Is it possible to achieve something similar on bare-metal Z2JH deployment, as shown in the Figure below? I’m thinking of revamping our cluster and I would offering a selection of docker images, where users could pick any number of GPUs and reserve a certain amount of CPU cores.
Take a look at wrapspawner which lets you define predefined profiles and expose them in Server Options.
You can create any spawner type at runtime with wrapspawner. If you are using K8s, kubespawner supports profiles. Did you look into it?
@yuvipanda has been working on profiles for Kubespawner as well. Check out this repo that lets you generate profiles via React app.
Oh, I read it too quickly. Thanks. I will look into it.
1 Like