Stratis 3.6 Released For Improving Linux Storage Management

[phoronix]

Phoronix: Stratis 3.6 Released For Improving Linux Storage Management

Red Hat engineers continue working on Stratis Storage as a modern Linux storage solution that leverages the Rust programming language and built atop the proven XFS file-system and LVM. Stratis continues to strive for ZFS and Btrfs like functionality although its use in the wild still seems rather limited...

https://www.phoronix.com/news/Stratis-Storage-3.6

[lproven]

I would probably like the sound of Stratis more if it reverted to EVMS instead of LVM. I have never much liked Linux LVM or its tooling. It's never too late to admit that one made a mistake and go back and fix it.

[espi]

I tried Stratis once and was pleasantly surprised, its pretty decent. Of course, all it really does is automate the setup of LUKS, LVM with XFS on top, but the cli tool is extremely intuitive and easy to use.

Now, no matter how many layers you pile below XFS it still isn't as featureful as btrfs. It doesn't feature compression (even though RH has their VDO compression layer, but its not upstream and Stratis doesn't use it), snapshots take forever and what nagged me the most was that if I didn't TRIM the pool it ran out of space by itself because the LVM overprovisioning isn't aware of files being deleted on the XFS layer.

[evert_mouw]

I still fail to see how Stratis is better than btrfs / zfs / bcachefs / ...

[S.Pam]

I still fail to see how Stratis is better than btrfs / zfs / bcachefs / ...

Easy: sell consultancy time, training and parnerships...

[Chugworth]

Stratis has always struck me as wasted effort, and kind of an embarrassment when coming from a large company like Red Hat. They should have engineers that could contribute to existing filesystem projects or design something totally new themselves. Not take the old XFS filesystem and try to strap some modern features on it with duct tape and bailing wire.

[pWe00Iri3e7Z9lHOX2Qx]

Last I checked they hadn't even implemented dm-verity or whatever they are going to use for hashing / checksumming. It's hard to believe RH is taking this seriously.

[phoenix_rizzen]

Stratis has always struck me as wasted effort, and kind of an embarrassment when coming from a large company like Red Hat. They should have engineers that could contribute to existing filesystem projects or design something totally new themselves. Not take the old XFS filesystem and try to strap some modern features on it with duct tape and bailing wire.

You mean like all the btrfs developers they've employed for several years now, without making a lot of headway on btrfs features/stability?

[Chugworth]

You mean like all the btrfs developers they've employed for several years now, without making a lot of headway on btrfs features/stability?

Do they still have Btrfs developers? I thought they don't even support Btrfs anymore.

I use Btrfs as the OS partition on several Linux workstations and servers and I'd say it's super stable. But it doesn't seem to be advancing much and I wonder, is that due to fundamental issues in its design, or lack of developers? OpenZFS on the other hand is advancing quite rapidly. It has awesome encryption features, and now that it supports reflinks there's even less of a reason to use Btrfs. I have been experimenting with Arch Linux on a ZFS RAIDZ2 (RAID-6) boot pool and it actually works pretty good.

[SyXbiT]

Stratis has always struck me as wasted effort, and kind of an embarrassment when coming from a large company like Red Hat. They should have engineers that could contribute to existing filesystem projects or design something totally new themselves. Not take the old XFS filesystem and try to strap some modern features on it with duct tape and bailing wire.

I think you're missing the point. btrfs was that attempt, and it has failed. So Red Hat, seeing that failure, has instead decided to build features on to an existing, stable filesystem. I get their reasoning.

Like all of you, I'm eagerly awaiting bcachefs