Announcement

Collapse
No announcement yet.

Linus Torvalds Doesn't Recommend Using ZFS On Linux

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Originally posted by carewolf View Post
    On traditionally spinning disks it will, because the read mechanism uses statistical error-correction and detection, so it can tell if the signal is muddy and ambigious and warn you. Chances of a flipped bit or two passing the error-correction are astronomically small. It mostly fails when entire sections are damaged physically.
    SMART is just an API from the OS perspective. The stuff that you are mentioning is done regardless of whether SMART is used and is necessary to get the drives to work at all. This is ECC stored alongside the data as part of the write. However, it does not protect against misdirected writes.

    A vibration can cause the head to go over the wrong track at the start of a write and perfectly overwrite it. In that case, the hardware ECC will look fine, but the data will be wrong. This has been observed in production in the past and is one of the reasons that ZFS’ end to end design is necessary to ensure data integrity.

    There are other possible causes of corruption too that are similarly not handled by the drive’s ECC. One would be a disk controller sending corrupted (but valid) write commands to the disks. Corruption from controllers has also been observed in the past and ZFS is able to detect it. The drive’s ECC is oblivious to it.

    If it counts for anything, I have more than a hundred commits in the ZFSOnLinux source tree and a few commits in Linus’ tree. I am speaking in the context of that experience.
    Last edited by ryao; 13 January 2020, 01:30 PM.

    Comment


    • Originally posted by Cape View Post
      I don't give a shat about data safety. I could loose 90% of all the files in my /home and not even notice it.

      Really important stuffs are copied on a couple of computers and another couple pendrives.

      Fuck ZFS, XFS and Btrfs.
      Performance is King.
      XFS is actually the single drive performance king as far as traditional measures of performance are concerned. ZFS will outperform it when you consider the effect of backups on system performance due to its practically immediate snapshots and ability to do incremental replication. ZFS will often outscale it too when parity RAID and volume management is needed.

      That being said, while you might not care about integrity, I can tell you that ZFS is protecting things like medical records where data corruption might kill people. Data integrity is important there. It is also used in places (e.g. supercomputers running classified simulations on which decisions are made that can affect human life) where downtime from the filesystem being corrupted could potentially kill people too. The integrity features are very important in such places.
      Last edited by ryao; 13 January 2020, 01:21 PM.

      Comment


      • Originally posted by jo-erlend View Post

        That doesn't explain anything to me. Are you trying to say that Linux doesn't provide the source code free of charge? Otherwise, why mention it as some kind of evidence that Linux is incompatible with CDDL?
        No that is not what I said and I even clarified it after the quote. How on earth did you miss that?

        CDDL requires that the ZFS source code is included as CDDL and CDDL only (aka the license of the source code files of ZFS cannot be relicensed) at the same time GPLv2 requires that "the work is to be licensed as a whole .. under the terms of this License".

        So if you include ZFS you no longer can release the Linux kernel source code as a whole under the terms of the GPLv2. For that to happen you would have to either break the license of the CDDL and relicense the ZFS code to GPL or you would have to break the GPL, neither of which is something that the Kernel developers are willing to do since that would open some really ugly consequences when preventing other parties from violating the GPL of the kernel.

        Comment


        • Originally posted by blackiwid View Post

          Sorry but nobody gives a shit about FreeBSD, it's used by 5 People around the world, and don't come now with some Playstation or other devices that might use it, because there you will never use ZFS. So nobody gives a shit to be compatible with them, if that OS would be so great nobody would care to bring ZFS to Linux in the first place because they just would use that System and move on. But people can forgo ZFS but they can't forgo Linux.

          So now let's Talk about Apple and to be complete also about MS, both don't offer YOU any solution to even have something between windows and Apple that fullfills your wishes. So if they don't give a shit about your problem why does the Linux community have to care about apple or windows support?

          So the only shiny feature FreeBSD has over other OSes especially Linux is native ZFS support so if you ask for a alternative to ZFS there is literally no reason to use this OS period. So then let's talk apple + linux support, Veracrypt seems to be the most suggested solution. you have then a encrypted file on your device that get's mounted as FS, the Filesystem below would probably easiest to use HFS+ because both Oses support that.
          No need to shit all over FreeBSD. The correct answer to his post is that Linux kernel developers should not have to intentionally break copyright licenses just because he happens to have a user case that depends on them doing that, aka you don't get to break the law to make it more convenient to you.

          Comment


          • Originally posted by F.Ultra View Post

            No that is not what I said and I even clarified it after the quote. How on earth did you miss that?

            CDDL requires that the ZFS source code is included as CDDL and CDDL only (aka the license of the source code files of ZFS cannot be relicensed) at the same time GPLv2 requires that "the work is to be licensed as a whole .. under the terms of this License".

            So if you include ZFS you no longer can release the Linux kernel source code as a whole under the terms of the GPLv2. For that to happen you would have to either break the license of the CDDL and relicense the ZFS code to GPL or you would have to break the GPL, neither of which is something that the Kernel developers are willing to do since that would open some really ugly consequences when preventing other parties from violating the GPL of the kernel.
            https://www.softwarefreedom.org/reso...rnel-cddl.html

            That being said, you can always request the ZFS source code. The CDDL might be more like the LGPL in terms of letting you mix in proprietary code, but as far as I know, no one actually does.

            You probably want to take up your complaints about not being able to get kernel source code with Nvidia, IBM (GPFS), Broadcom, etcetera. They are the ones not releasing their code as OSS. Ironically, this is a situation that the mainline kernel created via the module interface to allow ports of drivers because it helped it become popular in the early days. The GPL is concerned with derived works and being from other systems, those modules are not derived from Linux, even if they were adapted to work with it through an interface made to allow for them.
            Last edited by ryao; 13 January 2020, 02:38 PM.

            Comment


            • Originally posted by jo-erlend View Post
              That doesn't explain anything to me. Are you trying to say that Linux doesn't provide the source code free of charge? Otherwise, why mention it as some kind of evidence that Linux is incompatible with CDDL?
              Free of Charge and legally restricted are two different things.

              Linux kernel license Gplv2 and ZFS CDDL is provided "free of charge" as in you don't need to pay money. But they are only legal to use if you obey the terms of license so are legally restricted.

              https://rosenlaw.com/wp-content/uplo...Litigation.pdf This is a old 2004 document that it would pay you to read.

              Originally posted by F.Ultra View Post
              CDDL requires that the ZFS source code is included as CDDL and CDDL only (aka the license of the source code files of ZFS cannot be relicensed) at the same time GPLv2 requires that "the work is to be licensed as a whole .. under the terms of this License".
              Sorry no there is a bug in CDDL that comes into play as well due to what was removed that has not been tested in court.

              3.6. Larger Works
              You may create a Larger Work by combining Covered Software with other code not governed by the terms of this License and distribute the Larger Work as a single product. In such a case, You must make sure the requirements of this License are fulfilled for the Covered Software.
              The highlighted bit is it the following:
              1) CDDL have to be declared in the whole works license and declare was it not CDDL.
              2) Can you have large work license not be CDDL and declare where CDDL is.
              3) both of above
              4) None of above no allowed license overlap so no whole of work license allowed at all.
              That line does not exactly say.

              So legally both GPLv2 and CDDL could be cat fighting for the master Larger work license there is no court ruling for exactly what that line in CDDL in fact means and that is a big problem.

              So the idea that CDDL is only limited to files may be wrong. MPL 1.1 had other parts that are removed from CDDL limiting the meaning of that bit.

              Comment


              • Linus is such a boomer.

                Comment


                • Originally posted by k1e0x View Post
                  Linus clearly doesn't understand the problem.

                  OracleZFS != OpenZFS in the same way MySQL != MariaDB
                  Oracle does not own OpenZFS. More than 50% of OpenZFS is not copyright to Oracle. It's copyright to the OpenZFS developers
                  Even Oracle owning 0.0001% copyright would give me the willies given the current API copyright nonsense going on.

                  Comment


                  • Originally posted by Slartifartblast View Post

                    Even Oracle owning 0.0001% copyright would give me the willies given the current API copyright nonsense going on.
                    Oracle owns at least that much of the copyright of the Linux kernel. This is also true for glibc, MAC OS X, FreeBSD, etcetera. If you want to avoid their copyrights, your choices would be Minix, Haiku and Windows, although they might own portions of those of which I am unaware.

                    Comment


                    • Originally posted by ryao View Post
                      Oracle owns at least that much of the copyright of the Linux kernel. This is also true for glibc, MAC OS X, FreeBSD, etcetera. If you want to avoid their copyrights, your choices would be Minix, Haiku and Windows, although they might own portions of those of which I am unaware.
                      Its not just that is a Oracle copyright. Its a Oracle copyright on something that is a suspect license that you cannot be sure you are following it correctly. Please note Oracle is also sitting in the patents over ZFS as well.

                      Comment

                      Working...
                      X