Announcement

Collapse
No announcement yet.

Reiser5 Issues New Development Release, Performance Numbers For Scaling Out

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

  • #31
    They should change the name. The association with a convicted murderer is a problem.

    Comment


    • #32
      Originally posted by hoohoo View Post
      They should change the name. The association with a convicted murderer is a problem.
      Knowing the open source community they'll rename it to EpsteinFS, you know where else would you install The GIMP to?

      Comment


      • #33
        **Knocks on Door**

        Good morning Brothers and Sisters, I'm here to talk to you about your lord and savior ZFS.

        **Gets Stabbed**

        But, for reals, I checked the Reisier5 patches and it supports Zstd compression now. No shit. When did that happen? IMHO, that gives it a slight advantage over XFS and Ext4 now.

        Code:
        +#if !defined( __FS_REISER4_COMPRESS_H__ )
        +#define __FS_REISER4_COMPRESS_H__
        +
        +#include <linux/types.h>
        +#include <linux/string.h>
        +
        +/* transform direction */
        +typedef enum {
        + TFMA_READ, /* decrypt, decompress */
        + TFMA_WRITE, /* encrypt, compress */
        + TFMA_LAST
        +} tfm_action;
        +
        +/* supported compression algorithms */
        +typedef enum {
        + LZO1_COMPRESSION_ID,
        + GZIP1_COMPRESSION_ID,
        + ZSTD1_COMPRESSION_ID,
        + LAST_COMPRESSION_ID,
        +} reiser4_compression_id;

        Comment


        • #34
          Originally posted by oleid View Post

          Not sure what you did back then, but there was no need to build your own kernel back then. Simply using the distribution default kernel was enough. I took my first baby steps with S.u.S.E. back in the 90s.
          I think it was Suse 7 or something like that. It was a known error/bug, a mistake from Suse. These things happen, and it was specifically because of ReiserFS, other file systems were not affected.

          Comment


          • #35
            Originally posted by skeevy420 View Post
            **Knocks on Door**

            Good morning Brothers and Sisters, I'm here to talk to you about your lord and savior ZFS.

            **Gets Stabbed**

            But, for reals, I checked the Reisier5 patches and it supports Zstd compression now. No shit. When did that happen? IMHO, that gives it a slight advantage over XFS and Ext4 now.

            Code:
            +#if !defined( __FS_REISER4_COMPRESS_H__ )
            +#define __FS_REISER4_COMPRESS_H__
            +
            +#include <linux/types.h>
            +#include <linux/string.h>
            +
            +/* transform direction */
            +typedef enum {
            + TFMA_READ, /* decrypt, decompress */
            + TFMA_WRITE, /* encrypt, compress */
            + TFMA_LAST
            +} tfm_action;
            +
            +/* supported compression algorithms */
            +typedef enum {
            + LZO1_COMPRESSION_ID,
            + GZIP1_COMPRESSION_ID,
            + ZSTD1_COMPRESSION_ID,
            + LAST_COMPRESSION_ID,
            +} reiser4_compression_id;
            Interesting. If it also supports encryption I'm listening.

            Comment


            • #36
              Originally posted by skeevy420 View Post
              **Knocks on Door**

              Good morning Brothers and Sisters, I'm here to talk to you about your lord and savior ZFS.

              **Gets Stabbed**

              But, for reals, I checked the Reisier5 patches and it supports Zstd compression now. No shit. When did that happen? IMHO, that gives it a slight advantage over XFS and Ext4 now.
              WTF?

              XFS/EXT4 are traditional filesystems, that fragment less, so got better performance in long time. Especially XFS is more performant, while EXT4 got user-focused features like case-insenitivity(wine/proton) and encryption (less performant than LUKS and with security bugs in the past)

              Reiser5 is alternative to BcacheFS, ZFS, BTRFS, because it's Copy-On-Write filesystem, with built-in volume manager, etc. Implementing compression, snapshotting, etc would be hard otherwise.
              All above got bitrot protection, but performance drops drastically over time, if filled over 50%(when it reaches 80% it becomes really bad). They also got filesystem-corrupting bugs(BTRFS when there's less than 1% of free space, while ZFS sometimes corrupts checksums when sending raw, encrypted subvolumes. They lowered chance with ZFS v2.1.4, but people confirmed it still happens ).
              So much more complicated code means more bugs, nothing new. but I hope that at least ZFS will be fixed soon, because BCacheFS/Reiser5 are far away from being ready for any serious storage. While BTRFS is getting redesigned now, with important parts rewritten, so it will need probably years to gain stability.

              Comment


              • #37
                Originally posted by tunnelblick View Post

                I think it was Suse 7 or something like that. It was a known error/bug, a mistake from Suse. These things happen, and it was specifically because of ReiserFS, other file systems were not affected.
                What error/bug?

                About kernel it could be:
                - too old or reiseirfs not compiled
                - built-in(so should work if bootloader passed correct options)
                - built as module (kernel rebuilt shouldn't be needed, only proper initrd/initramfs generation, to include such module)

                It covers omly simple cases, because if you've used LVM/LUKS/MDADM, then there could be bugs in init scripts and all this initrd/initramfs generation machinery.

                But if you ever needed to build your kernel (if it wasn't your wrong assumption), then it probably means ReiserFS wasn't supported by SuSe then.
                If installer allowed to graphically mkreiserfs partition and select it as rootfs, then it's some kind of bug, but in the installer (not kernel per se).

                Anyway your original post about abandoning reiserfs, after all this shenanigans with building custom kernel and blaming reiserfs/kernel/SuSe for that is suspicious at least. Custom kernel doesn't sound like "supported" configuration by any means.

                Comment


                • #38
                  Originally posted by evil_core View Post
                  What error/bug?
                  Huh? I never asked for help. Also this happened way more than a decade ago. I just wanted to tell my experience with reiser from the past, I don't need help from others.

                  Comment


                  • #39
                    Originally posted by tunnelblick View Post

                    Huh? I never asked for help. Also this happened way more than a decade ago. I just wanted to tell my experience with reiser from the past, I don't need help from others.
                    But your bad experience doesn't sound it was caused by reiserfs itself at all, but something else.
                    ReiserFS doesn't deserve bad mouth from you.
                    It was simple, stupid filesystem(without volume-manager built-in) and if it has been pointed to proper device containing it, it would mount.

                    My intention wasn't personal attack on you, but pointing that it wasn't reiserfs fault (but most probablt user fault).

                    In reiserfs times, I was PLD linux distro developer and fixed init scripts generation, related to LVM over LUKS and few other things, so I really know how it worked then(but I feel stupid and lost with SystemD machinery now ;-)

                    Comment


                    • #40
                      Originally posted by F.Ultra View Post

                      The proper acronym for those are SIC: Social Injustice Crusaders.
                      I like that one.

                      Originally posted by F.Ultra View Post

                      Actually the 90% developers that don't want anything to do with it is not because of the murder, that happened long before due to Hans being an outright difficult person to discuss things with, the flame wars on LKML between him and the other kernel devs where long and hard. The murder was just the final straw that broke the camels back.
                      As is most often the case. The use case also probably got less interesting with changes in the tech world and other emerging alternatives, tho I don't really know the details of reiserfs to be sure.

                      Originally posted by F.Ultra View Post
                      And yet we would probably be able to find some name that would have you wanting to rename it. Also renaming something is not cancelling anything.
                      We could actually take this as a chance to use a name that describes the filesystem instead of its author. It's generally useful to name things after a reduction of what they do.

                      Comment

                      Working...
                      X