The future of HAMMER1
kusumi.tomohiro at gmail.com
Fri Jul 20 21:40:00 PDT 2018
Predrag Punosevac <punosevac72 at gmail.com>:
> I was wondering if Matt and the other developers involved can shed some
> light onto the future of HAMMER1 and ultimately of DragonFly.
> I started dicking with DragonFlyBSD 7-8 years ago mostly due to my
> curiosity about HAMMER. I tried few times in the past to use DF as a
> main file server at the place of my employment just quickly to reverse
> back to ZFS and FreeBSD but my home file server is still happily running
> dfly# uname -a
> DragonFly dfly.bagdala2.net 5.2-RELEASE DragonFly v5.2.2-RELEASE #17:
> dfly# mount
> ROOT on / (hammer, noatime, local)
> devfs on /dev (devfs, nosymfollow, local)
> /dev/serno/B620550018.s1a on /boot (ufs, local)
> /pfs/@@-1:00001 on /var (null, local)
> /pfs/@@-1:00002 on /tmp (null, local)
> /pfs/@@-1:00003 on /home (null, local)
> /pfs/@@-1:00004 on /usr/obj (null, local)
> /pfs/@@-1:00005 on /var/crash (null, local)
> /pfs/@@-1:00006 on /var/tmp (null, local)
> procfs on /proc (procfs, local)
> DATA on /data (hammer, read-only, noatime, local)
> BACKUP on /backup (hammer, noatime, local)
> /dev/serno/5QG00XF0.s1e at DATA on /test-hammer2 (hammer2, local)
> /data/pfs/@@-1:00001 on /data/backups (null, local)
> /data/pfs/@@-1:00002 on /data/nfs (null, NFS exported, local)
> I feel the time to physically rebuild my home file server slowly
> approaching so I would like to be ready for that day.
> In the light of the fact that HAMMER2 is out and recommended for root
> partition does HAMMER have any future? I see Tomohiro Kusumi still
> having a commit or two a month which touches HAMMER but it seems that
> even his commits are minor bug fixes related to his work/poking with
> HAMMER2. Is everyone else done playing with HAMMER1?
HAMMER1 still has several known bugs, and at least one of them (ENOSPC
issue) is hard to fix due to its design.
Also note that several features that were not written by Matt are
incomplete or unstable.
> Is HAMMER2 now stable enough to guarantee consistency of my data? My
> concern is that it typically takes at least 10 years for a file system
> to mature under lot of use and abuse by a much wider community than DF.
> BTRFS is vaporware even after major investment since 2007. With the size
> of the DF community I feel it might takes 20-30 years for HAMMER2 to be
> trully safe.
I agree with 10 years, but HAMMER1 isn't that huge in terms of code
size. It also has a quite simple design. It's about 30% or less than
the size of ZFS, Btrfs, or XFS(on Linux).
> One of the main advantages for me personally of HAMMER1 over ZFS besides
> fine grained history was that it seems to be ok-ish with non-ECC RAM.
> Scrub of death
> is real thing and I feel very uncomfortable using ZFS on my cheap home
> hardware. I know that under non-budgetary constrains I should use ECC
> memory with DF as well. How safe HAMMER2 is with non-ECC memory?
More information about the Users