Garage — a lightweight, S3-compatible distributed object store — plus its web admin panel, bundled into one container for Unraid. Both are the official upstream binaries, unmodified, wired together with s6-overlay. No manual CLI setup: single-node layout, S3 key and bucket are all created automatically on first boot.

Table of Contents

1. Why one container?
2. What is this?
3. Quick Start on Unraid
4. Connecting a client
5. Configuration
6. Backup
7. Support this project

1. Why one container?

Garage's own Community Applications listings are split across three separate templates from one maintainer: the core server (in two network variants) and the web admin panel as its own install. This template bundles the core server and the admin panel into one container — both official, unmodified upstream binaries, no rebuild from source — so one install gives you a working, browsable S3 store.

2. What is this?

Garage is a lightweight, S3-API-compatible object store built for small, geo-distributed self-hosted deployments — Rust, Apache-2.0, actively developed. This template runs it in single-node mode (Garage's own --single-node flag, introduced in v2.3.0), which auto-creates the cluster layout at startup — the manual garage layout assign / garage layout apply steps a bare install normally requires are not needed here. An S3 access key, secret key and bucket are pre-seeded from the template fields on first boot (Garage's --default-access-key / --default-bucket flags), so there is no second CLI step to create them either.

garage-webui runs alongside it, talking to Garage's admin API over localhost — buckets, keys and cluster health are all browsable without touching a terminal.

3. Quick Start on Unraid

1. Install from Community Applications, or add this template's URL directly: https://raw.githubusercontent.com/junkerderprovinz/garage/main/templates/garage.xml
2. Set Access Key and Secret Key — anything reasonably random works, e.g. generate a secret with openssl rand -hex 24.
3. Optionally set Bucket to a name you want created immediately (e.g. backups).
4. Start the container. The S3 endpoint is http://SERVER_IP:3900; the admin panel is at http://SERVER_IP:3909.

Watch the container log for GARAGE IS READY — first boot generates the config and layout, this takes a few seconds.

4. Connecting a client

Any S3-compatible tool works. A couple of common ones:

rclone (rclone config, or a config file entry):

[garage]
type = s3
provider = Other
access_key_id = YOUR_ACCESS_KEY
secret_access_key = YOUR_SECRET_KEY
endpoint = http://SERVER_IP:3900
region = garage

restic (repository URL):

export AWS_ACCESS_KEY_ID=YOUR_ACCESS_KEY
export AWS_SECRET_ACCESS_KEY=YOUR_SECRET_KEY
restic -r s3:http://SERVER_IP:3900/my-restic-repo init

5. Configuration

Setting	Container Variable	Default	Notes
S3 API Port	(port)	3900	The functional endpoint every client connects to.
S3 Website Port	(port)	3902	Advanced. Static-site hosting straight from a bucket.
Admin Panel Port	(port)	3909	Browser UI: buckets, keys, cluster health.
Data	(path)	/mnt/user/appdata/garage/data	Object data and metadata. Must be mapped.
Config	(path)	/mnt/user/appdata/garage/config	Generated garage.toml (RPC secret, admin token). Must be mapped — losing it changes your RPC/admin secrets on next boot.
Access Key	ACCESS_KEY	(empty)	S3 access key, pre-seeded on first boot. Required to write anything real.
Secret Key	SECRET_KEY	(empty)	S3 secret key, pre-seeded on first boot.
Bucket	BUCKET	(empty)	Bucket name to create automatically on first boot. Leave empty to create buckets later via the admin panel or any S3 client.
DB Engine	DB_ENGINE	sqlite	Garage's metadata engine. lmdb is faster but less tolerant of an unclean shutdown; sqlite is the safer default.

Ports 3901 (Garage's internal RPC) and 3903 (the admin API garage-webui talks to) are intentionally not published — nothing outside the container needs them directly.

6. Backup

Everything lives under the Data and Config mounts — back both up the same way you back up any other appdata folder. Losing Config does not lose your objects, but it does regenerate the RPC secret and admin token on next boot, which breaks the admin panel's stored session and (for a real multi-node cluster, not this single-node setup) cluster membership.

7. Support this project

Questions, bugs, ideas? GitHub issues →.

If this template saves you a setup hassle, consider buying me a coffee: