pve-automation

Fancy homelab automation bits.

Roadmap

• Create a script that will create a barebones Ubuntu Server base inside of PVE to base these builds on
• Plain Docker base image
• K3s base image
• Buildkite base image
• Packer:
  • Automated rebuilds? Github Actions? Buildkite?
• K3s:
  • Automated deployments? Flux? ArgoCD?
• Nomad:
  • Automated deployments???

Rough Instructions

1. Use scripts/create-ubuntu-template.sh to create a barebones Ubuntu Server template.
   • Previously used: ./create-ubuntu-template.sh -i 9001 -n pve-002 -r jammy -s current -m amd64 -t local -T local-lvm -F
2. Follow the packer README to create the base images.
3. Follow the terraform README to create the cluster base.
4. Follow the ansible README to configure the cluster.
5. Follow the fluxcd README to set up gitops.

Notes (k3s)

• Cluster ingress is handled by Traefik, which is a deployment built-in to k3s.
• There is a cluster "gateway" of sorts. The support node runs the following services:
  • postgres (as k3s leader datastore)
  • pgadmin (for administrating said postgres)
  • nginx
    • TCP proxy load balancer to each cluster node's traefik instance
    • Reverse proxying pgadmin
    • I have two internal DNS records that point to this node:
      • gateway.main.k8s.2811rrt.net - split horizon between Tailnet and my home network
      • *.main.k8s.2811rrt.net - CNAME to gateway.main.k8s.2811rrt.net
      • Note to self: if you rebuild the cluster, you need to update the Tailnet leg of the split horizon record, as the rebuild changes the instance's Tailnet IP address.

Notes (nomad)

• Cluster ingress is handled by Traefik
• Jobs are currently deployed via Terraform module, but is unreliable when a full redeploy is needed.
  • A series of targeted applies are needed to get all jobs deployed successfully, but I need to run that down later.