The 3-2-1 Rule: How to Organize Backups in a Small Company

Most companies that lost data “were making backups.” The copy just sat on the same disk, in the same room, or hadn’t been tested in years. The 3-2-1 rule is the shortest way to avoid that.

What 3-2-1 means

• 3 — three copies of your data: the working copy and two backups.
• 2 — on two different media: for example, the server disk and an external drive or another server.
• 1 — one copy off-site: in another building, in a data center, in the cloud.

The point is that no single event — fire, theft, ransomware, a failed disk controller, an admin’s mistake — should be able to destroy all copies at once.

What it looks like for 1C

The working database on the server (copy 1). A daily .dt export or a copy of the database files to another disk on the same server or to a NAS in the office (copy 2, different media). The same export is sent overnight via SFTP to a server in the data center (copy 3, off-site). Keep 30 daily versions: if the database was corrupted a week ago and you noticed today, you still have something to restore from.

For a file server

Files on the server (1). An incremental copy to a NAS or second disk — rsync, Synology Hyper Backup, Windows Server Backup (2). A nightly sync of changes only to the data center — rsync, Restic, Borg or Veeam (3). The first full copy takes a night or two; after that only new and changed files are sent — minutes.

For a website

The site on a VPS (1). A daily snapshot of the virtual machine, taken on the provider’s side (2). A daily database dump and file archive to another server or to object storage (3). The snapshot saves you from “an update broke it”; a separate dump saves you from “deleted the database and that ended up in the snapshot.”

Two mistakes that void the whole scheme

The backup is writable from the production server. If ransomware gains access to the server, it will encrypt the attached backup disk and the network share as well. The off-site copy should accept data under a separate account and must not allow deleting old versions — append-only mode is available in Borg, Restic and Veeam (immutable repository).

Nobody has tested a restore. Once a quarter, restore the 1C database from a backup to a test server and open it. If the procedure takes a day and requires someone who has since left the company, you don’t have a backup.

What a copy in the data center costs

A backup server at MyCloud is a minimal virtual machine (1 vCPU, 1 GB RAM — 3 000 UZS per day including IP) plus disk at 50 UZS per GB per day. 200 GB for 1C exports with history comes to 13 000 UZS per day, about 390 000 per month. 1 TB for a file server — about 1.6 million per month. Data from an office in Tashkent travels over TASIX free of charge, at the speed of your connection.

For employees’ documents, the ready-made MCDrive storage with versioning is more convenient; for servers and databases, there’s a backup server you connect your own tools to. We’ll help you choose a scheme and set up the schedule — free of charge.