The problem with btrfs subvolumes is that you have to use btrfs. I’m good with Ext4. It’s nice and reliable.
I wanted btrfs for ages for snapshots… Then I moved to nixos with Declarative config… Then I needed to move partitions to extend my drive into old root, decided to move to btrfs from ext4 cos why not at the time… Only afterwards do I realize snapshots aren’t that helpful when you’re on nixos… Or am I missing something?
Transparent fileystem compression and deduplication (btrfs feature not in ext4) compresses data while still having it be accessible normally. This leads to big space savings.
You can use the tool
compsizeto check it out.They are still useful for user data. For the nixos system itself, not so much.
btrfs subvolumes are cool, but I’ll still take partitions for separation. The whole point of having a separate home partition is that it’s separate from / if we have to, say, reinstall the OS, or the filesystem breaks or whatever.
sure it might be possible to install an OS into a btrfs subvolume without wiping other subvolumes, but do I wanna risk it? Nah.
– Frost
I have installed an OS onto just the btrfs root subvolume, leaving the home directory intact. This is how I originally swapped from Manjaro to Arch. The arch manual install instructions helped.
But this should be a feature of the graphical installers imo.
sure it might be possible to install an OS into a btrfs subvolume without wiping other subvolumes, but do I wanna risk it?
There is neither a “might” nor a risk. Subvolules behave like partitions with the only actual difference being that they are not fixed in size but all share the same space.
That… Is literally how you do it. You install the system onto a subvolume. Or many, in fact - the way I do. Root, var, srv, home, opt all get their own subvolume. Only boot stays as a separate partition.
Ignoring
/boot, what is the benefit of putting everything else in different subvolumes? As opposed to just one subvolume for/and one for/home, which is what I currently have. It just looks to me like it’d be extra work, but I’m probably missing somethingYou can snapshot them independently. E.g. I snapshot
/on every update and boot,/homeevery boot, and temporary file directories such as/tmp&/var/tmpdon’t get snapshot at all and are also mounted withnodev,nosuid,noexecflags.For one thing if you’re snapshotting your subvolumes for backup purposes then it will ignore nested subvolumes. Eg you dont want to preserve snapshots of huge installed game directories, cache directories, logs or docker / podman images, usually. Saves a lot of space.
Also you can tune copy on write on a subvolume, which is great for some use cases for performance.
You snapshot them separately, with snapshots stopping where another subvolume starts. Have a problem booting? Copy the latest /-snapshot. But if for example /var/log is a separate subvolume it persists and you can look up what was wrong.
My setup right now has subvolumes
root(mounted to/),home(mounted to/home) ,logs(mounted to/var/log),snapshots(mounted to/.snapshots),pacman-cache(mounted to/var/cache/pacman) andswap(for the swapfile obviously).I do snapshots of
root,homeandlogs(landing insnapshots) regularly that don’t require much space (only the difference between on subvolume and its snapshot uses real space in btrfs) and which can all be restored together as well as separately, while losing the temporary data incacheandswapis not problem.And you can also transfer the snapshots somewhere else as backup (into another brtfs filesystem or as a file), including just transfering the difference from the last one as an incremental backup.
Might be possible? It’s not only possible but rather easy to do, in fact I often have multiple installs side by side in different subvolumes and I can switch between them, reinstall as needed, snapshot them. The only partition I have separate is /boot because it has to be vfat for the esp. Tbh accidentally blowing up a subvolume isn’t all that much more or less difficult to do than accidentally blowing up a partition. IMO if you’re paranoid about wrecking something the only safe separation is multiple drives and disconnecting the one you want to avoid touching. I keep home in a separate subvolume specifically for the separation you mention, it can persist across different installs because it’s separate. Subvolumes are fully isolated from most failures with the exception of filesystem corruption.
I like to create partitions with specific max sizes, you can’t do that with btrfs subvolumes unfortunately. It is possible to set a quota but that can have a major performance impact with btrfs so it is not really a valid Option.
Thats why I shy away from using btrfs subvolumes.
I always do LVM now. Virtual partitions are the way to go, personally.
The amount of times that the root partitions have been 20-25GB but have still gotten completely full such that updates break… Well I could count it on one hand but still difficult enough that I would rather just use virtual partitions, especially on a server.
I’ll take ZFS datasets, thank you very much.
Does btrfs still slow down the system until freezing, when doing snapshots of the sys subvolume? I never used btrfs again. Imagine, you set a cron for hourly snapshots, and what you get is 2 minutes of freezing every full hour 🙈 I never had that kind of trouble with other snapshotable file systems as sys volume.
Never experienced that before. I have been on openSUSE for 9 years now, snapshots are pretty much instantaneous.
This can happen when gquotas are enabled, thats why gquotas are not recommended with snapshots
I’ll dig into that.
Never had this. It’s pretty much instanteneous.
I may try again some day, then…
deleted by creator










