Announcement

Collapse
No announcement yet.

OpenZFS 2.2 Released With Block Cloning, Linux Container Support & Better Performance

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

  • #21
    Originally posted by Chugworth View Post
    What would be the advantage of selecting blake3 rather than the default fletcher4?
    Hello, I am trying to understand the benefits of blake3 vs. fletcher4 checksums. But I dont get it. What is a good reason to move to blake3? Performance wise I believe fletcher4 is still faster. Is...

    Comment


    • #22
      Originally posted by mbod View Post

      Hello, I am trying to understand the benefits of blake3 vs. fletcher4 checksums. But I dont get it. What is a good reason to move to blake3? Performance wise I believe fletcher4 is still faster. Is...
      Well yeah, I've seen vague answers like "used for encryption" but that doesn't really explain how encryption would take advantage of it. Wouldn't the blocks of data and their checksums be at a lower level than everything the encryption is doing? And how would it be a benefit to dedup?

      Comment


      • #23
        Originally posted by Chugworth View Post
        Well yeah, I've seen vague answers like "used for encryption" but that doesn't really explain how encryption would take advantage of it. Wouldn't the blocks of data and their checksums be at a lower level than everything the encryption is doing? And how would it be a benefit to dedup?
        Why dont you just participate in the discussion on https://github.com/openzfs/zfs/discussions/15421 ?
        You can ask all your questions there.

        Comment


        • #24
          Originally posted by wapsi View Post
          That "Block cloning" is very welcome improvement, and a game changer / trigger to me to migrate my BTRFS volumes to ZFS:

          Before this OpenZFS version and that "Block cloning" feature, the cp --reflink'ing didn't work with ZFS when I was copying data within NFS shares which were ZFS volumes at the server.

          And also: I verified that NFS Server-Side Copying is working with this OpenZFS version (it didn't work before). Though, I had to update my Debian 12 coreutils package from the stock 9.1 version to 9.4, and then:
          Code:
          cp /mnt/nfs/zfs-volume0/testdata1 /mnt/nfs/zfs-volume1/testdata1
          used the NFS Server-Side Copying on OpenZFS correctly: the copied data didn't round trip via the NFS client anymore! ... Plus: I verified the same oCn KDE's Dolphin file manager (the stock version on Debian 12): it worked correctly too: the copied data didn't round trip via the NFS client anymore.

          Really cool!
          Thanks for the info re. NFS, appreciated!

          Please note some particulars regarding datasets you may find useful: https://github.com/openzfs/zfs/pull/...ent-1742172842

          Comment


          • #25
            Originally posted by Ranguvar View Post

            Thanks for the info re. NFS, appreciated!

            Please note some particulars regarding datasets you may find useful: https://github.com/openzfs/zfs/pull/...ent-1742172842
            Thanks.

            I also recommend to read this thread: https://www.truenas.com/community/th...th-zfs.103309/ It's about the Linux coreutils' cp and its reflink related changes / improvements in the recent coreutils versions. The --reflink=auto over NFS on this OpenZFS 2.2 version fallbacked to normal copying on Debian 12 coreutils version 9.1 (https://packages.debian.org/source/bookworm/coreutils). But after I compiled/upgraded the coreutils package manually to 9.4 version the --reflink=auto really used reflinking over NFS+OpenZFS 2.2.

            Comment

            Working...
            X