Skills
▥NYC
skills/smithery/ai/new-homelab-guest

new-homelab-guest

SKILL.md

Claude Code Instructions for hummel.casa Terraform Infrastructure

This repository manages Proxmox LXC guests using Terraform for the hummel.casa homelab infrastructure.

Primary Workflow: Adding New LXC Guests

Prerequisites (Manual Steps Outside Repository)

  1. Reserve MAC Address: Get next virtual MAC address from MAC address table in Notion doc (format: 0A:BC:DE:00:00:XX)
  2. Reserve IP Address: Get next private IP in homelab CIDR range (10.20.71.XXX) from Notion doc
  3. Create DHCP Reservation: Add MAC/IP pair to Unifi controller
  4. Plan Application: Select docker image/tag and subdomain (e.g., my-app -> my-app.hummel.casa)
  5. Create Ansible Playbook: Add playbook to ansible repository at https://gitea.hummel.casa/hummel.casa/ansible

Repository Structure

  • main.tf: Core infrastructure with locals.guests definitions and shared resources
  • Individual {guest-name}.tf files: Each guest has its own file with random_password, proxmox_lxc, and cloudflare_record resources
  • setup-ansible-pull-cron.sh: Provisioner script that configures new guests

Steps to Add New Guest

1. Add Guest Definition to main.tf

Add new entry to locals.guests in main.tf:

guest_name = {
  mac              = "0A:BC:DE:00:00:XX"
  dhcp_reservation = "10.20.71.XXX"
  domain           = "subdomain.hummel.casa"
  target_node      = "pveX"           # Available: pve, pve2, pve4, pve5, pve6, pve7, pve8, pve9
  cpu_cores        = 1                # Typically 1
  ram_mib          = 512              # Common: 256, 512, 1024, 2048, 4096
  root_fs_size     = "8G"             # Common: 5G, 8G, 12G, 16G, 24G, 25G
  playbook         = "playbook.yml"   # Ansible playbook name
}

2. Create Individual Guest File

Create {guest-name}.tf with three resources:

resource "random_password" "guest_name" {
  length           = 30
  special          = true
  override_special = "_%@"
}

resource "proxmox_lxc" "guest_name" {
  target_node  = local.guests.guest_name.target_node
  start        = true
  onboot       = true
  hostname     = local.guests.guest_name.domain
  ostemplate   = local.debian_12_bookwork_lxc_template
  password     = random_password.guest_name.result
  unprivileged = true

  memory = local.guests.guest_name.ram_mib
  cores  = local.guests.guest_name.cpu_cores

  features {
    nesting = true  # Required for Docker
  }

  rootfs {
    storage = "local-lvm"
    size    = local.guests.guest_name.root_fs_size
  }

  ssh_public_keys = <<-EOT
    ${local.public_keys.lxc_guest_ssh}
    ${local.public_keys.atlantis_root_2}
    ${local.public_keys.gitea_hummel_casa}
  EOT

  network {
    name   = "eth0"
    bridge = "vmbr0"
    ip     = "dhcp"
    hwaddr = local.guests.guest_name.mac
  }

  provisioner "file" {
    source      = "setup-ansible-pull-cron.sh"
    destination = "/tmp/script.sh"
  }

  provisioner "remote-exec" {
    inline = [
      "chmod +x /tmp/script.sh",
      "/tmp/script.sh ${local.guests.guest_name.playbook}"
    ]
  }

  connection {
    type        = "ssh"
    user        = "root"
    private_key = file("/root/.ssh/id_ed25519")
    host        = local.guests.guest_name.dhcp_reservation
  }

  lifecycle {
    ignore_changes = [ostemplate]
  }
}

resource "cloudflare_record" "guest_name" {
  zone_id = data.cloudflare_zone.hummel_casa.zone_id
  name    = element(split(".", local.guests.guest_name.domain), 0)
  content = local.guests.guest_name.dhcp_reservation
  type    = "A"
  ttl     = 1
  proxied = false
  comment = local.dns_comment
}

3. Git Workflow

git checkout -b add-guest-name-guest
git add main.tf guest-name.tf
git commit -m "add guest-name guest"
git push -u origin add-guest-name-guest
# Create PR via Gitea web interface

4. Post-Deployment Monitoring Setup

After the guest is successfully deployed and the application is running:

  1. Access Uptime Kuma: Navigate to the monitoring dashboard
  2. Add HTTP Monitor: Create a new monitor for the guest's public endpoint
    • Monitor Type: HTTP(s)
    • Friendly Name: Guest application name (e.g., "Notes App")
    • URL: Full application URL (e.g., https://notes.hummel.casa)
    • Heartbeat Interval: 60 seconds (recommended)
    • Max Retries: 1
    • Timeout: 48 seconds
  3. Configure Notifications: Ensure the monitor is connected to appropriate notification channels
  4. Test Monitor: Verify the monitor can successfully reach the application

This ensures all public services are properly monitored for uptime and availability.

Current Infrastructure Details

Domain: hummel.casa Terraform Backend: PostgreSQL at pg.hummel.casa:5432/terraform_backend LXC Template: debian-12-standard_12.7-1_amd64.tar.zst Ansible Repository: https://gitea.hummel.casa/hummel.casa/ansible Provisioner: setup-ansible-pull-cron.sh (runs every 60 minutes)

Available Proxmox Nodes

  • pve: 10.20.71.10
  • pve2: 10.20.71.20
  • pve4: 10.20.71.40
  • pve5: 10.20.71.50
  • pve6: 10.20.71.60
  • pve7: 10.20.71.15
  • pve8: 10.20.71.25
  • pve9: 10.20.71.35

SSH Public Keys Included

All guests receive these SSH public keys for access:

  • lxc_guest_ssh: General LXC access key
  • atlantis_root_2: Atlantis/automation access
  • gitea_hummel_casa: Gitea integration key

Provisioner Script Details

The setup-ansible-pull-cron.sh script automatically configures each new LXC guest:

#!/bin/bash

ANSIBLE_PLAYBOOK=$1

apt-get update
apt-get install ansible -y
apt-get install git -y

cat > /etc/systemd/system/ansible-pull.timer <<EOF
[Unit]
Description=Run ansible-pull every few minutes

[Timer]
OnBootSec=5min
OnUnitActiveSec=60min
Unit=ansible-pull.service

[Install]
WantedBy=timers.target
EOF

cat > /etc/systemd/system/ansible-pull.service <<EOF
[Unit]
Description=ansible-pull

[Service]
ExecStart=/usr/bin/ansible-pull -U https://gitea.hummel.casa/hummel.casa/ansible.git -C main -i localhost, $ANSIBLE_PLAYBOOK
EOF

ansible-galaxy install geerlingguy.docker,7.0.2
ansible-galaxy install git+https://github.com/tphummel/ansible-role-caddy-tls-dns.git,main
ansible-galaxy role install geerlingguy.node_exporter,2.1.0

systemctl daemon-reload
systemctl enable ansible-pull.timer
systemctl start ansible-pull.timer
# run ansible-pull once, adhoc, right now
systemctl start ansible-pull.service

# block until ansible-pull is done
while systemctl is-active --quiet ansible-pull.service;
do
  echo "Waiting for adhoc run of ansible-pull to finish...";
  sleep 2;
done

Script Functions:

  1. System Setup: Updates packages, installs Ansible and Git
  2. Timer Configuration: Creates systemd timer running every 60 minutes after initial 5-minute delay
  3. Service Configuration: Sets up ansible-pull to fetch from gitea.hummel.casa/hummel.casa/ansible
  4. Role Installation: Installs common Ansible roles (Docker, Caddy, Node Exporter)
  5. Initial Run: Executes the specified playbook immediately after guest creation

Example Guest Configurations

Typical Web Application Guest

example_app = {
  mac              = "0A:BC:DE:00:00:XX"
  dhcp_reservation = "10.20.71.XXX"
  domain           = "example-app.hummel.casa"
  target_node      = "pve8"
  cpu_cores        = 1
  ram_mib          = 512
  root_fs_size     = "8G"
  playbook         = "example-app.yml"
}

Resource-Heavy Application Guest

heavy_app = {
  mac              = "0A:BC:DE:00:00:XX"
  dhcp_reservation = "10.20.71.XXX"
  domain           = "heavy-app.hummel.casa"
  target_node      = "pve2"
  cpu_cores        = 1
  ram_mib          = 4096
  root_fs_size     = "25G"
  playbook         = "heavy-app.yml"
}

Troubleshooting

Common Issues

  • Guest won't start: Check target_node availability and LXC template existence
  • Ansible fails: Verify playbook exists in ansible repository and syntax is valid
  • DNS not resolving: Confirm Cloudflare record was created and propagated
  • SSH access issues: Verify DHCP reservation matches guest IP assignment

Required Environment Variables

# Proxmox API access
export PM_API_TOKEN_ID="terraform-prov@pve!mytoken"
export PM_API_TOKEN_SECRET="token"

# Cloudflare DNS management
export CLOUDFLARE_API_TOKEN="token"

# PostgreSQL backend access
export PGUSER="user"
export PGPASSWORD="password"

Best Practices

  1. Resource Allocation: Start with minimal resources, scale up as needed
  2. Node Distribution: Spread guests across available Proxmox nodes for load balancing
  3. Naming Convention: Use descriptive, URL-friendly names for domains
  4. Branch Strategy: Always create feature branches for new guests
  5. Documentation: Update ansible playbooks with service-specific configurations
  6. Monitoring: Always add new public endpoints to Uptime Kuma after deployment for availability monitoring
Weekly Installs
1
Repository
smithery/ai
First Seen
11 days ago
Installed on
claude-code1