Improved Btrfs Scrub Code Readied For Linux 6.4, ~10% Faster

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • caligula
    Senior Member
    • Jan 2014
    • 3309

    #11
    Originally posted by S.Pam View Post

    I suspect its a comment made in jest. In my experience, Btrfs works really well, both on my private systems and on servers at work.
    Yes, but the scrub has been around 50% slower than ZFS. So.. it takes ages with larger arrays of 60+ TB.

    Comment

    • S.Pam
      Senior Member
      • Oct 2018
      • 676

      #12
      Originally posted by caligula View Post

      Yes, but the scrub has been around 50% slower than ZFS. So.. it takes ages with larger arrays of 60+ TB.
      How so? Are you perhaps trying to scrub a RAID56 filesystem?

      Comment

      • guglovich
        Senior Member
        • Jul 2017
        • 289

        #13
        A drop in the sea. We need to work towards fast partition mounting.

        Comment

        • cynic
          Senior Member
          • Oct 2011
          • 1082

          #14
          Originally posted by guglovich View Post
          A drop in the sea. We need to work towards fast partition mounting.
          they're working on it. Extent tree v2 should solve the issue, but it won't come too soon.

          Comment

          • waxhead
            Premium For Life
            • Jul 2014
            • 1138

            #15
            Originally posted by guglovich View Post
            A drop in the sea. We need to work towards fast partition mounting.
            Already solved with the block tree feature.

            http://www.dirtcellar.net

            Comment

            • caligula
              Senior Member
              • Jan 2014
              • 3309

              #16
              Originally posted by S.Pam View Post

              How so? Are you perhaps trying to scrub a RAID56 filesystem?
              Yeah, it's a waste of space to use RAID1 in a home NAS.

              Comment

              • zxy_thf
                Senior Member
                • Mar 2012
                • 621

                #17
                Originally posted by GreenReaper View Post
                Perhaps, but can you convert from one to another, and will people actually do that? If not there will still be a lot of old filesystems around.
                It's not a huge trouble since hard drives have lifespans
                I "converted" my XFS volumes to btrfs when replacing my old hard drive, due to lack of data checksum, and I'm planning to change the default hashing algorithm to xxhash when replacing my current btrfs volumes, which will happen a few years from now.

                Comment

                • fitzie
                  Senior Member
                  • May 2012
                  • 672

                  #18
                  Originally posted by waxhead View Post

                  Already solved with the block tree feature.
                  yes, you need to compile btrfs userspace with an enable-experimental flag to create/convert btrfs filesystems. I did that, and it took my 50+TB filesystem mount from 4 minutes to a few seconds. Btrfs is pretty good these days.

                  Comment

                  • geearf
                    Senior Member
                    • Dec 2011
                    • 2147

                    #19
                    Originally posted by fitzie View Post

                    yes, you need to compile btrfs userspace with an enable-experimental flag to create/convert btrfs filesystems. I did that, and it took my 50+TB filesystem mount from 4 minutes to a few seconds. Btrfs is pretty good these days.
                    Oh wow, maybe I won't need to use bcache to speed mounting time anymore.

                    Comment

                    • S.Pam
                      Senior Member
                      • Oct 2018
                      • 676

                      #20
                      Originally posted by caligula View Post

                      Yeah, it's a waste of space to use RAID1 in a home NAS.
                      Then I suggest you scrub each drive individually instead of the whole filesystem. It is an unfortunate 'bug', as when you scrub the filesystem, btrfs will initiate a scrub on each device in parallel, but with striped profile, it needs to read all stripes, which are on the other drives. This will cause a massive slowdown, especially on rotating disks. The solution is to scrub one device at the time.

                      Comment

                      Working...
                      X