Announcement

Collapse
No announcement yet.

XFS Feature Used For Online Fsck Graduates From Experimental

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

  • XFS Feature Used For Online Fsck Graduates From Experimental

    Phoronix: XFS Feature Used For Online Fsck Graduates From Experimental

    Last week the main pull request of XFS file-system driver updates for the Linux 4.16 was sent in and referred to by XFS maintainer Darrick Wong as having great scads of new stuff. The Oracle engineer has now sent in a secondary pull request of XFS for Linux 4.16...

    http://www.phoronix.com/scan.php?pag...x-4.16-Round-2

  • #2
    Regarding these awesome and ongoing developments for XFS, I always wondered if, when I created my XFS stores on a 4.4 kernel, how much benefit and updates does that filesystem receive when I install a newer kernel? Does the existing filesystem receive some sort of updates also after being mounted? What happens if I go back to an older kernel after doing some writes in a drastically new kernel?

    Comment


    • #3
      Originally posted by stiiixy View Post
      Regarding these awesome and ongoing developments for XFS, I always wondered if, when I created my XFS stores on a 4.4 kernel, how much benefit and updates does that filesystem receive when I install a newer kernel? Does the existing filesystem receive some sort of updates also after being mounted? What happens if I go back to an older kernel after doing some writes in a drastically new kernel?
      I don't know about XFS in particular but most filesystems use feature flags in their on-disk format so that the kernel and userspace tools can detect what features are enabled and refuse mounting (or operating on) partitions with unknown feature flags.

      Newer kernels will not alter the filesystem configuration or do anything that isn't retro-compatible. Which means that they won't enable new features.

      If you want to change the filesystem settings, you must do so with userspace tools on an unmounted filesystem. For XFS it seems this is the one https://linux.die.net/man/8/xfs_admin
      While for ext2-3-4 you would use tune2fs http://www.linuxdevcenter.com/cmd/cm...path=t/tune2fs

      Comment


      • #4
        Okay, that starts to clear things up a bit. Thanks mate.

        Comment


        • #5
          How about data checksums, that's all I really care about besides reliability and usable performance. I want protection from bitrot and to the extent possible data corruption and efficient mirroring. I'm not sure ZFS is really going anywhere, I found BTRFS to be a useless toy, and right after I finally switched to Win 10 for my primary photo processing computer just so I could add REFS mirroring MS dropped full REFS support. I don't really care about dedup because it's not worth the risk for me but could see how it would be a very big deal. I do care about portability and OS compatibility, would be great to process and store with the same file system and reliably access data from MS, Linux and BSD. Thought about Mac and ZFS for now switching to APFS after it's proven but their stance on Data checksums is arrogant and idiotic.

          Comment

          Working...
          X