What actually counts as a copy under the 3-2-1 rule, why a RAID mirror or a snapshot only protects against one kind of failure, and the one-page inventory that exposes the gaps.
Free weekly email with the week’s posts, videos and one tip: https://chartmann1590.github.io/homelab-the-right-way/newsletter/
Disclosure: this is my own article (Charles Hartmann, Homelab the Right Way). Written with AI help and edited by me; the setup described is mine.
I feel like nothing I do is “right”.
Also, I feel like I can’t afford to back things up. Not with the prices of hardware and cloud storage.
Yeah, it does feel like that. I wanted to backup my photos but eventually realized that instead of just uploading them to a NAS, I could use something like Immich to get a nice gallery. Immich even calls uploading from your phone backup. So my photos are backed up now, right? Nope, you have to back up Immich too. Ah, ok, it has these database backups, that’s fine right? Nope, that’s just a db dump, you need to backup the actual photo files. Ok, I will setup RAID so they are mirrored to another hard drive. RAID IS NOT BACKUP!!! Ok, fine, I will rsync them to another external hard drive periodically. Nope, what if your place burns down? Ok, I will also spend 13€ a month on a Hetzner storage box and upload an encrypted backup there with borg. Where are you storing your borg passphrase? Fine, I’ll write it down on a piece of paper and give that to my parents to store at their house. Am I done? Is that 3-2-1? Yeah, but it’s not automated, you should have a backup server. Aaaahhhhhh!
Totally get that. You don’t have to back up everything, just the stuff you can’t replace: photos, documents, configs. That set is usually small enough for an old USB drive plus a cheap or free offsite option. Rebuildable stuff (media, ISOs, containers you can redeploy) can skip it. Imperfect backups of the important 5% beat a perfect plan you never get to.
Problem I have is so much of my data is “projects” where like, only god and my client know what random shit they’ll ask for.
That’s the hard case. When you can’t predict what a client will ask for, backing up the whole project folder is simpler than trying to pick files, and moving finished projects to an archive keeps the active set small.
But it feeeeels like a backup, and isn’t that what really matters? In my setup I have 3 drives (2x 8Tb, 1x 14 Tb) and I finally have the opportunity to do some consolidation and not just have 3 single drives plugged into a server. Zfs, raid , a dedicated backup drive, but I’m still figuring out what the layout should be.
Ha, it absolutely feels like one, right up until a bad sync or a fat-fingered rm hits every disk at once. With 2x8 TB plus a 14 TB, one layout people often land on is mirroring the two 8s (ZFS mirror) for live data and using the 14 TB as a separate backup target (snapshots replicated to it), plus something offsite for the stuff you can’t lose. That gets you redundancy and an actual backup out of the same three drives.
