initial commit: Dawarich LXC role (CT 459 on pve02, .159)

Self-hosted location history. 4-container compose: Rails 8 app + Sidekiq
+ PostGIS 16-3.4 + Redis 7, plus watchtower. Authentik OIDC end-to-end.
Image pinned at freikin/dawarich:1.7.11 (OIDC support requires >= 1.7.8).

PostGIS DB lives in this LXC, not on the central DB VM (.172) — central
image is postgres:16-alpine without postgis, swapping it carries broader
blast radius than colocating here. Convention exception captured in
homelab-docs project_dawarich memory.

Roles:
  - dawarich: system + Docker + compose + weekly prune timer
  - alloy:    logs+journald → Loki, node metrics → Prometheus

Bring-up sequence proven 2026-06-01. README documents the 5-trap build
chain (image version, entrypoint scripts, solid_cache SQLite bind mount,
APPLICATION_HOSTS+localhost, force_ssl+healthcheck).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Your Name
2026-06-01 21:24:09 -04:00
commit e7b8d4df17
20 changed files with 1126 additions and 0 deletions
+5
View File
@@ -0,0 +1,5 @@
---
- name: restart alloy
command: docker compose -f /opt/alloy/docker-compose.yml up -d --force-recreate
args:
chdir: /opt/alloy
+37
View File
@@ -0,0 +1,37 @@
---
- name: Create Alloy directories
file:
path: "{{ item }}"
state: directory
owner: cbalders
group: cbalders
mode: '0755'
loop:
- /opt/alloy
- /var/lib/alloy
- name: Deploy Alloy compose
template:
src: docker-compose.yml.j2
dest: /opt/alloy/docker-compose.yml
owner: cbalders
group: cbalders
mode: '0644'
notify: restart alloy
- name: Deploy Alloy config
template:
src: config.alloy.j2
dest: /opt/alloy/config.alloy
owner: cbalders
group: cbalders
mode: '0644'
notify: restart alloy
# Bring Alloy up (idempotent — docker compose up -d is a no-op if running
# and config hasn't changed). The handler force-recreates on config edit.
- name: Ensure Alloy is running
command: docker compose -f /opt/alloy/docker-compose.yml up -d
args:
chdir: /opt/alloy
changed_when: false
+161
View File
@@ -0,0 +1,161 @@
// Alloy — Docker container logs + journald → Loki. River syntax.
// Per-LXC config; only the `host:` label changes via alloy_host_label var.
// ============================================================
// Docker — discover + scrape container logs
// ============================================================
discovery.docker "containers" {
host = "unix:///var/run/docker.sock"
refresh_interval = "15s"
}
discovery.relabel "containers" {
targets = discovery.docker.containers.targets
rule {
source_labels = ["__meta_docker_container_name"]
regex = "/(.+)"
target_label = "container"
}
rule {
source_labels = ["__meta_docker_container_log_stream"]
target_label = "stream"
}
rule {
source_labels = ["__meta_docker_container_label_com_docker_compose_service"]
target_label = "service"
}
rule {
source_labels = ["__meta_docker_container_label_com_docker_compose_project"]
target_label = "compose_project"
}
rule {
target_label = "job"
replacement = "docker"
}
rule {
target_label = "host"
replacement = "{{ alloy_host_label }}"
}
}
loki.source.docker "containers" {
host = "unix:///var/run/docker.sock"
targets = discovery.relabel.containers.output
forward_to = [loki.process.docker.receiver]
}
loki.process "docker" {
// Replay protection on first boot — Loki rejects > 7d.
stage.drop {
older_than = "24h"
drop_counter_reason = "older_than_24h"
}
// pulse-agent broken-pipe noise on hosts running pulse-agent --enable-host.
stage.drop {
expression = ".*broken pipe.*"
drop_counter_reason = "broken_pipe_noise"
}
forward_to = [loki.write.default.receiver]
}
// ============================================================
// Journald — LXC systemd units
// ============================================================
loki.source.journal "host" {
path = "/var/log/journal"
max_age = "1m"
forward_to = [loki.process.journal.receiver]
relabel_rules = loki.relabel.journal.rules
labels = {
job = "journald",
host = "{{ alloy_host_label }}",
}
}
loki.relabel "journal" {
forward_to = []
rule {
source_labels = ["__journal__systemd_unit"]
target_label = "unit"
}
rule {
source_labels = ["__journal__hostname"]
target_label = "instance"
}
rule {
source_labels = ["__journal_priority_keyword"]
target_label = "severity"
}
}
loki.process "journal" {
forward_to = [loki.write.default.receiver]
}
// ============================================================
// Loki — push
// ============================================================
loki.write "default" {
endpoint {
url = "{{ alloy_loki_url }}"
}
}
// ============================================================
// Prometheus — embedded node_exporter (replaces standalone)
// ============================================================
prometheus.exporter.unix "node" {
// Container view of host filesystems — bind-mounted in compose.
rootfs_path = "/host/rootfs"
procfs_path = "/host/proc"
sysfs_path = "/host/sys"
// Default collectors match upstream node_exporter set.
}
prometheus.scrape "node" {
targets = prometheus.exporter.unix.node.targets
forward_to = [prometheus.relabel.node.receiver]
scrape_interval = "15s"
job_name = "{{ alloy_prom_job | default('node_lxc') }}"
}
// Rewrite the instance + job labels to match the historical
// node_exporter scrape, so existing dashboards continue to slice by
// the same instance string. Alloy's exporter component injects
// job="integrations/unix" on its targets — we have to overwrite it
// here; `job_name` in prometheus.scrape only names the scrape pool,
// it doesn't relabel the metrics.
prometheus.relabel "node" {
forward_to = [prometheus.remote_write.observe.receiver]
rule {
target_label = "instance"
replacement = "{{ alloy_prom_instance | default(ansible_default_ipv4.address + ':9100') }}"
}
rule {
target_label = "job"
replacement = "{{ alloy_prom_job | default('node_lxc') }}"
}
// Match the static labels Prom previously injected at scrape time
// (see homelab-ansible-vm-observability/roles/configs/files/prometheus.yml).
rule {
target_label = "group"
replacement = "{{ alloy_prom_group | default('lxc') }}"
}
rule {
target_label = "hostname"
replacement = "{{ alloy_prom_hostname | default(alloy_host_label) }}"
}
}
prometheus.remote_write "observe" {
endpoint {
url = "{{ alloy_prom_remote_write | default('http://observe.lan.balders.ca:9090/api/v1/write') }}"
}
}
@@ -0,0 +1,33 @@
---
# Self-contained Alloy stack on /opt/alloy. Runs independently from the
# host's main service (kestra/infisical/mcp/etc) so a service-side compose
# down doesn't take logging with it.
services:
alloy:
image: grafana/alloy:latest
container_name: alloy
restart: unless-stopped
command:
- run
- /etc/alloy/config.alloy
- --storage.path=/var/lib/alloy/data
- --server.http.listen-addr=0.0.0.0:12345
ports:
- "12345:12345"
# Share the host's PID namespace so prometheus.exporter.unix reads
# /proc with the host kernel's cgroup view (cgroup-aware MemAvailable).
# Without this, /proc/meminfo returns hybrid values: MemTotal from the
# host cgroup but Cached/SReclaimable from the container, leading to
# a ~25% MemAvailable inflation. See docs/audit/alloy-consolidation-2026-05-21.md.
pid: host
volumes:
- /opt/alloy/config.alloy:/etc/alloy/config.alloy:ro
- /var/lib/alloy:/var/lib/alloy
- /var/log/journal:/var/log/journal:ro
- /run/log/journal:/run/log/journal:ro
- /etc/machine-id:/etc/machine-id:ro
- /var/run/docker.sock:/var/run/docker.sock:ro
# Host metric collection for prometheus.exporter.unix (node_exporter replacement)
- /:/host/rootfs:ro,rslave
- /proc:/host/proc:ro
- /sys:/host/sys:ro
@@ -0,0 +1,8 @@
[Unit]
Description=Prune dangling docker images (Watchtower leaves them behind)
After=docker.service
Requires=docker.service
[Service]
Type=oneshot
ExecStart=/usr/bin/docker image prune -af
+9
View File
@@ -0,0 +1,9 @@
[Unit]
Description=Weekly docker image prune
[Timer]
OnCalendar=Sun 03:30
Persistent=true
[Install]
WantedBy=timers.target
+4
View File
@@ -0,0 +1,4 @@
---
- name: reload systemd
systemd:
daemon_reload: true
+207
View File
@@ -0,0 +1,207 @@
---
# ---------- System setup ----------
- name: Set timezone
copy:
content: "{{ timezone }}"
dest: /etc/timezone
mode: '0644'
register: tz_file
- name: Apply timezone
command: dpkg-reconfigure -f noninteractive tzdata
when: tz_file.changed
- name: Set hostname
copy:
content: "dawarich"
dest: /etc/hostname
mode: '0644'
- name: Apply hostname
command: hostname dawarich
changed_when: false
- name: Ensure hostname in /etc/hosts
lineinfile:
path: /etc/hosts
regexp: '^127\.0\.1\.1'
line: "127.0.1.1 dawarich"
- name: Update apt cache
apt:
update_cache: true
cache_valid_time: 3600
register: apt_cache_result
failed_when: false
changed_when: apt_cache_result.changed | default(false)
- name: Install base packages
apt:
name: "{{ packages }}"
state: present
- name: Create user
user:
name: "{{ item.name }}"
groups: "{{ item.groups }}"
shell: "{{ item.shell }}"
append: true
loop: "{{ users }}"
- name: Deploy SSH authorized keys
authorized_key:
user: cbalders
key: "{{ item }}"
state: present
loop: "{{ ssh_authorized_keys }}"
# ---------- Docker ----------
- name: Check if Docker is installed
command: docker --version
register: docker_check
changed_when: false
failed_when: false
- name: Install Docker
when: docker_check.rc != 0
block:
- name: Download Docker install script
get_url:
url: https://get.docker.com
dest: /tmp/get-docker.sh
mode: '0755'
- name: Run Docker install script
command: /tmp/get-docker.sh
- name: Remove install script
file:
path: /tmp/get-docker.sh
state: absent
- name: Ensure Docker service is running
systemd:
name: docker
state: started
enabled: true
- name: Add cbalders to docker group
user:
name: cbalders
groups: docker
append: true
# ---------- Docker registry mirror ----------
- name: Ensure /etc/docker exists
file:
path: /etc/docker
state: directory
mode: '0755'
- name: Read existing daemon.json
slurp:
src: /etc/docker/daemon.json
register: daemon_json_raw
failed_when: false
- name: Configure Docker registry mirror
copy:
content: "{{ ((daemon_json_raw.content | b64decode | from_json) if daemon_json_raw.content is defined else {}) | combine({'registry-mirrors': ['http://registry.lan.balders.ca:5000']}) | to_nice_json }}"
dest: /etc/docker/daemon.json
owner: root
group: root
mode: '0644'
register: docker_daemon_config
- name: Restart Docker if mirror config changed
systemd:
name: docker
state: restarted
when: docker_daemon_config.changed
# ---------- Dawarich stack ----------
- name: Create app directories
file:
path: "{{ item.path }}"
state: directory
owner: "{{ item.owner | default('cbalders') }}"
group: "{{ item.group | default('cbalders') }}"
mode: '0755'
loop:
- { path: /opt/dawarich }
# Postgis container runs as uid 70 (alpine 'postgres' user). Pre-chown
# so the first boot doesn't fail with "directory is not empty / wrong
# permissions" the way Nextcloud did (feedback_nextcloud_db_perms).
- { path: /opt/dawarich/postgres, owner: '70', group: '70' }
# Redis runs as uid 999 on bind volumes (feedback_redis_uid_999).
- { path: /opt/dawarich/redis, owner: '999', group: '999' }
# Dawarich Rails app writes uploads/cache here as uid 1000.
- { path: /opt/dawarich/public }
- { path: /opt/dawarich/watched }
- { path: /opt/dawarich/storage }
# solid_cache + solid_queue + solid_cable SQLite files (writable by the
# rails uid inside the container — image runs as uid 1000).
- { path: /opt/dawarich/db_data }
- name: Deploy docker-compose
template:
src: docker-compose.yml.j2
dest: /opt/dawarich/docker-compose.yml
owner: cbalders
group: cbalders
mode: '0640'
register: compose_changed
- name: Pull images
command: docker compose pull
args:
chdir: /opt/dawarich
changed_when: false
- name: Recreate containers if config changed
command: docker compose up -d --force-recreate
args:
chdir: /opt/dawarich
when: compose_changed.changed
changed_when: true
- name: Ensure containers are running
command: docker compose up -d
args:
chdir: /opt/dawarich
when: not compose_changed.changed
changed_when: false
# Dawarich runs migrations on boot (entrypoint includes db:migrate). First
# boot can take ~60-90s because the initial schema + PostGIS extension load
# is heavy. Don't fail the play if it's not ready yet — Watchtower-driven
# recreates will converge.
- name: Wait for Dawarich to be healthy
uri:
url: "http://localhost:{{ dawarich_port }}/api/v1/health"
method: GET
register: dawarich_health
until: dawarich_health.status is defined and dawarich_health.status == 200
retries: 30
delay: 5
ignore_errors: true
# ---------- Weekly docker prune (Watchtower side-effect cleanup) ----------
- name: Deploy docker-prune systemd units
copy:
src: "{{ item }}"
dest: "/etc/systemd/system/{{ item }}"
owner: root
group: root
mode: '0644'
loop:
- docker-prune.service
- docker-prune.timer
notify: reload systemd
- name: Enable docker-prune timer
systemd:
name: docker-prune.timer
enabled: true
state: started
daemon_reload: true
@@ -0,0 +1,187 @@
# {{ ansible_managed }}
# Dawarich — self-hosted location history. Four containers + watchtower:
# dawarich_db postgis/postgis:16-3.4-alpine (PostGIS-enabled Postgres)
# dawarich_redis redis:7-alpine (cache + Sidekiq queue)
# dawarich_app freikin/dawarich (Rails web tier, port 3000)
# dawarich_sidekiq freikin/dawarich (background jobs)
# DB lives here (not the central DB VM) because the central image is plain
# postgres:16-alpine without the postgis extension. See memory project_dawarich.
services:
dawarich_db:
image: "{{ postgis_image }}"
container_name: dawarich_db
restart: unless-stopped
shm_size: 1g
environment:
POSTGRES_USER: "{{ dawarich_db_user }}"
POSTGRES_PASSWORD: "{{ dawarich_db_password }}"
POSTGRES_DB: "{{ dawarich_db_name }}"
volumes:
- /opt/dawarich/postgres:/var/lib/postgresql/data
# No published ports — only the app + sidekiq reach it over the compose net.
healthcheck:
test: ["CMD-SHELL", "pg_isready -U {{ dawarich_db_user }} -d {{ dawarich_db_name }}"]
interval: 15s
timeout: 5s
retries: 10
start_period: 30s
dawarich_redis:
image: "{{ redis_image }}"
container_name: dawarich_redis
restart: unless-stopped
volumes:
- /opt/dawarich/redis:/data
# No published ports — internal-only.
healthcheck:
test: ["CMD", "redis-cli", "ping"]
interval: 15s
timeout: 3s
retries: 5
dawarich_app:
image: "{{ dawarich_image }}"
container_name: dawarich_app
restart: unless-stopped
depends_on:
dawarich_db:
condition: service_healthy
dawarich_redis:
condition: service_healthy
ports:
- "{{ dawarich_port }}:3000"
volumes:
- /opt/dawarich/public:/var/app/public
- /opt/dawarich/watched:/var/app/tmp/imports/watched
- /opt/dawarich/storage:/var/app/storage
# Solid_cache + solid_queue + solid_cable SQLite files. Without this
# bind, the queue DB is ephemeral and in-flight Sidekiq jobs are lost
# on container recreate. Path matches the entrypoint's expectation.
- /opt/dawarich/db_data:/dawarich_db_data
# The image ships a `web-entrypoint.sh` that runs `db:prepare` (creates
# the solid_cache + solid_queue SQLite files in /var/app/storage, runs
# Postgres migrations) before exec'ing the command. We MUST use it —
# without it, Rails 8 boots fail with "No database file specified" on
# solid_cache's first lookup. The image default entrypoint is `bundle exec`
# alone, which skips the prep.
entrypoint: ["/usr/local/bin/web-entrypoint.sh"]
command: ["bin/rails", "server", "-p", "3000", "-b", "::"]
environment:
RAILS_ENV: production
SELF_HOSTED: "true"
# localhost + ::1 needed for Docker healthcheck (Rails 8 host
# authorization rejects any host not in this list, including loopback).
APPLICATION_HOSTS: "{{ dawarich_domain }},localhost,127.0.0.1,::1"
APPLICATION_PROTOCOL: "https"
DOMAIN: "{{ dawarich_domain }}"
TIME_ZONE: "{{ timezone }}"
DISTANCE_UNIT: "km"
# Database
DATABASE_HOST: dawarich_db
DATABASE_PORT: "5432"
DATABASE_USERNAME: "{{ dawarich_db_user }}"
DATABASE_PASSWORD: "{{ dawarich_db_password }}"
DATABASE_NAME: "{{ dawarich_db_name }}"
RAILS_MAX_THREADS: "5"
# Redis
REDIS_URL: "redis://dawarich_redis:6379"
# Secrets
SECRET_KEY_BASE: "{{ dawarich_secret_key_base }}"
OTP_ENCRYPTION_PRIMARY_KEY: "{{ dawarich_otp_primary_key }}"
OTP_ENCRYPTION_DETERMINISTIC_KEY: "{{ dawarich_otp_deterministic_key }}"
OTP_ENCRYPTION_KEY_DERIVATION_SALT: "{{ dawarich_otp_salt }}"
# Registration policy — OIDC-only when enabled. Anything else and a
# local user form silently appears on /users/sign_up.
ALLOW_EMAIL_PASSWORD_REGISTRATION: "{{ dawarich_allow_email_password_registration | string | lower }}"
{% if dawarich_oidc_enabled %}
# ---- Auth: Authentik OIDC ----
OIDC_CLIENT_ID: "{{ dawarich_oidc_client_id }}"
OIDC_CLIENT_SECRET: "{{ dawarich_oidc_client_secret }}"
OIDC_ISSUER: "{{ dawarich_oidc_issuer }}"
OIDC_REDIRECT_URI: "{{ dawarich_oidc_redirect_uri }}"
OIDC_PROVIDER_NAME: "{{ dawarich_oidc_provider_name }}"
OIDC_AUTO_REGISTER: "{{ dawarich_oidc_auto_register | string | lower }}"
OIDC_PKCE_ENABLED: "true"
{% endif %}
# TCP socket check, not HTTP. Rails 8 force_ssl (APPLICATION_PROTOCOL=https)
# redirects every plain-HTTP path to https — curl/wget can't satisfy a 2xx
# without TLS, and Puma is plain-HTTP (Caddy terminates TLS upstream). The
# socket check confirms Puma is listening, which is what we actually care
# about for orchestration.
healthcheck:
test: ["CMD-SHELL", "ruby -rsocket -e 'TCPSocket.new(\"localhost\",3000).close' || exit 1"]
interval: 30s
timeout: 10s
retries: 5
start_period: 120s
dawarich_sidekiq:
image: "{{ dawarich_image }}"
container_name: dawarich_sidekiq
restart: unless-stopped
depends_on:
dawarich_db:
condition: service_healthy
dawarich_redis:
condition: service_healthy
dawarich_app:
condition: service_started
volumes:
- /opt/dawarich/public:/var/app/public
- /opt/dawarich/watched:/var/app/tmp/imports/watched
- /opt/dawarich/storage:/var/app/storage
# Solid_cache + solid_queue + solid_cable SQLite files. Without this
# bind, the queue DB is ephemeral and in-flight Sidekiq jobs are lost
# on container recreate. Path matches the entrypoint's expectation.
- /opt/dawarich/db_data:/dawarich_db_data
# Same entrypoint pattern as the web tier — sidekiq's script ensures the
# DB is migrated before workers start consuming jobs.
entrypoint: ["/usr/local/bin/sidekiq-entrypoint.sh"]
command: ["sidekiq"]
environment:
RAILS_ENV: production
SELF_HOSTED: "true"
# localhost + ::1 needed for Docker healthcheck (Rails 8 host
# authorization rejects any host not in this list, including loopback).
APPLICATION_HOSTS: "{{ dawarich_domain }},localhost,127.0.0.1,::1"
APPLICATION_PROTOCOL: "https"
DOMAIN: "{{ dawarich_domain }}"
TIME_ZONE: "{{ timezone }}"
DATABASE_HOST: dawarich_db
DATABASE_PORT: "5432"
DATABASE_USERNAME: "{{ dawarich_db_user }}"
DATABASE_PASSWORD: "{{ dawarich_db_password }}"
DATABASE_NAME: "{{ dawarich_db_name }}"
REDIS_URL: "redis://dawarich_redis:6379"
SECRET_KEY_BASE: "{{ dawarich_secret_key_base }}"
OTP_ENCRYPTION_PRIMARY_KEY: "{{ dawarich_otp_primary_key }}"
OTP_ENCRYPTION_DETERMINISTIC_KEY: "{{ dawarich_otp_deterministic_key }}"
OTP_ENCRYPTION_KEY_DERIVATION_SALT: "{{ dawarich_otp_salt }}"
BACKGROUND_PROCESSING_CONCURRENCY: "{{ dawarich_sidekiq_concurrency }}"
watchtower:
container_name: dawarich_watchtower
image: containrrr/watchtower:latest
restart: unless-stopped
dns:
- 192.168.1.1
ports:
- "8088:8080"
volumes:
- /var/run/docker.sock:/var/run/docker.sock
environment:
- DOCKER_API_VERSION=1.40
- WATCHTOWER_HTTP_API_UPDATE=true
- WATCHTOWER_HTTP_API_TOKEN={{ watchtower_api_token }}
- TZ={{ timezone }}
- WATCHTOWER_CLEANUP=true
- WATCHTOWER_INCLUDE_RESTARTING=true
- WATCHTOWER_NO_STARTUP_MESSAGE=true
- WATCHTOWER_NOTIFICATION_REPORT=true
- WATCHTOWER_NOTIFICATIONS=shoutrrr
- WATCHTOWER_NOTIFICATION_URL={{ watchtower_gotify_url }}
- WATCHTOWER_NOTIFICATIONS_HOSTNAME=DAWARICH
- WATCHTOWER_NOTIFICATION_TITLE_TAG=DAWARICH
# Render empty when nothing was updated/failed — see feedback_watchtower_pushover.
- 'WATCHTOWER_NOTIFICATION_TEMPLATE={% raw %}{{- if .Report -}}{{- with .Report -}}{{- if or .Updated .Failed -}}{{ len .Updated }} updated, {{ len .Failed }} failed{{- range .Updated}} | updated {{.Name}}{{- end -}}{{- range .Failed}} | FAILED {{.Name}}{{- end -}}{{- end -}}{{- end -}}{{- end -}}{% endraw %}'