Skip to content


Upgrade compute nodes

Upgrade GPU driver

  1. Verify that the target driver is available at If not, obtain the driver file from, and transfer it to the pod repo in the kernel namespace, in directory /repo/nvidia. Make sure the name of the driver file is in the format of NVIDIA-Linux-<architecture>-<driver-version>.run. For example:
  2. Pull the latest ansible playbook repo from, update the nvidia_version variable. The nvidia_version variable can be under a single host in nautilus-hosts.yaml, or in the group variable files under group_vars directory, depending on which nodes will get GPU driver updated.
  3. Run command ansible-playbook upgrade.yml -l <hode> -t gpu -e netbox_token=<your-netbox-token>. With the extra variable netbox_token, a record regarding node rebooting due to GPU upgrade will be generated in, but it's optional.

Upgrade all packages

Pull the latest ansible playbook repo and run ansible-playbook upgrade.yml -l <hode> -t os -e netbox_token=<your-netbox-token>.

Upgrade kubernetes

Pull the latest ansible playbook repo, and edit the file group_vars/all with the matching versions of kubernetes, kubernetes_repo and crio.

Here's an example of the configs:

kubernetes_version: "1.26.11"
kubernetes_repo_version: "v1.26"
crio_version: "1.26"

The value of kubernetes_repo should include the patch version and match the version of kubernetes installed on the master node, for example, 1.26.11. The value of crio_version only applies to the nodes that are runnning crio runtime, instead of containerd.

After updating the versions, run ansible-playbook upgrade.yml -l <hode> -t kubernetes.