How to Manage the Large, Medium, and Small Workspace Sizes for Users¶

As per the Creating Worskpaces Guide, GPU Cluster Manager provides default values for the Large, Medium, and Small workspaces that can be assigned. At times, an administrator may wish to increase or decrease these default sizes to meet the requirements of their organization's need.

GPU Cluster Manager provides a mechanism for it's Administrator role to manage the default workspace sizes viewed by Users.

User Experience¶

When a user wishes to create a new workspace, they will see the following within GPU Cluster Manager:

User Workspace Creation View

A User will first choose the environment from several listed, then choose the Compute Resources they wish to have within that environment.

The small, medium, large, fractional, and cpu-only instance size values can be modified by the admin using the following steps executed on the cluster management host:

Modifying Files on the Cluster Management Host¶

1. Get all the `compute crd` properties and values¶

kubectl get compute -n cattle-system

Example Output:

$ kubectl get compute -n cattle-system
NAME                DISPLAY NAME
cpu-only            CPU Only
fractional-nvidia   Fractional
large-nvidia        Large
medium-nvidia       Medium
small-nvidia        Small

2. Edit the size(s) of the desired compute CRD¶

kubectl edit compute <name> -n cattle-system

Note

Change the .spec.resources field when editing this file so that changes are reflected on the front end too!

Example output from changing the nvidia-small CRD:

$ kubectl edit compute small-nvidia -n cattle-system

 # Please edit the object below. Lines beginning with a '#' will be ignored,
 # and an empty file will abort the edit. If an error occurs while saving this file will be
 # reopened with the relevant failures.

  apiVersion: workload.memverge.ai/v1alpha1
 kind: Compute
 metadata:
  annotations:
    meta.helm.sh/release-name: mvai-ctrl-resources
    meta.helm.sh/release-namespace: cattle-system
  creationTimestamp: "2025-06-10T23:36:58Z"
  generation: 1
  labels:
    app.kubernetes.io/instance: mvai-ctrl-resources
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/name: mvai-ctrl-resources
    app.kubernetes.io/version: v0.4.0
    helm.sh/chart: mvai-ctrl-resources-0.4.0
  name: small-nvidia
  resourceVersion: "39939805"
  uid: 5f11b17b-b6ed-4500-9228-e0a96a478792
 spec:
  displayName: Small
  resources:
    cpu: "1"
    memory: 1Gi
    nvidia.com/gpu: "1"