Not a tech support question, I’m just curious. I recently installed it. Everything is working great, feels like I got a whole new laptop compared to my previous setup. I haven’t tried out any of btrfs’s unique features, so I dunno, nothing special I can report about it. Coming from Debian I was just surprized by how different Fedoras installer defaults are. Do you agree with btrfs being a default option?
BTRFS FUCKS; HARD
I think it was like fedora 32 or 33 they started using btrfs as default. You’re a lil behind.
I agree with it though I always have. Fedora has always integrated newer technologies into their distro, usually after opensuse still hahaha. But they were one of the first to default to Wayland and pipewire as well. If it wasn’t for distros like fedora and opensuse adopting these as default then, Linux on the desktop would be 10 years behind now
Does it do like openSUSE and create the subvolumes and snapshots for you? Because I dread configuring these freaking things from my days in Arch, and I fell in love with openSUSE when I realised it does it for you.
(And no, the
archinstallscript doesn’t autoconfigure BTRFS unless you wipe your entire disk)I was pleasantly surprised to learn this with openSUSE as well last week. It already saved my buttocks once when I messed up something with the greetd config file. Recovering from the latest working snapshot was easy. OpenSUSE is a very nice distro. Bought some merch to support them a bit.
Only @root and @home, no automatic snapshotting by default. I’ve been using “BTRFS assistant” to set them up
That’s been the default for years
The big reason is that btrfs has more features like copy on write, snapshots, subvolumes and data validation.
It used to eat data but that’s not been the case for a few years
Another great thing about BTRFS is that it can detect hardware problems sooner: if your BTRFS drive keeps losing data to corruption; that’s because it has detected a corruption that other FS’s would silently work with
It used to eat data but that’s not been the case for a few years
Isn’t that a RAID5/6 thing?
It used to eat data regardless even when it was supposedly stable
In newer kernels I believe raid5/6 are stable but the dangerousness thing is that it takes a huge amount of time to rebuild. I think this is true of raid10 as well.
I’m talking about the implementation of RAID5/6 for BTRFS specifically.
The RAID56 feature provides striping and parity over several devices, same as the traditional RAID5/6. There are some implementation and design deficiencies that make it unreliable for some corner cases and the feature should not be used in production, only for evaluation or testing. The power failure safety for metadata with RAID56 is not 100%.
Do you know if the documentation is outdated? Has this changed recently?
On Fedora, Btrfs has been the default for years now iirc. It’s modern and rock solid too (as long as you avoid Raid 5/6) and has some features I can’t live without nowadays:
- Copy-on-write (prevents file duplication)
- Snapshots (your systems broke? most easy rollback you will ever experience is with Btrfs in combination with Timeshift)
- on-the-fly compression (I’d recommend “–compression-force=zstd:3” as a mount option. Last I checked Fedora defaulted to using the lowest compression level, which is not the Btrfs default, making you lose some gains. FYI about the “force”: btrfs by default checks whether a file is compressible or not, this is redundant with zstd, which does the same thing but quite a bit faster AFAIK)
CoW also prevents corrupt files in the case of power loss
I wouldn’t want to live without bootable snapshots anymore. Other stuff like compression and dedupping is handy but it’s the snapshots that are a killer feature for me.
Been using it on arch for three years, pretty solid fs if ask me. Even started using it for all my external drives. Never had a problem.
You can scrub and balance your fs with
brrfs -assistant. Snapper is awesome. Still use deja-dup for a bit of daily backups of personal working files in $USER.The big struggle I see folks have with BTRFS is reclaiming space. However, that is with something like snap-pac and snapper setup.
The features of BTRFS are nice. Just a bit of a learning curve from my experience.
I run a desktop and laptop with BTRFS.
Fedora, like OpenSUSE that has btrfs by default, should be running scheduled scripts to do scrub, maintenance and dump old snapshots (either by number of snapshots or by age) on its own. Others having issue with btrfs are probably manually setting it up and not knowing the options or don’t have scripts that run on the back end. Having said that I do notice the partitions suggested for root are highly conservative, I usually add 30% on top of suggested root size and haven’t had issues in 7+ years
If you’re running btrfs manually and don’t setup clean up scripts I’m slightly confused how you get into trouble in the first place since that also means there won’t be any automated snapshots.
You can assign auto snapshots or create on demand, but whether or not you have a maintenance tool that does scrub, cleaning whatever is another story. I guess my point was something that has Btrfs as default install will also have some curation around the tools that optimize that system
I guess I just feel like if you’re manually configuring BTRFS you’ll either use it like a regular FS, or you’ll set it up to make use of the features in which case you’ll probably setup both automatic snapshots and cleanup. Possibly with auto scrubbing too. I don’t really see a situation where someone who manually opts to use it sets up snapshotting manually and then doesn’t setup any form of cleanup.
If you browse the webs so many people with “Help my drive is full of snapshots, what do I do?”. If there is a failure mode people will find it. Whereas a curated distro is OOTB ready to go without user intervention.
Oh, absolutely not saying it isn’t user misconfiguration. I need to revisit my autoscrub setup and my maintenance tasks.
Just something to be aware of.
I’m using universalblue, a Fedora derivative, that defaults to, and maybe requires btrfs, and it’s been good. Because it’s important for the immutability and atomic updates I kind of feel like I shouldn’t do anything custom with it. I haven’t seen performance issues or other weirdness. If I had a choice though I would still use btrfs. It makes backups and snapshots very easy.
Ostree should be filesystem independent
