Goodbye FreeNAS

[Jailer]

I don't like this dramatic "I'm leaving X for Y" type of posts. If you're going to switch to something else, just do it.

I was just about to post the same thing.

Why all the drama, just leave if you don't like it.

 

[danb35]

I tend to disagree. FreeNAS 11 has had a much more rational development cycle.

Fair enough. I guess I'd feel better if I saw Kris address what went wrong, culturally, to allow something like Corral to happen (and again, IMO, Corral is simply the latest and worst example of a problem that's been going on for at least a few years), and what they're doing to keep it from happening again. Of course, doing that would probably involve applying undercoating to a few busses, so I can understand not wanting to do so, but it would give some confidence that they actually learned something from this mess.

 

[Spearfoot]

@Jailer @m0nkey_

 

[danb35]

I'm so mad I actually uploaded by accident a bode plot for a filter. Look:

"In the heat of composition I find that I have inadvertently allowed myself to assume the form of a large centipede."

 

[Stux]

So, in summary, I hope you like RAID1 if you're moving to BTRFS.

Please correct me if BTRFS' RAID6 support is suddenly production ready.

 

[danb35]

I'm probably going a little opposite-over-adjacent, but I'm also baffled that Rockstor is using btrfs. The biggest advantage of btrfs, IIRC, is that you can add drives "safely"--that is, you can turn a three-disk RAID5 into a four-disk RAID5. That is pretty big, especially for the home market, and it's something that ZFS will probably never see. But other than that, I can't imagine why anyone would prefer btrfs over ZFS.

 

[Mlovelace]

What could possibly compel someone to trade ZFS for btrfs?

Especially when there is ZFS on Linux, which I also use everyday on several other servers. :)

 

[Ericloewe]

I'm probably going a little opposite-over-adjacent, but I'm also baffled that Rockstor is using btrfs. The biggest advantage of btrfs, IIRC, is that you can add drives "safely"--that is, you can turn a three-disk RAID5 into a four-disk RAID5. That is pretty big, especially for the home market, and it's something that ZFS will probably never see. But other than that, I can't imagine why anyone would prefer btrfs over ZFS.

The B-tree supposedly allows for a few neat things that Block pointer rewrite would do on ZFS, but their development methodology is dubious at best, as evidenced by the whole "RAID5/6 doesn't actually work outside of tightly controlled demos" talk.

 

[cyberjock]

IMO BTRFS has a ways to go before it will compete with ZFS. A few years ago I put it at 2020 where BTRFS will be a significant player and may compete with ZFS. I'm a ZFS fan. That may change someday if BTRFS keeps chugging along doing what its doing and provides me with benefits that ZFS cannot compete with. However, today I don't feel that BTRFS is as reliable, trustworthy, and malleable to be well-used. The RAID5/6 is a major, major game change if/when that gets fixed though. That's one of the top few reasons I have why BTRFS is clearly not "trust it with my data yet".

I do welcome the competition though. If BTRFS proves superior to ZFS, then let the better file system win. Its as simple as that.

 

[styno]

Synology uses BTRFS, but 'only' as a filesystem. Last time I checked the raid portion was still mdadm-based. Implementations like this should be rock solid, stable and definitely not brand new.
(Just dropping the info, hail ZFS! )

 

[danb35]

Please correct me if BTRFS' RAID6 support is suddenly production ready.

Not according to Rockstor's FAQ as of last night.

Edit: Linky. Also:

However a proviso here is that The BTRFS community consensus is that raid5 and raid6 levels of btrfs support are not yet fully stable and so are *not recommended for production use*. Please see the btrfs wiki for up to date information on all btrfs matters.

And the btrfs wiki notes:

Single and Dual Parity implementations (experimental, not production-ready)

And the filesystem is ten years old. Wow. From an outsider's perspective, I'd say that either (1) they just don't care about RAID5/6, or (2) some early design decisions have unexpectedly made RAID5/6 much harder than had been anticipated.

 

[Ericloewe]

And the filesystem is ten years old. Wow. From an outsider's perspective, I'd say that either (1) they just don't care about RAID5/6, or (2) some early design decisions have unexpectedly made RAID5/6 much harder than had been anticipated.

I think it's mostly (1), since the current implementation clearly has had zero thought put into it.

 

[scrappy]

I dont see why yes its a newer file system but is now officially marked as stable and is backed by a lot of big company's/organisations.

Pretty much what others here recently stated. Using BTRFS at this point in time is like playing with a BETA release. ZFS is far more mature and battle tested. While no file system is perfect, I would not be willing to trust my personal data on BTRFS when ZFS is available for both Linux and *BSD systems. At least find yourself a Linux NAS solution with ZFS.

 

[danb35]

I dont see why yes its a newer file system but is now officially marked as stable

"Stable" is, of course, a relative term. As noted throughout this thread, neither the btrfs devs not the Rockstor devs consider the RAID5/6 code to be at all stable, so unless you're planning on using mirrors, you will not have a stable filesystem/RAID solution on which to store your data. ZFS on Linux is pretty stable and mature, so if you just have to have Linux for some reason (like Docker), you can still have a stable, reliable, modern filesystem/RAID manager--it just won't be btrfs.

But if Docker is the issue, there's no need to leave FreeNAS for that--build a VM with your preferred Linux distro (boot2docker, RancherOS, etc.) and run your dockers there. You can do that in FN11, and there are already guides here on the forums. No, it won't be running natively on FreeNAS, but it wasn't doing that under Corral either.

Ultimately, of course, it's your server, and your data, and nobody here is going to tell you to stay with FreeNAS if FreeNAS doesn't meet your needs--but I think we'd all encourage you to choose wisely if you're going to make a switch.

 

[blaco]

I was thinking about going to Linux and btrfs 6 months ago, because of Linux and KVM, but they introduced bhyve.
I liked the ability to add a drive to a RAID5/6, but if it's unstable...

Zfs is stable and I love my data. I thought about ZFS on Linux, but I haven't found a nice System with GUI like in FreeNAS.

For everyone who has a look into this thread and is interested in btrfs:
https://btrfs.wiki.kernel.org/index.php/Status
Even Compression, Raid-1 and Device replace (which are basic functions) are only marked as "mostly OK" - and not "OK"!

I hope everyone who decides to go with btrfs reads the status page...

 

[Ericloewe]

I was thinking about going to Linux and btrfs 6 months ago, because of Linux and KVM, but they introduced bhyve.
I liked the ability to add a drive to a RAID5/6, but if it's unstable...

Zfs is stable and I love my data. I thought about ZFS on Linux, but I haven't found a nice System with GUI like in Freenas.

For everyone who has a look into this thread and is interested in btrfs:
https://btrfs.wiki.kernel.org/index.php/Status
Even Compression, Raid-1 and Device replace (which are basic functions) are only marked as "mostly OK" - and not "OK"!

I hope everyone who decides to go with btrfs reads the status page...

• "Auto repair and compression may crash"
• Scrub + RAID5/6: "will verify but not repair"
• Device replace: "gets stuck on devices with bad sectors" (Isn't that the whole point of a disk replacement!?)
• RAID1: "Needs at least two available devices always. Can get stuck in irreversible read-only mode if only one device is present." (Irreversible!?)
• RAID0: OK (Great! You're almost at Intel fakeRAID levels of utility!)
• RAID5/6: "Write hole still exists" (What do you mean "still"? It should never have been there!), parity not checksummed (WTF????!?!!?!??!!??)
• Quotas, qgroups: mostly OK (WTF? It's not rocket science!)
• Free space tree: Too much to copy, but it boils down to "this thing is as fragile as a data structure can be, and our repair tools only make it worse".

I hadn't actually been following btrfs closely and this is much worse than I expected.

Oh, and it doesn't actually "work" on big-endian systems, which I'm guessing is the default mode of operation for Linux on ARM.

 

[Arwen]

I agree with others on the many of the faults of BTRFS. However, I used it at home for
years and on my laptop until quite recently. Worked just fine, and as far as I could tell,
no data loss. Did not use mirroring, just used the snapshotting for alternate boot
environments for easy OS software upgrades. Now all 3 use ZFSonLinux.

It's not a complete copy on write file system. I ran across a quirk where BTRFS would
over-write a file move / rename, not COW, (Copy On Write). So they had to document the
behaviour on a crash. The results varied dramatically. From not being done, to erasing
the file. I can't find the reference right now. Nothing to worry about in general. But, for
production data, nope.

It does appear that Linux kernel 4.12 will have the RAID-5/6 issue resolved. That kernel
is now a release canidate. However, I tend to run LTS kernels, (Long Term Support), so
I would not run 4.12 unless I had to. I prefer to wait for more bugs to be squashed.

Edit: Found the reference to the rename issue;

https://btrfs.wiki.kernel.org/index.php/FAQ#What_are_the_crash_guarantees_of_rename.3F

 

[garym]

Nothing wrong with a Linux based NAS but I would never trust my data on BTRFS.

It seems people don't want a bullet proof NAS to protect their data any longer. They want a Swiss army knife appliance to do everything but securely store their data.
I hope that FreeNAS stays true to the core of safe data storage and hard drive management, only secondarily adding features for those that want them.

 

[danb35]

It seems people don't want a bullet proof NAS to protect their data any longer.

This probably isn't quite fair. The (or "a") problem is that "bullet proof" is both hard to define and hard to measure, as well as boring, while features are easy to see and sexy. So, while I might value data protection, it's hard to tell what does it well, and squirrel!

 

[Mlovelace]

It seems people don't want a bullet proof NAS to protect their data any longer. They want a Swiss army knife appliance to do everything but securely store their data.
I hope that freenas stays true to the core of safe data storage and hard drive management, only secondarily adding features for those that want them.

While I agree with your sentiment, there is no such thing as a "bullet proof NAS". People should be following the 3-2-1 rule for truly irreplaceable data. Raid is not a backup and disasters happen.