Teardown

Reuse the same exports from Prerequisites (CLUSTER_NAME, AWS_REGION, AWS_PROFILE, LAB_DIR, and Argo CD vars if you created the capability).

export AWS_PROFILE=sandbox
export AWS_PAGER=""
export AWS_REGION=ap-southeast-2
export AWS_DEFAULT_REGION="$AWS_REGION"
export CLUSTER_NAME=kueue-lab
export LAB_DIR="$(pwd)/demo"   # from repo root; or export LAB_DIR="$(pwd)" if already in demo/
export ARGOCD_CAPABILITY_NAME=lab-argocd
export ARGOCD_ROLE_NAME=ArgoCDCapabilityRole

Work from demo/:

cd "$LAB_DIR"

Stop KueueViz port-forwards if they are still running:

pkill -f 'port-forward.*kueueviz' 2>/dev/null || true

Remove batch work and Kueue

Delete Jobs first so Auto Mode can drain Spot nodes, then remove queues, the NodePool, and the Argo CD Application:

kubectl delete jobs -n project-a --all --ignore-not-found
kubectl delete jobs -n project-b --all --ignore-not-found
kubectl delete jobs -n project-c --all --ignore-not-found

kubectl delete -f examples/kueue/queues.yaml --ignore-not-found
kubectl delete -f examples/kueue/batch-nodepool.yaml --ignore-not-found
kubectl delete -f examples/argocd/kueue-application.yaml --ignore-not-found
kubectl delete -f examples/argocd/in-cluster-secret.yaml --ignore-not-found

# Helm/Application usually removes kueue-system; force if needed
kubectl delete namespace kueue-system --ignore-not-found --wait=false

If a manifest file is missing, delete by name instead:

kubectl delete application kueue -n argocd --ignore-not-found
kubectl delete secret in-cluster -n argocd --ignore-not-found
kubectl delete clusterqueue shared-batch --ignore-not-found
kubectl delete resourceflavor spot on-demand --ignore-not-found
kubectl delete nodepool batch --ignore-not-found
kubectl delete namespace project-a project-b project-c --ignore-not-found

Delete Argo CD capability

If you created the lab capability:

export ARGOCD_ROLE_ARN="arn:aws:iam::$(aws sts get-caller-identity --query Account --output text):role/${ARGOCD_ROLE_NAME}"

aws eks disassociate-access-policy \
  --cluster-name "$CLUSTER_NAME" \
  --principal-arn "$ARGOCD_ROLE_ARN" \
  --policy-arn arn:aws:eks::aws:cluster-access-policy/AmazonEKSClusterAdminPolicy \
  --region "$AWS_REGION" 2>/dev/null || true

eksctl delete capability -f config/argocd-capability.yaml

# wait until gone (eksctl has no --wait on delete capability)
while eksctl get capability --cluster "$CLUSTER_NAME" --region "$AWS_REGION" 2>/dev/null | grep -q lab-argocd; do
  echo "waiting for capability delete..."
  sleep 15
done

# Role may already be gone if CloudFormation removed it with the capability stack
aws iam delete-role --role-name "$ARGOCD_ROLE_NAME" 2>/dev/null || true

If config/argocd-capability.yaml is gone, delete by name:

eksctl delete capability \
  --cluster "$CLUSTER_NAME" \
  --name "$ARGOCD_CAPABILITY_NAME" \
  --region "$AWS_REGION"

This lab attaches no extra IAM policies to ArgoCDCapabilityRole. delete-role returning NoSuchEntity after a successful capability delete is fine.

Destroy cluster

eksctl delete cluster -f config/cluster.yaml --wait

If the file is missing:

eksctl delete cluster --name "$CLUSTER_NAME" --region "$AWS_REGION" --wait

Confirm the cluster is gone:

aws eks describe-cluster --name "$CLUSTER_NAME" --region "$AWS_REGION" 2>&1 | head

Expect ResourceNotFoundException.