ZFS and other FS's as Filesystem

at least nobody mention Btrfs :slight_smile:
I used ZFS with FreeNAS, Promox and OpenMediaVault I said used because

  1. it’s a pain in the ass when it fail and don’t support encryption.
  2. It’s also, like lvm, Btrfs and any abstraction to facilitating the hardware management consume CPU and Memory, sometimes too much.
  3. it’s harder to extend than a RAID10 (here a how to make a RAID10 with only 2drives)
1 Like


I never liked ReiserFS, even though SuSE, one of my main distros, used it as default, I always changed that. XFS was always better!
Maybe Hans Reiser listened too much to AC/DCs Dirty Deeds… - anyway, it’s almost history now…

My 2 cents

Something interesting:


  • ZFS was designed for Solaris (BSD) not Linux.
  • Ports of that functionality are ZFS/Linux and BtrFS.

Both don’t reach quite the stability of ZFS on BSD, but there are other issues too, like resource hungry (RAM & CPU), enlargement of volumes and rebuild. iXsystems don’t even include automatic rebuild of ZFS in their FreeNAS series, only in the more expensive TrueNAS series.

The situation for linux is suboptimal at the moment…

My 2 cents

1 Like

yep, licence conflict, and refusal of linus, I heard about recently


It looks like Linus is running his mouth about something he knows nothing about again. I do wish he’d stop that.

ZFS on Linux is OpenZFS, and they’re in the process (if they aren’t already there) of moving to where the Linux code will be the foundation for OpenZFS. btrfs, well… It’s broken by design, to the point where RedHat gave up on it after dumping heaven only knows how much money into it.

I think this issue is exaggerated due to FreeNAS (which has its own requirements), but it’s true that ZFS likes its RAM. Is this really a problem? It’s 2020, ffs–8+ GB of RAM in a server is not a lot.

No idea what either of these is referring to; ZFS rebuilds just fine and has always allowed for pool expansion.

What does this (even if true, which it isn’t completely) have to do with the state of ZFS on Linux?

True, and it’s largely a result of core Linux devs setting out to deliberately sabotage ZFS.


Don’t misunderstand me, I really like ZFS, and would like it stable as a linux foundation!
I know it’s called OpenZFS.

Pool enlargement is not the same as volume enlargement…


My 2 cents!

Then why are you tossing out a bunch of random half-truths which you aren’t even bothering to explain, much less defend?

So what are you talking about, then? Strictly speaking, a “volume” is a zvol, a virtual block device, which aren’t used a lot other than for VMs–which was why I didn’t think you were referring to them. But if that’s what you meant, expanding a zvol is even more trivial than expanding a pool.

The main issue with ZFS is still the license. I’m not a lawyer, and do not know of the legal issues possibly still in OpenZFS…
Quote: “As Oracle’s code is not open source, the OpenZFS wish to maintain compatibility with Solaris ZFS pool versions 29–35 is difficult to realize.”
The others stuff are smaller “issues” or whatever, but no BIG issue.

RAM is still often enough an issue with SME clients (Small and Medium Enterprises).
Clients buying undersized servers, or older servers with limited RAM.
Example: HP Microserver G8 with a max of 16 GB RAM. Actually enough, depending what you’re trying to run.
Proxmox with ZFS, a Windows Server and a NethServer is RAM-Wise still a squeeze on such Hardware…

iXsystems are a major contributer, along with others like ProxMox to OpenZFS…

My main use for ZFS at the time was for VMs…

My 2 cents

That’s definitely an issue if the objective were to build ZFS into the kernel–but nobody’s wanting to do that. Does the license prevent a distro from including ZFS? Proxmox doesn’t think so, nor does Canonical (Ubuntu). I am a lawyer, but I practice in a completely different field–but their lawyers don’t seem to have a problem with it. I do understand the concern–Oracle is, if anything, even more litigious than SCO–but there hasn’t been a word from them about suing over ZFS in either Proxmox or Ubuntu.

To be more precise, OracleZFS after version 28 is not open source–versions before that were (yes, the CDDL is an open-source license). The only relevance to this discussion is if someone wanted to import a recent OracleZFS pool on an OpenZFS system, which won’t work, and is highly unlikely to ever work.

Only to the extent they’re true–you still haven’t said what you think the issues are with “enlargement of volumes and rebuild” (and FreeNAS is utterly irrelevant to this discussion–the way it manages rebuilds is a matter of its own middleware; it has nothing to do with the underlying ZFS code, as is proved by the fact that TrueNAS behaves differently). Are you talking about expansion of parity RAID vdevs?

I’ve bolded the important part. That server is more than enough to run Neth (and Neth on ZFS) in any environment where it would be remotely appropriate. Even in your virtualization example, 16 GB should be adequate, depending on how much RAM Windows needed.

Sure they are. And the design decisions in FreeNAS (and TrueNAS) are constrained by what ZFS will allow. But if TrueNAS does “automatic rebuild” (whatever that is), doesn’t that pretty effectively show that ZFS allows that capability?


Quite agree with all your points.

Oracle ZFS Version is already stated as 29-35. Including that would mean backporting or reverse engineering or a SHIM or worse. Legal headaches…

But I agree with your summary about it mainly concerning those wanting to import an existing Oracle ZFS > v28…
Similiar with MySQL / MariaDB.
It’s not something a “techie” can solve, the issues behind these are mainly legal and restraints by such.

I can understand why Larry Ellison, as owner of Oracle, would want to hem MySQL development as not to compete with his prime product.
I can also understand the Dev-Teams frustration. At the time it was bought by Sun, they had free hand in development. When Oracle bought Sun, all that changed.
Recall what happened with OpenOffice and their team?

Smaller issues are - in my opinion of the english language - anything else besides legal issues.
That includes resizing and rebuilding - at least in my case. IT is for me like a well behaved dog. Almost all OpenSource, and moct closed source, I can handle easily!
There aren’t that many filesystems i haven’t confronted in my 30+ years as self employed IT…

Depending what you need to run…
Yeah, then the client needs something like a MS-SQL based “App” Server, that needs 8-12 GB RAM to run (slowly), 16 would be OK.
That’s what I mean by squeezed…

Even living in Switzerland doesn’t mean everything’s gold…
Our high exchange rate relevant to the $ / £ / € doesn’t really help the economy as a whole. Swiss economy is mainly export based, be it manufacturing, heavy industries (Dam turbines), medical or financial services. Swiss banks don’t get rich by managing the finances of your average Joe here in Switzerland, employed as any other…

Really like that comparison:


My 2 cents

You keep mentioning these, but what do you mean by them? Pool resizing (well, expansion) is trivial, as is zvol expansion. Resilvering works.

As to the licensing issues, the question regards the legality of running OpenZFS on Linux with varying levels of integration. OracleZFS isn’t in view–OpenZFS can’t use OracleZFS code (at least, any code written since the fork), OpenZFS won’t be compatible with OracleZFS pools > v28, and nobody’s talking (at least with any seriousness) about trying to change that. IOW, compatibility with OracleZFS pools really isn’t a goal of OpenZFS–it’d be nice to have, I guess, but it isn’t there and I don’t think anybody’s working to change that.

Again, the license issue deals with integration in Linux, and there are a few flavors:

  • Include OpenZFS in the Linux kernel–this would be a licensing problem, but so what? Nobody (certainly not the OpenZFS devs) is asking for this.
  • Package ZFS in a Linux distro–I’ve heard it suggested that this could be a licensing problem, but I don’t see why, and Canonical’s lawyers apparently don’t see it either. And if Oracle is litigious enough to sue over an API, it would be professional suicide for Canonical’s lawyers to OK this if they weren’t really sure.
  • Install and use ZFS on Linux as an end user–definitely not a licensing problem, any more than installing closed-source software on Linux would be.

Like I say, these work for me too.
It’s akin like having to explain to certain people who believe that all RAIDs, even RAID0, are redundant in some form.
RAID0, even though containing the Acronym RAID does also contain the numeral 0 (ZERO), which means no redundancy!

Some people just don’t get simple things.

ZFS and OpenZFS are (today) quite different beasts. True that they share a common history.
They have several things even today in common (Featureset), but are neither compatible nor the same thing. And - complete different licensing terms!
You also compare OpenZFS in the Linux Kernel with ZFS in a Linux Distro like Canonical?
These guys use OpenZFS, not ZFS! Using ZFS, not OpenZFS would open up a legal nightmare, I’m sure you’ld agree!

Just because we can easily read arabic numerals (as well as “roman” numerals) doesn’t imply in any form that any of us can also read “written” arabic…


Yes, they are, but it’s beside the point. Any discussion of ZFS on any OS other than Oracle Solaris is talking about OpenZFS; the only ones using OracleZFS are Oracle.

No, I don’t. Nobody (other than Linus for some reason) is talking about any flavor of ZFS in the kernel (and now he’s saying he meant OracleZFS, which makes his comments make even less sense). And yes, of course Canonical is using OpenZFS–I thought that was obvious enough to not require the distinction.

I apologize if my use of “ZFS” has confused you–I thought it would be obvious that any mention of ZFS in connection with Linux would be OpenZFS.


Using the terminology OracleZFS and OpenZFS makes things absolutly clear…