FUTURE

Overhyped shit that got a beating before it left unstable. Should not be trusted on a reliable system.

Shame is that its in much better shame now (RAID1 has been fine for a long time) but apparently it'll never shake the bad reputation it got by being pushed too hard before it was ready.

btrfs + snapper is awesome

>slow
>still unstable
>doesn't do anything really special in the long run
"Future" my ass.

Based.

What's so special about btrfs?

Attached: gril.jpg (991x1024, 66K)

It's like ZFS, but with more flexible volumes instead of vdev shite, and it's licensed such that it is able to be directly in the Linux kernel source.

COW

Rude

I had problems when btrfs was new. I haven't had any in a long time except one: making many subvolumes and snapshots inevitably slows down the whole disk for some reason. Back when I used distros with standard package managers I thought it'd be a great idea to put everything on a btrfs raid, and keep my system files in a subvolumes which I snapshotted every time there was a package change. Ideally I could've rolled back any update that went bad or even configuration changes that I didn't like. Unfortunately, I found that keeping more than just a few snapshots slowed it down too much. On top of that, I had docker running on the same btrfs raid using the btrfs backend which made tons of subvolumes. These turned out to also kill performance. In the end I found that anything more than 10-20 subvolumes caused me problems. I came to btrfs with the impression that tens of thousands of subvolumes could be made with no performance impact (in fact, I thought there'd be a performance gains if the subvolumes meant keeping each subvolumes file tree smaller.)

These days I get the same functionality from using nix and lvm, and avoiding unnecessary subvolumes and btrfs snapshots.

The only other gotcha is making sure to set the +C attr (NoCoW) on files like databases and disk images. You're gonna have a bad time if you put a 10 gig virtual machine on btrfs with CoW turned on. Applications pretty much never know to set this flag when it's appropriate. If you're running a database or anything that uses a massive monolithic file like a VM or some games, I'd use a separate partition. With virtual machines, I learned how to use an lvm volume as the disk, which dramatically improves performance vs a large file because you cut out layers of every disk transaction.

problem is that nocow disables checksumming and therefore that nice self-healing stuff.

>You're gonna have a bad time if you put a 10 gig virtual machine on btrfs with CoW turned on.
I was going to do this. But it's an SSD, will it still have performance problems?

Depends very much on what you use the VM for and how it uses its virtual disk.

If you can I'd make a scratch partition and test it out and see how it goes

Are there any noticeable advantages in moving from ext4 to some other filesystem (btrfs, zfs, xfs?) for someone who uses PC mainly for programming, multimedia and some vidya?

>vidya
steam may not function properly on btrfs

It's also not as hardware-hungry as ZFS and scales better to more drives than ZFS.

(That is ZFS under Linux/BSD, I'm not actually sure if it didn't work a bit better under Solaris or not).

>But it's an SSD, will it still have performance problems?
It can work, but it's not a brilliant idea even then; I know from a multi-GB sqlite database.

You really might want nodatacow or a different filesystem.

Isn't f2fs the future since the future is without spinning rust?

Quite possibly not for you, no. Ext4 is a really rather good filesystem.

You still might want some variant on RAID5/6 on your backup storage, and that could be hosted on mdadm, btrfs, snapraid, zfs, ...

F2fs is okay, but it's pretty rigid. 16TB maximum volume size, 255 character filenames, and so on. Also it's not very space-efficient for smaller files, at least it isn't here.

Besides spinning rust is going to be here for a while. And maybe the future is more like Ceph, even for hobbyists (at least once that becomes easier to manage and more stable).

Can you elaborate?

I see. I'm not really itching to change my filesystem, I'm just exploring options. If ext4 is good enough, I'm fine with that. I do backups on 2 separate disks and I don't use RAID.

I usually just use ext4, is there any reason to use something else? I'm about to get a X1 E so I'll have nvme drives.

xfs is ancient, deprecated tech no longer used by anything other than RHEL
zfs is solaris pajeetware that even freebsd is finally moving away from
btrfs has massively improved but it still has an underserved bad reputation from its early days.

the one filesystem everyone agrees is excellent is HAMMER, but its stuck on dragonfly only

Crypt + btr FS + bcache

But it's still pain in the ass to set this up

and browsers
and databases
and virtual machines
it's useless

>btrfs
DOA MEME
>flif
DOA MEME
>av1
DOA MEME
>rust
DOA MEME

Why should I use this over ext4 for RAID 1?

Ñeñeñe

>*loads 100% of your CPU*
Nothing personal kiddo

I've never had problems with btrfs, not even once. Why do you guys suck at computerstuff and get so many problems? Just stop using a computer, dumbfucks.

My Synology NAS runs BTRfs and I have never had a problem.

What is all the fuss?

the absolute state of Jow Forums in 2019

More like dead on arrival ahah

when is it coming to windows 10

>bcachefs
>one developer

>However the future is obviously bcachefs and if you don't start using it when it becames stable, you will be a moron.
>when it becames stable

So in 2047?

Youre stuck with nigger technology faggot system

its already on reactos

it doesn't really work though.

>(((Oracle))) product
Into the trash it goes.

The future is bcachefs where you can cache writes to your spinning rust using your fast SSD
Imagine SSD speeds on your HDD

I'll stick with ext4

so... fusion drive? that's an Apple invention, you know