UNOFFICIAL/EXPERIMENTAL
This is a dated record of one run. It is not a Guide. Every line that starts with $ is a command that was run, and the text under it is the output it printed. Times are UTC.
Versions and setting
- cub v0.8.7 with the installer plugin, in a home directory that holds no ConfigHub credentials. No account was used.
- kind v0.31.0 with a node at Kubernetes v1.35.0, kubectl client v1.36.0, kustomize v5.8.1, Docker 29.4.0.
- The cluster had one node and no GPUs.
- The script was fetched from the live site and is byte-for-byte the file at
site/sh/nvidia-gpu-operator-v26-7-1/default/try.shin this repository. - The scratch directory is written
<work>. The cluster had its own kubeconfig file there, so the machine's default kubeconfig was not touched.
The answer in brief
The script ran from start to finish in 8 seconds and exited 0. All eight CRDs were established before any other object was applied. The operator and the three node-feature-discovery workloads were running 47 seconds after the script ended. The ClusterPolicy reported ready. The operator created none of its GPU DaemonSets, because the cluster had no GPU nodes.
Make the cluster
$ kind create cluster --name cwtest-1010-gpuop --kubeconfig <work>/kubeconfig --wait 120s
$ kubectl get nodes
started 2026-10-10T20:34:36Z
kubectl cluster-info --context kind-cwtest-1010-gpuop --kubeconfig <work>/kubeconfig
Thanks for using kind! 😊
NAME STATUS ROLES AGE VERSION
cwtest-1010-gpuop-control-plane Ready control-plane 22s v1.35.0Fetch the script and run it
New to cub? Install the cub CLI first. Public catalog packages pull and render anonymously, and you sign in only once a command saves or changes ConfigHub data.
$ curl -s -o try.sh https://confighub.github.io/helm-expt/site/sh/nvidia-gpu-operator-v26-7-1/default/try.sh
$ bash ./try.sh
started 2026-10-10T20:35:11Z
>> Pull the package and render the default base variant into ./nvidia-gpu-operator-v26-7-1-default
Wizard wrote <work>/nvidia-gpu-operator-v26-7-1-default/out/spec/selection.yaml and inputs.yaml
Collector produced 2 fact(s) in <work>/nvidia-gpu-operator-v26-7-1-default/out/spec/facts.yaml
Base: default; components: []
Namespace: gpu-operator
Rendered 28 manifest(s) to <work>/nvidia-gpu-operator-v26-7-1-default/out/manifests
Next: cub installer upload --work-dir <work>/nvidia-gpu-operator-v26-7-1-default --space <slug>
>> Read what was rendered; nothing has touched the cluster yet
clusterpolicy-cluster-policy.yaml
clusterrole-gpu-operator-node-feature-discovery-gc.yaml
clusterrole-gpu-operator-node-feature-discovery.yaml
clusterrole-gpu-operator.yaml
clusterrolebinding-gpu-operator-node-feature-discovery-gc.yaml
clusterrolebinding-gpu-operator-node-feature-discovery.yaml
clusterrolebinding-gpu-operator.yaml
configmap-gpu-operator-gpu-operator-node-feature-discovery-master-conf.yaml
configmap-gpu-operator-gpu-operator-node-feature-discovery-worker-conf.yaml
customresourcedefinition-clusterpolicies-nvidia-com.yaml
customresourcedefinition-computedomaincliques-resource-nvidia-com.yaml
customresourcedefinition-computedomains-resource-nvidia-com.yaml
customresourcedefinition-gpuclusters-nvidia-com.yaml
customresourcedefinition-nodefeaturegroups-nfd-k8s-sigs-io.yaml
customresourcedefinition-nodefeaturerules-nfd-k8s-sigs-io.yaml
customresourcedefinition-nodefeatures-nfd-k8s-sigs-io.yaml
customresourcedefinition-nvidiadrivers-nvidia-com.yaml
daemonset-gpu-operator-gpu-operator-node-feature-discovery-worker.yaml
deployment-gpu-operator-gpu-operator-node-feature-discovery-gc.yaml
deployment-gpu-operator-gpu-operator-node-feature-discovery-master.yaml
deployment-gpu-operator-gpu-operator.yaml
namespace-gpu-operator.yaml
role-gpu-operator-gpu-operator-node-feature-discovery-worker.yaml
role-gpu-operator-gpu-operator.yaml
rolebinding-gpu-operator-gpu-operator-node-feature-discovery-worker.yaml
rolebinding-gpu-operator-gpu-operator.yaml
serviceaccount-gpu-operator-gpu-operator.yaml
serviceaccount-gpu-operator-node-feature-discovery.yaml
>> Ensure the gpu-operator namespace exists
namespace/gpu-operator created
>> Check 8 CRDs included with this package
customresourcedefinition.apiextensions.k8s.io/clusterpolicies.nvidia.com serverside-applied
customresourcedefinition.apiextensions.k8s.io/computedomaincliques.resource.nvidia.com serverside-applied
customresourcedefinition.apiextensions.k8s.io/computedomains.resource.nvidia.com serverside-applied
customresourcedefinition.apiextensions.k8s.io/gpuclusters.nvidia.com serverside-applied
customresourcedefinition.apiextensions.k8s.io/nodefeaturegroups.nfd.k8s-sigs.io serverside-applied
customresourcedefinition.apiextensions.k8s.io/nodefeaturerules.nfd.k8s-sigs.io serverside-applied
customresourcedefinition.apiextensions.k8s.io/nodefeatures.nfd.k8s-sigs.io serverside-applied
customresourcedefinition.apiextensions.k8s.io/nvidiadrivers.nvidia.com serverside-applied
customresourcedefinition.apiextensions.k8s.io/clusterpolicies.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/computedomaincliques.resource.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/computedomains.resource.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/gpuclusters.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/nodefeaturegroups.nfd.k8s-sigs.io condition met
customresourcedefinition.apiextensions.k8s.io/nodefeaturerules.nfd.k8s-sigs.io condition met
customresourcedefinition.apiextensions.k8s.io/nodefeatures.nfd.k8s-sigs.io condition met
customresourcedefinition.apiextensions.k8s.io/nvidiadrivers.nvidia.com condition met
>> Apply the rendered objects
clusterpolicy.nvidia.com/cluster-policy created
clusterrole.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-gc created
clusterrole.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery created
clusterrole.rbac.authorization.k8s.io/gpu-operator created
clusterrolebinding.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-gc created
clusterrolebinding.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery created
clusterrolebinding.rbac.authorization.k8s.io/gpu-operator created
configmap/gpu-operator-node-feature-discovery-master-conf created
configmap/gpu-operator-node-feature-discovery-worker-conf created
Warning: resource customresourcedefinitions/clusterpolicies.nvidia.com is missing the kubectl.kubernetes.io/last-applied-configuration annotation which is required by kubectl apply. kubectl apply should only be used on resources created declaratively by either kubectl create --save-config or kubectl apply. The missing annotation will be patched automatically.
customresourcedefinition.apiextensions.k8s.io/clusterpolicies.nvidia.com configured
... (7 more warnings of the same kind cut, one for each of the other CRDs)
customresourcedefinition.apiextensions.k8s.io/computedomaincliques.resource.nvidia.com configured
customresourcedefinition.apiextensions.k8s.io/computedomains.resource.nvidia.com configured
customresourcedefinition.apiextensions.k8s.io/gpuclusters.nvidia.com configured
customresourcedefinition.apiextensions.k8s.io/nodefeaturegroups.nfd.k8s-sigs.io configured
customresourcedefinition.apiextensions.k8s.io/nodefeaturerules.nfd.k8s-sigs.io configured
customresourcedefinition.apiextensions.k8s.io/nodefeatures.nfd.k8s-sigs.io configured
customresourcedefinition.apiextensions.k8s.io/nvidiadrivers.nvidia.com configured
daemonset.apps/gpu-operator-node-feature-discovery-worker created
deployment.apps/gpu-operator-node-feature-discovery-gc created
deployment.apps/gpu-operator-node-feature-discovery-master created
deployment.apps/gpu-operator created
namespace/gpu-operator configured
role.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-worker created
role.rbac.authorization.k8s.io/gpu-operator created
rolebinding.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-worker created
rolebinding.rbac.authorization.k8s.io/gpu-operator created
serviceaccount/gpu-operator created
serviceaccount/node-feature-discovery created
>> Done. The cluster received exactly the files in ./nvidia-gpu-operator-v26-7-1-default/out, with the packaged lifecycle steps checked.
[exit 0]
finished 2026-10-10T20:35:19ZWatch what the cluster did
started 2026-10-10T20:35:29Z
$ kubectl -n gpu-operator rollout status deployment/gpu-operator --timeout=300s
Waiting for deployment "gpu-operator" rollout to finish: 0 of 1 updated replicas are available...
deployment "gpu-operator" successfully rolled out
$ kubectl -n gpu-operator get deploy,ds
NAME READY UP-TO-DATE AVAILABLE AGE
deployment.apps/gpu-operator 1/1 1 1 34s
deployment.apps/gpu-operator-node-feature-discovery-gc 0/1 1 0 35s
deployment.apps/gpu-operator-node-feature-discovery-master 0/1 1 0 35s
NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE
daemonset.apps/gpu-operator-node-feature-discovery-worker 1 1 0 1 0 <none> 35s
$ kubectl get clusterpolicy cluster-policy -o jsonpath=...
ready Ready=True(NFDLabelsMissing) Error=False(Ready)
$ kubectl get crd | grep -c nvidia.com\|nfd
8
$ kubectl -n gpu-operator get pods
NAME READY STATUS RESTARTS AGE
gpu-operator-7486b8b7d9-wwwrm 1/1 Running 0 34s
gpu-operator-node-feature-discovery-gc-d4887b8cf-rg8kq 0/1 ContainerCreating 0 34s
gpu-operator-node-feature-discovery-master-74f7f6fc45-579t6 0/1 ContainerCreating 0 34s
gpu-operator-node-feature-discovery-worker-4b6pf 0/1 ContainerCreating 0 34s
deployment "gpu-operator-node-feature-discovery-master" successfully rolled out
daemon set "gpu-operator-node-feature-discovery-worker" successfully rolled out
$ kubectl -n gpu-operator get pods
NAME READY STATUS RESTARTS AGE
gpu-operator-7486b8b7d9-wwwrm 1/1 Running 0 47s
gpu-operator-node-feature-discovery-gc-d4887b8cf-rg8kq 1/1 Running 0 47s
gpu-operator-node-feature-discovery-master-74f7f6fc45-579t6 1/1 Running 0 47s
gpu-operator-node-feature-discovery-worker-4b6pf 1/1 Running 0 47s
$ clusterpolicy
ready Ready=True(NFDLabelsMissing) Error=False(Ready)
$ kubectl -n gpu-operator get ds
NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE
gpu-operator-node-feature-discovery-worker 1 1 1 1 1 <none> 48s
observed 2026-10-10T20:36:06ZThe first look came 10 seconds after the script ended, while three images were still being pulled. The second look shows all four pods running.
What went wrong in the first run
- The script applies the CRDs twice. It first applies the packaged CRD file with server-side apply and waits. It then applies every rendered object with a plain
kubectl apply, and the rendered objects include the same eight CRDs. Kubernetes prints one warning for each CRD about a missinglast-applied-configurationannotation. Nothing failed. A reader will still wonder what the eight warnings mean. - The closing line says more than the script did. It ends with "with the packaged lifecycle steps checked". The script carried out the install steps. It did not run the Job the plan names for before an upgrade, or the Job it names for after a delete.
The script was corrected, and run again
Both faults above were fixed in the script generator the same day. The corrected script applies each CRD once, and its last lines say only what it did. It was run twice more, on a second throwaway cluster, with the system's bash 3.2.
The first of the two runs was on the empty cluster. It printed no warning and exited 0 in 6 seconds. Its last line named the steps it had not run, in an earlier wording that also listed pre-apply. That was wrong, because the script does run the CRD step, and the wording was corrected before the second run. This is that first run from the namespace step on.
$ /bin/bash ./try.sh
>> Ensure the gpu-operator namespace exists
namespace/gpu-operator created
>> Check 8 CRDs included with this package
customresourcedefinition.apiextensions.k8s.io/clusterpolicies.nvidia.com serverside-applied
customresourcedefinition.apiextensions.k8s.io/computedomaincliques.resource.nvidia.com serverside-applied
customresourcedefinition.apiextensions.k8s.io/computedomains.resource.nvidia.com serverside-applied
customresourcedefinition.apiextensions.k8s.io/gpuclusters.nvidia.com serverside-applied
customresourcedefinition.apiextensions.k8s.io/nodefeaturegroups.nfd.k8s-sigs.io serverside-applied
customresourcedefinition.apiextensions.k8s.io/nodefeaturerules.nfd.k8s-sigs.io serverside-applied
customresourcedefinition.apiextensions.k8s.io/nodefeatures.nfd.k8s-sigs.io serverside-applied
customresourcedefinition.apiextensions.k8s.io/nvidiadrivers.nvidia.com serverside-applied
customresourcedefinition.apiextensions.k8s.io/clusterpolicies.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/computedomaincliques.resource.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/computedomains.resource.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/gpuclusters.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/nodefeaturegroups.nfd.k8s-sigs.io condition met
customresourcedefinition.apiextensions.k8s.io/nodefeaturerules.nfd.k8s-sigs.io condition met
customresourcedefinition.apiextensions.k8s.io/nodefeatures.nfd.k8s-sigs.io condition met
customresourcedefinition.apiextensions.k8s.io/nvidiadrivers.nvidia.com condition met
>> Apply the rendered objects, except the CRDs already settled above
clusterpolicy.nvidia.com/cluster-policy created
clusterrole.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-gc created
clusterrole.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery created
clusterrole.rbac.authorization.k8s.io/gpu-operator created
clusterrolebinding.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-gc created
clusterrolebinding.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery created
clusterrolebinding.rbac.authorization.k8s.io/gpu-operator created
configmap/gpu-operator-node-feature-discovery-master-conf created
configmap/gpu-operator-node-feature-discovery-worker-conf created
daemonset.apps/gpu-operator-node-feature-discovery-worker created
deployment.apps/gpu-operator-node-feature-discovery-gc created
deployment.apps/gpu-operator-node-feature-discovery-master created
deployment.apps/gpu-operator created
namespace/gpu-operator configured
role.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-worker created
role.rbac.authorization.k8s.io/gpu-operator created
rolebinding.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-worker created
rolebinding.rbac.authorization.k8s.io/gpu-operator created
serviceaccount/gpu-operator created
serviceaccount/node-feature-discovery created
>> Done. The cluster received the files in ./nvidia-gpu-operator-v26-7-1-default/out. Its CRDs were settled first, from the package.
>> Not run by this script: the pre-apply and pre-upgrade and pre-delete and post-delete steps of this base's plan. Those belong to an upgrade or a delete.
[exit 0]
finished 2026-10-10T21:31:56ZThe second run used the final script on the same cluster, where everything already existed. It left the eight CRDs alone, reported every other object unchanged, printed no warning and exited 0 in 5 seconds.
$ /bin/bash ./try.sh
>> Ensure the gpu-operator namespace exists
namespace/gpu-operator configured
>> Check 8 CRDs included with this package
>> The required CRDs already exist; leave them under their current owner
customresourcedefinition.apiextensions.k8s.io/clusterpolicies.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/computedomaincliques.resource.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/computedomains.resource.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/gpuclusters.nvidia.com condition met
customresourcedefinition.apiextensions.k8s.io/nodefeaturegroups.nfd.k8s-sigs.io condition met
customresourcedefinition.apiextensions.k8s.io/nodefeaturerules.nfd.k8s-sigs.io condition met
customresourcedefinition.apiextensions.k8s.io/nodefeatures.nfd.k8s-sigs.io condition met
customresourcedefinition.apiextensions.k8s.io/nvidiadrivers.nvidia.com condition met
>> Apply the rendered objects, except the CRDs already settled above
clusterpolicy.nvidia.com/cluster-policy unchanged
clusterrole.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-gc unchanged
clusterrole.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery unchanged
clusterrole.rbac.authorization.k8s.io/gpu-operator unchanged
clusterrolebinding.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-gc unchanged
clusterrolebinding.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery unchanged
clusterrolebinding.rbac.authorization.k8s.io/gpu-operator unchanged
configmap/gpu-operator-node-feature-discovery-master-conf unchanged
configmap/gpu-operator-node-feature-discovery-worker-conf unchanged
daemonset.apps/gpu-operator-node-feature-discovery-worker configured
deployment.apps/gpu-operator-node-feature-discovery-gc configured
deployment.apps/gpu-operator-node-feature-discovery-master configured
deployment.apps/gpu-operator configured
namespace/gpu-operator configured
role.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-worker unchanged
role.rbac.authorization.k8s.io/gpu-operator unchanged
rolebinding.rbac.authorization.k8s.io/gpu-operator-node-feature-discovery-worker unchanged
rolebinding.rbac.authorization.k8s.io/gpu-operator unchanged
serviceaccount/gpu-operator unchanged
serviceaccount/node-feature-discovery unchanged
>> Done. The cluster received the files in ./nvidia-gpu-operator-v26-7-1-default/out. Its CRDs were settled first, from the package.
>> Not run by this script: the pre-upgrade, pre-delete and post-delete steps of this base's plan. The entry page says who runs each one.
[exit 0]
finished 2026-10-10T21:34:45ZAfter it, the four pods were running and the ClusterPolicy reported ready. The second cluster was deleted in the same way as the first.
What the walk did not cover
- No GPU. The ClusterPolicy's
Readycondition carries the reasonNFDLabelsMissing, and no driver, toolkit or device-plugin DaemonSet was created. - The upgrade and delete steps of the plan were not run.
- One run, on one machine, with one node.
- Only the
defaultbase of one version. The other bases of this chart have their own scripts, and they were not run here. - The plain-objects bundle was not used. The script renders the installer package locally.
Clean-up
$ kind delete cluster --name cwtest-1010-gpuop --kubeconfig <work>/kubeconfig
Deleted nodes: ["cwtest-1010-gpuop-control-plane"]No cluster or container with the walk's name remained, and the checksum of the machine's default kubeconfig was the same as before the walk.