Homelab

Documentation for Ansible Scripts, Kubernetes Manifest and other files needed to run my homelab.

K3S Cluster

My k3s cluster is currently running a set of raspberry pi model 4bs with different RAM amounts.

Setup the dependencies for a master or worker node:

1. Install the Raspberry Pi 64 bit Lite OS onto an SD Card. Make sure you enable ssh-key only authentication
2. Boot the pi with the created sd card, and ssh into the default user.
3. Modify the /etc/dhcpcd.conf to set a static IP address for the node. Add this IP to the k3snodes group in the ansible hosts file. Current with how you have it setup on your mac, that file is located .ansible/hosts.ini. Example config for the hosts.ini file can be found in the next section.
4. Copy the ssh key the raspberry pi image burner tool adds to ~/.ssh/authorized_keys to the root ssh key directory. Run: sudo mkdir -p /root/.ssh && sudo cp ~/.ssh/authorized_keys /root/.ssh. This so that ansible can run commands with sudo priveledges on the raspberry pi.
5. Reboot the raspberry pi.
6. Run the k3s/playbooks/setupraspberrypi.yml playbook. That will update the os, the cgroup memory config, add config so that the ican USB to SATA convert mounts SSDs as "storage devices", and other k3s specific requirements

Example hosts.ini

[k3smaster]
192.168.4.200 var_disk=sda var_blkid=asd

[k3snodes]
192.168.4.201 var_disk=sda var_blkid=asd


[cube:children]
k3smaster
k3snodes

[k3smaster:vars]
ansible_user=root

[k3snodes:vars]
ansible_user=root

Setup of a Raspberry Pi with External Storage

This is needed for the Longhorn manager. Any node that you want to have persistent storage will need to follow these steps.

1. Verify what path the SSD is attached to by running ansible cube -b -m shell -a "lsblk -f". In most cases this will be /dev/sda.
2. Specify that path for the node's ip in the hosts.ini file as the variable var_disk.
3. Wipe the SSD disk by running ansible <HOST IP OR GROUP> -b -m shell -a "wipefs -a /dev/{{ var_disk }}"
4. Format SSD as an ext4 partition with ansible <HOST IP OR GROUP> -b -m filesystem -a "fstype=ext4 dev=/dev/{{ var_disk }}"
5. Get the block id of the partition with ansible <HOST IP OR GROUP> -b -m shell -a "blkid -s UUID -o value /dev/{{ var_disk }}", add that id to the host.ini file as the variable var_blkid
6. Mount the block with ansible <HOST IP OR GROUP> -m ansible.posix.mount -a "path=/storage01 src=UUID={{ var_blkid }} fstype=ext4 state=mounted" -b

Once the node has come online, The longhorn controller should pick this up and setup the needed containers for that node.

Setup k3s

1. To setup a master node run the k3s/playbookssetupmaster.yml playbook. To create a worker node, run the k3s/playbookssetupnodes.yml. Note that for this playbook you will need to specify the server token from the master node. This can be fetched by running ssh node@<MASTER IP> "sudo cat /var/lib/rancher/k3s/server/node-token"
2. Copy the kubectl config file from the k3s master to your working machine: scp root@<MASTER_IP>:/etc/rancher/k3s/k3s.yaml ~/.kube/config.
3. Modify ~/.kube/config to specify the raspberry pi master ip as the server ip

Adding a Service

All k3s services are defined in k3s/service. The name of the directory containing the manifests for the service needs to be the same name as the samespace for that service. Example structure of a service's directory:

ntfy
├── ns.yml 
├── deployment.yml
├── certificate.yml
├── issuer.yml
├── service.yml
├── ingress.yml

For services that require ssl/tls termination on the ingress, you must create a cloudflare api key secret like kubectl create secret generic -n <NAMESPACE> cloudflare-api-key-secret --from-literal=api-key=<VALUE>. This allows dns-01 challenges to happen within cert-manager. The namespace must be created first, followed by the issuer, certificate, deployment, service and ingress in that order.

Longhorn Notes:

Longhorn was instantiated with:

helm repo add longhorn https://charts.longhorn.io
helm repo update
helm install longhorn longhorn/longhorn --namespace longhorn-system --create-namespace --set defaultSettings.defaultDataPath="/storage01"

The rest of longhorn (service/certs) can be managed with the manifests in the longhorn directory.

Setup The IP Config For the Server

sudo apt-get -y install net-tools network-manager

Edit the /etc/netplan/00-installer-config.yaml file to look like this. Change the IP address to the one you want to use.

network:
  ethernets:
    eno1:
      dhcp4: true
      addresses: [192.168.4.202/24]
      nameservers:
        addresses: [1.1.1.1,8.8.8.8,192.168.4.1]
      routes:
      - to: 255.255.255.0/24
        via: 192.168.4.1
  version: 2
  wifis: {}

Home Assistant

Home assistant is running in a docker container on the k3s cluster. The config for it is in the /config directory which will be created as the container spins up.

To use home assistant with the nginx ingress, you will need to add the cluster IP to the trusted_proxies list in the home assistant config, as well as enable the x_forwarded_for option.

The config you need to add should look like:

http:
  use_x_forwarded_for: true
  trusted_proxies: 10.42.1.3 

To add the config stream sh to stdin/stdout

 kubectl exec --stdin --tty home-assistant-845b7cdc7d-65xkl -n home-assistant -- /bin/sh
 /config # 

Install nano

apk add nano