ZFS FreeBSD moving to ZFS-on-Linux

Beastie7

Well-Known Member

Thanks: 152
Messages: 386

#26
It's nice of you to assume I hadn't read the entire thread before weighing in - I won't entertain your snark though. Ports of ZFS to other platforms came from upstream changes in illumos. You don't need to reaffirm what I'm already aware of. Thank you.

Anyway, we'll see what the future holds for FreeBSD.
 

Hakaba

New Member

Thanks: 7
Messages: 19

#27
it explicitly says that those same developers will continue to work on ZFS, just in the different repository, trying to be OS-agnostic.
I think a good way to turn the OpenZFS chapter is to rename ZoL in ZFSon*X (or something like that, I prefer LibreZFS like the story is similar to OpenSSL/LibreSSL ?).
Or use the name OpenZFS, but it can be confusing.
Let Linux in the name means linuxisme is ok, no ?
And what about the license ?
 

cynwulf

New Member

Thanks: 15
Messages: 18

#28
With a Linux centric upstream, isn't it feasible, when considering other projects ported from Linux, such as KMS/DRM, that FreeBSD could end up in a position of perpetually playing catch up, e.g. lagging behind?

As I understand it the reference implementation was the upstream, now the Linux downstream is to be the FreeBSD upstream, with some verbal guarantees of Linux devs being mindful of FreeBSD compatibility when they commit changes?
 

aht0

Active Member

Thanks: 36
Messages: 115

#29
I'm not sure that HAMMER/HAMMER2 integration is that simple. Requires some drastic changes to the kernel as I recall.

(edit: though I remember reading that HAMMER2 addresses some of this?)
Matthew Dillon himself commented on this somewhere, claiming that porting it would be rather hard because some of the H/H2 code is very DragonFly-specific. I expect him to know about it.
 

olli@

Active Member
Developer

Thanks: 144
Messages: 156

#30
I could see a future where for pure reasons of efficiency and cost savings one might want to just use ZOL on Linux. Would there by any advantage to using it on FreeBSD once both ZFS implementations are at parity?
Of course, quite a few. For example, FreeBSD's VFS is ARC-aware (Linux' isn't), and FreeBSD's TRIM support is much better than Linux'. And yes, all of that will be kept when we switch from Illumos to ZoL. The discussion on the mailing list has made it absolutely clear that no regressions will be accepted.

Nowadays, the active development of ZFS does not happen on Illumos anymore – instead, ZoL is the leading development branch within OpenZFS. So, by switching over, FreeBSD's ZFS support can only get better.
 

tingo

Daemon

Thanks: 359
Messages: 1,915

#32
And (if I understand this correctly) that is exactly what is going to happen - via the developers that make ZFS work in FreeBSD. You (as in "you all") don't have to trust in Linux, you only have to have faith in (our) ZFS developers.

Of course, if it for some reason doesn't work out they can always switch to something else, or switch back.
The worst that can happen is that FreeBSD doesn't get any new features from outside
The best that can happen is that ZFS development get's all the new features directly from the main development branch of ZFS today. Wait - that is ZoL, right?
 

grahamperrin

Active Member

Thanks: 26
Messages: 111

#33
… trying to undo any Linuxisms …
Think less of undoing; think more of #ifdef statements.

Somewhere this week I read clear guidance about distinguishing code that's to be for FreeBSD. I didn't bookmark it, unfortunately; if I find it I'll revisit this post.
 

Sevendogsbsd

Well-Known Member

Thanks: 115
Messages: 271

#34
This is a good point and I am one of those, too, who has never used ZFS.
I only use ZFS on my build "server" because it has spinning SAS drives and poudriere supposedly works better with it. On my workstation, I need trim support because I am 100% SSD so use UFS and it works very well.

ZFS is not why I came to FreeBSD either: I came here because the Linux camp is a mess + systemd, and I appreciate that FreeBSD is designed as a whole by a team and not kludged together with duct tape and bailing wire. Not disparaging Linux because I appreciate its "rebel" roots, but it has strayed so far from that as to almost be unrecognizable.
 

rigoletto@

Daemon
Developer

Thanks: 794
Messages: 1,717

#35
They are not exactly totalitarian but more like parasites, if they can't suck your code into GPL they can't drain it out, and they hate this situation.

[EDIT]

I mean, if Sun had released ZFS written in a super tied way with Solaris, like DragonFlyBSD Hammer v1, to make clear they don't want it running in any other system, Linux would have ZERO problems to suck it if the ZFS license was GPL.
 

phoenix

Administrator
Staff member
Administrator
Moderator

Thanks: 1,196
Messages: 4,034

#36
It's better / more accurate to see this change as "OpenZFS migrating to ZFS-on-LInux" rather than "FreeBSD moving to ZoL".

Currently, the illumos source tree for ZFS is the "canonical" upstream source repo for OpenZFS. Other projects take that as the starting point, do their own developments, and then request their features be added to the illumos-gate repo. However, there's a lot of politics and waiting around to get those changes into illumos-gate. Meanwhile, there's lot of new development happening outside of illumos-gate that's not getting pushed back upstream, and not flowing back downstream to others. This is leading to a collection of not-always-compatible implementations of ZFS on different OSes. (People complain about Android fragmentation all the time ... well, this is starting to happen with ZFS.)

So, the OpenZFS folks are looking to rebase the "canonical" upstream source repo on the ZoL sources. Then split that OpenZFS repo into "OS independent" and "OS dependent" directories for each OS (FreeBSD, Linux, Delphix, Illumos, Windows, MacOS X, etc). And moving to a more open / standardised way for developing features in the downstream OS repos and pushing those changes upstream, in order for other projects to get access to them.

This will still be done under the OpenZFS umbrella. It's still the same set of developers working on things. And there's work underway to get the FreeBSD-specific features (like TRIM support and NFSv4 ACLs) into the ZoL tree. Nothing will be changing in the FreeBSD releases until there's feature and stability parity between "old" ZFS and "new" ZFS (which won't happen until at least 13.0).

So, there's no reason to get your knickers all in a twist.

For more info on this, check out the OpenZFS website, the Monthly ZFS Leadership Meeting minutes and videos, and the Feature Ownership spreadsheet. It's an exciting time to be a ZFS user, as there's a lot of nifty features coming down the pike. Features that would take years to get into FreeBSD if we had to wait around for things to get into illumos-gate first. :)
 

ralphbsz

Daemon

Thanks: 839
Messages: 1,367

#37
So, there's no reason to get your knickers all in a twist.
ABSOLUTELY! Thank you for saying that clearly.

There are a lot of Windows- / Microsoft-Haters in the open source community. OK, for historical reasons I can understand that, but it is simply WRONG, both for ethical and for technical reasons. Today, hating Microsoft is obsolete. But what really disturbs me: how many Linux haters there are. Clearly, Linux has some problems (the interaction style of senior developers, starting with Linus himself, the coding style, systemd, and so on). But hating Linux, and assuming evil motives behind everything it does, and guessing that everything that Linux gets close to will be damnation is just stoopid. Don't do that. If you want to be a lemming and follow group pressure and emption, just climb up on a cliff and jump down, and leave us alone.

In my opinion, the rebasing of ZFS for the various FOSS operating systems is a good thing, but a difficult thing. Let the developers deal with it. They are competent and responsible people.
 

Beastie7

Well-Known Member

Thanks: 152
Messages: 386

#38
If you want to be a lemming and follow group pressure and emption, just climb up on a cliff and jump down, and leave us alone.
I can equally say the same thing for apologists who marginalize the harm said companies and/or developers have done to the greater software ecosystem. It's almost as if you haven't been paying attention, or you choose you roll with whatever status-quo is being shoved down our throats.

Language like this is even more harmful than what we criticize here.
 

Sevendogsbsd

Well-Known Member

Thanks: 115
Messages: 271

#39
There are a lot of Windows- / Microsoft-Haters in the open source community. OK, for historical reasons I can understand that, but it is simply WRONG, both for ethical and for technical reasons. Today, hating Microsoft is obsolete
Microsoft and product hater since 1993. Since Microsoft has been under their new CEO, instead of their previous nazi bully, the company has seemingly embraced open source. No clue whether to trust them but it doesn't matter to me because IMHO, their products are junk and I don't have to rely on them because I choose not to. No opinion is wrong or obsolete because it is subjective and personal.
 

drhowarddrfine

Son of Beastie

Thanks: 1,038
Messages: 2,937

#40
I don't trust Microsoft because they haven't earned my trust. I don't trust Linux because their technical direction is off base. These people may know what they're doing but what they're doing is in their own interests and not mine...technically speaking but as businesses, too.
 

Trihexagonal

Daemon

Thanks: 725
Messages: 1,295

#41
But their Optical Trackball Mouse is the epitome of Microsoft Excellency in Engineering!

Now you, yes you, can buy one at newegg.com for the low, low sum of only $599.

bargains_galore.png

I have 2 and could make somebody a heck of a deal on that, and unlike newegg I'll cover shipping. How can you afford not to buy it?
 

Crivens

Moderator
Staff member
Moderator

Thanks: 723
Messages: 1,678

#42
I hear that ZoL is hitting a large snag on Linux with versions 5+ due to some kernel interface being allowed for GPL code only. So maybe this simply makes it easier for the devs to come to FreeBSD?
 

ralphbsz

Daemon

Thanks: 839
Messages: 1,367

#43
You are being optimistic. The new of the death of ZFS on Linux are premature.

It is true that Greg K-H (who is one of the top decision makers about the Linux kernel) has had a Linus-style tantrum about ZFS using kernel interfaces that are obsolete. In reality, his tantrum is not about obsolete interfaces, but about the fact that ZFS uses a different license, which he doesn't like. But because ZFS relies on fast vector operations to calculate checksums and RAID parity, and probably because its checksum format is different from that supported in the Linux kernel (not 100% sure on that), ZFS needs to use special kernel interfaces. Whether Greg K-H likes it or not.

My expectation is: Greg K-H's salary is paid by a certain entity (we don't need to go into the details of that), and that entity is funded by certain other entities, which is in merger talks with yet other entities. Those other entities make really good money by selling Linux to people who need ZFS, because there is no free Linux native file system / RAID layer with comparable functionality and quality (neither BtrFS nor XFS are anywhere close). Greg K-H will either change his bad attitude towards ZFS, or he will become unemployed and lose the ability to make decisions about Linux. This will take time; Rome wasn't built in one day. Patience please.
 

ralphbsz

Daemon

Thanks: 839
Messages: 1,367

#45
Thank you for finding that! The Heise article is mostly a german translation of Greg K-H's e-mail, with a little background info added.

If the Linux people want to cripple ZFS on Linux, they are stupid. Stupid people don't make money. The future of Linux is driven by money (as is the future of everything on the planet, fortunately or unfortunately).
 

PMc

Active Member

Thanks: 39
Messages: 117

#46
ABSOLUTELY! Thank you for saying that clearly.

There are a lot of Windows- / Microsoft-Haters in the open source community. OK, for historical reasons I can understand that, but it is simply WRONG, both for ethical and for technical reasons. Today, hating Microsoft is obsolete. But what really disturbs me: how many Linux haters there are.
I gave up on linux in 1995. For quality reasons.
It has not improved.

I don't hate Microsoft. I ignore it, it's not an OS.

If the Linux people want to cripple ZFS on Linux, they are stupid. Stupid people don't make money. The future of Linux is driven by money
Those people get paid?? Thats cute...

(as is the future of everything on the planet, fortunately or unfortunately).
Kali Yuga Surf Club.
 

rigoletto@

Daemon
Developer

Thanks: 794
Messages: 1,717

#47
Thank you for finding that! The Heise article is mostly a german translation of Greg K-H's e-mail, with a little background info added.

If the Linux people want to cripple ZFS on Linux, they are stupid. Stupid people don't make money. The future of Linux is driven by money (as is the future of everything on the planet, fortunately or unfortunately).
I suppose they are willing to impose bcache as the new best FS ever.
 

semi-ambivalent

Active Member

Thanks: 3
Messages: 124

#48
I suppose they are willing to impose bcache as the new best FS ever.
From your link: "Bcachefs is the hot new thing: a next generation..."
A long time ago I saw this on some BSD vs. Linux argu-thread:
Linux - All the hot new s**t right away.
BSD - When it absolutely, has to work, every time.

I was already on FreeBSD (anonymous ftp install = magic) but if I had not been I would have changed, just on the audacity of that remark.
 

pyret

Member

Thanks: 27
Messages: 50

#49
neither BtrFS nor XFS are anywhere close
After 10 years of btrfs futility, it is pretty much dead. From Redhat:
⁠Btrfs has been deprecated

The Btrfs file system has been in Technology Preview state since the initial release of Red Hat Enterprise Linux 6. Red Hat will not be moving Btrfs to a fully supported feature and it will be removed in a future major release of Red Hat Enterprise Linux.
The Btrfs file system did receive numerous updates from the upstream in Red Hat Enterprise Linux 7.4 and will remain available in the Red Hat Enterprise Linux 7 series. However, this is the last planned update to this feature.
 

Beastie7

Well-Known Member

Thanks: 152
Messages: 386

#50
I suppose they are willing to impose bcache as the new best FS ever.
This will go through another hype cycle, then fail into irrelevancy like Btrfs.

Supposedly Red Hat is developing (yet another) storage project called 'Stratis'. Touted as 'next generation' storage. These re-inventions just don't seem to end don't they?

From your link: "Bcachefs is the hot new thing: a next generation..."
A long time ago I saw this on some BSD vs. Linux argu-thread:
Linux - All the hot new s**t right away.
BSD - When it absolutely, has to work, every time.
Ironically, no systems innovation came from Linux, and a lot core technologies in Linux came from Red Hat acquisitions.
 
Top