search

opendatahub-io-contrib / jupyterhub-singleuser-profiles

MIT License
11 stars 24 forks source link

Get Maximum Available GPU using Prometheus Data #216

Closed VaishnaviHire closed 2 years ago

VaishnaviHire commented 2 years ago

This PR implements changes that include using Prometheus metrics from dcgm-exporter to obtain maximum available GPUs for users at a given time.

Related PR: https://github.com/opendatahub-io/odh-manifests/pull/551

VaishnaviHire commented 2 years ago

Testing Steps:

  1. Ensure you have GPU Operator installed and GPU nodes enabled in the cluster
  2. Install ODH Operator
  3. Scale the Operator to 0 and Update Jupyter Hub DC with custom image.
  4. Update jupyterhub-cluster role with permissions given here - https://github.com/opendatahub-io/odh-manifests/pull/551/files
  5. Verify available GPUs in Spawner UI.