Red Hat's Stratis Storage Project Reaches Its 1.0 Stable Milestone

[gnufreex]

Put bcachefs instead XFS, and Stratis would be totally best solution out there... after some more development. Red hat would need to hire Kent Overstreet.

[Niarbeht]

Then some wise fella decided it's a good idea to converge to zero. Like we have some major libraries and applications that haven't reached 1.0 for a friggin' decade (NetworkManager, GStreamer, and so on). Heck, GStreamer's first release was called 0.0.0. Does that make any sense? No, it doesn't.

https://semver.org/

EDIT: The point here being that there are a lot of ways to handle versioning, and some projects might have a philosophy on version numbers that makes perfect sense, if only you'd take the time to look at it.

[anarki2]

So......... no one is concerned that they developed it for 2 years, and never had it available for testing on Fedora or any other distro until their 1.0 release? And now they want yo ram it down Linux users throats after getting ZERO input on its design outside of Red Hat?

But they're releasing it exactly for testing? Just because it's rightfully called 1.0 instead of the gimmicky sh*t like 0.0.3.2 and the likes that's so popular with OSS, that doesn't mean it's not being released "early".

Back in the day, 1.0 was the VERY FIRST release you'd publish (except with Oracle of course, they start with 2.0 hahaha). Then some wise fella decided it's a good idea to converge to zero. Like we have some major libraries and applications that haven't reached 1.0 for a friggin' decade (NetworkManager, GStreamer, and so on). Heck, GStreamer's first release was called 0.0.0. Does that make any sense? No, it doesn't.

Stratis won't even be available in RHEL 8.0 I'd assume, they'll most likely release it in 8.1 at the soonest, even then only as a tech preview, after some dogfooding in Fedora.

And as someone else explained, Stratis is nothing else but some wrapper cr@p around existing building blocks. At least for now. So there's really not that much to test anyway.

[kallisti5]

> "Stratis filesystems are formatted with XFS, but managed on behalf of the user"

NOPE. I just noped right out of Stratis. XFS is pretty horrible with larger disks. Fragmentation is a massive issue with modern heavy workloads and disk sizes. In past jobs we pretty much had to run defragmentation continuously or the XFS kernel driver would start erroring out about being unable to obtain massive blocks of contiguous memory.

Mix in xfs_fsr core dumping randomly and it was a horrible thing to manage. For some reason, older Ops guys advertise that XFS is superior, but all I see is a stable, but woefully outdated filesystem which can't keep up with modern disks.

[starshipeleven]

Stratis really looks as a duck-taped stop-gag measure to quickly come-up with a snapshotting solution. Not a valid contender in my opinion. It's basically a gerbill with angry brows painted on it.

This. It's basically a wrapper over features that exist already in other Linux systems.

[starshipeleven]

So......... no one is concerned that they developed it for 2 years, and never had it available for testing on Fedora or any other distro until their 1.0 release? And now they want yo ram it down Linux users throats after getting ZERO input on its design outside of Red Hat?

Please state in detail where they claimed they want to ram down Linux user throats.

They will of course offer it in their own distros, and that's it.

[Anvil]

lets say you use XFS instead of EXT4 on all , would stratisd work better?

[Raven3x7]

in my experience BTRFS is mostly there (give or take a few features that needs fine tuning, mostly RAID5/6 not being production ready, but progressively getting there over the latest patches), basically it's a nearly-top dog.

BCacheFS would eventually have a good chance the above two top dogs. (but still hasn't finished implementing most of the key features), and has been started with the explicit goal of achieving this. It's basically a nice puppy running a bit behind.

Stratis really looks as a duck-taped stop-gag measure to quickly come-up with a snapshotting solution. Not a valid contender in my opinion. It's basically a gerbill with angry brows painted on it.

In my experience BTRFS is feature rich but buggy and slow.

[RahulSundaram]

yes, exactly.

I think they made a big mistake with it. I don't have an use case for such filesystem, because I just use rhel/centos on vps and for development, but I wouldn't used this stratis because of lack of wider testing.

I think noone is reading anything anymore before commenting. For one, this is not a filesystem at all. Stratis is a userspace manager that reuses many of widely used components like xfs and devicemapper. More details at

https://stratis-storage.github.io/FAQ.html

https://stratis-storage.github.io/StratisSoftwareDesign.pdf

https://lwn.net/Articles/755454/

[darkbasic]

Since I moved to zfs I don't look back. So much better than btrfs.