Although the ZFS on Linux It will not be a novelty in 2020, it has been news for months. Part of the fault is Canonical, a company that develops one of the most famous distributions, to add initial support on Ubuntu 19.10 and promise full support on Ubuntu 20.04. At first it was all laughs… until Linus Torvalds, the main person in charge of the Linux kernel, has stepped up to say, literally, «Don't use ZFS (on Linux). It's that easy".
The story is recent. Last Monday, a user complained that ZFS on Linux had screwed up his operating system. The professional de Torvalds did not wait, assuring that the nucleus is not responsible for what has happened to him. Basically the culprit is the file system, one that kernel developers can't get their hands on and include all the support they would like because it's owned by Oracle.
Linus Torvalds letter on why not use ZFS on Linux
Note that "we do not separate users" is literally about user space applications and the core that I maintain. If someone adds a kernel module like ZFS, they are alone. I can't keep it up, and I can't be bound by other people's kernel changes. And honestly, there is no way to merge any of ZFS's efforts until you receive an official letter from Oracle signed by your senior legal counsel or preferably Larry Ellison himself saying yes, it's okay to do so and treat the end result as GPL'd.
Other people think that it may be okay to merge the ZFS code into the kernel and that the module interface does it just fine, and that's their decision. But considering the litigious nature of Oracle and the licensing questions, there is no way I can feel safe doing so. And I'm also not interested in a kind of "ZFS wedge layer" that some people seem to think would isolate the two projects. That doesn't add any value to our side, and given Oracle's interface copyright claim (see Java), I don't think it's a real license gain either.
Don't use ZFS. It's that easy. It was always more of a buzzword than anything else I think, and the licensing issues just make it not a start for me.
The benchmarks I've seen don't make ZFS look that good. And as far as I can tell, it no longer has any real maintenance, so from a long-term stability point of view, why would you want to use it in the first place?
What is the problem
The problems that Torvalds sees with ZFS in Linux are mainly two:
- You will not work with it until Larry Ellison gives you written permission to treat it as a GPL. Without working with it, ZFS on Linux is not officially endured.
- Performance is not the best it could be.
After reading this letter, will you be using ZFS on Linux?
Linus is right. Very few can find ZFS useful; it would make zero difference to the bulk of users. So putting it in the kernel and getting into a legal fight with Oracle is pointless. But it was also very good to tell you that if you don't have a specific reason for choosing it, you are wasting your time.
It's the old story cdo they want to "sell" you a filesystem, they always come up with the "performance" verse. And the fact is that there is no one who is good at everything. Everyone is good at one thing and lousy at something else.
At one time I was with the "hype" of comparing file systems, looking for the best: in performance, security and features. I was studying the Phoronix benchmarks. At the end? After a thousand turns, I ended up with the same thing I had at the beginning: ext2 / 4 and btrfs.
Your system is not going to "fly" by putting another file system. And for the majority of users, almost any FS will do the job for them (ext4 x default).
In short, whether you understand what you are doing or not, the filesystem you use will always be a more or less "personal" choice (Linus or not LInus :-)