Manifest reference

Lab manifests are generated under demo/ at the repository root during Deploy and Operate. That directory is gitignored — do not create config/ or examples/ at the repo root.

Relative paths in deploy pages (config/…, examples/…) always mean $LAB_DIR/….

Layout

Directorydemo/
- Directoryconfig/
  - cluster.yaml
  - argocd-capability.yaml
  - Directoryiam/
    argocd-trust-policy.json
- Directoryexamples/
  - Directoryargocd/
    in-cluster-secret.yaml
    kueue-application.yaml
  - Directorykueue/
    queues.yaml
    batch-nodepool.yaml
  - Directoryworkloads/
    sample-jobs.yaml
    burst-jobs.sh

demo/config/

File	Created on	Purpose
`cluster.yaml`	Cluster (Auto Mode)	eksctl Auto Mode cluster (`kueue-lab`, Kubernetes 1.36)
`argocd-capability.yaml`	Argo CD capability	Managed Argo CD capability (`lab-argocd`)
`iam/argocd-trust-policy.json`	Argo CD capability	Trust policy for `ArgoCDCapabilityRole`

demo/examples/argocd/

File	Created on	Purpose
`in-cluster-secret.yaml`	Kueue via GitOps	Register this EKS cluster by ARN as an Argo CD destination (not `kubernetes.default.svc`)
`kueue-application.yaml`	Kueue via GitOps	Argo CD Application for the Kueue Helm chart + KueueViz

Notable Helm values in kueue-application.yaml (lab defaults):

Value	Why
`enableKueueViz: true`	Installs KueueViz with the controller
`controllerManager` / `kueueViz.*` `nodeSelector` + `CriticalAddonsOnly`	Land on Auto Mode system pool
`kueueViz.backend.ingress.enabled: false`	Lab uses port-forward, not Ingress
`kueueViz.backend.ingress.host: localhost:8081`	Chart still uses this host to write `ws://localhost:8081` into the frontend ConfigMap
`KUEUEVIZ_ALLOWED_ORIGINS`	CORS for `http://localhost:8080` and `http://127.0.0.1:8080`
`syncPolicy.automated.selfHeal: true`	Do not kubectl-patch ConfigMaps/Deployments — edit Helm values instead

Chart pin used in this walkthrough: KUEUE_CHART_VERSION=0.19.1.

demo/examples/kueue/

File	Created on	Purpose
`queues.yaml`	Queues and flavors	Namespaces `project-a/b/c`, ResourceFlavors `spot` / `on-demand`, ClusterQueue `shared-batch`, LocalQueues — not Argo CD AppProjects
`batch-nodepool.yaml`	Queues and flavors	Auto Mode NodePool `batch` allowing Spot (+ On-Demand) for Spot-flavored Jobs

Default quota exports when you generate queues.yaml:

Export	Default	Used for
`SPOT_CPU_QUOTA`	`10`	Spot flavor CPU in `shared-batch`
`SPOT_MEMORY_QUOTA`	`20Gi`	Spot flavor memory
`ON_DEMAND_CPU_QUOTA`	`4`	On-Demand flavor CPU
`ON_DEMAND_MEMORY_QUOTA`	`8Gi`	On-Demand flavor memory

demo/examples/workloads/

File	Created on	Purpose
`sample-jobs.yaml`	Queues and flavors	One `sample-job` per project namespace (busybox, 1 CPU, queue label)
`burst-jobs.sh`	Verify batch	Clones each tenant’s `sample-job` with `jq` into `burst-N` Jobs (default sleep 180s)

Key labels

Key	Value example	On
`kueue.x-k8s.io/queue-name`	`project-a`	Job `metadata.labels` — must match a LocalQueue name in that namespace

Objects the lab creates (cluster)

Kind	Name(s)	Source
Namespace	`project-a`, `project-b`, `project-c`, `kueue-system`, `argocd`	queues / Helm / capability
ResourceFlavor	`spot`, `on-demand`	`queues.yaml`
ClusterQueue	`shared-batch`	`queues.yaml`
LocalQueue	`project-a` / `b` / `c` (in matching ns)	`queues.yaml`
NodePool	`batch`	`batch-nodepool.yaml`
Application	`kueue` (ns `argocd`)	`kueue-application.yaml`
Secret	`in-cluster` (ns `argocd`)	`in-cluster-secret.yaml`

Upstream docs