authentik: boot-persistent quadlet conversion with deploy handler

The compose-era authentik trio died at every boot (rootless podman does not
auto-start bare containers). Convert the trio to Quadlet units installed by
deploy-remote.sh into ~/.config/containers/systemd/, enabled + WantedBy=
default.target:

- 3 unit files under dankclaw/authentik/quadlet/ (postgres -> server -> worker
  ordering via Wants=/After=; secrets stay in the git-ignored host .env,
  expanded by systemd at start time exactly like the gitea units)
- deploy_authentik() handler: install units, daemon-reload, retire stale
  compose-era containers only while units are inactive, enable once, then
  start in dependency order with podman-health gates between stages
- compose.yaml: authentik service definitions retired to a comment pointing at
  the quadlet dir (same pattern as searxng/open-webui/gitea/caddy)
- tests/authentik.bats rewritten from compose assertions to unit-file
  contract assertions (images, Exec roles, ordering, loopback-only publish,
  pg_isready probe, secret hygiene, dispatch routing)
This commit is contained in:
Frank
2026-08-28 03:09:19 -07:00
parent 5829a0bfe8
commit 914e251ebc
6 changed files with 211 additions and 120 deletions
@@ -0,0 +1,37 @@
# Quadlet unit: Authentik PostgreSQL 16 (rootless, boot-persistent).
#
# Install path on dankclaw: ~/.config/containers/systemd/authentik-postgres.container
# Unit name: authentik-postgres.service (podman 5.x drops the container- prefix)
#
# Replaces the compose service retired when this stack went boot-persistent
# (rootless Podman does not auto-start bare containers at boot). Secrets stay
# out of this file: systemd expands ${AUTHENTIK_PG_PASS} from the git-ignored
# host .env at start time (same pattern as gitea-postgres).
[Unit]
Description=Authentik PostgreSQL container
After=network.target
[Container]
ContainerName=authentik-postgres
Image=docker.io/library/postgres:16-alpine
Network=dankclaw_default
EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
Environment=POSTGRES_DB=authentik
Environment=POSTGRES_USER=authentik
Environment=POSTGRES_PASSWORD=${AUTHENTIK_PG_PASS}
Volume=/home/dank/authentik/postgres:/var/lib/postgresql/data
HealthCmd=/bin/sh -c "pg_isready -d authentik -U authentik"
HealthInterval=30s
HealthTimeout=10s
HealthRetries=5
HealthStartPeriod=20s
[Service]
EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
TimeoutStartSec=900
Restart=always
RestartSec=5
[Install]
WantedBy=default.target
@@ -0,0 +1,46 @@
# Quadlet unit: Authentik core + embedded Caddy outpost (rootless, boot-persistent).
#
# Install path on dankclaw: ~/.config/containers/systemd/authentik-server.container
# Unit name: authentik-server.service (podman 5.x drops the container- prefix)
#
# The embedded outpost serves /outpost.goauthentik.io/* including the
# /auth/caddy forward_auth endpoint the Caddyfile chains to — no outpost token
# exists for this deployment by design. Resolved by name (A2A DNS) from every
# container on dankclaw_default, including caddy. The 9000 publish is
# loopback-only on purpose: LAN clients reach Authentik through Caddy, never
# around it; the loopback publish exists so host-side tooling can probe
# /-/health/. Runs as the image default user (same userns mapping the compose
# era used, so the existing /data ownership carries over unchanged).
[Unit]
Description=Authentik server container
After=network.target authentik-postgres.service
Wants=authentik-postgres.service
[Container]
ContainerName=authentik-server
Image=ghcr.io/goauthentik/server:2026.8.0
Exec=server
Network=dankclaw_default
EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
Environment=AUTHENTIK_POSTGRESQL__HOST=authentik-postgres
Environment=AUTHENTIK_POSTGRESQL__NAME=authentik
Environment=AUTHENTIK_POSTGRESQL__USER=authentik
Environment=AUTHENTIK_POSTGRESQL__PASSWORD=${AUTHENTIK_PG_PASS}
Environment=AUTHENTIK_SECRET_KEY=${AUTHENTIK_SECRET_KEY}
PublishPort=127.0.0.1:9000:9000
HealthCmd=python -c "import urllib.request;urllib.request.urlopen('http://localhost:9000/-/health/')"
--shm-size=512m
HealthInterval=30s
HealthTimeout=10s
HealthRetries=5
HealthStartPeriod=30s
[Service]
EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
TimeoutStartSec=900
Restart=always
RestartSec=5
[Install]
WantedBy=default.target
@@ -0,0 +1,37 @@
# Quadlet unit: Authentik worker — background tasks, email, webhook/log outbox
# (rootless, boot-persistent).
#
# Install path on dankclaw: ~/.config/containers/systemd/authentik-worker.container
# Unit name: authentik-worker.service (podman 5.x drops the container- prefix)
#
# Same image as the server; the Exec= command selects the worker role. Runs as
# user 0:0 (rootless: maps to the invoking host user, matching the compose-era
# semantics). No published ports — the worker talks to PostgreSQL in-network
# and has no listener of its own.
[Unit]
Description=Authentik worker container
After=network.target authentik-postgres.service authentik-server.service
Wants=authentik-server.service
[Container]
ContainerName=authentik-worker
Image=ghcr.io/goauthentik/server:2026.8.0
Exec=worker
Network=dankclaw_default
EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
Environment=AUTHENTIK_POSTGRESQL__HOST=authentik-postgres
Environment=AUTHENTIK_POSTGRESQL__NAME=authentik
Environment=AUTHENTIK_POSTGRESQL__USER=authentik
Environment=AUTHENTIK_POSTGRESQL__PASSWORD=${AUTHENTIK_PG_PASS}
Environment=AUTHENTIK_SECRET_KEY=${AUTHENTIK_SECRET_KEY}
Volume=/home/dank/authentik/data:/data
User=0:0
[Service]
EnvironmentFile=/home/dank/serverAppConfigs/dankclaw/.env
Restart=always
RestartSec=5
[Install]
WantedBy=default.target
+7 -62
View File
@@ -246,69 +246,14 @@ services:
retries: 3
# --- Authentik (SSO / OIDC provider) -------------------------------------
# 3-service stack: PostgreSQL + server + worker. Redis was removed in
# 2025.10 (sessions/caching moved to PostgreSQL). The Caddyfile's
# forward_auth vhosts proxy to authentik-server:9000 (compose network).
# Now boot-persistent Quadlet units, not compose services: authentik-
# postgres -> authentik-server -> authentik-worker start in dependency order
# via ~/.config/containers/systemd/*.container on the host (repo source:
# dankclaw/authentik/quadlet/; deploy-remote.sh deploy_authentik installs
# and starts them). Redis was removed in 2025.10 (sessions/caching moved to
# PostgreSQL). The Caddyfile's forward_auth vhosts proxy to
# authentik-server:9000 over the dankclaw_default network (A2A DNS).
# See ../docs/Infrastructure/Authentik.md.
authentik-postgres:
image: docker.io/library/postgres:16-alpine
container_name: authentik-postgres
restart: unless-stopped
environment:
POSTGRES_DB: authentik
POSTGRES_USER: authentik
POSTGRES_PASSWORD: ${AUTHENTIK_PG_PASS}
volumes:
- /home/dank/authentik/postgres:/var/lib/postgresql/data
healthcheck:
test: ["CMD-SHELL", "pg_isready -d authentik -U authentik"]
interval: 30s
retries: 5
start_period: 20s
authentik-server:
image: ghcr.io/goauthentik/server:2026.8.0
container_name: authentik-server
restart: unless-stopped
command: server
environment:
AUTHENTIK_POSTGRESQL__HOST: authentik-postgres
AUTHENTIK_POSTGRESQL__NAME: authentik
AUTHENTIK_POSTGRESQL__USER: authentik
AUTHENTIK_POSTGRESQL__PASSWORD: ${AUTHENTIK_PG_PASS}
AUTHENTIK_SECRET_KEY: ${AUTHENTIK_SECRET_KEY}
ports:
- "9000:9000"
volumes:
- /home/dank/authentik/data:/data
shm_size: 512mb
depends_on:
authentik-postgres:
condition: service_healthy
healthcheck:
test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:9000/-/health/')"]
interval: 30s
timeout: 10s
retries: 5
start_period: 30s
authentik-worker:
image: ghcr.io/goauthentik/server:2026.8.0
container_name: authentik-worker
restart: unless-stopped
command: worker
environment:
AUTHENTIK_POSTGRESQL__HOST: authentik-postgres
AUTHENTIK_POSTGRESQL__NAME: authentik
AUTHENTIK_POSTGRESQL__USER: authentik
AUTHENTIK_POSTGRESQL__PASSWORD: ${AUTHENTIK_PG_PASS}
AUTHENTIK_SECRET_KEY: ${AUTHENTIK_SECRET_KEY}
user: "0:0"
volumes:
- /home/dank/authentik/data:/data
depends_on:
authentik-postgres:
condition: service_healthy
# Caddy (the homelab's single reverse proxy) left compose on 2026-08-26:
# it is boot-persistent via its Quadlet unit (dankclaw/caddy/quadlet/
+45 -1
View File
@@ -359,7 +359,7 @@ deploy_compose_service() {
return 0
fi
case "$service" in
comfyui|hermes|hermes-dashboard|hermes-webui|prometheus|grafana|loki|alloy|node-exporter|authentik-postgres|authentik-server|authentik-worker)
comfyui|hermes|hermes-dashboard|hermes-webui|prometheus|grafana|loki|alloy|node-exporter)
case "$service" in
comfyui|loki|alloy)
build_compose_image "$service"
@@ -505,6 +505,47 @@ relink_unit() {
run_privileged ln -sfn "$REMOTE_DIR/scripts/$service.service" "/etc/systemd/system/$service.service"
}
# Authentik runs as three boot-persistent Quadlet units (postgres -> server ->
# worker). The trio deploys as one unit: unit files are installed, stale
# compose-era containers holding the claimed names are retired, then the chain
# starts in dependency order with health gates between stages. Re-running the
# handler on an already-active stack is a deliberate no-op (no DB bounce).
wait_container_healthy() {
local name="$1" limit="$2" status="unknown"
for ((i = 1; i <= limit; i += 5)); do
status="$(podman inspect -f '{{.State.Health.Status}}' "$name" 2>/dev/null || echo missing)"
[[ "$status" == "healthy" ]] && {
echo "$name healthy after approximately ${i}s"
return 0
}
sleep 5
done
echo "FAIL: $name did not become healthy in ${limit}s (status: $status)" >&2
return 1
}
deploy_authentik() {
local unit
mkdir -p "$quadlet_unit_dir"
for unit in authentik-postgres authentik-server authentik-worker; do
cp "dankclaw/authentik/quadlet/$unit.container" "$quadlet_unit_dir/"
done
systemctl --user daemon-reload
# Retire compose-era containers only while the units are not yet active;
# unit-managed containers self-heal via Restart=always.
if ! systemctl --user is-active --quiet authentik-postgres.service; then
podman stop authentik-postgres authentik-server authentik-worker 2>/dev/null || true
podman rm authentik-postgres authentik-server authentik-worker 2>/dev/null || true
fi
systemctl --user enable authentik-postgres.service authentik-server.service authentik-worker.service
systemctl --user start authentik-postgres.service
wait_container_healthy authentik-postgres 120 || return 1
systemctl --user start authentik-server.service
wait_container_healthy authentik-server 300 || return 1
systemctl --user start authentik-worker.service
}
deploy_llm_large() {
relink_unit frank-default-large
# NPU is intentionally not restarted here. It is a separate, retired
@@ -555,6 +596,9 @@ deploy_service() {
pihole) deploy_pihole ;;
gitea-postgres) deploy_gitea_unit gitea-postgres gitea-postgres ;;
gitea-server) deploy_gitea_unit gitea-server gitea-server ;;
authentik-postgres|authentik-server|authentik-worker)
deploy_authentik
;;
frank-default-large) deploy_llm_large ;;
frank-default-small) deploy_llm_small ;;
frank-default-npu) deploy_llm_npu ;;
+39 -57
View File
@@ -19,86 +19,68 @@ COMPOSE="$REPO_ROOT/dankclaw/compose.yaml"
CADDYFILE="$REPO_ROOT/dankclaw/Caddyfile"
ENV_EXAMPLE="$REPO_ROOT/dankclaw/.env.example"
# --- Compose: the 3-service stack ---
# --- Quadlet units: the 3-service stack ------------------------------------
# The trio left compose: they are boot-persistent Quadlet units installed to
# ~/.config/containers/systemd/ by deploy-remote.sh (deploy_authentik).
@test "compose defines authentik-postgres" {
grep -qE '^ authentik-postgres:' "$COMPOSE"
}
UNIT_DIR="$REPO_ROOT/dankclaw/authentik/quadlet"
@test "compose defines authentik-server" {
grep -qE '^ authentik-server:' "$COMPOSE"
}
@test "compose defines authentik-worker" {
grep -qE '^ authentik-worker:' "$COMPOSE"
@test "the three authentik quadlet units exist" {
for unit in authentik-postgres authentik-server authentik-worker; do
[ -f "$UNIT_DIR/$unit.container" ]
done
}
@test "authentik-postgres uses the official postgres image" {
grep -q 'image: docker.io/library/postgres:' "$COMPOSE"
grep -q 'Image=docker.io/library/postgres:16-alpine' "$UNIT_DIR/authentik-postgres.container"
}
@test "authentik-server uses the official goauthentik image" {
grep -q 'image: ghcr.io/goauthentik/server:' "$COMPOSE"
@test "server and worker use the official goauthentik image" {
grep -q 'Image=ghcr.io/goauthentik/server:' "$UNIT_DIR/authentik-server.container"
grep -q 'Image=ghcr.io/goauthentik/server:' "$UNIT_DIR/authentik-worker.container"
}
@test "authentik-server runs the 'server' command" {
# The same image runs both server and worker; the command selects the role.
grep -qE '^\s+command:\s*server' "$COMPOSE"
@test "the same image runs both roles; the Exec command selects the role" {
grep -qE '^Exec=server$' "$UNIT_DIR/authentik-server.container"
grep -qE '^Exec=worker$' "$UNIT_DIR/authentik-worker.container"
}
@test "authentik-worker runs the 'worker' command" {
grep -qE '^\s+command:\s*worker' "$COMPOSE"
@test "startup order is postgres -> server -> worker via unit ordering" {
grep -q 'After=network.target authentik-postgres.service' "$UNIT_DIR/authentik-server.container"
grep -q 'Wants=authentik-postgres.service' "$UNIT_DIR/authentik-server.container"
grep -q 'Wants=authentik-server.service' "$UNIT_DIR/authentik-worker.container"
}
@test "authentik-server depends on a healthy authentik-postgres" {
local block
block=$(awk -v s="^ authentik-server:\$" '
$0 ~ s { inblk = 1; next }
inblk && /^ [a-z]/ { inblk = 0 }
inblk' "$COMPOSE")
echo "$block" | grep -q 'condition: service_healthy'
}
@test "authentik-worker depends on a healthy authentik-postgres" {
local block
block=$(awk -v s="^ authentik-worker:\$" '
$0 ~ s { inblk = 1; next }
inblk && /^ [a-z]/ { inblk = 0 }
inblk' "$COMPOSE")
echo "$block" | grep -q 'condition: service_healthy'
}
@test "authentik-server exposes port 9000" {
local block
block=$(awk -v s="^ authentik-server:\$" '
$0 ~ s { inblk = 1; next }
inblk && /^ [a-z]/ { inblk = 0 }
inblk' "$COMPOSE")
echo "$block" | grep -qE '"9000:9000"'
@test "the server publishes 9000 on loopback only; LAN reaches it through Caddy" {
grep -q 'PublishPort=127.0.0.1:9000:9000' "$UNIT_DIR/authentik-server.container"
}
@test "authentik-postgres healthcheck uses pg_isready" {
local block
block=$(awk -v s="^ authentik-postgres:\$" '
$0 ~ s { inblk = 1; next }
inblk && /^ [a-z]/ { inblk = 0 }
inblk' "$COMPOSE")
echo "$block" | grep -q 'pg_isready'
grep -q 'HealthCmd=.*pg_isready' "$UNIT_DIR/authentik-postgres.container"
}
# --- Compose: environment (secrets from .env) ---
# --- Units: environment (secrets stay in the host .env, expanded at start) --
@test "authentik services consume AUTHENTIK_PG_PASS from .env" {
grep -q 'AUTHENTIK_PG_PASS' "$COMPOSE"
@test "the units consume AUTHENTIK_PG_PASS via systemd expansion, not literals" {
grep -q 'POSTGRES_PASSWORD=${AUTHENTIK_PG_PASS}' "$UNIT_DIR/authentik-postgres.container"
grep -q 'PASSWORD=${AUTHENTIK_PG_PASS}' "$UNIT_DIR/authentik-server.container"
grep -q 'PASSWORD=${AUTHENTIK_PG_PASS}' "$UNIT_DIR/authentik-worker.container"
}
@test "authentik services consume AUTHENTIK_SECRET_KEY from .env" {
grep -q 'AUTHENTIK_SECRET_KEY' "$COMPOSE"
@test "the units consume AUTHENTIK_SECRET_KEY via systemd expansion, not literals" {
for unit in authentik-server authentik-worker; do
grep -q 'AUTHENTIK_SECRET_KEY=${AUTHENTIK_SECRET_KEY}' "$UNIT_DIR/$unit.container"
done
}
@test "no real authentik secret is committed in compose" {
# The compose must reference the secret via ${VAR}, never a literal.
! grep -qE 'AUTHENTIK_SECRET_KEY:\s*["\x27]?\w{16,}' "$COMPOSE"
@test "no secret material is committed in the quadlet units" {
! grep -rqE 'SECRET_KEY=[^$]' "$UNIT_DIR"
}
@test "deploy-remote.sh routes the authentik trio to the quadlet handler, not compose up" {
grep -q 'authentik-postgres|authentik-server|authentik-worker)' "$REPO_ROOT/scripts/deploy-remote.sh"
grep -q 'deploy_authentik()' "$REPO_ROOT/scripts/deploy-remote.sh"
! grep -qE '^[[:space:]]+comfyui\|.*authentik' "$REPO_ROOT/scripts/deploy-remote.sh"
}
# --- Caddyfile: the auth vhost ---