Announcement

Collapse
No announcement yet.

Ubuntu Developers Discuss Dropping ReiserFS

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

  • Ubuntu Developers Discuss Dropping ReiserFS

    Phoronix: Ubuntu Developers Discuss Dropping ReiserFS

    Ubuntu developers are presently deciding what to do with support for the ReiserFS file-system, up to and including dropping kernel support for the aging but stable file-system...

    http://www.phoronix.com/vr.php?view=MTQ2MDM

  • #2
    No good answers...

    The demise of reiserfs is a sad story. Here you have the first widespread used journaled fielsystem. It was for the longest time the only filesystem that work effeciently under LVM allowing ultrafast growing while online.

    If somebody asked me for an enterprise class filesystem, reiserfs would be on the short list....

    ... if it weren't for the fact that it's not being actively maintained by anyone.

    Hans was rude, crude and of course, a murderer... but in his madness there was certainly genius.

    Reiserfs (v3) you will be missed.

    Meanwhile, Linux filesystems have regressed... not saying that ext4 isn't "better" than ext3 in some ways, but it still doesn't resize quickly and I believe you still have the separate inode (out of inodes) problem. So, it's not reiserfs in those respects.

    My choice has been to use XFS, but again, it's not reiserfs either....

    I think the world is hoping for a "good" BTRFS, but that seems to still be a ways off...

    Losing reiserfs has hurt enterprise Linux... but it will recover....

    Comment


    • #3
      Originally posted by cjcox View Post
      ...

      Losing reiserfs has hurt enterprise Linux... but it will recover....
      I do not think ReiserFS was ever meant to be use for enterprises. ReiserFS is designed as a better filesystem for desktops than EXT3 at the time. EXT2, EXT3, EXT4 and XFS is meant for enterprises. When I used ReiserFS, it sometimes becomes unstable compared to other filesystems. I only use ReiserFS for junk data and video recordings because it too unstable and too unreliable to use it for the main setup.

      Probably, someone will take ReiserFS and create a fork of it. Hopefully ReiserFS can get better being stable and reliable when that person does do that.

      This bad new is only for people that use Ubuntu. The kernel developers still have it listed in the kernel. Until kernel developers completely gets rid of ReiserFS and Reiser4 support out of the kernel, users can still re-compile the kernel to include ReiserFS and Reiser4 support.

      Comment


      • #4
        Reiserfs

        I still use reiserfs v3 in production on my Desktops. My fileserver moved to XFS because I wanted a battle tested filesystem for my RAID setup. I dont like and will probably never trust ext2/ext3/ext4. I've been burnt too many times in the past by extfs instability. Sad day, I wish Ubuntu would go away from Linux. It used to be an ok system, but now I find that Ubuntu is too broken at handling upgrades and they are going down the crazy path of tablet desktops. 30" monitors will never work well with a UI designed for fingers.

        Comment


        • #5
          Reiserfs has some unique features that are really useful in some situations.
          One of this feature is tail merge that is really helpful when you have a huge number of small files and performance are not a big concern.
          I use reiserfs on the repository of BackupPC; the variable number of inode and the tail merge feature allows for packing a huge number of files and an even bigger number of physical links on a single filesystem without wasting space.
          In the past, when HR was arrested and reiser had an uncertain future, I tried to migrate a 100GB volume from reiserfs to ext3. It used more than 130GB of space when the copy failed for inode exaustion.
          The only linux filesystem that is comparable with reiserfs is btrfs, BUT it's not stable yet (even the disk format is subject to changes) and it's slower for many type of workload.
          Loosing reiserfs would be a big deal, at least until btrfs will become a mature fs.
          Last edited by zipman; 09-12-2013, 05:43 PM.

          Comment


          • #6
            Originally posted by tecknurd View Post
            Probably, someone will take ReiserFS and create a fork of it.
            They'd better change the name then. Maybe just to RFS or something like that. Otherwise, no matter how good it will get, it'll always be remembered as "the murderer's filesystem". There is such a thing as bad publicity...

            Comment


            • #7
              Stay in prison !

              Comment


              • #8
                ReiserFS (v3) on Linux served me faithfully for something like 10 years with acceptable perfromance and never losing a byte. Since the Big Kernel Lock changes after 2.6.32 it was made unstable and this still hasn't been fixed. And btrfs ate a server full of data the first and last time I tried it. So now to new old things: ZFS for me.
                Last edited by stevenc; 09-12-2013, 07:52 PM.

                Comment


                • #9
                  Drop kernel module? No way

                  I can understand dropping support in the installer, but they don't really have to do anything to support the kernel module as that's part of the upstream kernel. Dropping the kernel module will break systems that use ReiserFS when they apt-get dist-upgrade

                  Comment


                  • #10
                    Originally posted by gigaplex View Post
                    I can understand dropping support in the installer, but they don't really have to do anything to support the kernel module as that's part of the upstream kernel. Dropping the kernel module will break systems that use ReiserFS when they apt-get dist-upgrade
                    Despite that the article mentions that the kernel module will not be dropped, even if it would be dropped only systems of administrators break that are dumb enough to upgrade without reading the release notes.

                    Comment


                    • #11
                      Originally posted by Vim_User View Post
                      Despite that the article mentions that the kernel module will not be dropped, even if it would be dropped only systems of administrators break that are dumb enough to upgrade without reading the release notes.
                      No, the article mentions they're considering dropping the kernel module. Current discussions suggest they probably won't, but it's not set in stone.

                      Comment


                      • #12
                        Drop .. no drop... it probably doesn't matter

                        The fact is that everyone is encouraged to move away from reiserfs because it lacks a true maintainer as far as I know.

                        For those that didn't trust reiserfs, I will say this, Jeff M. did a good job of making patches for SuSE/SUSE... and so it was really the only platform for reiserfs IMHO, but given it was the only enteprise distro with journaled filesystem and lvm for such a long time, etc., etc. The SUSE folks made reiserfs work and work even better by adding extended attributes and the like... good job. But alas, that support ended some time ago.

                        Reiserfs is much more of a complex data structure written to disk and therefore bad sectors did tend to make things quite bad. But one could argue that a bad sector in the right place on ext3/4 does equally as bad things (could even be worse), it's just easier to hit a bad place with reiserfs. As long as the disks were good and you didn't cut power, reiserfs has been very stable. But again, on an distro (like Red Hat, for example) where there is disdain for the filesystem, sure... stay as far away from their implementation as possible (unpatched, uncared for, etc...).

                        I've certainly seen my share of weird anomalies with ext2/3/4.... and at least with reiserfs, I could to a rebuildtree and usually get everything back... can't say the same for the others. But in all fairness to compare ext2/3/4 to reiserfs is very much apples to oranges. While a few nice features like extents made it in... those filesystems still show their ancient (and I mean that in a fuddy duddy way) roots.

                        So... love reiserfs, hate reiserfs, the fact is, without good maintenance, it has to die... that's just the way it is. Sure, the kernel devs can keep it limping along, but eventually, it's going to rot.

                        Who knows, maybe Reiser4 will be the "answer"?

                        Comment


                        • #13
                          Originally posted by cjcox View Post
                          Who knows, maybe Reiser4 will be the "answer"?
                          Answer to what? It seems very unlikely to be even be merged upstream. Ext4 works fine for most users. XFS if you are using large amounts of data. Btrfs if you want more of the fancy features. Don't see much of a chance for anything els atm. I am interested in seeing where tuxfs goes however.

                          Comment


                          • #14
                            I understand dropping it from the installer and even from grub and the default kernel image.
                            But for backwards compatibility it should remain as an installable extra module. I still have old disks with reiserfs on them, and being able to plug them and have them work 5-10 years from now is an important feature.

                            Comment


                            • #15
                              Originally posted by stevenc View Post
                              ReiserFS (v3) on Linux served me faithfully for something like 10 years with acceptable perfromance and never losing a byte. Since the Big Kernel Lock changes after 2.6.32 it was made unstable and this still hasn't been fixed. And btrfs ate a server full of data the first and last time I tried it. So now to new old things: ZFS for me.
                              Kconfig snippet for btrfs:

                              Btrfs is highly experimental, and THE DISK FORMAT IS NOT YET
                              FINALIZED. You should say N here unless you are interested in
                              testing Btrfs with non-critical data.
                              Yeah, that doesn't sound like something you should trust your data to. Not that I don't sympathize... we all want the shiny things (automatic integrity checks on read, nearly free snapshots, etc.). Although btrfs is rapidly getting better (heh!), IMO it's not yet ready for production use. (Especially if your production includes multiple disks.)

                              Comment

                              Working...
                              X