login
Search
Dev's.
What are the chances of getting a port of ZFS to OpenBSD? I can't
quite bring myself to run solaris since it lacks so much of what I
love about OpenBSD and Linux is back to square one because of the
reasons I moved to OpenBSD.Khalid
I aked Theo about this at OpenCon 2008. To cut a long story short...
No
--
Best Regards
Edd
Have you ruled out FreeBSD?
Why are so many people so hot for ZFS? From what little I've read
about it, it sounds very complex, which means bugs and a nasty learning
curve. Not something I'm interested in trusting my data to.
If you begin to read more than a little about ZFS, you'll understand why
it's valued so highly.--
Jason Dixon
DixonGroup Consulting
http://www.dixongroup.net/
Then again, you give the zfs a name, and throw it a list of raw devices
and you can have that filesystem 'newfs'ed, writeable, nfsexported and
running in a more than usual fault-tolerant raided mode in mere seconds
regardless of size.Some admins value that. (except the nfs part, but it's optional)
I like nfs!
The amount testing they told us about is pretty incredible. The
design looks pretty interesting. After a couple of years of testing
in the wild, there hasn't been too many serious horror stories. The
one bit flipflop starts becoming more common at the terabyte level.And it is pretty sexy to bit.
--
http://www.glumbert.com/media/shift
http://www.youtube.com/watch?v=tGvHNNOLnCk
"This officer's men seem to follow him merely out of idle curiosity."
-- Sandhurst officer cadet evaluation.
"Securing an environment of Windows platforms from abuse - external or
internal - is akin to trying to install sprinklers in a fireworks
factory where smoking on the job is permitted." -- Gene Spafford
learn french: http://www.youtube.com/watch?v=j1G-3laJJP0&feature=related
Given the Dev's have answered this one before, and have better things to
do, I shall take it upon myself. I'm sure they will correct me if I'm
mistaken.The ZFS code is under a license which the OpenBSD team have deemed
incompatible with the BSD License they use. [0]Whilst there could be a FUSE-based implementation of ZFS on OpenBSD, and
indeed I think one may have already been started, to properly take
advantage of the strength of ZFS the code would have to be in the
kernel. Performance will suck if nothing else.If its not BSD-licensed code, its not going in the kernel. End of
discussion. Said policy has been a universal truism of OpenBSD since it
began.I would suggest, if you want XFS in OpenBSD[1], set about persuading Sun
to re-release the ZFS code as BSD-licensed. Indeed, I think given it
would have to be kernel code, I really do mean BSD-licensed rather than
BSD-license-compatible.And when you do, be sure to remember to buy Satan some mittens.
--
SD
[0] http://marc.info/?l=openbsd-tech&m=110806948606417&w=2
[1] Which I personally would love to see.
The hammer FS seems promising from the BSDtalk Will & Matthew did.
Sevan / Venture37
Outside of a single person who's doing porting (to an unknown OS),
there's not been much in the way of updates on the status. It's a BETA
filesystem at best, and still being tested with the 2.0 release of
DragonFly.2.1 seems promising, but HAMMER doesn't seem as well developed as one
might hope.http://gitweb.dragonflybsd.org/dragonfly.git?a=search&h=HEAD&st=commit&s...
FWIW, there's no license restriction I spotted in DragonflyBSD that
would prevent it being ported, if one were motivated to.
Extracted from http://www.dragonflybsd.org/hammer/index.shtml:
"If you are interesting in porting HAMMER to another OS, please drop me
a line at dillon at backplane.com. I will be creating a new DragonFly
mailing list specifically for HAMMER porting as well as a git or
mercurial repository (I haven't decided which yet) separate from the
DragonFly repository. "--
Thanks,
Jordi Espasa Clofent
HAMMER is a completely different beast. From my understanding, it
attempts to avoid failures through clustering. But it doesn't have the
resiliency to avoid data corruption, nor the recovery functionality, of
ZFS.There was a thread recently on dragonfly-users that covered much of
this.http://leaf.dragonflybsd.org/mailarchive/users/2009-01/msg00058.html
--
Jason Dixon
DixonGroup Consulting
http://www.dixongroup.net/
On Thu, Jan 15, 2009 at 10:55 AM, Dave Wilson
I have to ask - if you're not copying the code, but only copying the
concept/technical requirements over (ie, a rewrite), that new code
would be bsd licensed, right?--
http://www.glumbert.com/media/shift
http://www.youtube.com/watch?v=tGvHNNOLnCk
"This officer's men seem to follow him merely out of idle curiosity."
-- Sandhurst officer cadet evaluation.
"Securing an environment of Windows platforms from abuse - external or
internal - is akin to trying to install sprinklers in a fireworks
factory where smoking on the job is permitted." -- Gene Spafford
learn french: http://www.youtube.com/watch?v=j1G-3laJJP0&feature=related
Yea but sun have patents over ZFS `concepts` [see wikipedia, google].
And the CDDL grants you a license. so you can't use another.
As marco already stated, it could be a kernel module. But it won't.
Why? Because nobody will write it. End of discussion, the rest is
noise.P.S. Personally, as much as I'd love to see ZFS in OpenBSD, I think
dtrace would be much more useful.--
Jason Dixon
DixonGroup Consulting
http://www.dixongroup.net/
Jason Dixon wrote:
> As marco already stated, it could be a kernel module. But it won't.
> Why? Because nobody will write it.Who is nobody anyway? I see he has an account on quite a lot of
computers, but I've never met the guy himself. He must be extremely
lazy if you're already saying that it won't be done *because* nobody
is writing it; this sounds like it should offend nobody very much,
possibly to the degree that he will write ZFS support into OpenBSD
simply to despite you.Where does nobody live? I'd love to buy him a beer!
I refer you to
http://www.sing365.com/music/lyric.nsf/Nowhere-Man-lyrics-The-Beatles/A4...
and http://uk.youtube.com/watch?v=dHLjYBsl2zAAnyway, we've already had a fairly conclusive ZFS answer. It can't be
integrated into base OpenBSD because the license is incompatible. The
license is
unlikely to change. Implementing it from the specifications has licence
issues, is hard and bug prone.A kernel module can use the code - all it needs is someone to volunteer
(sorry, not me : not interested). There's then nothing stopping someone
creating
a minor OpenBSD fork with the integrated module ZFS support, if they
need it from installation time that badly.PK
He's a french guy. I've seen him during p2k8 in Budapest ;-)
(did anyone take a picture of him?)
Ciao,
Kili
*sigh*
--
Antoine Jacoutot <ajacoutot@bsdfrog.org>
If you really really want ZFS, run an OpenSolaris box. It's not going
to kill you.It's clear there will be no ZFS in OpenBSD. It's not a priority of the
developers.If running an OpenSolaris box is against your ethics, religion, or
common sense, then no ZFS for you!BTW, Everything I've read about ZFS on FreeBSD says that it sucks, at
least until FreeBSD 8.0 is out. Panics under high load are common.Right now, the most complete and reliable ZFS experience is going to
be on OpenSolaris. I'm running Nevada Build 105 and have no problems.
"ZFS-like" functionality could be added to softraid. This would be more
realistic.--
Best Regards
Edd Barrett
(Freelance software developer / technical writer / open-source developer)
I'm using ZFS (two pools - data and root) on FreeBSD 7.1 and it seems
to be quite stable. The load is not too high but there's no problem
when compiling entire xorg and xfce with a qemu session running in
background.--
O< ascii ribbon campaign - stop html mail - www.asciiribbon.org
I doubt that nobody is french ;)
--
Gilles Chehade
http://www.poolp.org/~gilles/
Please, contribute to my happiness ;)
http://www.openbsd.org/want.html
About the same as porting anything substantial to OpenBSD. The
developer must be passionate about the project and have the resources
(time, money) to accomplish it[1]. There is no way to quantify the odds
of this happening. I see no code attached to your email, so I presume
you're not volunteering.[1] Not even taking into account any licensing implications.
--
Jason Dixon
DixonGroup Consulting
http://www.dixongroup.net/
Why don't you write a port?
You can do the kernel pieces using .ko (like kqemu for example).
Ha you flatter me! You really want my code in the OpenBSD kernel!
Yeh just been reading that actually before I emailed misc@....
If you just sit and wait for it, I'd say: "zero to very-little".
| Sunil Naidu | Re: Linux 2.6.20-rc6 |
| Alan Cox | Re: Dual-Licensing Linux Kernel with GPL V2 and GPL V3 |
| Chris Snook | Re: init's children list is long and slows reaping children. |
| Greg Kroah-Hartman | [PATCH 001/196] Chinese: Add the known_regression URI to the HOWTO |
git: | |
| Jarek Poplawski | [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| David Miller | [GIT]: Networking |
| Gerrit Renker | [PATCH 27/37] dccp: Integration of dynamic feature activation - part 2 (server side) |
| Eric W. Biederman | Re: [PATCH 10/11] avoid kobject name conflict with different namespaces |
