Fedora 37 Looks To Ship With Stratis Storage 3.1 Support

[RahulSundaram]

Oh please. None of us care if it's a new file system or slick glue between existing capabilities.

You absolutely should. Expecting a user space tool to compete with a full filesystem on things like snapshotting is pointless regardless of some whitepaper says. It is never going to have the performance characteristics of a cow fs. I am guessing none of you have tried out the tool.

I know you work at RH and I like the insight you provide in many posts, but you don't have to try and defend every single thing they do.

Umm, I haven't worked for Red Hat is well over a decade and I am explicitly saying ignore this project and it is going nowhere. That would be a very strange way of defending a project.

What was the previous solution and what will be the future solution?

They tried out Btrfs and it didnt work out for them. Who knows what will be the future solution? I am confident it won't be Stratis.

[RahulSundaram]

The second is that for pretty well the entire life of this project they have had a notice telling people no to use it because it would corrupt their data.

Can you provide the source for this? I haven't seen that.

[skeevy420]

Can you provide the source for this? I haven't seen that.

You have to go very far back. I just happened to remember reading that years back. The 1.0 release notes could have made early adopters nervous. By version 2.0, November 2019, it looks like it should have been good enough for folks like us.

I don't think anyone should be surprised that the .5 release of a file system setup has a warning.

[RahulSundaram]

You have to go very far back. I just happened to remember reading that years back. The 1.0 release notes could have made early adopters nervous. By version 2.0, November 2019, it looks like it should have been good enough for folks like us.

I don't think anyone should be surprised that the .5 release of a file system setup has a warning.

Indeed, an early warning before the first stable release is nowhere close to the nearly the entire life of the project that MadeUpName claimed. I had read that same warning from years back as well and it is clearly not the same thing. I am skeptical of the value of it obviously (and it is still only a tech preview and not supported for both RHEL 8 and RHEL 9 which isn't a good sign for the longevity of the project) but I wouldn't expect high level wrappers to eat your data. XFS is a very solid filesystem and all the individual components including device mapper etc are already very mature and widely used as well.

[Thaodan]

I'm using LVM plus Ext4 or XFS right now. Does stratis do it much better that it is worth a switch?

[pipe13]

I'm using LVM plus Ext4 or XFS right now. Does stratis do it much better that it is worth a switch?

That kinda depends on what "it" is. If you are comfortable with the simplicity of xfs and ext4, don't need to shrink xfs, or to shrink or move ext4 partitions frequently, I'd stay with what you know and what works well for you.

I do use stratis. I've used stratis since I installed Fedora34 on this new workstation last summer. Stratis works well, does snapshots, and can be thinly provisioned. All of which are attractive and Really Neat, and none except thin provisioning I've actually ended up using. So for me, I'd probably be slightly better off if I'd stuck with ext4. (Naked xfs isn't a fit because it won't shrink, which in the past I've done about yearly. I'd have to recheck whether thinly provisioned Stratis partitions can be shrunk -- I thought they could but I'm not going to look it up today.) Stratis has a straightforward command-line interface that I find simpler than LLVM + [your chosen filesystem].

So I'm not as pessimestic as Rahul. Stratis 3.1.0 is on both Fedora 36 and CentOS Stream 9, from the System repository on the former and appstream on the latter. So I think stratis is officially supported in that respect.

What is not yet officially supported is installing a root stratis filesystem. It can be done and I have done it and have used it as my daily Fedora driver for nearly a year. But neither gparted-gui nor blivet-gui can yet create stratis partitions and anaconda can't yet install CentOS Stream 9 to them either. That's pretty important and until they do I agree with Rahul: Stratis may be Really Neat, but is not yet fully supported. I understand these features are being worked on, but until CentOS Stream 9 announces anaconda support for stratis, and I test a root fs install on both a VM and real hardware, I can't really recommend stratis for critical use.

[RahulSundaram]

So I'm not as pessimestic as Rahul. Stratis 3.1.0 is on both Fedora 36 and CentOS Stream 9, from the System repository on the former and appstream on the latter. So I think stratis is officially supported in that respect.

It is available in both. Fedora doesn't have commercial support and uses Btrfs by default in most places. Red Hat however doesn't commercially support it and considers it a tech preview in both RHEL 8 and RHEL 9. If your expectations are to get the convenience and aesthetics of an integrated tool for existing functionality that is already provided outside of it, sure Stratis may fit your needs just fine. Things like VDO is heavily under advertised and if Stratis is what it takes to bring more visibility into that, so be it. What I am specifically pessimistic is any notion that it is going to fully replace the functionality and performance of a cow filesystem. Now if Red Hat offers feature parity, full commercial support and shows benchmarks with similar performance for snapshots, encryption etc, I can happily admit I was wrong. Until then, I am going to bet on btrfs (with experimental features disabled) or bcachefs (when it gets merged upstream which appears to be close) before I count on Stratis.

[Waethorn]

I'm using LVM plus Ext4 or XFS right now. Does stratis do it much better that it is worth a switch?

Let's not forget that a lot of NAS systems in use (QNAP especially) already use mdadm with ext4, and overlayfs for storage checkpoints. Some companies have balked at btrfs for being immature, and can't license OpenZFS within their Linux-based products. Combining the goals of those into a single project would benefit companies building commercial products built on open-source software.

I want to see more information about multi-drive redundancy with parity features. Call it "software-RAID5/6" or whatever you want, but ZFS/Open-ZFS has it built in, and nothing else does. Btrfs's support is still unstable even after years of development. If Stratis does error correction with respect to the filesystem on multi-drive parity arrays like ZFS does, that'd get my interest.

[MadeUpName]

You have to go very far back. I just happened to remember reading that years back. The 1.0 release notes could have made early adopters nervous. By version 2.0, November 2019, it looks like it should have been good enough for folks like us.

I don't think anyone should be surprised that the .5 release of a file system setup has a warning.

They revmped their web site less than a year ago and it was still on there then and it was after the 1.0. No I can not point to some thing that no longer exists.

[RahulSundaram]

They revmped their web site less than a year ago and it was still on there then and it was after the 1.0. No I can not point to some thing that no longer exists.

You can use things like https://web.archive.org/web/*/https:...age.github.io/ and the source repo for the website