The future of HAMMER1

Jasse Jansson jasse at yberwaffe.com
Sat Jul 21 11:57:43 PDT 2018



On 2018-07-21 01:26, my123 (@never_released) wrote:
> Hi,
>
> It's worth noting that the cheaper Pentium parts support ECC such as 
> the Pentium G4560.
Yeah, I know. I got the i3 kinda cheap, that's why I bought it.
>
> Personally, I believe that the ZFS scrub story is a bit overblown for 
> non-ECC systems, never had an issue with it.
I agree with this, and the "myth" that ZFS RaidZ1 ain't safe and you 
should use RaidZ2 or higher.
There's some paranoia running wild in the FreeNAS community.
>
>
> On 21/07/2018 00:12, Jasse Jansson wrote:
>>
>>
>> On 2018-07-20 23:12, Xin LI wrote:
>>> On Fri, Jul 20, 2018 at 2:07 PM Rickard von Essen
>>> <rickard.von.essen at gmail.com> wrote:
>>>> Both "scrub of death" and "you have to use ECC ram when you use 
>>>> ZFS" are myths.
>>> +1.  ECC is not a requirement for ZFS, and ZFS only rewrites data when
>>> it sees checksum mismatch AND the replica were correct, so the chance
>>> of scrub damaging data is extremely low because you need at least two
>>> flipped bits in the first place.
>>>
>>> You do want to use ECC RAM if the data is valuable though; the small
>>> price difference will definitely worth it if you ever see a corruption
>>> that you can't recover from.
>>
>> In my experience "I" am the biggest threath to my stored pics and 
>> other stuff I stored in my servers in my basement.
>> I have accidently countersunk an external HD used for backups twice 
>> in the last 10 years, and I doubt I'll have ever been so close to a 
>> lethal heart attac, twice again.
>> Both times all the files could be restored using a data recovery 
>> program I can't remember the name of right now.
>> I don't take that risk anymore.
>>
>> My main server runs FreeNAS with a 5-disk raidZ volume (ECC RAM), 
>> scrubbed twice every month.
>> My backup server also runs FreeNAS with a 5-disk RaidZ volume (ECC 
>> RAM), also scrubbed twice every month.
>> Both servers are attached to UPS power.
>> The backup server rsynch's the whole volume every third night.
>> I log in regularly to my servers to check they are OK.
>> I also make sure to replace the HD's every 3 years.
>>
>> You really think I would risk another near death experience by 
>> running NON-ECC RAM for the little price difference there is today.
>> But it's your pics/data/files and your choice.
>>
>> Oh, btw, if you are going to buils a new server with ECC memory, make 
>> sure the CPU supports it, the latest i3's don't.
>>
>>>
>>>> One good discussion on the topic can be found on:
>>>> https://reddit.app.link/QOrzkoipIO
>>>>
>>>> On Fri, Jul 20, 2018, 22:37 Predrag Punosevac 
>>>> <punosevac72 at gmail.com> wrote:
>>>>> Hi,
>>>>>
>>>>> 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?
>>>>>
>>>>> 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.
>>>>>
>>>>> 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
>>>>>
>>>>> https://forums.freenas.org/index.php?threads/ecc-vs-non-ecc-ram-and-zfs.15449/ 
>>>>>
>>>>> 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?
>>>>>
>>>>> Cheers,
>>>>> Predrag
>>
>
>



More information about the Users mailing list