Is it worth switching to BTRFS for the average user?
I'm having trouble understanding all the benefits of BTRFS and how they'll apply to me.
Copy on Write and auto-compression seem like they will free up a bit of space.
What other practical benefits will I see from using BTRFS? Are there any noticeable performance benefits?
I use my computer to dual-boot. I don't need snapshots because I have a custom script for a fresh install. I use my PC for gaming and work. I've got an NVMe, two SSD's and one HDD.
Thanks in advance!
Transparent compression, snapshots, copy-on-write, deduplication, and data checksumming (and healing via read or scrub) are the main things you might notice as an end-user. BTRFS is mostly on par or slightly slower than other simpler filesystems for speed (can be faster on HDDs due to compression), so if you're using it you probably want to be taking advantage of the features it offers, not because it's going to be faster.
I thought compression would not help much with disk space as well. I believe it depends on the use case. After switching to btrfs and enabling zstd compression, my Arch install reduced from 100GB to 60 GB in terms of used disk space. Most of the savings are from documentations of development packages.
Just for reference
::: spoiler spoiler ~> compsize -x / Processed 699693 files, 766975 regular extents (791577 refs), 360356 inline. Type Perc Disk Usage Uncompressed Referenced
TOTAL 57% 39G 68G 69G
none 100% 23G 23G 23G
zstd 35% 15G 44G 45G
prealloc 100% 69M 69M 104M
~> compsize -x /var/lib/flatpak Processed 340412 files, 115619 regular extents (256345 refs), 209687 inline. Type Perc Disk Usage Uncompressed Referenced
TOTAL 55% 6.7G 12G 24G
none 100% 3.6G 3.6G 6.6G
zstd 36% 3.1G 8.5G 18G
~> compsize -x /home/user/.local/share/Steam Processed 219633 files, 1097250 regular extents (1111566 refs), 57457 inline. Type Perc Disk Usage Uncompressed Referenced
TOTAL 84% 249G 295G 296G
none 100% 203G 203G 203G
zstd 50% 46G 91G 92G
prealloc 100% 36M 36M 36M
~> compsize -x /home/user/.local/share/bottles Processed 18582 files, 33406 regular extents (33406 refs), 2366 inline. Type Perc Disk Usage Uncompressed Referenced
TOTAL 53% 1.8G 3.3G 3.3G
none 100% 959M 959M 959M
zstd 36% 907M 2.4G 2.4G :::
So it's 29G(43%) from / + /home, 5.3G(45%) from flatpak packages, 46G(16%) from Steam, 1.5G(47%) from Bottles, ~82G total out of 380G(22%) which is nice
As a (semi) power user I also use btrfs subvolumes to create "partitions" (single disk system, @root, @home, @docker), allows for making snapshots only for system or user data, etc.
All around, I love btrfs and I am never going back to journaling fs like ext4
If you don't think you want snapshotting then dont worry too much about it and use whatever fs you like. if you want a nice btrfs experience try tumbleweed or leap, it is already fully integrated with system and grub , and maintenance with cleanup etc. cobbling together btrfs on your own is fine too but if you miss sometging and mess things up it is usually user error not btrfs error. .
I use BTRFS simply because I run a rolling distro of Linux. For the average user, I don't think it is quite as necessary but the snapshots are nice. Of course, you could use timeshift to make snapshots as well.
I love btrfs. Mostly, because I had a couple of instances of data loss with ext4, and because it's far more user friendly than xfs. Btrfs has sane defaults, and when you do want to use the tooling, there's a lot of good documentation, and it's mostly straightforward.
That said, unless you have some reason to be mucking around with the fs or want to play with snapshots or such, then I'd say just use your OS defaults.
[This comment has been deleted by an automated system]
For me, BTRFS is worth it for the transparent compression and file integrity checks alone.
I am not sure what Fedora does for virtual machines via boxes (like transparently disabling COW), but one caveat with BTRFS is that you have to be aware that COW is a problem for usage patterns of virtual machines on their images. Just for this one example alone BTRFS is probably not a good fit for the average user.
Another thing which I just learned very recently, is to mount BTRFS volumes with the noatime option significantly speeds up a lot of workloads on my machine (machine has a decent SSD). Again, using the noatime option might create some problems in corner cases, so one should be aware what one is doing.
Yes, snapshots nice to have and compression reduces the disk usage for the operating system and software quite a bit.
This is hearsay or you didn't properly configure your snapshots, no way around that.
I found the "best of both worlds" setup is xfs for root fs and then btrfs for /home.
Is this arch?
Yeah, I assumed something similar. Arch (may be endeavor also) is very much a la carte build your own system, so coyld be some tweaks needed. I found Leap and Tumbleweed have no issues with btrfs because its integral to the whole distro. My NVME use on desktop and laptop have been going steady for 7 years.
There is a 2 page zypper cheat sheet online that is helpful for learning the commands, but they also have Yast2-GTK GUI you can launch; installing packages is just clicking the checkboxes of what you want and unchecking for what you dont want, or selecting certain package versions and click the lock icon. Then hit apply. The nice part of the GUI is you can see what files it provides and read descriptions etc all in one place. I agree, the installer is confusing because there are so many ways to configure your partitioning-either manually or with auto, or semi auto., It is a bit overwhelming at first, I was lost when first coming to linux, and reinstalled it a few times to suit as I learned what it (and myself) was doing. Now that I'm used to it it provides an amazing amount of power and flexibilty.
Also something not glaringly obvious is when you get to the summary page of the installer you can click the software title and it brings you to the package install section, you can uncheck any or all groups/patterns or click detail and choose individual packages that suit your needs.
Yes, a few examples of advantages here: https://lemmy.world/post/3197228
It you have a functioning dual boot don't worry about it. Next install it's worth looking in into and pretty easy to setup. It has trim support for ssds and is way easier to resize on a live system than lvm with another filesystem
Yes. I dual-boot and need everything to be accessible from Windows so I gave Windows 400 GiB while Linux only 75 GiB. Without compression I would be out of disk space. I think I hold about 95-100 GiB of data in 70 GiB. Snapshots are to revert broken updates, not to setup fresh installations. If you use a rolling release distro, they become invaluable. You can boot into snapshots if an update goes sideways.
I use ZFS, not BTRFS, but both have snapshots, and if you're someone who likes to tinker with their system like me, it lets you do so without worrying that it will break.
I use btrfs only for secondary and external drives, while keeping the main system on ext4.
Yes and no. It is good and I did one restore of some files that worked fine, but in my case it was noticeably affecting my boot up times, and I reverted to ext4 (boot ups were fast again to less than one minute). For some reason, BTRFS was resulting in quick login, but about 18 mins before my actual desktop was responsive after login. I spend many days trying to troubleshoot that. Maybe you won't have this problem. I had my SSD system drive on ext4 with Timeshift backups, and my /home partition on BTRFS.
So I'm back on my ext4 doing a daily automated backup to a second drive with rsync (LuckyBackup app). I think there are further kernel improvements coming to BTRFS later in this year. But I'll probably only retry it again end of 2023 or in2024. So if you decide to move, just benchmark your boot times, so you can judge if it affects them badly or not.
That said, BTRFS has some great features, lost no data for me, and I think has a great future.
Copy on write is likely to introduce significant performance decreases in cases where large or medium size files have a couple bytes changed. It’s usually recommended to turn CoW off on those files; I found it to be more hassle than it’s worth for a root filesystem. It is still a reasonable file system for file storage that looks more like archival - files land there and seldomly or never change. If you don’t have a specific need in mind though, I wouldn’t bother - in my opinion, it’s not great as a general purpose filesystem.
Do you happen to have a source or benchmark for this? My understanding of CoW is that the size of the file does not matter, as BTRFS works with blocks and not files. When a block is changed, it's written to a new location. All the old blocks that are not changed are not written again - this wouldn't even make sense in the context of how BTRFS deduplicates blocks anyway.
So:
10 kB base file
modify 1kB of the content
== 11kB total "used" space, and 1kB of new written blocks.
that old 1kB that is no longer part of the file will eventually be cleaned up if needed, but there's no reason to delete it early.
Why would OpenSuse/Fedora choose it as their default filesystem if it was as bad you describe?
Thanks for the advice!
How do you define medium or large files? What examples of these exist on root?