# MadBase Storage & Persistence

MadBase uses a tiered storage approach combining local persistence, distributed databases, and S3-compatible object storage.

## 1. Local Block Storage
Nodes use local NVMe/SSD storage for high-performance service data:
- **Database**: PostgreSQL data stored in `/var/lib/postgresql/data`.
- **Metrics**: VictoriaMetrics data stored in `/victoria-metrics-data`.
- **Logs**: Loki chunks stored in `/loki`.

## 2. Object Storage (S3)
Used for backups, static assets, and long-term state retention.
- **Backups**: Database backups are automatically piped to S3-compatible buckets.
- **Provider**: Works with Hetzner Bucket Storage, AWS S3, or MinIO.

### Configuration
```bash
S3_ENDPOINT="https://fsn1.your-storage-endpoint.com"
S3_ACCESS_KEY="your_key"
S3_SECRET_KEY="your_secret"
S3_BUCKET="madbase-backups"
```

## 3. Backup & Restore
### Manual Backup
```bash
curl -X POST http://localhost:8001/api/v1/cluster/database/backup
```
### Manual Restore
```bash
curl -X POST http://localhost:8001/api/v1/cluster/database/restore \
  -d '{"backup_url": "s3://..."}'
```

## 4. Planning & Capacity
- **DB Nodes**: Scale vertically (RAM) for active datasets; scale horizontally (Nodes) for read-throughput and HA.
- **Retention**: Configure VictoriaMetrics and Loki retention periods in the System Node config to manage disk usage.