
The profiles library can also read configuration from file - although this feature is more for testing purposes (to not require OpenShift/Kubernetes for developlment), it is possible to use file as a source of configuration in production.
JupyterHub deployed on OpenShift is not very flexible regarding configuration of particular servers. One use case we are tackling with this library is when a set of users need to have credentials preloaded in their servers to avoid pasting them directly to notebook - think access credentials for object storade.
Obviously anything else you need/can share in environment variables is a good fit - endpoint URLs, team wide configuration etc.
You can imagine different workflows will require different resources - submitting a Spark job will not be as resource heavy as running actual analysis on data - for that, you can use combination of images and resources section to set specific resource quotas for specific image(s).