driver-vault/DriverVault
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
main
DriverVault/Docker/docker-compose.prod.s3.yml
T
tajniak81andClaude Opus 5 9a2a4ab72e The files leave the volume the database sits on 
Every attachment — a document scan, a fuel receipt, a workshop invoice, a
photo of a part's box — has lived inside pb_data, in a directory beside the
SQLite file. One volume held both, so neither could be sized, backed up or
moved without the other. PocketBase can keep those bytes in an S3 bucket
instead, and now it is told to.

Nothing on the way to a client changes, because an attachment was never a
storage URL to begin with: it is fetched from GET /api/{records}/{id}/file,
which re-checks car access and asks PocketBase for the bytes as the service
account. PocketBase streams from the bucket through that same endpoint rather
than redirecting to it, so the web app, the phone and the plugin cannot tell
which side of the switch they are on.

The bootstrap that already creates the collections now writes PocketBase's
files-storage settings too, from PB_S3_*, on every boot and only when they
differ from what is already there — then asks PocketBase to prove it can reach
the bucket, and says so in the log either way. Two asymmetries are deliberate.
A read of the settings masks the stored secret, so a rotation of the secret
alone is invisible from here and needs another PB_S3_* to move with it. And it
never turns S3 back off: files already written to a bucket are reachable only
while PocketBase still points at it, so dropping the configuration would strand
them rather than undo anything.

Each deployment shape is one compose file with an .env example of the same
name, not a base plus an overlay to remember — six of each per folder, for
Docker and Docker-AIO alike: the plain one, .seaweedfs, .s3, and the three prod
twins. The SeaweedFS files run master, volume, filer and gateway as one process
and a one-shot init container beside it, because PocketBase never issues a
CreateBucket and SeaweedFS will not conjure one on first upload. The credentials
do double duty there — the gateway's only identity is also what PocketBase
authenticates with. In the all-in-one that gateway is a second container rather
than a fourth process under supervisord: keeping the object store inside the
image, on the volume the files are being moved off, would have defeated the
point and would have meant rebuilding.

Files uploaded before the switch are not carried across; PocketBase copies
nothing, and both READMEs say so where an operator will read it.

The TLS overlay and its Caddyfile go. The section they served stays, without
them: nothing in the stack terminates TLS any more, so it now names the four
variables to set in front of whichever proxy already does — TRUST_FORWARDED_PROTO
being the one that decides whether a charger is believed about how it arrived.

Unexercised: this was written on a machine without Docker, so the pinned
SeaweedFS image, the bucket-create and the settings write have not been run
against a live stack. The Go side builds, vets and tests clean.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-09-04 19:32:19 +02:00

175 lines
9.0 KiB
YAML
Raw Permalink Blame History

name: drivervault
# Production DriverVault stack, with external S3 — pulls prebuilt images from
# the registry instead of building from source. Self-contained: one file, no
# overlays. Everything an operator needs to set lives in .env.
#
# 1. cp .env.prod.s3.example .env (then edit it — PB_S3_* especially)
# 2. docker compose -f docker-compose.prod.s3.yml pull
# 3. docker compose -f docker-compose.prod.s3.yml up -d
#
# This is docker-compose.prod.yml pointed at an S3 endpoint that already exists
# somewhere else — its own host, another compose project, or any S3-compatible
# service. PocketBase keeps its record files — document scans, service and
# refill receipts, workshop invoices, part photos — in that bucket instead of on
# the pb_data volume. The database and PocketBase's own backups stay on PB_DATA.
# Clients cannot tell the difference: an attachment has always been fetched
# through the API Server, never from a storage URL.
#
# The bucket must already exist, and nothing here runs the gateway. For a
# SeaweedFS that comes up with the stack, use docker-compose.prod.seaweedfs.yml.
#
# Before turning this on for a stack that already has uploads: PocketBase does
# NOT copy existing files into the bucket. See README.md.
#
# Traffic flow (browser): Web App BFF --/api--> API Server --> PocketBase.
#
# On first boot:
# • PocketBase upserts the superuser from PB_ADMIN_* (create-if-missing).
# • the API Server creates any missing collections, reconciles existing ones,
# and creates the DriverVault super-admin from DRIVERVAULT_SUPERADMIN_*.
# Both steps are idempotent, so restarts and upgrades are safe.
services:
pocketbase:
image: "${PB_IMAGE:-10.2.1.10:5500/admin/drivervault-pocketbase:latest}"
container_name: drivervault-pocketbase
restart: unless-stopped
extra_hosts:
# Lets PB_S3_ENDPOINT name the Docker host as host.docker.internal.
# Harmless when the endpoint is somewhere else entirely.
- "host.docker.internal:host-gateway"
environment:
# The superuser is created/updated on boot (the API Server authenticates
# with it). This is the only place the first superuser can be created — the
# REST API cannot bootstrap it.
PB_ADMIN_EMAIL: "${PB_ADMIN_EMAIL:?set PB_ADMIN_EMAIL in .env}"
PB_ADMIN_PASSWORD: "${PB_ADMIN_PASSWORD:?set PB_ADMIN_PASSWORD in .env}"
volumes:
# Named volume by default; set PB_DATA to a host path in .env for a bind mount.
- "${PB_DATA:-pb_data}:/pb/pb_data"
ports:
# Bound to localhost by default — the admin UI (/_/) is reachable only on
# the host. Set PB_BIND=0.0.0.0 in .env to expose it on the network.
- "${PB_BIND:-127.0.0.1}:${PB_PORT:-8070}:8070"
healthcheck:
test: ["CMD-SHELL", "wget -qO- http://127.0.0.1:8070/api/health || exit 1"]
interval: 10s
timeout: 3s
retries: 12
start_period: 10s
api-server:
image: "${API_IMAGE:-10.2.1.10:5500/admin/drivervault-api-server:latest}"
container_name: drivervault-api
restart: unless-stopped
extra_hosts:
# Lets PB_S3_ENDPOINT name the Docker host as host.docker.internal.
# Harmless when the endpoint is somewhere else entirely.
- "host.docker.internal:host-gateway"
depends_on:
pocketbase:
condition: service_healthy
environment:
API_ADDR: ":8080"
# Reach PocketBase by its service name on the internal network. Override
# POCKETBASE_URL in .env to point the API Server at a database outside
# this stack — that is also how you make a retarget done from the panel
# permanent, since the panel's change lasts only for the container's life.
POCKETBASE_URL: "${POCKETBASE_URL:-http://pocketbase:8070}"
POCKETBASE_ADMIN_EMAIL: "${PB_ADMIN_EMAIL}"
POCKETBASE_ADMIN_PASSWORD: "${PB_ADMIN_PASSWORD}"
# Probed by the panel status page. This is a server-to-server call inside
# the compose network, so the default is the service name — plain
# localhost:8090 would resolve to this container itself. Override
# WEBAPP_URL in .env to make a change from the panel's Web App screen
# permanent; the panel alone only holds it for the container's life.
WEBAPP_URL: "${WEBAPP_URL:-http://web-app:8090}"
CORS_ALLOW_ORIGINS: "${CORS_ALLOW_ORIGINS:-http://localhost:8090}"
AUTH_USERS_COLLECTION: "${AUTH_USERS_COLLECTION:-users}"
# Schema + super-admin bootstrap (idempotent). Leave this ON: a release can
# add collections or fields the server needs, and a stack that skips the
# bootstrap never gets them. The API Server self-heals exactly one thing —
# app_settings, the collection holding the plugin settings, which it
# creates on demand because it cannot serve the plugin panel without it.
# Every other schema change still depends on this flag. Turn it off only
# for a database you know already matches the release.
PB_BOOTSTRAP: "${PB_BOOTSTRAP:-true}"
DRIVERVAULT_SUPERADMIN_EMAIL: "${DRIVERVAULT_SUPERADMIN_EMAIL:-}"
DRIVERVAULT_SUPERADMIN_PASSWORD: "${DRIVERVAULT_SUPERADMIN_PASSWORD:-}"
DRIVERVAULT_SUPERADMIN_NAME: "${DRIVERVAULT_SUPERADMIN_NAME:-Administrator}"
# OCPP charger control (Anker Solix). Chargers are rejected unless they
# reach the server over TLS. Behind a TLS-terminating reverse proxy, set
# OCPP_PUBLIC_URL to the public wss:// base and API_BIND so the proxy can
# reach this port; only drop OCPP_REQUIRE_TLS on a trusted network.
OCPP_REQUIRE_TLS: "${OCPP_REQUIRE_TLS:-true}"
OCPP_PUBLIC_URL: "${OCPP_PUBLIC_URL:-}"
# Diagnostic: set ANKER_MQTT_FRAME_LOG=1 in .env to log every frame the
# charger publishes over Anker's broker, decoded ones and unreadable ones
# alike, with their bytes. It is how a frame nobody has named gets named —
# do something in the Anker app while a control read holds the connection
# open, and read the frames back out of the log. Off by default: with it on
# a charger under a live trigger writes a line every few seconds.
ANKER_MQTT_FRAME_LOG: "${ANKER_MQTT_FRAME_LOG:-}"
# --- File storage --------------------------------------------------
# Read by the API Server's bootstrap, which writes them into PocketBase's
# settings on every boot, idempotently. Only record files move — scans,
# receipts, invoices, part photos. The database and PocketBase's own
# backups stay on PB_DATA. The bucket must already exist: nothing here
# creates it.
PB_S3_ENABLED: "true"
PB_S3_BUCKET: "${PB_S3_BUCKET:-drivervault}"
PB_S3_ENDPOINT: "${PB_S3_ENDPOINT:?set PB_S3_ENDPOINT in .env}"
PB_S3_REGION: "${PB_S3_REGION:-us-east-1}"
PB_S3_ACCESS_KEY: "${PB_S3_ACCESS_KEY:?set PB_S3_ACCESS_KEY in .env}"
PB_S3_SECRET: "${PB_S3_SECRET:?set PB_S3_SECRET in .env}"
# true for SeaweedFS and MinIO, false for AWS S3 proper.
PB_S3_FORCE_PATH_STYLE: "${PB_S3_FORCE_PATH_STYLE:-true}"
ports:
# Localhost-only by default (the Web App reaches it over the internal
# network). Set API_BIND=0.0.0.0 to expose the API panel — and the
# /ocpp/{serial} endpoint chargers dial into — on the network.
- "${API_BIND:-127.0.0.1}:${API_PORT:-8080}:8080"
# No volume: the API Server keeps no state on disk — every setting it owns,
# plugin settings included, lives in PocketBase under PB_DATA.
healthcheck:
# Declared here rather than relying only on the image's HEALTHCHECK, so the
# depends_on gate below still works against an older pulled image.
test: ["CMD-SHELL", "wget -qO- http://127.0.0.1:8080/healthz || exit 1"]
interval: 10s
timeout: 3s
retries: 12
start_period: 20s
web-app:
image: "${WEB_IMAGE:-10.2.1.10:5500/admin/drivervault-web-app:latest}"
container_name: drivervault-web
restart: unless-stopped
depends_on:
# The image now ships a HEALTHCHECK, so wait for the API Server to be
# serving rather than merely started.
api-server:
condition: service_healthy
environment:
# The BFF reverse-proxies /api/* — and /ocpp/*, the address chargers are
# told to dial — to the API Server over the internal network.
API_BASE: "http://api-server:8080"
# Believe an inbound X-Forwarded-Proto. The API Server reads it to decide a
# charger arrived over TLS, so leave this off unless a TLS-terminating
# proxy in front of the stack is the only way in: otherwise a charger could
# claim wss over a plaintext connection. Set it to true when TLS ends at
# that proxy and OCPP_PUBLIC_URL names a wss:// base through it.
TRUST_FORWARDED_PROTO: "${TRUST_FORWARDED_PROTO:-false}"
ports:
# The public front door. Bound on all interfaces so browsers can reach it.
- "${WEB_PORT:-8090}:8090"
healthcheck:
test: ["CMD-SHELL", "wget -qO- http://127.0.0.1:8090/healthz || exit 1"]
interval: 10s
timeout: 3s
retries: 12
start_period: 10s
volumes:
pb_data:
Reference in New Issue View Git Blame Copy Permalink