Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 06 Feb 2021 17:33:03 +0100
From:      Walter von Entferndt <walter.von.entferndt@posteo.net>
To:        Abner Gershon <6731955@gmail.com>
Cc:        freebsd-geom@freebsd.org
Subject:   Re: gjournal, turn off automatic synchronization clarification
Message-ID:  <2023081.RhTPgMbj8J@t450s.local.lan>
In-Reply-To: <mailman.63.1612612801.46298.freebsd-geom@freebsd.org>
References:  <mailman.63.1612612801.46298.freebsd-geom@freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
At Samstag, 6. Februar 2021, 13:00:01 CET Abner Gershon 
<6731955@gmail.com> wrote:
> Appreciate your advice. There seems to be overwhelming enthusiasm for
> ZFS.

This enthusiasm is justified.  ZFS offers many advantages for your 
setup.

> Maybe I am swimming against the current leaning toward UFS. My
> reasons are: 1. Have relied on dump backups to LTO tape for the past
> decade and am very comfortable with dump and restore. Sure, tar would
> not be hard to learn but will it reliably handle the samba files with
> names like "Bob's ideas about marketing.doc" correctly?

You can use find(1) & cpio(1)'s -0 option to use zero-terminated 
filenames.

> 2. I have 72 GB ram but am planning to run a Windows guest and Linux
> guest with Oracle database on behyve as well as a few jails.
> Concerned ZFS will eat up too much memory.

Adjust the appropiate sysctl(8) knobs to restrict ZFS ARC cache size.  
ZFS dedup & cloning features are especially useful when working with 
jails and VMs.  For Linux guests you don't need a VM anymore, Linux-
branded jails are now in STABLE, i.e. about to come latest with 13-REL.  
Maybe they are already in 12.2, I don't know.


> 3. Not really convinced that "bit-rot" should be a concern. I
> understand it is real. But, in the past 15 years I can't recall
> coming across a single corrupted data (pdf, word doc, ledger, mp3,
> etc) file.

I understand that walking over the street when the traffic lights show 
red is dangerous.  But I never had an accident for over 40 years doing 
that...

> I currently manage about 4TB of data and it grows by about
> 500 or 600 GB per year. Have been using ext3 and ext4 on debian linux
> for the past 15 years.

Use 3+-way mirrors or RAID with double/triple parity on disks >8 TB.  
The reason is that on such large disks, the likelyhood of unrecoverable 
& undetected bit errors is astonishingly high.  Without a checksum, 
these are undetected.  ZFS offers strong checksums from platter to 
applications memory.
-- 
=|o)  "Stell' Dir vor es geht und keiner kriegt's hin." (Wolfgang Neuss)





Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?2023081.RhTPgMbj8J>