NAS Backups, Part 1: Overview
After building my nice NAS, the next thing on my sysadmin todo list was to ensure it is backed up. In this miniseries (probably 2 posts), I'm going to talk about the backup NAS that I've built. As of the time of typing, it is sucessfully backing up my Btrfs subvolumes.
In this first post, I'm going to give an overview of the hardware I'm using and the backup system I've put together. In future posts, I'll go into detail as to how the system works, and which areas I still need to work on moving forwards.
Personally, I find that the 3-2-1 backup strategy is a good rule of thumb:
- 3 copies of the data
- 2 different media
- 1 off-site
What this means is tha you should have 3 copies of your data, with 2 local copies and one remote copy in a different geographical location. To achieve this, I have solved first the problem of the local backup copy, since it's a lot easier and less complicated than the remote one. Although I've talked about backups before (see also), in this case my solution is slightly different - partly due to the amount of data involved, and partly due to additional security measures I'm trying to get into place.
For hardware, I'm using a Raspberry Pi 4 with 4GB RAM (the same as the rest of my cluster), along with 2 x 4 TB USB 3 external hard drives. This is a fairly low-cost and low-performance solution. I don't particularly care how long it takes the backup to complete, and it's relatively cheap to replace if it fails without being unreliable (Raspberry Pis are tough little things).
Here's a diagram of how I've wired it up:
I use USB Y-cables to power the hard rives directly from the USB power supply, as the Pi is unlikely to be able to supply enough power for mulitple external hard drives on it's own.
Important Note: As I've discovered with a different unrelated host on my network, if you do this you can back-power the Pi through the USB Y cable, and potentially corrupt the microSD card by doing so. Make sure you switch off the entire USB power supply at once, rather than unplug just the Pi's power cable!
To do the backup itself I'm using the fact that I store my data in Btrfs subvolumes and the
btrfs send /
btrfs receive commands to send my subvolumes to the remote backup host over SSH. This has a number of benefits:
- The host doing the backing up has no access to the resulting backups (so if it gets infected it can't also infect the backups)
- The backups are read-only Btrfs snapshots (so if the backup NAS gets infected my backups can't be altered without first creating a read-write snapshot)
- Backups are done incrementally to save time, but a full backup is done automatically on the first run (or if the local metadata is missing)
While my previous backup solution using Restic for the server that sent you this web page has point #3 on my list above, it doesn't have points 1 and 2.
Restic does encrypt backups at rest though, which the system I'm setting up doesn't do unless you use LUKS to encrypt the underlying disks that Btrfs stores it's data on. More on that in the future, as I have tentative plans to deal with my off-site backup problem using a similar technique to that which I've used here that also encrypts data at rest when a backup isn't taking place.
In the next post, I'll be diving into the implementation details for the backup system I've created and explaining it in more detail - including sharing the pair of scripts that I've developed that do the heavy lifting.