Linux Founder Linus Torvalds Draws Ire for Criticizing Oracle ZFS

[cyklondx]

Never know ya IBM is the monkey wrench. They could decide paying for ZFS is less hassel. But Red Hat has been defaulting XFS for RHEL server for 3 years now... and the continue unlining XFS improvements. They as far as I know are still following their XFS roadmap.

https://www.phoronix.com/scan.php?page=news_item&px=XFS-2019-Copy-On-Write-Better
This is from almost a year ago now... but ya so far the work on xfs continues. Hopefully IBM doens't abandon it and jump in with Oracle.

its possible they'll implement most of the zfs stuff into xfs
IBM is known from locking down and making things inaccessible to people. Remember blackice? After IBM bought it its gooone. Plenty projects like that. (they were bought, implemented deeply into base functionality of their most expensive systems - and became black unattainable boxes unless you buy their stuff at 100k/year)

(don't forget IBM plan for RHEL is to make windows competitor - it will get locked down like windows in due time.)

 

[Vermillion]

What other solutions do copy on write, checksum everything, have snapshots that don't take up additional space, don't lose a ton of storage space to parity, and are free?

btrfs, XFS, and bcachefs are three that are free. btrfs is already mainlined and it's usable (I considered it for my Plex NAS but went Snapraid with Mergerfs using ext4 and XFS) but I don't know anyone who would use it in an enterprise production environment. bcachefs has great potential especially if it gets mainlined in the next year. XFS also has great potential and is already mainlined. However, until some of these reach more feature parity with ZFS, ZFS wins for what it can do. Sadly it's bolted onto Linux instead of being integrated with it.

 

[mikeo]

One day maybe ZFS will get block pointer rewrites working.

 

[Zarathustra[H]]

One day maybe ZFS will get block pointer rewrites working.

Yep.

That is the only technical limitation of ZFS compared to it's competitors as far as I am concerned.

Everything else is licensing nonsense, which while I understand is important, frustrates me that it gets in the way.

They partially addressed block pointer rewrites in OpenZFS 8 I think. It is now possible to remove a vdev after creating it.

It's only intended to remove accidental additions shortly after they are added - however - as it utilizes redirects in the index and is generally kind of slow.

This is to help with the "oops I added a single drive to my pool instead of a redundant vdev and now I have compromised the redundancy of the entire pool" problem. So if you remove the drive immediately after adding it, there are no ot very few redirects and it isn't a problem. You can still use it to remove more mature vdevs (provided you have ebeoigh space) but the overhead in dealing with redirects would be bad.

 

[toast0]

The part I am curious about is if this CDDL nonsense predates Oracle or not.

What other people said here; Oracle picked CDDL for a reason, and that reason was most likely so that it would be hard to ship integrated with Linux; however, towards the end, they were making more and more overtones towards open source, and there's some chance they would have relicensed if they were convinced OpenSolaris wasn't going to take the world by storm.

I will agree with you that ZFS is a stupid choice for a single drive.

I don't think it's a terrible choice for a single drive; it's not great, you have to pay the performance cost of COW on updates, without getting the benefit of automatic recovery from data corruption; however, on a system that can really only have a single disk, like a laptop or some of the SFF pcs, you still get to have snapshots, and you do get to know when your data is corrupt. When/if the disk does do bad things, you're going to have a nasty problem on your hands, but you might be better off with early failure than late failure. Although, if you have my luck with SSDs, when they fail, they simply disappear from the bus, never to respond to commands again, so it doesn't matter what filesystem you have, I hope you have backups. You can set copies=2 on a single disk ZFS, which would give you some ability to recover from corruption, but with a significant penalty for writes and again, not a great amount of real ability to recover... I still don't think it's terrible for a laptop.

 

[OutOfPhase]

ZFS amazes me, from a CS perspective. It's beyond my mortal capabilities and needs, but the sheer number of real world problems they considered and solved right in the file system are amazing.

Some smart cookies designed it.

 

[Zarathustra[H]]

ZFS amazes me, from a CS perspective. It's beyond my mortal capabilities and needs, but the sheer number of real world problems they considered and solved right in the file system are amazing.

Some smart cookies designed it.

My thoughts exactly.

I just wish it weren't saddled with a stupid license.