Howdy,

I just completed a fresh install of FreeBSD 9 with ZFS root mirroring with 100% success. It's main usage is a rsync server to backup my home folder in gentoo.

Anyhoo, I was searching youtube for more info on managing ZFS on freebsd and I saw a video of someone demonstrating ZFS snapshot with ubuntu.

I'm not sure which type of ubuntu he was using i.e desktop or server edition. I though ZFS was not available for linux, at least not yet. Or maybe I have been under a rock and was unaware it is available for some linux distros.

If ubuntu has ZFS support, are there other distros that support ZFS as well?

I'll be much obliged

There are 2 options ZFS-FUSE (basically implemented outside the kernel) https://www.ibm.com/developerworks/l...zfs/index.html and an attempt to produce an in-kernel port http://zfsonlinux.org/.
Its basically a licensing issue; SUN's CDDL is incompatible with the GPL as far as the Linux kernel goes.

See also https://wiki.ubuntu.com/ZFS/

I'd love(!) to see it ported to the Linux kernel properly, preferably endorsed by RH.

I know what you mean. I am using it with freebsd9. ZFS rocks!!!

In a round about way. There is a Debian port that is using the FreeBSD kernal so you get both worlds.

I was impressed when I played with it on Solaris. Really has some advantages.

ZFS is one of the more advanced file systems out there. Would be nice to see it get a full featured Linux ports. I don't see why an amendment couldn't be made to the GPL license to allow limited use of CDDL licensed software provided all guidelines for the GPL license and CDDL licenses are meet equally and mutually.

When Solaris+ZFS was owned by SUN, they might have got around to opening the license after some persuasion, but now its owned by Oracle I wouldn't hold your breath.
I think we'll have to wait for the port by Lawrence Livermore National Laboratory above to be more fully developed and endorsed by eg RH.
I wouldn't want to use a FUSE FS on prod and I'd wait for it to be part of the std install from a major distributor for an in-kernel version.

Support? that needs further definition.

If you mean, "I want an enterprise distro, with a support contract, and want to be able to use that support contract to get ZFS issues fixed, using the normal support mechanism that I am paying for" then, AFAIK, the answer is no and is likely to stay no for the foreseeable future.

If you mean "I am happy to hack something, provided I can discuss the same with people on a forum (probably not officially supported by the distro), maybe with the help of a clear how-to document, and I don't expect anything formal from the distro itself, just informal discussions with other knowledgeable users" then chances seem a lot better.

If the second of those doesn't satisfy you (and there are certainly scenarios in which it shouldn't), I think that you should forget about ZFS. Although, except for performance and the kind of quasi-lock-in issue, ZFS is hugely impressive and it would be nice to have.

Also, be aware that BTRFS is an attempt to bring a ZFS-like filesystem to Linux natively, and, when mature, is really going to be a very worthwhile option (my current understanding is that there is some sort of BTRFS repair tool available, which was a show-stopper for many, it is just that it isn't fully capable yet...and I wouldn't yet be prepared to bet on all the corner cases being as well tested as you would like).

@ salasi

I never used ZFS before nor do I know anything about these licenses, I just begun using ZFS in freebsd9. When I saw a youtube video of someone using ubuntu and ZFS, I assumed ZFS has finally came to linux, but chrism01 pointed out it is probably a zfs-fuse implementation. If BTRFS is a native ZFS like filesystem, I can't hardly wait to use it in linux.

Last I recall, Gentoo and its derivatives (like Sabayon that's on the other laptop here) were able to install ZFS and use it. It's not a stock option out of the box, but it is indeed an option.

(Apparently, it is 'guess the word of the day' time; don't worry it isn't difficult to work out.)

ZFS-Fuse is indeed most likely, but the 'hack-and-ignore-the-licenses' solution is also possible. Neither of those is truly a desirable solution, but that's what there is at the moment.

Some of the more adventurous distros have gone down the BTRFS route already. IIRC, BTRFS was supposed to be the default FS type in the last two releases of Fedora (in that someone said 'next release we'll have BTRFS as the default'), although, in the end, they didn't quite go that far. It has become an option, but not the default (and, at least the first of those must have included the time when there wasn't any BTRFS repair utility, so I'll repeat the word adventurous).

Even Oracle, in Oracle Unbearable Linux, have BTRFS as an option (and the release was a few days before the repair tool existed in any kind of workable form!), which is more adventurous (that word, again) than I'd have expected an enterprise supplier to be.

That said:

• In effect, BTRFS was a somewhat Oracle sponsored project, in that Oracle employed Chris Mason, the lead developer; more recently Mason has left Oracle (and remember, this is the Oracle that, these days, owns Sun, who in turn own ZFS). I'm sure when the project started, Oracle would have been clear that having a ZFS competitor that they could use in their own Linux would have been an unambiguous good thing... Maybe their own interest in BTRFS made them more adventurous.
• ZFS really seems to be a solid piece of work; it is not just a filesystem, but it is also a series of utilities and tools to make it work. From a sys admin point of view, sometimes the most important thing is that you know that you can set up an storage system that does the job, reliably, and in a short time. If you know the ZFS system (ie, the tools as well as the filesystem), you can do that, and do it quickly and easily.
• As far as performance is concerned, the story is a bit more difficult. For BTRFS, there is a reasonable amount of testing around, and you can say that in some scenarios the performance is competitive, in others it is some way off the pace. How that works in a real-world scenario is a bit more difficult to asses, but one test (Distrowatch, Fedora) using BTRFS as the only FS resulted in a system that was noticeably sluggish, and that was cured by re-installing with ext4 (I think).
• BTRFS is not yet that mature, and every two or three kernel versions seems to be accompanied by either a (small) step forward or a (small) step back in performance. In addition, BTRFS is more difficult to configure optimally than more traditional filesystems, because it has more options (compression, space cache, etc), and, of course, more options means more difficult to test adequately.
• SUSE is doing some interesting work on using BTRFS with installing/removing software, so that, in principle at least, you can revert any system changes that you have made with the standard system management tools (ie, you don't go around the system tools with your own taballs). Apparently, you can even revert to conditions that system never was in, by reverting individual changes, which sounds mildly dangerous and mind-bending, but is probably necessary for this to be useful. Other distros are going down a similar path, but as far as I can tell, SUSE is the furthest down that path.
• The only 'level playing field' (in as much as there is one) test that I saw of ZFS vs native Linux file systems (Phoronix), ZFS didn't do too well on performance. While BTRFS had some good results and some less good results, ZFS more went with the less good results.
• That said, ZFS has real abilities in the area of RAID arrays and in multi-level storage systems that utilise the abilities of SSDs (and ZFS can make good usage of SSDs) and the simple testing on Phoronix was never likely to explore those options, because it never could constitute a level playing field, and the test program would have grown exponentially.
• And, to repeat, sometimes performance (access times, bandwidths) is not the only thing. Well, really, that type of performance is not the end of performance (reliability and robustness, costs (ML storage arrays), power consumption, time to configure) are all performance, too, but don't get measured in benchmarking tests.

My opinion (based on an amount of reading, and zero testing, so far) is that there are definite uses for one of these more advanced file systems, but just right now 'give the installer this big partition, and let it use it all for BTRFS/ZFS and let the installer put everything there' just isn't it, at least so far.

Much obliged for that detail explaination on tne status of ZFS and BTRFS in linux.