Two issues caused TLS to break on photos.carabosse.cloud over IPv6
(GrapheneOS + Immich app via Orange 5G NAT64):
1. Per-service vhosts only listened on IPv4 (listen 443 ssl). On IPv6,
nginx fell back to the first vhost loaded alphabetically and served
its certificate, breaking hostname verification on every other vhost.
2. /etc/letsencrypt/{live,archive} were 0700 root:root after certbot
created them, so the nginx worker (user http on Arch) could not read
the chained intermediates and served the leaf-only chain.
Changes:
- Add catch-all 00-default.conf default_server on :80 and :443 (v4+v6)
with a self-signed cert and 'return 444'. ACME challenges still
answered on :80.
- Add IPv6 listeners ([::]:80 and [::]:443 ssl) to immich, gitea, ntfy,
uptime_kuma vhosts and to the temporary ACME provisioning vhost.
- Apply 0755 on /etc/letsencrypt/live and /etc/letsencrypt/archive on
every run, not only at initial cert provisioning.
Homelab Ansible Playbooks
This repository contains Ansible playbooks and roles I use to manage my NAS and some VMs 👨💻.
This project is designed for personal/familial scale maintenance, if you find this useful for your use, want to share advises or security concerns, feel free to drop me a line.
This is a good playground to learn and I encourage you to adapt these roles to your needs. While they might not be production-ready for all environments, I'm open to adapting them for Ansible Galaxy if there's community interest!
Architecture Overview
Platform Support: Arch Linux, Debian/Ubuntu
Core Design:
- A unique system administrator (
{{ ansible_user }}) - Security hardened sshd
- Shared services pattern: Single PostgreSQL and Valkey (Redis) instances serve all services
- Rootless Podman: Containers run as
{{ ansible_user }}(daemonless,sudo podman psshows nothing) - User systemd services:
systemctl --user status <service>with lingering enabled - Nginx reverse proxy for web services
- IP Freebind when available (e.g. unbound does not wait for wireguard to be up to start resolving DNS)
Available Services:
| Service | Description |
|---|---|
| dns | Unbound caching DNS + Pi-hole ad blocking + VPN resolver |
| nfs | Network file system server |
| zfs | ZFS installation and management |
| uptime-kuma | Uptime monitoring |
| ntfy | Notification server |
| gitea | Git server |
| immich | Photo management |
| static-web | Static website hosting |
| vpn | WireGuard server |
Port Reservation Rules
Reserved ports that must not be used as role defaults:
| Port(s) | Protocol | Reserved for |
|---|---|---|
| 80 | tcp | Nginx |
| 443 | tcp | Nginx |
| 3000-3009 | tcp | Testing |
| 4430 | tcp | Testing |
| 8080 | tcp | Testing |
When adding a new role, pick a default port outside these ranges.
Requirements
Ansible >=2.15
Base tools:
# linux
apt-get install ansible ansible-lint ansible-galaxy
pacman -Syu ansible ansible-lint ansible-galaxy
# macos
brew install ansible ansible-lint ansible-galaxy
# windows
choco install ansible ansible-lint ansible-galaxy
Other roles:
ansible-galaxy collection install -r requirements.yml
Usage
If you have a password on your ssh key --ask-pass is recommended, --ask-become-pass is always asked in these roles, as most tasks require elevated privileges. These are dropped time to time when the default user privilege is enough.
ansible-playbook -i inventory/hosts.yml playbook.yml \
--ask-pass \
--ask-become-pass
You can also call you ssh agent to unlock your key prior to simplify your calls:
ssh-add ~/.ssh/my_key
# unlock it
ansible-playbook -i inventory/hosts.yml playbook.yml \
--ask-become-pass
Bootstrapping a new host
For fresh hosts (only root available, no admin user yet):
ansible-playbook playbooks/bootstrap.yml -l <hostname> --ask-pass
This installs Python and sudo, creates {{ ansible_user }} with sudo rights, and copies your local ~/.ssh/id_ed25519.pub. Supports Arch Linux and Debian/Ubuntu.
To use a different SSH key:
ansible-playbook playbooks/bootstrap.yml -l <hostname> --ask-pass \
--extra-vars 'bootstrap_ssh_public_key="ssh-ed25519 AAAA..."'
Then set a password for the new user (required for sudo --ask-become-pass):
ssh root@<hostname> passwd jambon
After that, run the host playbook normally:
ansible-playbook playbooks/<hostname>.yml --ask-become-pass
Developping
Linting:
ansible-lint
npx prettier --write .
Q&A
Immich crash loop: PostgresError: must be owner of extension vector
Immich tries to self-update the pgvector extension at startup, but its database user is intentionally NOSUPERUSER, so the ALTER EXTENSION vector UPDATE call fails and the microservices worker exits with code 1.
Fix it on the running host by updating the extension as the postgres superuser:
sudo -u postgres psql -d immich -c 'ALTER EXTENSION vector UPDATE;'
The Immich role also runs this automatically on subsequent playbook runs, so re-deployments after a pgvector package upgrade do not require manual intervention.