Oracle Talks Up Btrfs Rather Than ZFS For Their Unbreakable Enterprise Kernel 6

[S.Pam]

If you scrub after powerloss you are OK. But it is risky to have a double fault, say powerloss during writes and a broken disk. This could lead to dataloss.

[jegp]

I think the new raid modes of Btrfs (raid1c3 and raid1c4 ) are a kind of the new and safe replacement for raid 5 and raid6.

https://kdave.github.io/btrfs-hilights-5.5-raid1c34/

[zxy_thf]

But why is btrfs seen as so much inferior to ZFS? I mean, it seems to be well supported, it seems to have most features (ok, not all of them, but much more features than say ext4 and more than enough features for most users), it seems to be stable enough, LXD recommends ZFS but it offers more features over btrfs, and data losses are memories of the past... I really don't understand, but beware, I'm not using either one, I only use plain old ext4 for now.

Lacking of rock-solid RAID-Z/RAID-Z2 support forces NAS users to use ZFS over btrfs.

btrfs is actually, imo, in a stagnant state where
1. Enterprise(*) users don't like it because it's new;
2. Professional/Prosumers don't feel like to use it because their archive servers need RAID;
3. The vast majority features are not attractive to normal users because it's counter-intuitive to how they normally "use a hard drive".

RAID-Z is the killer feature for many users who have an array of hard drives and want a bit more over the classical file system,
but btrfs, for some reason I don't know, is really reluctant to stabilize this feature.

(*) "Enterprise" here refers to companies that would like to have IT service contact, not those few companies who can develop a next gen file system by themselves (e.g., Google).

[CommunityMember]

The NetApp vs Oracle suit was over patents .... If my memory serves me Sun sued first and NetApp only brought up the 7 patents as a defence mechanism.

As is usually the case with IP, the entire story is more complicated, apparently starting with discussions about licensing (apparently Sun thought Netapp owed them some money for using *their* IP) which eventually escalated to a lawsuit initiated by Netapp, and then counter claims by Sun, and ending in a whimper (lawsuit settled with all terms confidential). The general belief was that Netapp and (by then) Oracle cross licensed the various patented technologies (both have a large portfolio of filesystem related patents) used in their various platforms, and keeping their respective portfolios intact.

[Mark625]

.... Something that makes heavy use of available free space for normal function is bound to slow down as free space it requires, gets scarcer. If there's no mandatory free space to be found it will bog down drastically.

If that's the case, ("makes heavy use of available free space for normal function"), then it's not really free space, is it? The filesystem should reserve that 20% from the start, and report it that way in 'df' and other utilities. Basically, admins will need to add 25% more disk space (after subtracting RAID parity usage) to their initial configuration in order to maintain desired performance levels at planned capacity. That seems like a lot, especially when talking about enterprise-grade equipment.

[tuxd3v]

Sometimes Linux kind of goes towards the consumer desktop space (where it looks like it will never win). It will be good for other operating systems on that project to keep it in check and focused on servers.

that is what we have been experiencing from some time now, linux trying to go desktop, and ofcourse the server gets outofhand,
FreeBSD and others capitalize on that, and go server..were the money is..

[R41N3R]

I've never had any real issues with btrfs and I use it on all my devices, including a Raid1 for my home server. But you should be aware about the behaviour of CoW filesystem when using it. I still believe that Rad Hat was wrong when removing Btrfs support.

[skeevy420]

I've never had any real issues with btrfs and I use it on all my devices, including a Raid1 for my home server. But you should be aware about the behaviour of CoW filesystem when using it. I still believe that Rad Hat was wrong when removing Btrfs support.

The only times I've ever had problems were when using features marked as experimental and that GRUB/Zstd issue. Can't get too upset when experimental and new features are the cause of issues.

My preference is to use ZFS. When that's not an option I use BTRFS. When that's not an option then I figure the distribution isn't worth considering.

[intelfx]

I think the new raid modes of Btrfs (raid1c3 and raid1c4 ) are a kind of the new and safe replacement for raid 5 and raid6.
https://kdave.github.io/btrfs-hilights-5.5-raid1c34/

That's like saying that walking is a cheap and healthy replacement for a Mercedes S-class

It's cheap and healthy, sure, but it's not a replacement in any way.

[intelfx]

through volsize property.

https://docs.oracle.com/cd/E37838_01/pdf/E61017.pdf

Oh, this is absolutely not the thing that's being talked about What you brought is simply called quotas.

We were talking about actual underlying device resize and/or restripe. Like, with btrfs, I can create a raid0 or raid5 over four 2TB disks, fill it with some data and then progressively replace each disk with a 1TB one. Or maybe I want to take one of the disks out of the array completely. I can do that with btrfs, without any interruption in service.