Workbenches (Jupyter Notebooks)¶
Workbenches provide managed Jupyter notebook environments for data scientists. Each workbench is a container running JupyterLab with pre-installed ML libraries, persistent storage, and optional GPU access. Workbenches are managed through the RHOAI Dashboard.
Dependencies¶
| Requirement | Type | Path |
|---|---|---|
| RHOAI Operator | Operator | components/operators/rhoai-operator/ |
DSC workbenches: Managed |
DSC component | components/instances/rhoai-instance/ |
DSC dashboard: Managed |
DSC component | Required for the UI to manage workbenches |
| GPU Infrastructure (optional) | Operator + Instance | See gpu-infrastructure.md |
Workbenches run on CPU by default. GPU access requires the GPU infrastructure stack (NFD + GPU Operator).
Enable It¶
Workbenches are enabled in the dev and full overlays. The minimal overlay enables Dashboard (the prerequisite).
Deploy¶
Workbenches are enabled automatically when the rhoai-instance ArgoCD Application points to the full or dev overlay.
# 1. Install the RHOAI operator
oc apply -k components/operators/rhoai-operator/
oc get csv -A | grep rhods
# 2. Create DSC with workbenches enabled (dev or full overlay)
oc apply -k components/instances/rhoai-instance/overlays/dev/
# 3. Wait for DSC
oc wait --for=jsonpath='{.status.conditions[?(@.type=="Ready")].status}'=True \
datasciencecluster/default-dsc --timeout=600s
Verify¶
# Notebook controller should be running
oc get pods -n redhat-ods-applications -l app=notebook-controller
# RHOAI Dashboard should be accessible
oc get route rhods-dashboard -n redhat-ods-applications
Usage¶
- Open the RHOAI Dashboard (route above)
- Navigate to Data Science Projects and create a project
- Click Create workbench in your project
- Select a notebook image (e.g., Standard Data Science, PyTorch, TensorFlow)
- Choose container size and optional GPU
- The workbench starts as a pod with persistent storage
Pre-built notebook images¶
RHOAI ships several validated images: - Standard Data Science -- pandas, scikit-learn, matplotlib - PyTorch -- PyTorch + CUDA - TensorFlow -- TensorFlow + CUDA - Minimal -- JupyterLab only
Disable It¶
Set workbenches.managementState to Removed in the DSC.
Stop running workbenches first: