Register for the iXsystems Community to get an ad-free experience and exclusive discounts in our eBay Store.

FreeBSD moving to ZFS-on-Linux

Chris Moore

Super Moderator
Moderator
Joined
May 2, 2015
Messages
9,660
That sounds fairly significant, but I don't know how impactful it will be. I have storage pools at work that I have moved between ZoL using RedHat and FreeNAS with no difficulty beyond file access permissions.
 

cobrakiller58

FreeNAS Experienced
Joined
Jan 18, 2017
Messages
336
hmmm well that is interesting, it probably won't effect me much but I'll be sure to check and double check my back-ups before doing the upgrade that the transition to ZoL is a part of. I feel like @Arwen will be happy to see this lol
 

KrisBee

FreeNAS Guru
Joined
Mar 20, 2017
Messages
937

Elliot Dierksen

FreeNAS Guru
Joined
Dec 29, 2014
Messages
775

KrisBee

FreeNAS Guru
Joined
Mar 20, 2017
Messages
937

rvassar

FreeNAS Guru
Joined
May 2, 2018
Messages
511
Doesn't change the basic fact of licence problems for Linux. Reaction elsewhere on the web: https://illumos.topicbox.com/groups/developer/Td3185838722eb01a/wakeup
Oh man... They implemented the full Sun RTI process. That explains quite a bit.

RTI = "Request to Integrate", it was the process that you had to follow to change the Solaris ON code base. It dates back to the late 80's I believe, certainly pre-'96 when I hired on. Not just pre-Agile Waterfall, but the full Cathedral with the guys wearing the funny hats...
 

Chris Moore

Super Moderator
Moderator
Joined
May 2, 2015
Messages
9,660
Oh man... They implemented the full Sun RTI process. That explains quite a bit.
It is probably undoubtedly why the software has changed so little over the years. They can never get approval to integrate changes.
 

rvassar

FreeNAS Guru
Joined
May 2, 2018
Messages
511
It is probably undoubtedly why the software has changed so little over the years. They can never get approval to integrate changes.
I was thinking more in terms of what it takes to attract developer talent to an open source project. Not just lack of RTI approval, but that kind of rigid process could be actually running off the younger generation that might otherwise be interested.

Solaris ON has a lot of stuff stacked against it already, it's a fork of the original AT&T SysV Unix. From what I remember, at the core it's mostly pure C... Not a lot of C++. So the whole OO paradigm that's been pushed in college for the last 35+ years is lost, and then add in a bureaucratic Waterfall process when everyone else is using Agile.
 

Chris Moore

Super Moderator
Moderator
Joined
May 2, 2015
Messages
9,660
I was thinking more in terms of what it takes to attract developer talent to an open source project. Not just lack of RTI approval, but that kind of rigid process could be actually running off the younger generation that might otherwise be interested.

Solaris ON has a lot of stuff stacked against it already, it's a fork of the original AT&T SysV Unix. From what I remember, at the core it's mostly pure C... Not a lot of C++. So the whole OO paradigm that's been pushed in college for the last 35+ years is lost, and then add in a bureaucratic Waterfall process when everyone else is using Agile.
The language is a little bit of a problem, depending on what school a programmer went through. Some schools are teaching all OO from the start with languages that are designed for it, like Java, but the school I went through is still teaching C++ as the foundational language and they don't even start talking about objects until second or third year. They don't force it on you, because the department head realizes that not every solution fits into an OO paradigm. He actually worked for a bank programming their IBM (something) big-iron, before going back and becoming an instructor that ended up running the department. He knows, and tells his students, that some problems are more easily solved by linear programming, but he won't let you get away with using goto, at least not in C++.

Also, I went looking for some sort of guide for how to get started and didn't easily find any answers. So if you have not been involved, how would you get involved?
 

Arwen

FreeNAS Expert
Joined
May 17, 2014
Messages
1,138
hmmm well that is interesting, it probably won't effect me much but I'll be sure to check and double check my back-ups before doing the upgrade that the transition to ZoL is a part of. I feel like @Arwen will be happy to see this lol
Yes, and no. I do want to see better passing of features between the various OpenZFS implementations, like FreeBSD's TRIM support. (FreeBSD has had TRIM with ZFS for years. ZFS on Linux is STILL fighting to get it implemented.)

I did notice in the last year or so, (mostly because I have been paying more attention), that ZFS for MacOS, (and now ZFS for MS-Windows), seem to keep up with the feature set pretty well;

OpenZFS Feature Matrix

But that may be due to simply porting features over, and not originating new features, (which ZFS for Linux has done a lot).
 

cobrakiller58

FreeNAS Experienced
Joined
Jan 18, 2017
Messages
336
I was looking at the lack of trim as well I can't imagine ZoL would be implemented on FreeNAS without TRIM as iX sells some all flash hardware.
 

Arwen

FreeNAS Expert
Joined
May 17, 2014
Messages
1,138
@cobrakiller58, it appears that FreeBSD's implementation of TRIM is clumsy. So Linux wants to implement a cleaner code base. Nothing I can do, except watch and wait.
 

cobrakiller58

FreeNAS Experienced
Joined
Jan 18, 2017
Messages
336
@Arwen sadly all I can do is wait and see as well ATM but the idea of additional devs contributing to ZoL could certainly be beneficial. Frankly I'll probably be spending the next 10 or so days trying to figure out how this may or may not affect my systems.
 

KrisBee

FreeNAS Guru
Joined
Mar 20, 2017
Messages
937

Arwen

FreeNAS Expert
Joined
May 17, 2014
Messages
1,138
@cobrakiller58, in the short term after FreeNAS starts using ZoL tree, (I'd guess FreeNAS 12.x), I would not run any beta or nightlies on production servers. Further, unless you need a new feature, don't enable them. Let others be the guinea pigs.

In the long term, we may see the ZoL features sooner on FreeBSD/NAS, and others.

Ideally, people would choose to implement new features on one of the branches, like Linux. It's sometimes easier to test out on Linux. (I would not expect races, quirks and bug fixes to be developed on only one tree. Fix where found and able to test, then port over to the others via OpenZFS practices.)
 

xCatalystx

FreeNAS Experienced
Joined
Dec 3, 2014
Messages
104
What does this mean for the OS itself for FreeNAS? Is the plan to stick with FreeBSD forever? I know for a lot of people the draw of FreeBSD; was the near-native support for ZFS. But for others, giving up certain features of Linux was not worth it, so they would use ZoL or something else.

From a technical point of view, assuming the middleware does 90% of the interaction with the OS, this should be pretty easy to port over.
 

Ericloewe

Not-very-passive-but-aggressive
Moderator
Joined
Feb 15, 2014
Messages
16,088
What does this mean for the OS itself for FreeNAS?
Nothing. FreeNAS will stay on FreeBSD, and, in a way, be the trailblazer for FreeBSD releases with the "new" code.


From a technical point of view, assuming the middleware does 90% of the interaction with the OS, this should be pretty easy to port over
Linux is not a complete OS, it's just the kernel. Even something as fundamental as glibc is really only a matter of opinion. So, the first question is "what distro?" and the mere thought of going through that discussion is enough to cement the idea as "very bad", in my book.

Besides, jails are fundamentally better than any similar thing on Linux (Docker is mostly a lie), and that's one of the big draws of FreeNAS. I guess Joe Random couldn't care less about security, but that doesn't mean I don't mind seeing his server pwned by the Russian Mafia to serve spam because he was running this or that plugin.
 

xCatalystx

FreeNAS Experienced
Joined
Dec 3, 2014
Messages
104
Nothing. FreeNAS will stay on FreeBSD, and, in a way, be the trailblazer for FreeBSD releases with the "new" code.
I am perfectly fine with this. I was mostly asking because it was bound to come up. I wouldnt be suprised if a fork ends up happening at some stage (doesn't mean it will be successful)

Besides, jails are fundamentally better than any similar thing on Linux (Docker is mostly a lie), and that's one of the big draws of FreeNAS. I guess Joe Random couldn't care less about security.....
While I use containers heavily in production, also do I with jails heavily. While I agree for the most part jails are a better technology instead of lxc and docker (in this case i refer to a general container platform install with no tweaking, which would be most docker installs) doesn't mean people would still like to use it.
 
Top