Description of problem:

Hey all, over the weekend, Intel hit this (what appears to be a) bug

They attempt to spawn a notebook, small size, 2 CPU 8gb RAM, on a cluster that can easily satisfy that, and are hit with no CPU available errors (see screenshots)

I noticed that on a cluster that's totally empty, I had occasion to attempt to restart JH, and the pod was hit with the same scheduling errors, despite all workers being able to satisfy the requests.

They opened a RH case: https://access.redhat.com/support/cases/#/case/03023390

I was told it would be good to open here as well to ensure we didn't just find a GA blocker

Prerequisites (if any, like setup, operators/versions):

Steps to Reproduce

1. Go to JupyterHub and change the notebook size from default to small (or any other value)
2. Pod fails to spawn (reaches JH spawn timeout) saying no CPUs available, despite CPUs being available on workers

Actual results:

Expected results:

Reproducibility (Always/Intermittent/Only Once):

Always reproduces on Intel's environment

Build Details:

Additional info: