Kueue Visualizer for RHOAI Dashboard

A community plugin for the Red Hat OpenShift AI (RHOAI) Dashboard that provides visibility into Kueue workload scheduling — queue topology, capacity, preemption policies, workload status, cohort borrowing, and resource flavor mapping.

Features

Queue Infrastructure page

• Interactive topology graph: LocalQueue → ClusterQueue → Cohort, with ResourceFlavors as leaf nodes
• Click any node for a detail panel: capacity bars, preemption policy, flavor quotas
• Cohort borrowing ledger: nominal vs. used vs. borrowed per resource dimension
• Namespaces table with per-namespace LocalQueue breakdown

Workloads page

• Filterable workload table with queue position, borrow status, and priority
• Per-workload drawer: scheduling status, flavor assignments, lifecycle timeline, condition warnings

---

Quick Start

Deploy on an Existing Dashboard

Prerequisites: Helm 3, oc CLI access, Kueue installed on the cluster, access to redhat-ods-applications namespace (requires cluster-admin).

1. Install the plugin

helm install kueue-visualizer chart/ \
  --namespace cp-kueue-visualizer \
  --create-namespace

This creates a Deployment, Service, ServiceAccount, and an OpenShift Route.

2. Register with the RHOAI Dashboard

Retrieve the current Module Federation configuration, append the kueueVisualizer entry, and apply it directly to the dashboard Deployment:

oc get configmap federation-config \
  -n redhat-ods-applications \
  -o jsonpath='{.data.module-federation-config\.json}' \
| python3 -c "
import json, sys
config = json.load(sys.stdin)
config.append({
  'name': 'kueueVisualizer',
  'backend': {
    'remoteEntry': '/remoteEntry.js',
    'authorize': False,
    'tls': False,
    'service': {
      'name': 'kueue-visualizer',
      'namespace': 'cp-kueue-visualizer',
      'port': 8080
    }
  }
})
print(json.dumps(config))
" > /tmp/mf-config-extended.json

oc set env deployment/rhods-dashboard \
  -n redhat-ods-applications \
  "MODULE_FEDERATION_CONFIG=$(cat /tmp/mf-config-extended.json)"

New dashboard pods roll out automatically. After roughly two minutes, reload the RHOAI dashboard — Kueue > Queue Infrastructure and Kueue > Workloads appear in the sidebar under Community plugins.

3. Verify

oc set env deployment/rhods-dashboard -n redhat-ods-applications --list \
  | grep '^MODULE_FEDERATION_CONFIG=' \
  | python3 -c "import json,sys; d=json.loads(sys.stdin.read().split('=',1)[1].strip()); print('\n'.join(e['name'] for e in d))"

You should see kueueVisualizer in the output.

---

Build & Push

npm install
npm run build

# Build and push (auto-computes next version from git tags)
./scripts/build-push.sh

# Or with an explicit version
./scripts/build-push.sh 0.2.0

Or via Make:

make image-push VERSION=0.2.0

---

Deploy Custom Image

helm install kueue-visualizer chart/ \
  --namespace cp-kueue-visualizer \
  --create-namespace \
  --set image.repository=quay.io/<your-org>/kueue-visualizer \
  --set image.tag=0.2.0

---

Upgrade

helm upgrade kueue-visualizer chart/ \
  --namespace cp-kueue-visualizer \
  --set image.repository=quay.io/rh-ai-community-plugins/kueue-visualizer \
  --set image.tag=0.2.0

The MODULE_FEDERATION_CONFIG env var on the dashboard Deployment only needs to be re-applied if the service name or namespace changes.

---

Re-applying after operator reconciliation

If the RHOAI operator reconciles the rhods-dashboard Deployment (e.g. after a RHOAI upgrade), it may restore MODULE_FEDERATION_CONFIG from the ConfigMap, dropping the kueueVisualizer entry. Re-run step 2 above to restore it.

---

Development

npm install
npm run start:dev     # Dev server on port 9500
npm test              # Run tests
npm run lint          # ESLint + markdownlint
npm run typecheck     # TypeScript type check
npm run validate      # typecheck + lint + test

---

RBAC

The plugin uses the RHOAI dashboard's /api/k8s proxy, which makes API calls impersonating the logged-in user. No additional ClusterRole is created by the Helm chart.

Users need read access to Kueue resources to use this plugin. This is provided by Kueue's built-in kueue-batch-user-role ClusterRole, which grants get/list/watch on clusterqueues, localqueues, and workloads.

Granting access

In a future RHOAI release, this role will be granted automatically when a user is given distributed workloads access via the RHOAI dashboard. Until then, a cluster admin must grant it manually:

oc adm policy add-cluster-role-to-user kueue-batch-user-role <username>

Or for all users in a group:

oc adm policy add-cluster-role-to-group kueue-batch-user-role <group-name>

Users without this role will see an explanatory message in the plugin instead of an error.

---

Security

The plugin follows OpenShift restricted-v2 SCC requirements:

• runAsNonRoot: true (pod level)
• allowPrivilegeEscalation: false (container level)
• capabilities.drop: ["ALL"] (container level)
• seccompProfile.type: RuntimeDefault (container level)

---

License

Apache-2.0