It’s most useful for something like /home so you have a full backup of your home directories without having to worry if some app has dumped its settings on a folder you don’t have marked for backup. But also backing up / is useful because it gives you an easy recovery in the event of say disk failure. Just restore the entire system from a backup.
I use borg with borgmatic. I just back up / (which includes home) and exclude some folders I don’t want (like /mnt or /tmp).
It does the same as you just said.
I have 20 borg snapshots of my nearly full 1tb drive which takes about 400gb of space on my NAS.
I do it at the file structure level, not at the block device level as the article suggests. Why would I want to back it up at the block device level instead?
You can have btrfs subvolumes that contain only the stuff you care about. In the article, they have a subvolume for untracked files, and this doesn’t get backed up.
I have one for /home and /srv (my self-hosted stuff), and those are the ones I care about backing up. For everything else, I keep manual copies in either /home or /srv (e.g. edits to system configs and container configs are in a git repo). So I could back up those two subvolumes and get 99% of what I care about, and hope I’ve done a good job documenting the other 1%, which are reproducible w/ a bit of time searching how to configure stuff.
This keeps backups efficient since you can ask the filesystem to only send the changes instead of going over all the files and figuring out what has changed, so it’s probably a lot faster.
This keeps backups efficient since you can ask the filesystem to only send the changes instead of going over all the files and figuring out what has changed, so it’s probably a lot faster.
?? Maybe you don’t have enough data for it to matter, but when you have tons of files, manually looking through the filesystem can take a while, whereas the FS can keep track of what’s actually changed.
Backing up via snapborg allows you to see file structure, because actually it is a file-based backup. snapper here allows me to separate snapshot creation from actual backups.
that is a good point. Being about the mount the borg repo and pull out a specific file is very useful. Not sure I would do it as a bloc device since I don’t think there is an easy way to see specific files.
I don’t really understand the advantage of backing up the whole btrfs volume.
Snapshots are made atomically, so this workflow allows you to separate snapshot creation from actual backing up.
As subvolumes are dynamically sized, you can create as many subvols as you like and backup those, that need it.
It’s most useful for something like /home so you have a full backup of your home directories without having to worry if some app has dumped its settings on a folder you don’t have marked for backup. But also backing up / is useful because it gives you an easy recovery in the event of say disk failure. Just restore the entire system from a backup.
I use borg with borgmatic. I just back up / (which includes home) and exclude some folders I don’t want (like /mnt or /tmp).
It does the same as you just said.
I have 20 borg snapshots of my nearly full 1tb drive which takes about 400gb of space on my NAS.
I do it at the file structure level, not at the block device level as the article suggests. Why would I want to back it up at the block device level instead?
You can have
btrfssubvolumes that contain only the stuff you care about. In the article, they have a subvolume for untracked files, and this doesn’t get backed up.I have one for /home and /srv (my self-hosted stuff), and those are the ones I care about backing up. For everything else, I keep manual copies in either /home or /srv (e.g. edits to system configs and container configs are in a git repo). So I could back up those two subvolumes and get 99% of what I care about, and hope I’ve done a good job documenting the other 1%, which are reproducible w/ a bit of time searching how to configure stuff.
This keeps backups efficient since you can ask the filesystem to only send the changes instead of going over all the files and figuring out what has changed, so it’s probably a lot faster.
Aaaah!
?? Maybe you don’t have enough data for it to matter, but when you have tons of files, manually looking through the filesystem can take a while, whereas the FS can keep track of what’s actually changed.
With snapshots it costs nothing if the OS files don’t change much
You misunderstood my question, because what you said is true either way with borg.
The question is, what is the advantage of backing up the whole subvolume “block device” vs just / file structure.
Backing up via snapborg allows you to see file structure, because actually it is a file-based backup.
snapperhere allows me to separate snapshot creation from actual backups.Oh i see.
that is a good point. Being about the mount the borg repo and pull out a specific file is very useful. Not sure I would do it as a bloc device since I don’t think there is an easy way to see specific files.
deleted by creator