You don’t see any reason to use ZFS for a pure storage solution??
I was referring to BTRFS in that post.
The thing about some Linux projects is that they can and do bloom, then stagnate. Their was a big push for BTRFS to take the place of EXT3/4. But now bcachefs is the latest thing, except it is at least 5 years behind BTRFS and 15 years behind ZFS.
Some Open Source projects that have regular users and have wide appeal, do quite well. Ones like GIMP, LibreOffice, FFMPEG and Handbrake. Some like Firefox and Chrome end up fragmenting the market / user base, taking development time from both, with the result of worse products for both.
One issue I have with a long term development project that has kernel dependencies, is the need for it to be bundled with the kernel source. Both BTRFS & bcachefs, (as far as I know), are bundled with the kernel. Any changes to those file systems requires a new kernel. Or in the former case of RedHat, back porting the changes.
As much as Linus T. dislikes OpenZFS and wants file systems in the kernel source, I actually think this is wrong. At least for heavy development. Before RedHat dropped BTRFS support, I toyed with that idea. Now I think it is BEST for an actively developed file system, like OpenZFS, to be outside the kernel. And support multiple kernels.
Having to upgrade your kernel for new BTRFS and bcachefs features may make sense for distros that keep up with kernels. But the Enterprise users are not interested in the latest kernels, (with the latest bugs and security holes), mostly they just want stability. Then security, (after all, if the server is not stable, its useless for Enterprise use…).
2 Likes
I’m a very long time Linux user and have settled on OpenSuSE (now Tumbleweed) those last few years. So I’ve been using BTRFS as part of the package (on my root partition onlt though).
All in all it’s been both solid and quite convenient, provided you configure it so that it doesn’t overcrowd it with snapshots. It has saved my system a couple times, lastly a couple days ago when a laptop was updated with a broken intel wifi driver (yay, thanks intel!).
Does that configuration support alternate boot environments in Grub?
I mean if the new update is so bad, you want to go back completely, can you just reboot and simply select another Grub entry?
I didn’t even check. 
It seemed cleaner to me to just remove that driver.
OpenSUSE Tumbleweed/Leap have grub menu entries to boot from previous snapshot, so the answer to your 1st question is yes when you install on root BTRFS, which is the default for their YAST installer. SUSE is committed to BTRFS and it’s a fundamental part of their Micro OS - immutable container host OS - system.
1 Like
I have used btrfs on linux, a while back, however, I did not use it as any kind of RAID. That said, I remember Allan Jude saying on more than one occasion, that events that ZFS would recover from, and uses for testing (like writing zeroes to one drive in a pool), would destroy a BTRFS pool.
Maybe this is apocryphal or out of date, but that is what I remember about BTR.
My suggestion would be to listen to the 2.5Admins podcast. The cast consists of Allan Jude, Jim Salter, and Joe Ressington. Allan and Jim have expert-level experience with both ZFS and BTRFS, and Joe claims to be the “half” admin…They seem to give both a fair shake, as well as getting in to other tech news as well. Even Jim prefers ZFS’ performance over BTR.
1 Like
There are comments about performance and one thing is clear, their are no serious obvious winners. Meaning BTRFS does not have a serious lead in performance, over OpenZFS, when used in a similar configuration. Nor is the opposite true, OpenZFS is not ultra fast.
In some cases other Linux file systems, like EXT4 or XFS could be faster than OpenZFS, (and possibly BTRFS). However, my personal preference is that I would take the data integrity features of OpenZFS, (and if OpenZFS did not exist, BTRFS), over speed.
File systems are not just about serving up data very fast. If they serve up the WRONG data, and that HAS happened to me before using OpenZFS, then you might as well not bothered saving the data.
Anyone think BTRFS is noticeably faster / slower than OpenZFS?
(When compared to similar configurations only…)
I’m neither a Linux kernel developer nor a file system developer. I use OpenZFS because it is the only file system that I know how to use. Now it works well and there is no reason to learn a whole new file system, except the new one has some fascinating features.
btrfsis in bad fame and I don’t want to try it at all,bcachefsmay be very promising but it still needs more time to do testing and get the promotion.
However, each time I upgrade Debian’s kernel, I have to install the related linux-headers-$(uname -r) and reinstall zfs-dkms. It is so annoying.
BcacheFS is interesting (I tried it!) but it didn’t end up being stable for me. An sVDEV combined with the regular HDD ZFS pool likely brings close to bcacheFS speeds to datasets for my use case - but without the dataloss issues I experienced using bcacheFS.
I suggest learning about sVDEVs, dataset recordsizes, and compression before considering any other filesystem because the three can help address the biggest stumbling blocks re: ZFS performance in many use cases. (shameless plug, but here we are)