# Garage — Lightweight Self-Hosted S3-Compatible Storage

> Garage is a lightweight, self-hosted, geo-distributed object storage system that implements the S3 API and is designed for small to medium infrastructure.

## Install

Save in your project root:

# Garage — Lightweight Self-Hosted S3-Compatible Storage

## Quick Use
```bash
# Download and start a single-node instance
curl -Lo garage https://garagehq.deuxfleurs.fr/dl/v1.0/x86_64-unknown-linux-musl/garage
chmod +x garage

# Generate a config file
cat > garage.toml << EOF
metadata_dir = "/var/lib/garage/meta"
data_dir = "/var/lib/garage/data"
db_engine = "lmdb"
replication_factor = 1

[s3_api]
api_bind_addr = "[::]:3900"
s3_region = "garage"
root_domain = ".s3.garage.localhost"

[rpc_bind_addr]
# defaults

[admin]
api_bind_addr = "[::]:3903"
EOF

./garage -c garage.toml server
```

## Introduction
Garage is a self-hosted S3-compatible object storage server built in Rust. Unlike heavyweight systems like Ceph or MinIO, Garage is designed for small clusters spread across multiple locations. It handles geo-distribution natively, replicating data across zones with minimal coordination, making it suitable for homelab setups and small organizations.

## What Garage Does
- Serves objects over the S3 API, compatible with standard AWS SDKs and tools
- Replicates data across geographic zones for redundancy without a central coordinator
- Supports multi-tenancy with access keys, bucket policies, and per-bucket quotas
- Provides a static website hosting mode directly from S3 buckets
- Runs as a single static binary with minimal resource consumption

## Architecture Overview
Garage uses a peer-to-peer protocol where all nodes are equal — there is no master or coordinator. Metadata is stored in a Merkle-search-tree CRDT that synchronizes across nodes without consensus overhead. Data blocks are replicated across zones according to a configurable replication factor. The cluster layout maps partitions to nodes, and rebalancing happens automatically when nodes join or leave.

## Self-Hosting & Configuration
- Deploy as a single static binary on each node (Linux, ARM, or x86_64)
- Configure via a TOML file specifying data directories, RPC ports, and S3 endpoint
- Assign nodes to geographic zones for cross-site replication
- Create access keys and buckets via the `garage` CLI or the admin HTTP API
- Monitor with built-in Prometheus metrics and optional Grafana dashboards

## Key Features
- Designed for geo-distributed clusters across multiple physical locations
- No single point of failure — fully decentralized peer-to-peer architecture
- Low resource footprint suitable for Raspberry Pi and homelab hardware
- S3-compatible API works with rclone, restic, Nextcloud, and any S3 client
- Static website hosting directly from S3 buckets with custom domain support

## Comparison with Similar Tools
- **MinIO** — Higher performance for single-site deployments but heavier resource usage
- **Ceph (RGW)** — Enterprise-grade with POSIX and block support, but complex to operate
- **SeaweedFS** — Blob-oriented store with FUSE mount, not primarily S3-focused
- **CubeFS** — Cloud-native distributed filesystem, requires larger minimum cluster
- **Wasabi / Backblaze B2** — Managed cloud storage, not self-hosted

## FAQ
**Q: How many nodes do I need to run Garage?**
A: A single node works for testing. For production geo-distribution, 3 nodes across 3 zones is recommended with replication factor 3.

**Q: Can I use Garage as a backup target?**
A: Yes. Tools like restic, Duplicati, and rclone work with Garage's S3 API for encrypted backups.

**Q: Does Garage support versioning or lifecycle policies?**
A: Bucket versioning is supported. Lifecycle policies for automatic deletion are on the roadmap.

**Q: What are the hardware requirements?**
A: Garage runs on hardware as small as a Raspberry Pi 4 with 2 GB RAM. Production nodes benefit from SSDs for metadata storage.

## Sources
- https://github.com/deuxfleurs-org/garage
- https://garagehq.deuxfleurs.fr/documentation/

---
Source: https://tokrepo.com/en/workflows/asset-d0a7b7dd
Author: AI Open Source