Bcachefs Sends In "The Last Big On Disk Format Upgrade" For Linux 6.14

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • phoronix
    Administrator
    • Jan 2007
    • 67332

    Bcachefs Sends In "The Last Big On Disk Format Upgrade" For Linux 6.14

    Phoronix: Bcachefs Sends In "The Last Big On Disk Format Upgrade" For Linux 6.14

    Following the Bcachefs pull requests being rejected during the Linux 6.13 cycle by the kernel's Code of Conduct committee, the Linux 6.14 kernel cycle is kicking off with a big pull request so that the upstream kernel can get back into sync with the latest development code for this cycle. This pull also contains the last anticipated major on-disk format upgrade prior to the removal of the "experimental" flag for this copy-on-write file-system...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite
  • waxhead
    Premium For Life
    • Jul 2014
    • 1150

    #2
    Seriously: Can someone that has knowledge of bcachefs please list what features are done and safe to use, and what features are not ready yet.

    I suggest using the btrfs status page as a reference as that is usually what bcachefs is compared against.

    It would be good to know some facts before the flamewar begins...

    http://www.dirtcellar.net

    Comment

    • Jedibeeftrix
      Senior Member
      • Feb 2011
      • 328

      #3
      what is it even for?

      Comment

      • Quackdoc
        Senior Member
        • Oct 2020
        • 5063

        #4
        Originally posted by waxhead View Post
        Seriously: Can someone that has knowledge of bcachefs please list what features are done and safe to use, and what features are not ready yet.

        I suggest using the btrfs status page as a reference as that is usually what bcachefs is compared against.

        It would be good to know some facts before the flamewar begins...
        tbh, im not really sure, I just use reflinks and compression, I don't even use raid atm but I know a lot of people do, this will include a fix for 32bit applications on 64bit systems which is common with wine, so at the very least it's not safe for that yet until 6.14 xD

        Comment

        • spicfoo
          Senior Member
          • Nov 2023
          • 721

          #5
          Originally posted by waxhead View Post
          Seriously: Can someone that has knowledge of bcachefs please list what features are done and safe to use, and what features are not ready yet.

          I suggest using the btrfs status page as a reference as that is usually what bcachefs is compared against.

          It would be good to know some facts before the flamewar begins...
          It is an experimental filesystem. No part of it has been declared stable to use.

          Comment

          • Mitch
            Senior Member
            • May 2017
            • 372

            #6
            Originally posted by Jedibeeftrix View Post
            what is it even for?
            BCacheFS a multi-disk capable filesystem that includes disk-tierring (hot vs cold storage). Tierring is a unique feature that other filesystems included in the kernel cannot do without the assistance of something like LVM or BCache, included somewhere else in the stack.

            Like BtrFS and ZFS, BCacheFS is a multi-disk capable filesystem. It can also do various RAID-like configurations for big integrity, reliability, performance, and storage capacity features. This is actually a really big sell, especially when you compare filesystems that are included in the Linux kernel (and ZFS is NOT included in the kernel).

            Additionally, BCacheFS allows transparent compression, and includes the super-fast LZ4 codec as a choice. It has some built-in encryption features (I haven't read enough on encryption to say more), and reflink. BtrFS has transparent compression but doesn't offer LZ4, and IIRC, BtrFS's devs find LZ4 not to be worth doing on BtrFS due to technical reasons unique to BtrFS.

            BCacheFS can also also do something called checksumming and heal-on-read, which means the filesystem can detect integrity problems and maintain integrity transparently. BtrFS and ZFS can also do these.

            So BCacheFS is one of the heavy-weight filesystems like ZFS and BtrFS, where you can do a bevy of incredible and unique things due to being multi-disk capable. Each of these 3 Filesystems also has features the other 2 filesystems cannot do. They're also extremely diverse in their approaches and implementations and each with pros and cons. ZFS also has tierring/tierring-like features, but it's not included in the Linux kernel by default. Many distros do not include ZFS, unlike BtrFS and BCacheFS, which are both built-in to the kernel.

            I actively use all 3 of these filesystems for different things, but BCacheFS is still experimental, so I don't yet trust it with any data I can't afford to lose. Sorry if this is a big mess, I'm sure someone else has written a better comparison and explanation.

            Comment

            • Daktyl198
              Senior Member
              • Jul 2013
              • 1577

              #7
              Somehow, I just don't believe him when he says it's the last big on-disk format change. Not because I think he's lying, but because he said "before the experimental label comes off" when we all know that's never going to happen lol.

              Comment

              • lyamc
                Senior Member
                • Jun 2020
                • 526

                #8
                Originally posted by waxhead View Post
                Seriously: Can someone that has knowledge of bcachefs please list what features are done and safe to use, and what features are not ready yet.
                Whatever is not safe to use is marked not safe to use.



                Can't answer many of the other questions.

                Comment

                • ayumu
                  Senior Member
                  • Oct 2008
                  • 661

                  #9
                  It's hopeless. Just use OpenZFS, it works well, bypasses most of Linux block/fs layer, which is garbage anyway.

                  Note that any serious distro makes OpenZFS's out of tree experience seamless.

                  Comment

                  • --M--
                    Junior Member
                    • Nov 2023
                    • 2

                    #10
                    Originally posted by Daktyl198 View Post
                    Somehow, I just don't believe him when he says it's the last big on-disk format change. Not because I think he's lying, but because he said "before the experimental label comes off" when we all know that's never going to happen lol.
                    Developers are always optimistic. They hope that they finally thought of everything, and that there won't me any more changes that require on-disk updates. But even with something as mature as ZFS, there have been ongoing "feature-level" changes. These are minor tweaks and in the case of ZFS they are often optional, but they can break backwards compatibility. And even well-established filesystems such as XFS have had full-on hard format changes as recently as four years ago. And XFS is generally considered very mature. It's surprising that they'd have to take such a drastic step.

                    The XFS change was unusual in so far as it didn't allow an upgrade from the older to the more modern format. You couldn't even do this offline. Instead, you had to create a backup of all your data, wipe the existing disks, and then restore the backup. It is my understanding that BCacheFS has so far managed to do online upgrades of the on-disk format. Those are pretty much non-issues for most early adopters. Sometimes they are necessary. But other than no longer being able to boot into an older kernel, you won't even notice. For a young file system that is undergoing heavy development, that's honestly pretty admirable. If the BCacheFS can continue only doing having those types of on-disk changes, many people will be perfectly fine with it.

                    For serious production use, older filesystems are still a safer bet, but then, that's pretty normal. Filesystems take years to gain broad market share.

                    Comment

                    Working...
                    X