Announcement

Collapse
No announcement yet.

Btrfs For Linux 5.1 Brings Configurable Zstd Compression Level, A Number Of Fixes

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

  • #21
    Originally posted by k1e0x View Post
    RAID 5, your right it's dead due to exponential risk of URE and RAID 6 isn't far behind on this problem. Although RAID 7-3 + (RAIDZ3) is a solid replacement for that scheme.
    RAIDZ1/2 is also fine, as it's aware of what data was lost and won't just rescan the whole array for lulz like a block-level raid controller (or mdadm) will do.

    Just as is Btrfs RAID1 or RAID10. Still not scanning the whole array because it know what was lost and where is the parity for it.

    Comment


    • #22
      We're finally getting that compression level switch, it's been too long!
      If it's configurable per folder, I won't even need to compress any backup anymore this will do just fine and will auto decide if it's worth it or not.
      I'm finally excited for another Linux release, alas too many weeks away :/

      Comment


      • #23
        Originally posted by geearf View Post
        We're finally getting that compression level switch, it's been too long!
        If it's configurable per folder, I won't even need to compress any backup anymore this will do just fine and will auto decide if it's worth it or not.
        I'm finally excited for another Linux release, alas too many weeks away :/
        Indeed. And what's more: there's finally some work regarding that nasty performance regression (as in total lockup) when balancing a volume with qgroups (aka regular openSUSE install). Hopefully, no more laptop freezes at the whim of the scheduled btrfs maintenance task! I'm happy.

        Comment


        • #24
          Originally posted by useless View Post

          Indeed. And what's more: there's finally some work regarding that nasty performance regression (as in total lockup) when balancing a volume with qgroups (aka regular openSUSE install). Hopefully, no more laptop freezes at the whim of the scheduled btrfs maintenance task! I'm happy.
          Wooo that's pretty bad, I'm happy I have not suffered that yet.

          Comment


          • #25
            Originally posted by geearf View Post
            Wooo that's pretty bad, I'm happy I have not suffered that yet.
            I have 3 regular OpenSUSE installs on btrfs for years and I've yet to suffer total lockups of any kind.

            Comment


            • #26
              Originally posted by starshipeleven View Post
              I have 3 regular OpenSUSE installs on btrfs for years and I've yet to suffer total lockups of any kind.
              That's the thing about that: not everyone could reproduce it. Devs spent a great time replicating the circumstances that triggered the issue. My workstation doesn't suffer from it, neither my wife PC. The only computer in which I have ever seen it is my laptop.

              Comment


              • #27
                Originally posted by useless View Post
                That's the thing about that: not everyone could reproduce it. Devs spent a great time replicating the circumstances that triggered the issue. My workstation doesn't suffer from it, neither my wife PC. The only computer in which I have ever seen it is my laptop.
                Did you try to swap the drive? I guess you already looked at dmesg for HDD issues.

                Last time I had a PC lock up like that was a HDD issue.

                Comment


                • #28
                  Originally posted by starshipeleven View Post
                  Did you try to swap the drive? I guess you already looked at dmesg for HDD issues.

                  Last time I had a PC lock up like that was a HDD issue.
                  It's a confirmed bug: https://bugzilla.suse.com/show_bug.cgi?id=1063638
                  It happens when a balance is issued in a volume with lots of group quotas IIRC.

                  Comment


                  • #29
                    Originally posted by useless View Post

                    It's a confirmed bug: https://bugzilla.suse.com/show_bug.cgi?id=1063638
                    It happens when a balance is issued in a volume with lots of group quotas IIRC.
                    Lots of group quotas = Snapper snapshots in this case

                    As that's what it is used for in OpenSUSE and what anyone that isn't on a server is going to have in his system.

                    Still, how many are yours? These are mine from my work laptop, for example.

                    Code:
                    hpprobook:/# btrfs qgroup show /
                    qgroupid         rfer         excl
                    --------         ----         ----
                    0/5          16.00KiB     16.00KiB
                    0/257        16.00KiB     16.00KiB
                    0/258       937.14MiB    937.14MiB
                    0/259        16.00KiB     16.00KiB
                    0/260         1.36GiB      1.36GiB
                    0/261        60.00KiB     60.00KiB
                    0/262        16.58MiB     16.58MiB
                    0/263        61.25MiB     61.25MiB
                    0/264         3.38MiB      3.38MiB
                    0/265        16.00KiB     16.00KiB
                    0/266         6.16MiB      6.16MiB
                    0/267         9.63GiB    228.20MiB
                    0/583         9.42GiB      2.31GiB
                    0/584         9.58GiB    110.25MiB
                    0/630         9.74GiB    187.60MiB
                    0/633        10.21GiB    317.83MiB
                    0/711         9.59GiB      2.71MiB
                    0/716         9.59GiB      1.67MiB
                    0/718        10.05GiB    138.32MiB
                    0/721           0.00B        0.00B
                    0/722           0.00B        0.00B
                    0/723           0.00B        0.00B
                    0/726         9.70GiB     95.86MiB
                    0/727         9.70GiB    160.00KiB
                    0/728         9.71GiB    288.00KiB
                    0/729         9.70GiB     80.00KiB
                    0/730         9.70GiB     80.00KiB
                    0/731         9.70GiB    316.00KiB
                    0/732         9.70GiB      1.45MiB
                    0/733         9.68GiB      4.73MiB
                    0/734         9.63GiB    464.00KiB
                    0/735         9.63GiB    468.00KiB
                    0/736         9.63GiB    144.00KiB
                    0/737         9.63GiB    576.00KiB
                    0/738           0.00B        0.00B
                    0/739           0.00B        0.00B
                    0/740           0.00B        0.00B
                    0/741           0.00B        0.00B
                    0/742           0.00B        0.00B
                    0/748         7.84GiB      7.84GiB
                    1/0          19.45GiB     10.04GiB

                    Comment


                    • #30
                      Originally posted by starshipeleven View Post

                      Lots of group quotas = Snapper snapshots in this case

                      As that's what it is used for in OpenSUSE and what anyone that isn't on a server is going to have in his system.

                      Still, how many are yours?
                      Oh, yes, I know. I was just referring to it in a more generic way.

                      Regarding 'how many': I can't remember specifically but it was something like 20-30 groups (which translates to only a couple of snapshots). I don't have my laptop with me now. I tough it was my SSD (a Samsung 840 EVO: it had some performance issues which were fixed long time ago with a firmware update) so naturally I changed my drive but no improvement. The only thing that occurs to me is that thermal throttling is hitting hard on a balance: some OEMs really sucks at designing/building computers; and that's even not counting all the nasty thing done in the UEFI firmware.

                      EDIT: My dad's laptop (an ancient one, btw) never showed that system stall. A sane balance (-dusage=50 -musage=50 to start) would complete in 15 min and the only thing that pointed to a running balance was some stuttering in KDE, nothing unbearable.
                      Last edited by useless; 05 March 2019, 02:08 PM.

                      Comment

                      Working...
                      X