PD Pi-hole / Unbound / Keepalived Stack

Repo stack path: /home/fizzlepoof/repos/truenas-stacks/pihole
Live PD stack path: /mnt/docker-ssd/docker/compose/pihole
Node role: PD primary / MASTER
VIP target: 10.5.1.53/24

Primary-node stack for the Pi-hole HA plan.

This stack intentionally covers the PD primary plus PD-hosted sync:

NOMAD and RPi4 replicas still need their own deployment steps.

Files

docker-compose.yaml — primary-node stack
.env.example — required variables
unbound.conf — shared recursive resolver config
keepalived.conf.template — rendered into keepalived.conf
bin/render_keepalived_conf.py — generates keepalived.conf from .env
bin/prepare_pd.sh — creates appdata dirs, downloads root hints, renders config, validates compose, optional --up

cd /mnt/docker-ssd/docker/compose/pihole
cp .env.example .env
nano .env   # set real passwords + verify LAN_INTERFACE
/usr/bin/bash bin/prepare_pd.sh

If that validates cleanly, bring it up:

/usr/bin/bash bin/prepare_pd.sh --up

KEEPALIVED_PASSWORD must be 8 chars or fewer because VRRP PASS auth is annoying and old.
Verify LAN_INTERFACE on PD before deployment (ip link show).
FTLCONF_dns_listeningMode=LOCAL is set because TrueNAS binds loopback on port 53.
Pi-hole NTP is explicitly disabled here because PD host chronyd already owns UDP 123.
Nebula Sync runs on PD and uses NEBULA_SYNC_PRIMARY / NEBULA_SYNC_REPLICAS from .env.
Nebula Sync is intentionally set to selective mode with gravity sync toggles enabled for groups, adlists, and domain lists so replica-specific runtime/network settings are not clobbered.
keepalived.conf is generated and should stay local/untracked.