Announcement

Collapse
No announcement yet.

Btrfs Gets RAID 5/6 Fixes With Linux 4.12

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

  • Btrfs Gets RAID 5/6 Fixes With Linux 4.12

    Phoronix: Btrfs Gets RAID 5/6 Fixes With Linux 4.12

    There are a number of Btrfs fixes/clean-ups for the Linux 4.12 kernel...

    http://www.phoronix.com/scan.php?pag...rfs-Linux-4.12

  • #2
    Was getting ready to build a server sometime later this year (hopefully in 2ish months) based on a nice little ryzen5 processor. I wanted to use BTRFS, but with the instability of RAID5, was going to use ZFS or mdadm instead.

    https://btrfs.wiki.kernel.org/index.php/Status

    Should these listed BTRFS fixes bring it to a usable home-use state? I looked though the changelog/pull request, but I'm not really sure what I'm looking at.

    Comment


    • #3
      I honestly wouldnt use it just yet. Not until it's taken at least a couple of months of hard testing. I'm patiently waiting for this to become stable as well.

      Comment


      • #4
        Originally posted by tiwake View Post
        Was getting ready to build a server sometime later this year (hopefully in 2ish months) based on a nice little ryzen5 processor. I wanted to use BTRFS, but with the instability of RAID5, was going to use ZFS or mdadm instead.

        https://btrfs.wiki.kernel.org/index.php/Status

        Should these listed BTRFS fixes bring it to a usable home-use state? I looked though the changelog/pull request, but I'm not really sure what I'm looking at.
        Wow, given those "unstable"s I'd say definitely do not use except if you want to help test.

        Comment


        • #5
          RAID 1 or 10 has always worked really well for me. There isn't much need for RAID 5 in my opinion. RAID 6 would be nice though, for the two drive failure protection.

          Comment


          • #6
            Originally posted by Swiftpaw View Post

            Wow, given those "unstable"s I'd say definitely do not use except if you want to help test.
            BTRFS is already in a usable home-use state, just don't use raid5 or raid6.

            Comment


            • #7
              Originally posted by tiwake View Post
              Was getting ready to build a server sometime later this year (hopefully in 2ish months) based on a nice little ryzen5 processor. I wanted to use BTRFS, but with the instability of RAID5, was going to use ZFS or mdadm instead.

              https://btrfs.wiki.kernel.org/index.php/Status

              Should these listed BTRFS fixes bring it to a usable home-use state? I looked though the changelog/pull request, but I'm not really sure what I'm looking at.
              Don't use RAID5/6, it's still not ready.

              Other features work fine, if you don't overdo them (making hundreds of snapshots or something).

              Comment


              • #8
                Even raid 1 has problems. "Lost" 40 GB of data just a couple weeks ago. The data was for some reason unrecoverable, even though no hardware failure. It just corrupted itself magically after a while, me doing nothing exotic.

                Comment


                • #9
                  Originally posted by AndyChow View Post
                  Even raid 1 has problems. "Lost" 40 GB of data just a couple weeks ago. The data was for some reason unrecoverable, even though no hardware failure. It just corrupted itself magically after a while, me doing nothing exotic.
                  Sounds weird. I've been using BTRFS on my home server since 2012. I'm up to 30 TB of disk on six drives now using RAID10. I've replaced two drives in that time. It's been really good about catching errors.

                  I run weekly scrubs and weekly SMART long self-tests.

                  If you aren't running scrubs, and errors develop on two drives before you notice them, you can lose the mirror. But that's why BTRFS has scrub, MD has scrub and hardware RAID controller have scrub / patrol read. It's necessary.

                  Or if you have a RAM error, the checksums can all fail. If it is a single bit error, BTRFS will reconstruct from the other mirror. But if you have a whole RAM stick that is corrupting data output in multiple locations, it can convince BTRFS (or ZFS for that matter) that the whole mirror is corrupt because all checksums are bad. Although if your RAM is that screwed up I'd expect a complete system failure with tons of random program crashes and kernel OOPS.

                  Comment


                  • #10
                    Originally posted by AndyChow View Post
                    Even raid 1 has problems. "Lost" 40 GB of data just a couple weeks ago. The data was for some reason unrecoverable, even though no hardware failure. It just corrupted itself magically after a while, me doing nothing exotic.
                    Did you run scrub on that partition with some kind of schedule?
                    I come from the world of hardware RAID so I always set up for scheduled scrubbing, never had issues on a few TB of RAID1 or on my workstation's main/data partitions (also RAID1).

                    Comment

                    Working...
                    X