Linus Torvalds Doesn't Recommend Using ZFS On Linux

[Volta]

Hmmm, this doesn't shine a very positive light on Linus. I always thought he made sense, he knows what he's doing and what's going on but this clearly shows he has no clue.
It won't matter in practice because of the license but Linux is missing out by being so hostile against ZFS.

It's zfs that was hostile. Now it's nothing serious anymore. In the long term it's not an option like Linus said.

[timrichardson]

Hmmm, this doesn't shine a very positive light on Linus. I always thought he made sense, he knows what he's doing and what's going on but this clearly shows he has no clue.
It won't matter in practice because of the license but Linux is missing out by being so hostile against ZFS.

He was replying to someone who was upset at changes in his kernel daring to break zfs. The old Linus would have gone ballistic because that is an outrageous thing to say. He really has reigned in his temper based on this. zfs is not gpl2, you use it, it's on you, and that could not be clearer. And then he even goes on nicely to explain the obvious reasons why this is so.

[Almindor]

It won't matter in practice because of the license but ZFS is missing out by being so hostile against Linux.

There, fixed it for you...

[k1e0x]

Linus clearly doesn't understand the problem.

OracleZFS != OpenZFS in the same way MySQL != MariaDB
Oracle does not own OpenZFS. More than 50% of OpenZFS is not copyright to Oracle. It's copyright to the OpenZFS developers.

"Why would you ever want to use it in the first place?"

Oh I don't know, maybe because I like my data and want to keep it? Show me the Linux file system with 20 years of development, 10 years of enterprise reliability and feature sets like sending a encrypted dataset to a remote untrusted server in incrementals without decrypting it and without spending an hour calculating changes.

Where is that universal Linux file system I can use native on 5 different OS's? Linux, macOS, Windows, FreeBSD and Illumos? ZFS nearly has Fat32 level OS compatibility.

I think it's fine out of tree really. Clearly Linux doesn't want nice things and FreeBSD is more than happy with the users that Like and want ZFS I'm sure. ZFS doesn't even compete with Linux filesystems.. it competes with NetApp. (You know that thing that probably stores all your Redhat VM's.. wouldn't it be great if that was open source running Linux... no... ok then... I guess you can keep paying them then..)

[polarathene]

OMG, you think that too?

9 + 1 is not 10? 2 is bigger than 10 to you?

Learn about semantic version numbers, they're not a whole number, it's not decimal and that should be rather obvious with a bit of research(unless you're trolling for a laugh). Or do you zero pad numbers in general?(eg 09)

---

So, the real question is, why would you use BTRFS at all? ZFS is mature, stable and doesn't eat your data.

Btrfs is in the kernel but still under development and loses data.

Just curious, when was the last known case of BTRFS eating/losing data? What kernel was it? Or is this just periodic regurgitation of historical issues?

---

All I want is something that can detect bit rot on single drives, without causing some other kind of calamity.

You can get that with HDD and filesystems like BTRFS(and I assume ZFS) no? On an SSD it doesn't matter, or needs to be handled specially since having extra copies(of metadata blocks I think to detect bitrot), as SSDs can de-dupe internally such that if that data goes bad, then both "copies" will also be bad and thus bitrot would not be detected. Can't recall if it was detection or repair. Something like that.

[Kamilion]

Just curious, when was the last known case of BTRFS eating/losing data? What kernel was it? Or is this just periodic regurgitation of historical issues?

You can get that with HDD and filesystems like BTRFS(and I assume ZFS) no? On an SSD it doesn't matter, or needs to be handled specially since having extra copies(of metadata blocks I think to detect bitrot), as SSDs can de-dupe internally such that if that data goes bad, then both "copies" will also be bad and thus bitrot would not be detected. Can't recall if it was detection or repair. Something like that.

Uh, 4.14 about. https://btrfs.wiki.kernel.org/index.php/Gotchas

Although I've been running btrfs since about 2013 in RAID1. Lots of drive failures. 'btrfs restore' has gotten me back in short order every time. Haven't lost any data since switching from FreeNAS/ZFS5000 to UbuntuLTS+HWE/BTRFS. And I won't touch LVM/LVM2 with a 40 foot pole. Lost metadata and broken arrays too many times. No reasonable tools to fix LVM2 either; good luck using testdisk/imagerec against EXT4 in an LVM2 volume. XFS is great while it's running, but heartache to fix.

[pal666]

i wonder what is intersection of groups of braindead people using zfs and and brandead people hating systemd

[k1e0x]

i wonder what is intersection of groups of braindead people using zfs and and brandead people hating systemd

Here is a real use case.

I use macOS, FreeBSD and Linux.
I use removable drives from time to time to store data. (long term cold storage)
I'd like to access that disk from any of the OS's I use.
Those drives need to be encrypted.

What file system would you recommend for me? I use ZFS currently for this.

[polarathene]

Oracle does not own OpenZFS. More than 50% of OpenZFS is not copyright to Oracle. It's copyright to the OpenZFS developers.

Ok.. so what you're saying is that Oracle still has some copyright, which as Linus expressed concern for given Oracle's history, could lead to legal action? Would you be willing to take all financial liability and any repercussions if Oracle were to pursue legal action? You probably wouldn't, nor does Linus want to risk that on himself. Unless you're a lawyer and happen to know for sure that it's legally a non-issue to be concerned about.

Show me the Linux file system with 20 years of development, 10 years of enterprise reliability

Well... ZFS wouldn't count as that either as it's not been a linux file system for that long has it? Definitely not an official one. Features aside, you've got XFS, which at some point may be able to compete in that feature space too. I'm not against it personally, but I can understand why it's not being accepted for mainline inclusion in any form.

Where is that universal Linux file system I can use native on 5 different OS's? Linux, macOS, Windows, FreeBSD and Illumos? ZFS nearly has Fat32 level OS compatibility.

Really...Windows and macOS have out of the box support for it? It's not like BTRFS where you need to use a third-party project to make the filesystem support available?

For many, some of that OS support would have no value to them for selecting a filesystem, it could even be considered a detriment considering how much more complicated ZFS compared to an implementation for FAT32.

Just look at UDF and it's wide support across OS, but how poor that is(it's not really reliable to use cross-OS), BTRFS support on Windows isn't official via the BTRFS devs, it's someone elses implementation and isn't entirely reliable/compatible with sharing between Windows and Linux systems, known cases of corruption due to differences. Similar can be said about other filesystems with support on other OS.

If ZFS is unlike those, and has better support across OS that it's practically portable without problems, great!

Joliet, HPFS, NTFS all have wide OS support if you just want a filesystem for that claim.

[pal666]

Either Oracle or Microsoft (Members of Linux Foundations who pay Linux Foundation and hence pay Linus the salary) like it or not Canonical will be their competitor.

moron, canonical is a member of linux foundation who pays linux foundation and hence pays linus the salary. and oracle or microsoft will compete with redhat(ibm), not with some free cdrom mailing shop, which can't do much more than try to sell work of other people