# You Don't Need S3 to Back Up Your Home Server or Production Server

<div data-node-type="callout">
<div data-node-type="callout-emoji">💡</div>
<div data-node-type="callout-text"><strong>A practical backup strategy for small production servers using an ssd and Google Drive.</strong></div>
</div>

Most backup guides assume you're on AWS with a fat budget. This one doesn't.

If you're running a small production server — a VPS, a homelab that serves real users, a side project with actual data — you can build a solid, layered backup system with tools you already have. No S3, no paid services, no complexity you don't understand.

> pg\_dump, rsync and rclone

**Setup: I have 3 VM's setup inside my bare metal server.**

Here's the strategy I use across a Proxmox setup on baremetal HP server. The same pattern applies to any Linux server.

## The Three-Layer Strategy

Good backups have three properties: they're **automated**, **layered**, and **offsite**. This setup delivers all three.

```shell
PostgreSQL databases
        ↓ pg_dump,pgdump_all (compressed) + rsync for files
  Local SSD
        ↓ rclone copy
    Google Drive
```

**Layer 1 — Database dumps to local SSD** `pg_dump` with the `-Fc` flag produces compressed, restorable snapshots. A companion `pg_dumpall --globals-only` captures users, roles, and permissions that `pg_dump` misses. Both land on an attached SSD.

**Layer 2 — File sync to SSD via rsync** Application uploads and user files are `rsync`'d from the app server to the database server's SSD mount. The `--delete` flag keeps the destination clean — no stale files accumulating over time.

**Layer 3 — Offsite via rclone to Google Drive** `rclone copy` pushes everything to Google Drive. This is your disaster recovery layer — if the physical machine dies, your data is still there.

## The Tools

`pg_dump -Fc` — PostgreSQL's native backup tool. The custom format (`-Fc`) is compressed binary, supports selective table-level restores via `pg_restore`, and is faster than plain SQL dumps for large databases.

`rsync -avz --delete` — The gold standard for incremental file sync over SSH. Archive mode, compressed transfer, with automatic cleanup of deleted files.

`rclone` — Think of it as rsync for cloud storage. Supports 50+ backends. Set it up once with `rclone config`, authenticate via OAuth, and it works like any other file operation.

`cron` — The legend that made it all possible.

## The Architecture

![](https://cdn.hashnode.com/uploads/covers/65cf0b043f77edd06a52edca/b43563b6-8f59-4aba-8c55-51113b26b5c2.svg align="center")

## What You Need

| Component | Tool | Why |
| --- | --- | --- |
| Database backups | `pg_dump` + `pg_dumpall` | Complete snapshots including globals |
| File sync | `rsync` | Incremental, SSH-native, battle-tested |
| Cloud offsite | `rclone` + Google Drive | Free 15 GB, expandable, zero setup cost |
| Scheduling | `cron` | Native to every Linux distro |
| Retention | Shell variable `RETENTION_DAYS=35` | Automatic cleanup, ~2 months of history |
| Network | Tailscale (optional) | Zero-config VPN between nodes |

* * *

## Setting Up rclone in 3 Steps

```shell
# 1. Install
curl https://rclone.org/install.sh | sudo bash

# 2. Configure (interactive, browser-based OAuth)
rclone config

# 3. Test
rclone ls googledrive:   # list your Google Drive root
```

The colon after `googdrive` is not a typo — it tells rclone you're referring to the remote, not a local path named `googdrive`.

* * *

## The Cron Schedule

```shell
# On VM3 (database server) — as postgres user
X X X * *   /usr/local/bin/pg-backup.sh

# On VM1 (gateway) — as root
X X X * *   /usr/local/bin/send-ssd-backup.sh >> /var/log/backup.log 2>&1

X X X * *   /usr/local/bin/gdrive-backup.sh
```

* * *

## One Thing Most Guides Skip

A backup you've never restored is a bet, not a guarantee.

After setting this up, run a restore drill. Spin up a throwaway VM, pull a dump from Drive, and run:

```bash
pg_restore -d mydb /path/to/backup.dump
and
create scripts for your files/folders backup
```

If it works, you have a backup system. If it doesn't, you have a false sense of security. Test early.

* * *

## What This Costs

| Resource | Cost |
| --- | --- |
| Google Drive (15 GB) | Free |
| rclone | Free / open source |
| rsync | Pre-installed |
| Tailscale | Free |
| Your time | learning + implementation ( X days) |

The only real cost is the SSD — which you probably already have attached to your server or USB.

* * *

*full guide on:* https://gaurablearn.hashnode.dev/