Linux script questions - what do the options mean?

I have been using a Windows to Linux backup of urbackup for four years.

I am now setting up a Linux (unraid) server and running the linux client script. but don’t know answers to the install script. All I want to do is backup files, no “Images”.

Please select the snapshot mechanism to be used for backups:

3) btrfs filesystem snapshots (dattobd and LVM will automatically use btrfs snapshots for btrfs filesystems)

4) Linux device mapper based snapshots (supports image backups and changed block tracking)

5) Use no snapshot mechanism. Files will be backed up without creating snapshots of them. Images backups will be not supported.

Enter choice (number 1-5, then enter): 4

I picked 4? Was that right? I only want file backups of my four xfs disks.

Also

Convert root device into device mapper device on boot (initramfs)? This is required for root device/filesystem backup. [Y/n]

I guessed a no as I only want file backup?

I guess I picked the wrong option with “4” as I seem to have log errors from that.

|Errors|22/06/25 16:35|Creating snapshot of "/mnt/user/-data-" failed|
|Errors|22/06/25 16:35|File system fuse.shfs not supported|
|Errors|22/06/25 16:35|Creating snapshot of "-data-" failed.|

How do I fix this? Do I re-run the script and pick 5? Or can I change a settings file?

And talking of settings. Should I be copying anything from the client to a backup? Due to the way my server runs its file system in RAM nothing is kept between reboots unless I copy it manually. Should I be copying settings files after changes? Which ones should I be keeping?

It might be worth Googling things when you’re presented with a choice you don’t know.

Here’s the gist, anyway. These are all to do with your client’s capabilities and the setup decisions you have chosen, and it’s not something that’s necessarily easy to detect or choose good defaults.

They all have to do with the fact that your machine may be writing to files as it’s trying to do a backup. This would normally result in an inconsistent backup. The worst example of this would be databases, as they generally modify parts in the middle of the file and assume that nobody else is changing it, which will make it completely corrupted in almost all situations (unless the database engine is off at the time).

With that in mind:

  1. btrfs filesystems are copy-on-write (a.k.a. CoW). This gives them the almost unique ability to choose the version of a file at a particular point in time (a filesystem snapshot). That solves the problem because the urbackup client requests a snapshot and then reads from the snapshot rather than the current version at the point it got to the file. You also wouldn’t want your databases on there because every change results in a new copy of the whole file (almost, it’s block-oriented, but that’s still a lot of useless copying).

  2. Linux device mappers are a way for linux to abstract devices away from their locations through indirection. You don’t write to /dev/sda, you write to /dev/my/lvm/location and that points to where /dev/sda should be at that point in time. The most common of these is the Logical Volume Manager (LVM). What that means in practice is that a part of the drive can be left empty and then the filesystem is “frozen” while it makes a copy of the partition you’re backing up (a partition snapshot). urBackup then reads from the copy while you continue to use the original. Obviously this means that you need to set up LVM and keep enough space on your drive spare to make a copy of all your data at any given time. Sometimes, the process of snapshotting can make your machine slower or even impossible to use while the snapshot copying operation is taking place. LDMs are also mostly-agnostic to your filesystem because they’re below the filesystem layer, so you could use XFS or ext4 or even FAT16 and it would be able to snapshot it the same way.

  3. No snapshot mechanism works ok, too, but be prepared that your backups may be corrupted in parts. It depends on how sensitive you are to individual file corruption. Mostly this will affect logs but, like I said above, databases are the biggest headache here. Most other types of files will either be not worth worrying about, or will get backed up in a different interation of the backup anyway.

Converting your root device to a device mapper device is a bit ambiguous. I haven’t tried it because I do all my setups myself, but I believe it’s a convenient way of saying “We’ll do LVM for you.”

1 Like

Thank you @h6w, most helpful.

Google led to more confusions as references appear on sites who assume you are a Linux guru. I was trying to find answers relevant to urBackup. I could not find anything in the help files either.

The backups I am doing are of static media files sitting on xfs formatted disks. FLACs and MKVs. So none of them should be “open” like a database at the time of the backup.

After the first backup has run, giving log errors, I now think I need to change things to the plain “option 5” and avoid snapshots. Especially as I got errors about those snapshots.

Do you know if I can update the client config? Or will I need to scrap my first test and just re-run the script again?

Gives this a gentle bump.

Does anyone know how to correctly change the options on the Linux client? Windows I can make sense of as it is a GUI with a menu. My Linux setup I now want to change these snapshot options but have no idea how. The only guess is to attempt an install on top and see what happens.