I am using duplicati and thinking of switching to Borg. What do you use and why?
I use restic. For local backups, Timeshift.
Seconded, I use restic with a remote blob storage and works nicely
Using borg backup, just because there are some nice frontends for the gnome ecosystem (when I am using gnome, I love to use gnome apps), and it has a nice cmd for scripting when using something else (using it on servers)
And there is a nice graphical frontend for it too: Vorta
Personally more of Pika Backup user ;)
I just use
rsyncto backup my home folder to my NAS.I don’t have backups. :/
And I will regret it some day.
I use github for code so that’s backed up though.
There are two kinds of people.
Those who make backups and those who will.
Kopia has served me great. I back up to my local Ceph S3 storage and then keep a second clone of that on a raid.
Kopiahas good performance and miltiple hosts can back up tp it concurrently while preserving deduplication – unlike borgbackup.
I’ve been using Kopia on my desktop computer for a few years now to do cloud backups. It’s generally working well and I haven’t found anything else with the same combination of features yet.
That said, kopia-ui is still a bit finicky and I’ve managed to bork a repo beyond repair a few times (e.g. once because my cloud provider account ran out of space, leading to some kind of inconsistent state) and there are some oddities, like the regular “periodic maintenance” (it’s a bit weird that it’s needed in the first place) randomly failing or taking forever.
Kopia has been working great for me as well. It’s simple, versatile and reliable. I previously used Duplicati but kept running into jobs failing for no reason, backup configurations missing randomly and simple restores taking hours. It was a hot mess and I’m happy I switched.
I want to love kopia but the command line syntax feels unnatural to me. I don’t know why either. For the whole month I test drove it, I had to look up every single time how to do something. Contrast this with restic which is less featureful in some ways but a few days in it felt like I was just using git.
I never used the command line with Kopia besides starting it up in server mode and used the web based GUI to configure, it was pretty simple to get everything setup that way. You may want to give it another try using Kopia in that mode.
My use case is for headless machines which makes it a no go in that regard unfortunately.
You can use the web ui remotely.
Personally I use it from command line, though, and my only complaint is that it’s too easy to start a backup you didn’t intend to… Buut if you’re careful about usong the
kopia snapshotcommand then it’s fine.Oh I thought the webui was only for server mode.
I just quickly glanced through the manuals of both restic and kopia. I think my trouble with kopia is that its style feels kind of weird. I’m just not able to wrap my head around it well.
kopia snapshot create /diris shorter but more confusing thanrestic -r repo backup /dir
I’ve been using restic. It has built-in dedup & encryption and supports both local and remote storage. I’m using it to back up to a local restic-server (pointing to a USB drive) and Backblaze B2.
Restores for single or small sets of files is easy: restic -r $REPO mount /mnt Then browse through the filesystem view of your snapshots and copy just like any other filesystem.
I use NixOS so all my system configuration is already saved in my NixOS configs, which I save on GitHub. For dotfiles that aren’t managed by NixOS I use syncthing to sync them between my devices, but no real backup cause I can just remake them if I need to, and things like my Neovim and VSCode configs are managed by my NixOS configs so they’re backed up as well.
You can take this to the extreme too by erasing your root partition each boot: https://grahamc.com/blog/erase-your-darlings/
Using that method you isolate all important state on the system for backup with zfs send.
Yeah I have a full impermanence setup using tmpfs, which is really nice. I did it like on the NixOS wiki and it’s been helpful for organizing my dotfiles and keeping track of all the random stuff that programs put everywhere.
I actually have all my stuff in a separate /stuff folder kinda by accident so my /home only has dotfiles and things like that.
There is no such thing as the objectively best solution. Each tool has advantages and disadvantages. And every user has different preferences and requirements.
Personally, I am using Borg for years. And I have had to restore data several times, which has worked every time.
In addition to Borg, you can also look at Borgmatic. This wrapper extends the functionality and makes some things easier.
And if you want to use a graphical user interface, you can have a look at Vorta or Pika.
Rsync is great but if you want snapshots and file history rsnapshot works pretty well. It’s based on rsync but for every sync it creates shortcuts for existing files and only copies changes and new files. It saves space and remains transparent for the user. FreeFileSync is also amazing
I am old school. I just use GNU Tar with the Pax format and multiple external detachable encypted hard drives. Reason is it is simple and a well known tool that is very common with a standard archive format.
I’m curious - how much data are you backing up with that method and how frequently are you doing your backups? Doesn’t sound like it would scale well, but I’m also wondering if maybe this is perfect and I’ve justbeenoverthinking it.
There is not a size limit. Lot of these other methods actually use GNU Tar behind the scenes anyway. More then that GNU tar has been used for decades for this purpose. Pull out any Unix book from 2 decades ago and you will see “tar”, “cpio”, and “dump/restore” as the way. The new tool out there is Pax and in fact GNU Tar supports the new “pax” format. Moreover GNU Tar with Pax format can backup almost full disk structure including hard links, ACLs, and extended attributes which a lot of tools do not do. It is still useful to archive some things at a lower level like your partition table, and boot blocks of course. You also have to decide what run-level (such as rescue) you want to archive in, and/or what services you should stop, or provide separate to file system dumps for depending on your system. Databases, and things like ecryptfs take some special thought (thought it does for any tool). It is also good to do test restores to verify your disaster plan.
I use tar on many systems. My workstation is about 1TB of data. Backup is about 11 hours though I think it could be faster if I disabled compression (I currently use the standard gzip compression which is not optimal). I think the process is CPU bound by the compression at the moment. Going to uncompressed or using parallel gzip at level 2 is probably the fastest you can do and should really speed things up by 4X or more. I have played with this some for my wife and her raw backup is a lot faster now. My wife uses USB 3 external drives specifically plugged into USB 3 ports (the one with the SS symbol and the blue interior), and with a USB 3 related cable. I use 6TB naked SATA drives I insert into a hot mount enclosure and store in storage boxes. My backup system can theoretically do incrementals too, but it has some issues since I have moved to BTRFS so I do not use that at the moment. Did always use before. I have an idea how to fix, but need to debug and test incrementals now.
How often: I backup monthly. When my incrementals were working I use to do it weekly or whenever I got nervous. Other option for the BTRFS file systems would be to use their native backup tools. Not sure though, I like to use generic stuff. Lot to be said for generic.
Big downside of tar is the mind numbing man page. Getting the options correct takes some real thought. You also have to be comfortable with the shell and Bash scripting. Big upside you can customize exactly what you want.
tar dates all the way back to the 70s.
Yes, I actually did not know how far back, thanks. Wikipedia seems to say 1979. I know my system admin book dated 1992 talks about it and it was common then. I think my brother use to use it in the early 1980s for his job and maybe I did too a few times. Wikipedia says GNU Tar is newer and traces back to 1987. The formats have changed some and there are several. The PAX format is much newer which I think was standardized in 2001 but GNU Tar would have taken time to implement it. I do not know that date.
People seem to forget that tar worked well back then and still does.
I had the chance to play with late 70s Unix for a bit a few years ago. (Hardware on loan from a museum.) VERY minimal, but still recognizable. (Well, my Unix reflexes are old - I started in the mid 80s.)
Interesting. About then I was using a VAX. Somehow I spend most of my time on other stuff until I switched to Linux around 2000.
My first Unix was 4.3BSD on a VAX-11/750. (There was another 11/750 running VMS, but I didn’t like that nearly as much.)
I am using similarly “dumb” back-up system. I’ve two external USB HDDs, to which I copy my home folder every 4 to 6 months. The back-up folder currently has circa 250 GiB, but I don’t use any compression and I also probably do not have to back up my Steam library multiple times.
Yes, it doesn’t scale very well, but at the same time, I do not need to hoard 5 year old data. Yes, I should have an off-site back-up, but if my house burns down, I have bigger problems than losing my old photos.
What problem are you trying to solve? Please think about that, and about your backup strategy, before you decide on any specific tools.
For example, here are several scenarios that I guard against in my backup strategy:
- Accidentally delete a file, I want to recover it quickly (snapshots);
- Entire drive goes kablooie, I want my system to continue running without downtime (RAID)
- User data drive goes kablooie, I want to recover (many many options)
- Root drive goes kablooie, I want to recover (baremetal recovery tools)
- House burns down or computer is damaged/stolen (offsite backups)
For my Ubuntu desktop, I use the builtin backup tool to take backups on my NAS. For my homelab, I have everything running on Proxmox and my Proxmox backup server takes care of the homelab backups.
Just a reminder. Consider and test your restore process as well. Backups without restore testing are kind of questionable. Also think how the restore will go. Do you want to do a bare metal restore, or will you just reinstall, and restore certain things for example. Lot of these backup methods will not get a true bare metal restore set, nor can file system backups be “perfect” if they are done on a running system. Databases and things like cryptfs mounts for example can be problematic for example. Nor do all tools necessarily backup the full structure of the file system.
Not saying these are always issues, just be aware of them.
I just use a script on an systemd timer. Well two scripts on two timers really - one running daily, one weekly for different data. It’s just a bunch of rsync commands copying folders to an hdd in my system and I reroute the output into a simple log file, mainly to verify if it ran at all. I am a bit paranoid about that. I can also run it manually whenever I want. Oh and some of the data I also rsync again to a smb cloud drive from Hetzner. I do not keep multiple versions and I delete remote files that have been deleted locally. It’s just a 1:1 copy.
Oh and I use OpenSuse Tumbleweed so I have auto configured btrfs snapshots. Though I have not needed them yet and could not even say how I can use those. I figure that out once I need them.I’m currently using TimeShift to backup my desktop onto an external hard drive (the why is because of how simple it is to use) and I’ll be making a copy of anything I upload to my jellyfin server onto the external hard drive as well. I hope to eventually have a dedicated backup server and have a duplicate of it at a friend’s house for offside backup too
