ZFS For Linux Is Now Available To The Public!

[kebabbert]

ZFS is....
licensed by oracle (aka satan),

Oracle also known as Satan, you say? So Oracle is Satan?

And yet, Oracle is developing BTRFS. You are too funny, man!

Besides, have you read the forum on BTRFS? It seems incredibly unstable and flawed. Sure in five years, it will probably have matured enough. But during these five years, ZFS will have developed further (unless ZFS development is stopped). ZFS will always be five years ahead. BTRFS is just a cheap ZFS copy cat.

[devsk]

BTRFS is an original load of crap work by Oracle. Just go and read BTRFS developer lists and bug reports to get a feel for how unstable BTRFS really is. ZFS was and is a master piece of original Sun Solaris engineers who thought ahead and created a scalable enterprise grade FS. BTRFS can not reach that level of stability and feature set in a million years!

Here are the bugs you can not expect from a usable (and note, not necessarily enterprise ready by any stretch) FS:

1. Have block level checksums but no way to use them to detect or fix anything. A corruption occurs, stays hidden and then suddenly you find files missing, and also find that btrfsck doesn't do anything. NOTHING! You are fucked! Full restore from backup!

2. Create a large number of small files and few large files. You are fucked! With 39% space available, you have ENOSPC. You can't do anything with the FS. No operations. Only option is to restore from full backup.

3. Do tonnes of IO for no reason. A filesystem should be quiet when user space is not writing anything. But BTRFS writes GBs per day with userspace writing only a few MBs. I call it the SSD killer!

These are first hand experience from a BTRFS zealot: me! I ran it on my desktop for a long time and pushed it whenever I got a chance, before I realized how bad the code really was.

I invite you to go over the mailing list and just watch the list discussions for few days!

Its a shame that I have to run ZFS on Linux but that's the reality of Linux filesystems at this time. And I am happy that people are working on this kernel module!

[kazetsukai]

You're the one trolling.
You ever wonder why zfs can't be in the kernel? Right... CAN'T. Because the license makes it impossible. Selected by satan himself with the specific intention to make it USELESS. Hence SATAN LICENSE.

The CDDL isn't a 'satan license'. You're out of your mind. The GPL and the CDDL aren't compatible. Does this make GPL software released under a 'satan license' too? The CDDL hasn't prevented BSD from implementing ZFS, in fact that's old news to that crowd already.

The GPL has been criticized endlessly for its incompatibility with other source licenses. I don't ever really see the same criticism for CDDL code, EXCEPT in situations where CDDL code would actually be useful in GPL'd software (like this scenario). This just boils down to 'Open Source' versus 'Free Software', and we all hear that enough from Richard Stallman already.

I really looked forward to using OpenSolaris for some projects, but Oracle stepping in put those plans to a stop. Illumos looks like it may or may not take off, who knows. Open Office splitting off? And worst of all, MySQL looks like it is in for years of deprecated development bitrot in Oracle's hands. I hate Oracle, I really do. They just waltzed into Sun HQ and turned everything inside out, shafting everyone. It is clear that making friends in the Open Source / Free Software communities is not a priority for them, making money is.

I don't think they (Oracle) have the same power over Solaris and Java they would have liked to since Sun started opening up those technologies, but they certainly made Solaris advancement in education and small business take a 180 in the opposite direction. Does anyone else see this Sun opening up pattern from the past few years as a precaution to a company like Oracle buying them out?

[kebabbert]

katzekui,
Dont worry, OpenIndiana will release in jan or feb, a release you can use in production. OpenSolaris lives as OpenIndiana.

BTW, ZFS rocks.

[kazetsukai]

@kebabbert:

I'm not sure why you think OpenIndiana will be ready by Jan/Feb, or what metric you are using to define a release as such, but its hard to call a fork this young 'production ready'. There really is no guarantee that OpenIndiana won't fall out of sync with Solaris and break compatibility, and there is no guarantee that Solaris bugfixes will make it into OpenIndiana in a timely fashion.

[kebabbert]

kazetsukai, I read somewhere that OpenIndiana is targeting a stable release, in jan or feb.

When (if) Oracle releases the Solaris 11 source code (probably when Solaris 11 is released) then OpenIndiana, Illumos and the other distros can catch up. This means that compatibility will not be broken. Those working on Illumos are several ex-Solaris kernel developers. They know the Kernel inside out. They will not break compatibility. And later, the source code will probably be released, so they can merge the Oracle code with the open distros.