• SysEng Quick Kubernetes Ansible Repo
  • Basic Information
  • YouTube Series
  • Components
  • Requirements
    • What if I don't want to use devcontainers?
      • Using Docker Directly
      • Installing Dependencies Locally
        • Python Packages
        • Ansible Collections
        • k3sup
        • kubectl
        • helm
  • Instructions
    • Configuring the Cluster
      • Regenerating the Inventory
    • Deploying VMs on Proxmox
      • Adding Ansible Secret Values
      • Configure the Proxmox Role Defaults
        • proxmox_api
        • proxmox_image_delete
        • promox_image_dest
        • proxmox_node
        • proxmox_template
        • proxmox_timeouts
        • proxmox_use_become
      • Configuring the Cloud Images
      • create_cluster playbook
      • remove_cluster playbook
    • Deploying k3s on Your VMs
      • Configure the k3sup Role Defaults
        • k3sup_context
        • k3sup_debug
        • k3sup_extra_args
        • k3sup_iface
        • k3sup_ip_range
        • k3sup_k3s_version
        • k3sup_k3s_server_noschedule
        • k3sup_kube_vip_lease_XXX
        • k3sup_kube_vip_version
        • k3sup_local_path
        • k3sup_prefer_ip
        • k3sup_ssh_key
        • k3sup_use_kube_vip / k3sup_use_kube_vip_cloud_controller
        • k3sup_vip
      • deploy_k3s playbook
    • Deploying Rancher to Your Cluster
      • Configure the rancher Role Defaults
        • rancher_bootstrap_password
        • rancher_cert_manager_version
        • rancher_debug
        • rancher_hostname
        • rancher_lb_service_enable
        • rancher_lb_service_name
        • rancher_replicas
        • rancher_version
      • deploy_rancher playbook
  • Additional Resources

This repo contains ansible playbooks to help you bootstrap a kubernetes cluster. If you have proxmox, it can even build the VMs for the cluster.

I demonstrate the use of this repo in a playlist on my YouTube channel.

Kubernetes with Ansible Playlist

• At least one VM
  • k3s needs at least one node to start
  • At least 3 servers are required for HA
  • At least 2 workers are required for HA if you disable workloads on the servers
  • lxc containers might work, but probably not as these plays exist today
• Devcontainers (and Docker/Podman)
  • Not strictly required, but the requirements are bundled making it easier
• The ability to edit YAML files
  • The defaults will almost certainly need some tweaking

The bundled devcontainer will make it much easier to use these plays. However, I understand this may not be for everyone.

If you have docker, but don't want to use devcontainers, that's an option.

docker pull ghcr.io/sysengquick/k3s:latest
docker run --rm -it -v ".:/workspace" --name sysengquick-k3s ghcr.io/sysengquick/k3s:latest /bin/bash -c 'cd /workspace && sudo -u vscode bash'

You can build the image from the Dockerfile locally if you prefer.

docker build -t ghcr.io/sysengquick/k3s:latest .
docker run --rm -it -v ".:/workspace" --name sysengquick-k3s ghcr.io/sysengquick/k3s:latest /bin/bash -c 'cd /workspace && sudo -u vscode bash'

If you don't want to use the container, you'll need to install the dependencies.

Ansible only runs on Linux and macOS. If you're using Windows, you must use WSL.

The Dockerfile is a pretty good template for what you need. Parts of it are just for convenience.

I use poetry to manage python dependencies, but it's not required.

I've only tested this with python 3.12 and ansible-core 2.16. Older versions might work, but have not been tested.

Once python and pip are installed, you can install the required packages.

python3 -m pip install ansible-core pyyaml

If you are using the proxmox plays, you will need two additional packages.

python3 -m pip install proxmoxer requests

Now that ansible is installed, you need to install the required collections.

ansible-galaxy collection install -r collections/requirements.yml

Ansible looks for collections in specific places. We can add a symlink so ansible can find our sysengquick.k3s collection.

ln -s collections/sysengquick ~/.ansible/collections/ansible_collections/

We are using k3sup to install k3s, so we'll need that.

export K3SUP_VERSION=0.13.5
sudo curl -sLSo /usr/local/bin/k3sup \
    https://github.com/alexellis/k3sup/releases/download/${K3SUP_VERSION}/k3sup
sudo chmod 755 /usr/local/bin/k3sup

We'll need kubectl to interact with our cluster.

export KUBECTL_VERSION=v1.27.12
sudo curl -sLSo /usr/local/bin/kubectl \
    https://dl.k8s.io/release/${KUBECTL_VERSION}/bin/linux/amd64/kubectl
sudo chmod 755 /usr/local/bin/kubectl

We need helm to install cert-manager and rancher.

export HELM_VERSION=v3.14.2
pushd /tmp
curl -sLS https://get.helm.sh/helm-${HELM_VERSION}-linux-amd64.tar.gz | tar xz
sudo mv linux-amd64/helm /usr/local/bin
rm -r linux-amd64
popd

After helm has been installed, you need to add some helm repos.

helm repo add rancher-stable https://releases.rancher.com/server-charts/stable
helm repo add jetstack https://charts.jetstack.io
helm repo update

Make sure you are using the devcontainer or have the requirements installed.

The playbooks need to know how the cluster should be configured. Open the file collections/sysengquick/k3s/playbooks/group_vars/all/cluster.yml.

Update the cluster nodes with the names and IPs of your server and worker nodes. You need at least one server, and three for a true HA setup. Workers are only required if you prevent workloads from be scheduled on the servers. This is a best practice, but is not required.

Update the node_ssh_user with the privileged user account on your nodes. This user must have ssh key authentication configured. It must be root or have full passwordless sudo privileges as root.

If you are using the proxmox playbooks, cloud-init will set everything up for you.

NOTE: The remaining keys are not important if you are not using proxmox.

In proxmox_network, set gateway and cidr as appropriate for your cluster network.

Set proxmox_server to the IP or DNS name for your proxmox API server.

Set proxmox_ssh_user to be a user on your proxmox server that can run qm disk import as root. If you want to allow-list only this single command, it must be added with NOPASSWD. Otherwise, the user must have full sudo privileges as root.

Either one of these entries should be sufficient.

ansible ALL=(root) NOPASSWD: /usr/sbin/qm
ansible ALL=(root) ALL

After updating the cluster configuration, you must rebuild inventory.yaml. There is a helper script to do this.

python3 scripts/update_inventory.py

This repo contains plays for deploying VMs from cloud images to a proxmox cluster.

An API token capable of creating templates and VMs in proxmox is required. I applied the PVEAdmin role to my token on the root resource.

The proxmox playbooks require two secret values: the cloud-init user password and the proxmox API token secret.

The cloud-init user password can be anything you want. It is only used for serial console logins in case SSH is not working.

Open vault.example.yml in collections/sysengquick/k3s/playbooks/vars/proxmox. Copy the contents of this file over the contents of vault.yml. Update the values as appropriate.

In ansible.cfg, there is a vault password file. Write your vault password to a file and update the path as needed. Encrypt the vault.yml file with ansible-vault.

ansible-vault encrypt --vault-pass-file /private/ansible/vault/sysengquick/k3s vault.yml

Replace the path in the command as appropriate.

Several options may need changed in the proxmox role defaults. Open the file main.yml in collections/sysengquick/k3s/roles/proxmox/defaults.

In the proxmox_api dictionary, you may need to change the token_id or user to match your token.

Change this to true if you want ansible to delete the cloud image file after it creates the template.

This is where the cloud image disk file is saved.

This field must match one of the nodes in your proxmox cluster.

NOTE: I don't have a multi-node cluster. If you want to place VMs on different nodes, migrate them after creation.

This dictionary contains base settings the VMs will inherit. You can alter the settings after creation if you want to tweak them individually.

The proxmox_timeouts dictionary is how long (in seconds) to wait for certain actions.

proxmox_use_become should be set true if your user has full sudo privileges. If you just added the qm disk import command without a password, leave it false.

In the proxmox role vars, proxmox_images defines the usable cloud images. There are four images defined currently: alma_linux_9, debian_12, ubuntu_2204, and ubuntu_2204_minimal.

I have done the most testing with debian_12 (aka bookworm), but any of these shold work.

If you want to add another cloud image, you must fill out a new entry in this dictionary.

I was not able to import Oracle Linux 8 or 9 cloud images. Proxmox didn't seem to understand the disk format it was using. I'd love a pull request if you can make this work.

You should be able to concatenate the base property with the name or digest to get a full URL to the file. The base property must end with a trailing slash.

When you are done tweaking your settings, run the cluster_create playbook.

ansible-playbook sysengquick.k3s.create_cluster -K

If you don't need a sudo password, you can press enter when prompted or leave off the -K option.

If you want to tear down your cluster, run the remove_cluster playbook.

ansible-playbook sysengquick.k3s.remove_cluster

Once you have your VMs, you're ready to install k3s.

NOTE: You do not need proxmox for this part. Any VMs capable of running k3s should work.

There are some values that may need updated in the k3sup role defaults. Open main.yml in collections/sysengquick/k3s/roles/k3sup/defaults.

This is the kubeconfig context. It's largely cosmetic.

Set true to get debug output printed about the console commands run by the playbook.

Array of additional arguments to pass to k3s.

The default interface on the VMs.

This is needed for a few things in the playbooks. There is no provision for handling VMs with different default interfaces.

Rnage of IPs that kube-vip cloud controller will assign to service loadbalancers.

For a single IP, make start and end the same. This is untested, but it should work.

NOTE: This value is not used if kube-vip cloud controller is not used.

The version of k3s to install. To install rancher, make sure you pick a compatible version.

Set this to true to prevent scheduling workloads on control planes.

It adds the node taint node-role.kubernetes.io/control-plane:NoSchedule to the cluster.

These properties control how timeouts on the kube-vip virtual IP leadership elections. You probably don't need to change these.

Which version of kube-vip to install.

Where to instll the kubectl configuration file on the local host.

When true, k3sup will connect to the nodes by IP and not hostname.

Path to the SSH key k3sup will use to connect to the nodes.

When true, these components will be installed on your k3s cluster. Any combination should work (both, neither, or only one).

The virtual IP to share with kube-vip.

NOTE: This property is unused if kube-vip is not installed.

Once your values have been set, run the deploy_k3s playbook.

ansible-playbook sysengquick.k3s.deploy_k3s

Once your cluster is up, the final step is to deploy rancher.

Take a look at the rancher role defaults. Open main.yml in collections/sysengquick/k3s/roles/rancher/defaults.

This is the password used to connect to rancher the first time. You should change this after installation, so the bootstrap password isn't really important.

This is the version of cert-manager to install.

Similar to k3sup_debug. Displays command module output after commands run by the playbooks.

This is the default hostname of your rancher UI. You can change this later, but it might be easier to change it here.

Set to true if you want to use a service loadbalancer. This might be useful if you disable the default traefik in k3s.

This is the name of the service loadbalancer if enabled.

This is how many rancher replicas to run. 3 is a good default for an HA deployment.

This is the version of rancher to install. Ensure you pick a version compatible with your selected k3s version.

After you've gone through the defaults, run the deploy_rancher playbook.

ansible-playbook sysengquick.k3s.deploy_rancher

Check out the YouTube playlist in the YouTube Series link at the top.

If you need help or something isn't working, file an issue on the github repo.